摘要:

Knowledge-based Computer System for the Detection of Matrix Interferences in Atomic Absorption Spectrometric Methods* Journal of Analytical Atomic Spectrometry I I WIM PENNINCKX PETER VANKEERBERGHEN D. LUC MASSART AND JOHANNA SMEYERS-VERBEKE ChemoAC Pharmaceutical institute Vrije Universiteit Brussel Laarbeeklaan 103 1090 Brussels Belgium A knowledge-based computer system which assists with the detection of matrix interferences during the validation of atomic absorption spectrometry (AAS) is described. The system compares the slopes of a standard addition line and an aqueous calibration line. An experimental design is proposed which guarantees that the probability of not detecting important interferences (/I-error) is acceptably low. To permit a detection of interferences by the comparison of curved calibration lines a linearization is proposed. The system is developed in a Windows environment and is one module of a more complete hypertext system for the final validation of AAS methods.Keywords Atomic absorption spectrometry; method validation; matrix interferences; computer systems Validation is one of the most important stages in the develop- ment of atomic absorption spectrometry (AAS) methods. It is the process by which the suitability is demonstrated of a method for providing useful analytical data.' In practice it is necessary to specify the requirements of the method and then to demonstrate that the method meets these requirements. An essential part of the validation is verification that the method is not importantly biased.2 In order to achieve this one must specify which bias can be accepted and then carry out experi- ments to investigate whether the actual bias does not exceed this specification.The best way to estimate the total bias of a method is to analyse a reference sample with a known concen- tration and to compare the mean measurement result with the true value. Unfortunately for many sample types no reference materials are available so that frequently a material with a chemical composition approximating that of the sample to be analysed must be used. However use of a different material decreases the conclusive force of the test. Moreover reference materials generally only permit validation at one concentration level while frequently a method must be vali- dated over a wider range. To compensate for these short- comings the analysis of reference materials has to be combined with recovery experiments on real samples that are spiked at different concentration levels in an early stage of the sample treatment (before the digestion for example).Both of these methods determine the total systematic error at a limited number of concentration levels (generally one to three) and require replicate analyses (Le. the complete pre-treatment and measurement) at each of these levels which is clearly a time- consuming procedure. Before determining the total bias it is recommended to trace a number of specific error sources. This often permits the early detection of an important bias with only a limited number of experiments. One of the main problems in spectrometry is a different sensitivity for the sample solution and the matrix-free (aqueous) solution due to matrix interferences that introduce * Presented at the Seventh Biennial National Atomic Spectroscopy Symposium (BNASS) Hull UK July 20-22 1994.relative systematic errors. These interferences affecting the sensitivity can be traced by one-point recovery experiments or by comparing the slopes of a standard addition line and an aqueous calibration line. The advantage of the latter is that it allows the detection of relative errors over the complete calibration range. The approach identifies problems that occur during the AAS measurement so that the additions can be performed after the pre-treatment of a single sample. If matrix interferences introduce an unacceptable bias in the method a solution to this problem must be found before continuing with the validation.On the other hand when the sensitivity of the method for sample and matrix-free solutions is similar the validation experiments can be continued. Although the method for the comparison of slopes is com- monly applied little attention is paid to the experimental design i.e. the number of measurement points and the distri- bution of these points over the calibration range. In this work a strategy is presented that advises which experiments should be performed to detect matrix interferences with satisfactory confidence. Another problem is that no strategy is available for the detection of interferences when curved calibration lines are used although some authors3 demonstrated that working in the non-linear part of the calibration line can offer more precise results.Here a method for the exploratory comparison of curved calibration lines is proposed. In order to present the strategy for the detection of matrix interferences in an attractive and user-friendly way it was implemented in a knowledge-based computer system. The system advises the analyst on the experimental design to be used it enables experimental data to be entered and saved performs the statistical tests on these data and summarizes the results in a report. This permits the analyst to document the method validation which contributes to the quality assurance of the produced data. The presented system is one module of a more complete system for the final validation of AAS methods. INSTRUMENTATION All programming was performed on a Compaq ProLinea 4/25s Personal Computer.Visual Basic 3.0 (Microsoft) was used as the programming environment. A Perkin-Elmer Zeeman 3030 atomic absorption spectrometer equipped with an HGA-600 graphite furnace an AS-60 autosampler and a PR-100 printer were used for graphite furnace (GF) AAS determinations. For flame-AAS determinations a Perkin-Elmer 373 spectrometer with a PRS-10 printer sequencer were used. DEVELOPMENT OF A STRATEGY FOR THE DETECTION OF INTERFERENCES USING A STRAIGHT LINE (FIRST DEGREE) CALIBRATION MODEL The general strategy for the detection of matrix interferences which is based on the comparison of the slopes of a standard Journal of Analytical Atomic Spectrometry March 1995 Vol.10 207addition line and an aqueous calibration line is summarized in Fig. 1. The strategy starts from the assumption that the maximum systematic error that can be accepted is defined and that an appropriate calibration model is used. Preliminary Information When matrix interferences are traced the systematic error is generally expressed in a relative way. One can state for example that interferences are considered unacceptable if they introduce a bias of more than 5%. Of course this limit depends on the objective of the analytical method. For example the requirements will be more strict for a method for the determi- nation of an active compound in pharmaceuticals than for a method used to obtain a rough estimation of the nutrient concentration in food.Therefore advice on the specification of the maximum acceptable relative error is given in a separate module of the knowledge-based system which is not discussed in this paper. Here situations will be considered where a maximum relative error of 5 or 10% can be accepted. In order to check the suitability of the calibration procedure one must check in the first place whether an unweighted model can be used (this is the case when the variance is constant over the complete calibration range) or whether it is necessary to weight Subsequently the goodness-of-fit of the calibration model must be demonstrated. For an unweighted model one can compare the error due to lack-of-fit with the pure exper- imental error using an analysis of ~ a r i a n c e ~ while an alterna- tive test which compares the sum of the weighted squared residuals with a theoretical X2-value can be applied for a weighted model.' Although the validation of the calibration procedure is handled in another module of the knowledge- based system (which was not discussed here) it is important to note that the experiments that are performed for that purpose permit the estimation of the slope of the calibration line (b,) and the residual variance (sYx2).This information is used in the development of the experimental design for the detection of interferences. The calibration range within which interferences Objective of the method c Validation calibration prccedure Required performance parameters css - Development experimental design Performance experiments Conclusion on matrix interfaces Detection matrix interfaces Fig.1 AAS determinations when a straight line calibration model is used General strategy for the detection of matrix interferences in are traced is determined during the method development. Often calibration standards are prepared over the complete concen- tration range where the Lambert-Beer law is valid. Development of the Experimental Design a- and p-errors Matrix interferences affecting the sensitivity are traced by comparing the slopes of a standard addition line (B,) and an aqueous calibration line (PI). The true slopes P1 and P are estimated by b and bZ which are computed from a limited number of measurement results and are normally distributed around the true values. The comparison is made by a t-test (Ho p1 = P,; HI p # /I2).The test is two-sided since inter- ferences can result in an increased as well as a decreased sensitivity. The selection of the tabulated t-value depends on the probability a one allows of falsely detecting a difference between the slopes (a-error). However there is also a prob- ability B that a real difference between the slopes is not detected. Although it is very important the p-error is often neglected. In this work it is shown how the probability of obtaining a false test result due to an a- and p-error can be controlled by the selection of a suitable experimental design. Thus starting from the probability one allows of making a false positive and a false negative conclusion advice is given on the number of measurements that must be performed and on the distribution of the measurement points over the concen- tration range.t-Test for comparison of the slopes To test whether a significant difference between the two slopes exists the following t-value is calculated where s(bl - bz) = J s b 1 2 + sb22 (2) where sb12 and sb are the variances of bl and b respectively. In Appendix Ia and Ib it is explained how s ( ~ ~ - ~ ~ ) can be computed respectively for an unweighted and a weighted calibration model When tcal exceeds the theoretical t-value ta,2 (see Comparison of the slopes of the calibration lines) at a significance level a = 5% it can be concluded with 95% confi- dence that both lines have a different slope. Therefore if (b - b,) exceeds k = ta s(bl - b2) a significant difference in both slopes has been detected.This is illustrated in Fig. 2(a) which shows the distribution of the estimated difference (b - b,) when the real difference (PI - P2) equals zero. As represented in Fig. 2(b) which in addition shows the distribution of (b - b2) for the situation in which ( p1 - P,) differs from zero there is a probability j? of not detecting a real difference (6 = p1 - p 2 ) between both slopes. Therefore as follows from Fig. 2(c) if a certain difference dmaX must be detected with a specified a and P error (ta/,+ tp)*s(b1-b2)<lP1 -P2lmax (3) or (4) for an unweighted calibration model (see Appendix Ia) and 208 Journal of Analytical Atomic Spectrometry March 1995 Vol. 10I I I Fig. 2 (a) Distribution of the estimated difference between the slopes (b - b2) when the real difference (Dl - p2) is zero.Owing to the spread on (b - b2) which is measured by s ( ~ ~ - ~ ~ ~ a probability a exists of falsely detecting a significant difference in slope. (b) Distribution of the estimated difference between the slopes (b - b2) when no difference exists (PI -p2 = 0 full line) and when the slopes really differ ( p1 - flz # 0 dotted line). A probability /? exists that a real difference between slopes is not detected. (c) Detection of a difference (PI - P2)nax with a specified a and /3 error for a weighted calibration model (see Appendix Ib). These expressions can be used for the development of a suitable experimental design. Application for the development of the experimental design Heteroscedasticity was not detected in the linear calibration ranges measured in this laboratory thus expression (4) was used for the establishment of the experimental design.It was assumed that the residual variance (syx2) the slope of the calibration line (b,) and the calibration range (R) were known. Moreover an estimation of the concentration in the sample solution (CSJ that was used for the preparation of the standard addition line was available. While aqueous standards can be prepared over the whole calibration range the lowest level of the standard addition line is determined by Css. Since the samples are generally diluted (immediately or after digestion) prior to the analysis the addition of the standards can be performed during this dilution. It was decided to use an equal number of measurement points for the aqueous calibration line and the standard addition line thus n = n l = n 2 .The parameters that must be optimized then are the number of measurement points (n) and the distribution of these points over the calibration range. Three ways of symmetrically distributing the measurement points over the calibration range are considered. The first calibration scheme (Scheme 1) distributes all points evenly over the calibration range. If for example a calibration range between 0 and 50mg1-' is considered and 6 measurements have to be performed the standards will be positioned at 0 10 20 30 40 and 50 mg 1-I. In order to satisfy expression (4) with the lowest possible number of measurements to be performed the measurement points should be distributed in such a way that the variance of the slope is minimal.This is obtained when all standards are positioned at the extremes of the calibration range (Scheme 2). In order to keep the distri- bution symmetrical this scheme is only applied for an even number of measurement points. Thus for the given example three standards are prepared containing 0 mg 1-l and three standards containing 50 mg 1-l. However with this calibration scheme the linearity of the calibration line cannot be checked. One must assume that the goodness-of-fit to the straight line model that was demonstrated earlier is still valid. In a third calibration scheme (Scheme 3) all standards are evenly distrib- uted over three concentration levels namely at the extremes and in the middle of the calibration range.To obtain a symmetrical distribution only situations where n is divisible by three are considered. For example two standards containing 0 mg 1-' two standards containing 25 mg l-' and two stan- dards containing 50 mg 1-1 are prepared. In Appendix I1 it is explained how the variance of the slope (sb) can be expressed as a function of rz and R for the different calibration schemes. The minimum number of measurement points that is needed to satisfy expression (4) can be determined (for the different schemes) iteratively. At each iteration n is increased (with 1 2 and 3 units for Schemes 1 2 and 3 respectively) and tajz as well as tS are adapted until expression (4) is satisfied. The described strategy is illustrated for four determinations of which two are by flame-AAS and another two by graphite furnace AAS.The information that is needed for the develop- ment of the experimental design is summarized in Table 1. Table2 gives the number of measurements that should be performed in order to detect a difference in slope of lo% with an a and @ error of 5 and lo% respectively. It can be seen that a difference of 10% can be detected with an acceptable Table 1 Information that is used to develop a suitable experimental design for the detection of matrix interferences Zn Fe Pb A1 Technique Flame-AAS Flame-AAS GF-AAS GF-AAS Slope 0.1753 0.03325 2.920 3.957 Residual s 0.002659 0.00243 0.002217 0.003885 Calibration Css/mg 1 - 0.4 1 0.01 0.005 range/mg 1-l 0-1 0-5 0-0.05 0-0.06 Table 2 Number of measurement points that should be included in the aqueous calibration line and the standard addition line in order to detect a bias of 10% and 5% with different distributions of the points over the calibration range (see text) and an a- and /?-error of 5 and 10'7'0 respectively.Experimental conditions are given in Table 1 Ibl -b,I = 10% ]b - b2l= 5% Scheme 1 Scheme 2 Scheme 3 Scheme 1 Scheme 2 Scheme 3 Zn 11 6 9 43 16 24 Fe 7 6 6 27 12 15 Pb 7 6 6 29 12 18 A1 7 6 6 29 12 15 Journal of Analytical Atomic Spectrometry March 1995 Vol. SO 209number of measurements for all elements and with the three investigated calibration schemes. However a larger number of measurements is necessary to detect a difference in sensitivity of 5% (Table 2). For example for the lead determinations an aqueous calibration line and a standard addition line each containing 29 measurement points evenly distributed over the concentration range need to be measured to detect a difference in sensitivity of 5%.This number is not only unacceptably high but it is also laborious to prepare 29 solutions with different analyte concentrations. The other schemes are more suited. With Scheme 3 for example 6 measurements can be performed at the extremes and in the middle of the calibration range. When all measurement points are positioned at the extremes of the range both lines should contain only 12 measur- ement points (Scheme 2). However since Scheme 3 allows the evaluation of the goodness-of-fit to the straight line model the use of that design is advised. When this scheme requires an unacceptable number of measurements points the application of Scheme 2 can be considered.Evaluation of Experimental Results Quality of regression lines The evaluation of the experimental results is summarized in Fig. 3. Since the slopes of two calibration lines are compared one must guarantee that the measurement points are accurately described by the regression lines. Although the goodness-of-fit is assumed to be demonstrated unexpected problems can occur ( e g . due to contamination or a bad injection). Therefore Experiments - A ,n I (Qualrty of the curves) Y ~ I ~ ~oimportant interferences lnteferences Fig. 3 Strategy for the evaluation of the experimental results the quality coefficient [QC( "/.)I6 is computed QC( Yo) = x 100 d n - ' where Ai and Ai are the measured and estimated absorbances (i = 1 ...n) respectively; A is the mean of all measured absorbances and n is the number of measurement points. When a weighted calibration model is applied is replaced by Ai in eqn. ( 6 ) . Calibration lines for which the QC exceeds a value of 5% are considered as unacceptable and cannot be used for the detection of interferences. Comparison ofthe slopes of the calibration lines First the residual variances of the aqueous calibration line (syxlz) and the standard addition line (sYxz2) are compared with an F-test. As shown in Fig. 3 the approach that must be followed depends on the result of this test. In order to detect a significant difference between the slopes a t-value is calcu- lated [see eqn. (l)].As is explained in Appendix I the residual variances can be pooled if they are comparable. In that case rca is compared with the tabulated t-value with n + n2 - 4 degrees of freedom and a significance level 2 where n and n2 are the number of measurement points in the aqueous cali- bration line and standard addition line re~pectively.~ If the residual variances are significantly different teal must be com- pared with an alternative theoretical t-value:' This t'-value equals the classical one if the number of measure- ments are the same for both calibration lines (nt = nz). When the calculated t-value is higher than the theoretical one the slopes of the standard addition and aqueous calibration lines are statistically different which indicates the occurrence of matrix interferences affecting the sensitivity.DEVELOPMENT OF A STRATEGY FOR THE DETECTION OF INTERFERENCES USING A SECOND DEGREE CALIBRATION MODEL Within a limited concentration range namely where the Lambert-Beer law is valid a straight line calibration model can be applied to describe the relationship between the concen- tration and the absorbance. Above this range the sensitivity decreases with increasing concentrations predominantly as a result of stray light. Consequently to estimate a concentration which is situated above the Lambert-Beer range another calibration model must be applied. Generally in such a situation one uses a second degree model. Although the use of such a calibration model causes little problems in routine analysis it has some important consequences for the validation of the method.The previously described strategy cannot be applied to detect interferences based on a different sensitivity in sample and aqueous solutions. Therefore it is proposed to linearize the curved regression lines after which the slopes can be compared. Thus the use of a linearization as a tool to trace interferences during the method validation is investigated. Linearization Starting from the work of de Galan et a[.,' L'vov et al." expressed the theoretical absorbance ( A o ) which would be measured when the Lambert-Beer law is valid as a function of the measured absorbance ( A ) and the fraction of un- 210 Journal of Anulyticul Atomic Spectrometry March 1995 Vol. 10absorbable radiation (s) 1 loA A0 = - log l + s 1-s(1OA-1) Although this equation was originally formulated for conven- tional AAS," the applicability for the linearization of Zeeman AAS calibration curves was experimentally shown." The main problem of the linearization is to find the appropriate value of s.L'VOV'~~~' suggests the experimental determination of s as a function of the limiting absorbance value Alim ( 9 ) For conventional AAS Alim is the absorbance that is measured in the concentration range where the calibration curve is completely flat in other words where an increase in concen- tration does no longer result in an increase in absorbance. When Zeeman AAS is applied Alim equals the roll-over absorbance. Wang et a1.I2 use an expression which is similar to eqn. (8) but determine the s-value for which the corrected absorbances best fit a straight line model.The goodness-of-fit to the straight line model is evaluated by the quality coefficient6 and the Fibonacci search13 is used for the selection of the optimal s i t . the value for which the lower QC is obtained. To evaluate both strategies to obtain sin practice calibration lines for different elements were measured by flame-AAS (Fe Mn Zn and Mg) as well as by graphite furnace-AAS (Pb Al Cu Mn). Standards containing very high concentrations of the analyte were prepared in order to determine Alim. However the curves were only linearized up to concentration levels where the sensitivity was at least one half of the value near the origin. Determinations at higher concentration levels should be avoided because small differences in the measured absorbance can result in important differences in the estimated concentration.For all investigated situations a good lineariz- ation was obtained with eqn. ( 8 ) when the method of Wang12 for the determination of s was used. Acceptable quality coefficients (<5%) were found and the slope of the linearized lines were found to be comparable with the slopes that were observed in the lower concentration range where the Lambert-Beer law was valid. The method of L'vov could only be applied in some of the situations investigated. This is illustrated in Fig. 4 which shows curved flame-AAS calibration lines for Zn and Mg as well as the linearized curves. It can be seen that for Zn a satisfactory linearization was obtained with both methods.For Mg on the other hand the method of L'vov resulted in an incomplete linearization due to an undere- stimation of s. Since the method of Wang seems to be generally applicable it has been applied further. Detection of Matrix Interferences To investigate whether a comparison of linearized calibration lines could be used for the detection of interferences some simulations were carried out. Table 3 shows an aqueous cali- bration and two standard addition lines with a sensitivity equal to 90% and 80% of the sensitivity in the aqueous environment. The optimal s for the linearization of these curves was obtained and the corrected absorbances as well as the equation of the straight lines that describe these data are computed (see Table 4). Acceptable QCs were found for all the lines.The slopes of both linearized standard addition lines were found to be 90.5% and 81.0% of the slope of the aqueous calibration line which indicates the presence of interferences. The same behaviour was observed for other simulations. Additionally a number of real applications were investigated. For example standard solutions containing different Fe Zn and Mn concentrations were prepared in different nitric acid concentrations (1 10 and 20%) and the absorbance in all I I I 1 I I 1 L 1 2 3 4 5 6 7 8 1.0 1 I 1 1 I I 1 1 I 0 1 2 3 4 5 6 7 8 9 1 0 Concentration/mg r' Fig.4 (a) Curved Zn FAAS calibration line ( W ) as well as the corrected lines with the method of L'vov (0) and of Wang (0). Eqn. (8) is applied with s-values of 0.026 and 0.032 respectively.(b) Curved Mn FAAS calibration line (U) as well as the corrected lines with the method of L'vov (0) and of Wang (0). Eqn. (8) is applied with s-values of 0.009 and 0.087 respectively Table 3 Simulated curved aqueous calibration line and standard addition lines with a decrease in sensitivity of 10% (S.A. 1) and 20% (S.A. 2) respectively Concentration/ mg I-' 0.0 2.5 5.0 10.0 20.0 40.0 ~~ Absorbance Aqueous S.A. 1 S.A. 2 0.010 0.009 0.008 0.131 0.118 0.105 0.245 0.221 0.196 0.454 0.408 0.363 0.790 0.711 0.632 1.141 1.027 0.913 Table 4 Linearization of the calibration lines given in Table 3. Apart from the corrected absorbances the correction factor (s) the quality coefficient (QC) and the slope ( b ) of the regression lines are given Concentration/ mg 1-l 0.0 2.5 5.0 10.0 20.0 40.0 S QC (%) bl Aqueous 0.010 0.132 0.25 1 0.479 0.915 1.893 0.068 2.5 0.0468 Absorbance S.A.1 0.009 0.119 0.227 0.434 0.835 1.712 0.089 1.9 0.0423 S.A. 2 0.008 0.106 0.202 0.389 0.754 1.531 0.117 1.3 0.0379 solutions was measured by flame-AAS. First only the results in the range where the Lambert-Beer law was valid were considered. The equations of the (straight) calibration lines were computed and are given in Table 5. For Mn and Fe a significant decrease in sensitivity was observed in high nitric acid concentrations. The t-test for the comparison of slopes showed that the slopes of the calibration lines prepared in Journal of Analytical Atomic Spectrometry March 1995 Vol. 10 21 1Table 5 Mn and Fe determined by flame-AAS in nitric acid Slopes of the straight and linearized calibration lines of Zn Zn Mn Fe Nitric Slope Slope Slope Slope Slope Slope acid straight corrected straight corrected straight corrected (YO) line line line line line line 1 0.167 0.168 0.0471 0.0453 0.0261 0.0256 10 0.161 0.168 0.0455 0.0434 0.0241 0.0233 20 0.157 0.165 0.0445 0.0424 0.0234 0.0222 10% and 20% HN03 were significantly lower than the slope of the line in 1% HN03 (a = 0.05).The effect of HN03 on the Zn sensitivity was less important and only the calibration line prepared in 20% HN03 had a slope that was slightly different from the slope in 1% HN03 (teal = 2.65 ttab = 2.31 a = 0.05). All the measurement results were then used including those in the range where the Lambert-Beer law was not valid. All the absorbances were corrected for the unabsorbable radiation and the equations of the linearized calibration lines were computed (see Table 5).A significant decrease in sensitivity for Fe and Mn was also observed here while for Zn no significant decrease in sensitivity was observed. Although the comparison of linearized calibration lines seems to give a good indication of the occurrence of inter- ferences a number of problems have to be solved before the strategy described for the comparison of straight calibration lines can be applied on the linearized curves. A major problem with curved calibration lines is the heteroscedasticity of the data. Transformation (e.g. log transformation) of the data in order to obtain a constant variance might hamper the use of statistical tests that assume normally distributed measurement results.As is shown in Appendix Ib the t-test for the compari- son of two slopes can be extended to a situation where the calibration parameters are estimated by weighted least squares. However before this test can be included in the validation strategy additional research must be done on a number of issues such as the selection of a suitable weighting factor the investigation of the goodness-of-fit of a weighted calibration model the effect of the linearization on the variance and the development of the experimental design. Therefore for the moment the comparison of linearized calibration curves is only considered as an exploratory test where the ratio of the slopes is compared with the maximum allowable relative error.IMPLEMENTATION OF THE STRATEGY IN A KNOWLEDGE-BASED COMPUTER SYSTEM The Development Environment The validation system was developed in Visual Basic and runs in a Windows environment. A Visual Basic program can be considered as a collection of different screens. On these screens objects are placed which contain information (textual or graphical) or which make it possible for the user to enter information (text-box) or execute an action (button). By click- ing such a button with the mouse the user can for example move to another screen and can be guided from screen to screen in a non-linear way depending on the answer to a question or on the result of a computation. The Structure of the System The knowledge-based system contains three main parts which are built around a central screen.The first part of the system advises on the experimental design to be used the second part evaluates the experimental results and the third part gives information on the handling of interferences. At the beginning of the consultation the user is asked to open a file that contains information on the analysis method (sample treatment . . .) the required performance criteria (maximum allowable relative error . . .) and the validation of the calibration procedure (bo syx2). This file is created in two other modules of the validation system which are consulted prior to the matrix interference module. Table 6 illustrates the information used by the system. The strategy that is described under Development of the Experimental Design is followed.To advise the analyst on the preparation of the standard addition line the analyst is first asked for an estimation of the analyte concentration in the sample that is used. With this information the system calculates which standard concentrations should be used for the dilution of the sample in order to obtain the required measurement solutions for the standard addition. If the samples are analysed as such the system proposes to add small volumes of highly concentrated standards to the sample (e.g. 1 portion of standard added to 9 portions of sample) since dilution of the sample could mask matrix effects. Otherwise the standards are added to the sample during the dilution. For the analysis method described in Table 6 for example the system recommends to prepare 18 aqueous standards namely six standards containing 0,25 and 50 pg 1-' Pb respectively.The standard addition line is prepared by adding 1 part of standards containing 0 38 and 72 pg 1-l Pb to 1 part of sample six times for each concen- tration level. After the performance of the experiments the data are entered manually into the system or by opening a data file [Fig. 5(a)]. Before the slopes of the aqueous cali- bration line and the standard addition line are compared the quality of both lines is evaluated [Fig. 5(b)]. The equation and the quality coefficient of the calibration line are computed and a graphical representation is given. If the QC exceeds the level of 5% the user is asked to check the data and (if no errors are found in the data entry) to prepare a new calibration line.When an acceptable quality coefficient is found an F-test is performed to compare the residual variances of both Cali- bration lines. For the example given in Table 6 no difference in residual variance was found. Finally the slopes are compared both calibration lines are plotted [Fig. 5(c)] the ratio of their slopes is computed and the slopes are compared by a t-test [Fig. 5(d)]. When matrix interferences are detected the system gives some basic information on how to handle them (add or change modifier change sample treatment. . . .). The authors thank 'De Diensten voor de Programmatie van het Wetenschapsbeleid' for providing financial support. Table 6 design of experiments to detect matrix interferences Information required by the knowledge-based system for the Information specified by the user in the start module Analyte Pb Sample type Milk Sample treatment Dilution 1 part of sample + 1 part of diluent Technique GF-AAS Pyrolytic coated tube with platform NH4H2P04 + Mg(NO,) modifier Calibration range/pg 1-' 0-50 Maximum allowable relative error (YO) 5 Information obtained in the calibration module 0.00292 4.9 x bl SYX2 Information specified by the user in the matrix interference module Analyte concentration in sample/pg 1-' f 24 21 2 Journal of Analytical Atomic Spectrometry March 1995 Vol.10(Slandard addition 1671 I COMPARISON OF BOTH LINES .. E&um&uK P[;; I / / IBclucour Y - .01253 + .002817x 1.8 Addi(ion; P - .04569 + .002645 I 1.3 0 50 Inlerferencet occw. l f r ranl lo test il IhE rloper am *telitticaUy bUIlM.lerent click l b ~ <Statistics> I I Ewalion P - -01253 + .002817 I The cafibalmn line has o accaplaMs QC 1.8 x . . The slopes of the standard addition line and the aqueous calibration line are compared with a t-test t calculated 6.19 t theoretical 2.04 Conclusion; The slopes of Ihe standard addition line and the aqueous calibration line are significantly dillaent. This indicator the occulence 04 matrix mlmf erencet. Fig. 5 Screens from the knowledge-based system (a) the experimental results can be entered; (b) the quality of the calibration lines is evaluated; (c) the standard additions line and the aqueous calibration line are graphically compared; and ( d ) the slopes of the standard additions line and the aqueous calibration line are statistically compared REFERENCES 1 2 3 4 5 6 7 8 9 10 11 12 13 Taylor J.K. Anal. Chem. 1983 55 600A. Penninckx W. Analusis 1994 22 M22. de Galan L. van Dalen H. P. J. and Kornblum G. R. Analyst 1985 110 323. Massart D. L. Vandeginste B. G. M. Deming S . N. Michotte Y. and Kaufman L. in Chemometrics a textbook Elsevier Amsterdam 1988. Cooper B. E. in Statistics for Experimentalists Pergamon Press Oxford 1969. Vankeerberghen P. and Smeyers-Verbeke J. Chem. Intell. Lab. Syst. 1992 15 195. Massart D. L. Smeyers-Verbeke J. and Rius F. X. TrAC Trends Anal. Chem. (Pers. Ed.) 1988 8 8. Massart D. L. Smeyers-Verbeke J. and Rius F. X . TrAC Trends Anal. Chem. (Pers. Ed.) 1989 8 49. de Galan L. and Samaey G. F. Spectrochim. Acta Part B 1969 24 679.L'vov B. V. Polzik L. K. and Kocharova N. V. Spectrochim. Acta Part B 1992 47 889. L'vov B. V. Polzik L. K. Kocharova N. V. Nemets Y. A. and Novichikhim A. V. Spectrochim. Acta Part B 1992 47 1187. Wang X. Smeyers-Verbeke J. and Massart D. L. Analusis 1992 20 209. Massart D. L. Dijkstra A. and Kaufman L. in Evaluation and Optimization of Laboratory Methods and Analytical Procedures Elsevier Amsterdam 1978. Paper 4 104601 C Received July 27 1994 Accepted November 24 1994 APPENDIX Ia Computation of the standard deviation of the difference between two slopes which are estimated by an unweighted model where Sb12 and sbz2 are the variances of bl and b2 respectively These are computed from the residual variances If the residual variances are comparable syx2 is calculated Therefore Journal of Analytical Atomic Spectrometry March 1995 Vol. 10 213APPENDIX Ib Computation of the standard deviation of the difference between two slopes which are estimated by a weighted model s(blw - bzw) = where Sblw2 and sbZw2 are the variances of b and b2 respectively APPENDIX I1 Variance of the slope as a function of the residual variance (sy2) the calibration range (R) and the number (n) and the distribution of measurement points. Scheme 1 Data are evenly spread over the calibration range S V X 2 S V X 2 with The residual variances are computed as If these are comparable the pooled residual variance syx2 is computed (nl - 2)SYXlW2 + (n2 - 2)SYXzw2 syxw2 = n + n2 - 4 Therefore s ( ~ ~ ~ - ~ ~ ~ ) can be computed as when n is odd svx2 when n1 is even Scheme 2 All data situated at the extremes of the calibration range ( n is even) Scheme 3 All data are evenly spread over three concentration levels namely the extremes and the middle of the calibration range ( n is divisible by 3) 214 Journal of Analytical Atomic Spectrometry March 1995 Vol. 10

ISSN:0267-9477    

DOI:10.1039/JA9951000207

出版商:RSC    

年代:1995

数据来源: RSC

摘要:

Determination of Selenium in Human Hair and Nail by Electrothermal Atomic Absorption Spectrometry* Journal of Analytical Atomic Spectrometry IAIN HARRISON AND DAVID LITTLEJOHN? Department of Pure and Applied Chemistry University of Strathclyde Cathedral Street Glasgow UK GI I X L GORDON S. FELL Institute of Biochemistry Glasgow Royal Infirmary Castle Street Glasgow UK G4 OSF A microwave digestion procedure was used to prepare samples for determination of total Se by electrothermal atomic absorption spectrometry ( ETAAS) with palladium modification. Conversion of organoselenium compounds to inorganic selenium was achieved by heating 100 mg hair with 0.5 mf concentrated nitric acid for 1 min and then after the addition of 0.5 ml hydrogen peroxide (30 vols.) reheating for a further 8 min.The accuracy of the method was confirmed by analysis of certified hair reference materials containing 2.00 and 0.58 pg g-' Se. The detection limit is 0.02 pg 8-l Se. A variation of the procedure was used to analyse nail samples. Digestion was achieved by heating 100 mg of nail with 1 ml nitric acid for 2 min and after the addition of 1 ml hydrogen peroxide for a further 33 (3 x 11) min. The detection limit for this procedure is 0.03 pg g-' Se. Analysis of hair reference materials and both hair and nail samples by neutron activation analysis and ETAAS showed reasonable agreement between the techniques. The ETAAS procedure is more rapid and convenient. Analysis of 25 hair samples gave a range of 0.31-0.76 pg g-' Se (mean value 0.52 & 0.11 pg g-') and for 27 nail samples a range of 0.17-0.66 pg g-' Se (mean value 0.46 J.0.16 pg g-') was obtained. Keywords Selenium; hair; nail; microwave digestion; electrothermal atomic absorption spectrometry Symptomatic Se deficiency in man is known but rare.' Chronic Se dietary lack however is more common and has been associated with various degenerative diseases.' The daily diet- ary intake of Se by the UK population has been estimated at only 40% of the recommended level,' and elderly people may consume even smaller amounts.2 Several biochemical mechan- isms involving selenoproteins have been de~cribed,~ with one important group the glutathione peroxidases (GSH-Px) forming part of the antioxidant defence mechanism^.^ Biological markers of Se nutritional status5 include the Se concentrations in whole blood plasma and urine.These reflect recent intake. The determination of GSH-Px activity in red blood cells and platelets gives an indirect measure over a longer (3 month) period. Measurement of the Se content of the sulfur-rich protein keratin found in hair and nail offers an even longer-term marker. Chinese workers have previously related the Se content in hair to dietary intake,6 and in New Zealand the Se content of toe nail has been shown to increase with Se ~upplementation.~ The ease of obtaining hair and nail by non-invasive sampling and the long term stability of the samples makes the materials particularly suitable for surveys of populations at risk of Se depletion. Ranges of Se concen- * Presented at the Seventh Biennial National Atomic Spectroscopy 7 To whom correspondence should be addressed.Symposium (BNASS) Hull UK July 20-22 1994. trations found in hair and nail from several countries are reported in Table l.'-13 The accurate determination of Se in hair by electrothermal atomic absorption spectrometry (ETAAS) requires the total destruction of all organoselenium compounds. These com- pounds are the acid resistant s- and Se-containing amino acids methionine cysteine and cystine which are present in keratin. Conventional digestion methods with acid reflux are time consuming and labour intensive. Tetraalkyl ammonium hydroxides have also been used to dissolve toe nail,12 but the method is slow and inconvenient. Microwave heating in sealed PTFE vessels can be used for the rapid digestion of tissues14 and other biological materials.15 Many digestion procedures use nitric acid often in combi- nation with other reagents although not all are compatible with the determination of Se by ETAAS.Nitric acid alone gives incomplete digestion of organoselen- ium compounds.16 Perchloric and hydrofluoric acids are effec- tive,17 but shorten tube lifetimes and decrease the analytical sensitivity. Procedures with these acids are lengthy and require numerous heating cooling and evaporation steps which result in analyte loss. Spectral and chemical interferencesl8*'' during the atomization of Se are caused by hydrochloric and sulfuric and phosphoric acids respectively. Hydrogen peroxide" has also been used in hair analysis using open reflux which lengthens the digestion time.Digestion by nitric acid and hydrogen peroxide in sealed PTFE vessels with microwave heating has been shown to give incomplete digestion of hair and low recoveries of Cu and Pb,'l if short heating times are used. Kingston and Jassie,I5 however sug- gested that nitric acid absorbs microwave energy better than other mineral acids implying that the complete digestion of organoselenium compounds should be achieved if the heating time and the acid peroxide sample ratios are optimized. Methods for the determination of Se in human hair and nail are described. Microwave heating is employed after the separ- Table 1 using various techniques Geographical variation of selenium levels in human hair Selenium concentration/ Sample type Technique Country I% g-' Reference Hair Fluorimetry China NAA France ICP-AES USA Japan New Zealand Venezuela Nail ETAAS Bulgaria NAA New Zealand USA Venezuela 0.55-16.20 0.58-0.76 0.28-0.52 0.41-0.75 0.25-0.61 0.70-2.90 0.05-0.61 0.16-0.1 8 0.74-1.17 0.90-4.90 8 9 10 10 10 11 12 7 13 11 Journal of Analytical Atomic Spectrometry March 1995 Vol.10 215ate addition of nitric acid and hydrogen peroxide. The Se concentration is determined by ETAAS with Pd chemical modification.22 The method has been validated by analysis of two certified hair reference materials and by comparison with neutron activation analysis (NAA). The proposed method is faster gives better recoveries of added Se and lower detection limits than previous methods. EXPERIMENTAL Instrumentation A Perkin-Elmer 1 100B atomic absorption spectrometer equipped with a PE HGA700 graphite furnace programmer a PE AS70 furnace auto-sampler and an Epson FX800 printer was used to obtain integrated absorption signals.A selenium electrodeless discharge lamp operated at 5 W from an external power supply was used as the light source. A band-pass of 0.7nm was used to isolate the 196.0nm Se line. Pyrolytic graphite coated graphite tubes (PE B0121092) and L'vov platforms (PE B0121091) were used throughout. Solutions were injected as a mixture of the sample or standard and modifier (ie. 10+ 10 pl). The furnace programme was as shown in Table 2. Reagents All solutions were prepared from analytical-reagent grade chemicals or commercially prepared high concentration solutions supplied (unless otherwise stated) by BDH (now Merck) UK.A Se"- stock solution (40 mg 1-l) was prepared by dissolv- ing 0.00500 g of selenomethionine (C5H1102NSe 98% Sigma) in 50ml of de-ionized water. A Se" stock solution (1000 mg 1-I) was purchased in the form of a 'Spectrosol' standard solution. A SeV' stock solution (1000 mg 1-I) was prepared by dissolv- ing 0.8160g of sodium selenate (Na2Se04 96%) in 50ml of de-ionized water. A Pd modifier solution (2000 mg 1) was prepared by dissolv- ing 0.2523 g of palladium nitrate [Pd(N03)2.2H20 99%] in 1 ml concentrated hydrochloric acid (HCl Aristar grade) and diluting to 100 ml with de-ionized water. Nitric acid (HNO 69% v/v Aristar grade) and hydrogen peroxide (H,02 30% v/v Aristar grade) were used throughout. Human hair reference materials were purchased from the Laboratory of the Government Chemist (LGC UK) certified reference material (CRM) 397 (from the Community Bureau of Reference) contains a Se concentration of 2.00&0.08 (2s) pg g-' and GBW09-101 (from the Chinese Academy of Science) contains a Se concentration of 0.58 kO.05 (2s) pg g-'.Procedure Sample preparation For method development a pooled sample (approximately 800 g) of freshly cut hair was collected in a polythene bag and cut further into 3-5 mm lengths. The bag was shaken vigor- Table2 HGA700 furnace programme used to determine Se in hair and nail; injections of mixed sample (or standard) and 500 mg 1-1 Pd modifier (10 + 10 pl) Ramp Hold Ar gas flow/ Step TemperaturerC time/s time/s ml min-' Read on 1 150 5 20 300 2 1200 5 30 300 3 2600 2 4 0 0.0 4 2650 1 4 300 ously for 10-15min to mix the sample.This mixture was divided into several equal portions that were frozen with liquid nitrogen crushed (in a porcelain mortar and pestle) dried at room temperature (in a desiccator) re-shaken and stored in clean plastic containers. Individual hair samples were collected in small polythene bags and prepared by cutting into 3-5 mm sections on top of a filter paper. The samples were then sealed in plastic con- tainers which were labelled and dated and shaken to mix the sample. A pooled nail sample was not prepared owing to difficulties in obtaining a homogeneous mixture. Individual samples had visible contaminants removed prior to cutting into 1-2 mm2 pieces. These pieces were stored in plastic containers and mixed thoroughly before analysis.An important aspect to consider in trace element determi- nation is the effect of washing procedures when attempting to remove exogenous contamination from the surface of the sample. This has been considered by others,23 without a consensus being reached. Different wash reagents (e.g. water hexane triton X-100 and anionic and cationic detergents) were investigated to ascertain their effect on the Se concentration in the pooled hair sample. These investigations were inconclus- ive and so washing was not carried out prior to analysis. Sample digestion Digestion was carried out initially in open borosilicate glass tubes (10 mm id Corning) on a 'Techne Driblock DB3' heating block and latterly in PTFE vessels (5 ml Savillex USA) in a Profile 'Soft Touch' (650 W) domestic microwave oven.In the development of the digestion procedure nitric acid or nitric acid and hydrogen peroxide were used in various volumes (0.2-2 ml) and at different open tube digestion temperatures or microwave digestion times (1-60 min) to digest the hair or nail sample (100 mg) reference material or Se standard solution. The recommended procedures are as follows. Concentrated nitric acid (0.5 ml) is added to hair (100 mg) weighed directly into a PTFE vessel. The vessel is then capped and heated at full power in the microwave oven (1 min). After cooling in air to room temperature the vessel is opened and hydrogen peroxide (0.5ml) is added. The vessel is resealed and heated in the oven (8 min). After cooling to room temperature the digestion mixture is diluted to a known volume (2 ml) and analysed.For nail samples (100 mg) concentrated nitric acid (1 ml) is added to the vessel and heated at full power (2 min). After cooling hydrogen peroxide (1 ml) is added and the vessel is re-heated [33 min (as 3 x 11 rnin)] before cooling and diluting to 3 ml with de-ionized water. Up to 200mg of hair could be digested using the above method. Otherwise it is not recommended to exceed the volumes and masses given without taking adequate pre- cautions. The hydrogen peroxide was always added after a period of initial digestion with nitric acid. Perchloric acid was never used as an alternative oxidant. Neutron activation analysis Samples of hair or nail (100 mg) were irradiated in either the reactor at the Scottish Universities Research and Reactor Centre (SURRC) East Kilbride UK (75Se) or at 'Tracerco' ICI Billingham UK (77mSe).At the SURRC reactor samples were irradiated at 300 kW (with a neutron flux of 3.6 x 10I2 n cm-2 s-l) for 18 h (over 3 d). After cooling for 7 d Se concentrations were determined at 136 and 264keV using a Tennelec counter (for 20h per sample). 'Tracerco' used a 250kW reactor (neutron flux of 216 Journal of Analytical Atomic Spectrometry March 1995 VoL 102-3 x lo1 n cm-2 s-') and samples were irradiated for 30 s before determining the Se concentration. Other irradiation details were not supplied. Different irradiation times were required owing to the differ- ent half-lives (75Se = 120 d and 77mSe = 17.5 s) and abundances ("Se =0.87% and 77mSe = 9.02%) of the two species.RESULTS AND DISCUSSION Digestion Studies Hair Initial studies used the pooled hair sample throughout to allow a comparison of the various results obtained. Nitric acid (0.5 ml) was used to digest the sample (100 mg) in open tubes at various temperatures. There was low signal recovery for added Se" and poor precision [relative standard deviation (RSD) of approximately 33%]. Heating to Driblock set tem- peratures of 100 140 and 180°C gave signal recoveries for 100 or 200ng of added Se" of 34 20 and 8% respectively. The actual digestate temperatures were not measured. Neve et al.I6 have reported low recoveries of organic Se and suggested that the incomplete digestion of organoselenium compounds was responsible.In this case however the low recoveries could only be due to analyte loss by the volatilization of SerV as for example SeO,. To reduce Se loss sealed PTFE digestion vessels were used with microwave heating. The sample (100 mg) was digested with 0.5 ml of concentrated nitric acid for 1 2 5 or 10 min at full power. The precision was better (RSD = 13%) than for the open tube procedure and the signal recovery for the addition of 100 or 200ng of Se" was about 80% for each digestion time in agreement with results obtained by Matusiewicz et aL2' When undigested solutions of different Se compounds were analysed by ETAAS with Pd modification differences in sensi- tivity occurred between the inorganic and organic selenium compounds even although the increase in the maximum pyrolysis temperature for Se with Pd was the same (from 800 to 1300 "C).Equal masses (0.74 ng) of Se" and SeV1 produced similar integrated absorbance values of about 0.05 s but 0.74 ng of Sell- gave approximately double the signal (0.09). Even when microwave digested with nitric acid for 5 min at full power Se"- gave greater signal sensitivity than both Se" and Se". These observations confirm that (i) nitric acid cannot totally digest organoselenium compounds and (ii) differences in the vaporization/atomization mechanism occur between organic (Sell- ) and inorganic (Se" and SeV') selenium when Pd modification alone is used. Styris et ~ 1 . ~ ~ considered the possible mechanisms of Se atomization with Pd modification and proposed the formation of an intermediate [Pd Se 01 complex which reduces losses of volatile SeO,.Other workers have suggested that palladium forms a solution of the analyte which is released by diffusion through the Pd metal." Organic selenium compounds either do not decompose to produce SeO or do so at a temperature where the Pd is more successful at preventing losses of Se. It may also be possible that the extent of Se atom formation is greater for organoselenium compounds than inorganic compounds irrespective of SeO losses owing to the production of lower concentrations of Se,. Welz et a2.26 reported that the addition of 5 pg of copper and 1Opg of magnesium nitrate as a mixed modifier gave similar sensitivities for Se"- SeIV and Se". Krivan and K~ckenwaitz,~ also reported that the stabilization of Se is uniform for all three species if Pd-Mg modification is used.As Cu is sometimes determined in clinical samples it was preferable not to add large amounts of this element to the atomizer. Also it was noted that in the presence of large concentrations of Mg and phosphate (from clinical samples) there is a large increase in the background signal at the Se wavelength. This is thought to occur through the same mechan- ism by which Ca increases the background absorption signal when in the presence of phosphate viz. the production of P molecule^.'^ Hence there are some advantages in not using Cu and/or Mg(NO,) as modifiers in the determination of Se. The disadvantage is that it is necessary to ensure that all of the Se compounds are converted to inorganic Se in order to obtain the same response and hence sensitivity characteristics as inorganic Se calibrant solutions.As nitric acid alone was incapable of totally digesting organoselenium compounds hydrogen peroxide was added to complete the digestion. The duration of the heating step proved the most important factor. The Se absorbance signals obtained from hair digests are presented in Fig. 1. Complete digestion was achieved by heating 100 mg of hair with 0.5 ml nitric acid for 1 rnin and then for a further 8 min after adding 0.5 ml hydrogen peroxide. No difference was apparent when the ratio of acid to peroxide was changed from 1 1 to either 2 1 or 1:2. In each case the final digest volume was made up to 2 ml. When the optimized digestion procedure was applied to the pooled hair sample to which the Set'- or SetV standard had been added satisfactory recoveries were obtained (Table 3).The precision (RSD) of the Se absorbance signals was <5% (integrated absorbance values of 0.028+ 0.001-0.452 f 0.014 s; n = 3; one standard deviation). Digests of the reference materials the pooled hair sample and hair with added SeIV were analysed by standard additions and the gradients compared with that of a standard calibration curve produced using Se" solutions (see Table 4). The gradi- ents were similar indicating no multiplicative interferences from the hair digest matrix. The concentration of Se in the v) 0.08 $ 3 g 0 0.06 -0 a .I- = 0.04 0 0 0 0 0 0 O O . A 0 0 0 0 B 0 n ns 4 20 40 Digestion timdmin 60 Fig. 1 Effect of hydrogen peroxide digestion time on the integrated Se AAS signal A hair; and B nail.Samples (100 mg) were initially digested in nitric acid for 1 min (hair) or 2 min (nail) and after peroxide digestion were diluted to 2 or 3 ml respectively. Approximate solution concentration of Se is 40 pg 1-' Table 3 Percentage signal recoveries for different Se standards added to a pooled hair sample prior to digestion; samples (100 mg) digested with HN0,-H,O in a microwave oven (see experimental section for details) Se standard oxidation state I1 - IV I1 - IV 11 - IV I1 - IV Mass of Se addedlng 0 0 37 37 74 74 148 148 Mass of Se detectedlng Recovery (%) 40 40 73 89 71 84 114 100 115 101 191 102 190 101 - - Journal of Analytical Atomic Spectrometry March 1995 Vol. 10 21 7Table 4 Gradients of standard additions and Se" calibration graphs Sample type Gradient k SD t* Pooled hair sample 1.63 k 0.06 2.04 Spiked hair sample 1.59 k 0.08 0.55 Hair reference material CRM 397 1.53 k0.05 1.53 Spectrosol Se" standard 1.57 f 0.04 - * Values for comparisons with Se" standards; tcrit = 2.23 (95% n = 10).Table 5 n = 3 and errors are k 2s Se concentrations in two hair reference materials; in all cases -~ Approximate Final digest [Se] in [Se] in sample mass/ volume/ CRM 397*/ GBW09-101t/ mg ml g-' g-' 100 2 1.92k0.06 0.54 f 0.06 50 2 1.98 f 0.08 0.61 k 0.06 100 3 1.93 -t 0.06 0.56 f 0.08 * Certified Se levels in CRM 397 2.00&0.08 pg g-' Se. t Certified Se levels in GBW09-101 0.58 f0.05 pg g-' Se. two hair reference materials (Table 5) showed good agreement with the certified values.Nail A similar study of the variation of absorbance signal with nitric acid-peroxide digestion time was carried out for nail samples (see Fig. 1). Total digestion was achieved after 35 min (2 min in nitric acid and 33 rnin in acid-hydrogen peroxide) if 1 ml of both acid and peroxide were used. Low RSDs (<5%) and high signal recoveries (97-101% for 0.37-1.48 ng Se added) showed the method to work as well as the hair procedure although no CRMs are available to validate the method. Comparison of Selenium Determination by NAA and ETAAS Samples (hair nail and hair reference materials) were analysed by both NAA (see experimental section for details) and ETAAS. The NAA results (see Table6) obtained from both 75Se and 77mSe compared relatively well with each other for the samples but not for the reference materials. Indeed only two of the five results for the reference materials agreed with the certified values.The results obtained by ETAAS from the hair samples were lower than those obtained by NAA. However the values obtained for the nail samples compared well with those of NAA and the results for the reference materials were within the limits of the stated ranges. There is no obvious reason why the NAA results for the reference materials should be in error. Interferences from other elements during counting is unlikely. The reference materials were in powder form where as the hair samples were only cut into short lengths. This physical difference may have contributed to the apparent error in the analysis of the reference materials.The difference between the concentrations of Se in hair determined by NAA and ETAAS were much smaller than for the reference material results. Table 6 Comparison of Se concentrations in hair samples and reference materials as determined by ETAAS and NAA; all errors are 1 2 s NAA (75Se) Sample type Hair Male Female Pooled CRM 397t Reference material GB W09- 101 1 Nail Male Female 136 keV 0.48 _+ 0.05 0.37 & 0.06 0.36 k0.09 2.10 k 0.14 1.22 k 0.11 0.61 k 0.04 - 246 keV 0.47 k 0.08 0.41 _+ 0.06 0.3? k 0.10 2.4l k0.12 1.48 fO.10 0.55 & 0.04 - NAA (77mSe) 0.54 _+ 0.08 0.40 & 0.07 0.38 & 0.06 1.87 f 0.24 - - 0.42 k 0.08 ETAAS 196.0 nm 0.34 f 0.08 0.29 f 0.08 0.22 f 0.08 1.93 & 0.08 0.58 & 0.04 0.57 k 0.04 0.39 t 0.06 * For 75Se n = 6; ETAAS n = 3; 77mSe not provided.t Certified Se concentration CRM 397 (2.00k0.08 pg g-'). 1 Certified Se concentration GBW09-101 (0.58 f0.05 pg g-'). Table 7 Analysis of samples from healthy adults in the Glasgow area using the proposed procedure [Se]/pg g-' Gender Male Female* Female Overall Values Range Median Mean Range Median Mean Range Median Mean Range Median Mean Hair 0.56-0.76 0.67 0.68 f 0.08 ~t = 9 0.49 0.53k0.14 n=7 0.49 0.53f0.12 n=9 0.53 0.52k0.11 n=25 0.34-0.72 0.39-0.71 0.34-0.76 Nail Finger 0.27-0.64 0.38 0.43k0.15 n=7 0.40-0.60 0.47 0.51 f 0.08 IZ = 5 0.44 0.47 20.13 n = 12 0.27-0.64 ~ ~~- Toe 0.17-0.53 0.27 0.32k0.15 n=6 0.37-0.66 0.56 0.53+0.10 n=9 0.47 0.44k0.16 n= 15 0.1 7-0.66 * Chemically treated. 218 Journal of Analytical Atomic Spectrometry March 1995 Vol.10The detection limits for the ETAAS method (based on 3s of replicate analyses of blank digests) were 0.02 pg g- ' Se (hair) and 0.03 pg g-' Se (nail) for a 100 mg sample. The character- istic mass was calculated to be 66 & 20 pg Se (n = 10 i.e. values calculated on 10 different occasions). Applications The ETAAS method was used to analyse samples of scalp hair and finger and toe nails obtained from healthy adults in the Glasgow area (Table 7). There was no significant difference in the mean concentrations of Se in the hair from male and female volunteers although the range was narrower for the male samples. Chemical treatment of female hair (such as bleaching dyeing or perming) did not affect the range of Se concentrations.Although there were slight differences in the concentrations of Se in the finger and toe nails of men and women the overall ranges and mean concentrations were similar. In general there was good correlation between the Se concentrations in hair and nail (r=0.89; n= 18). The authors would like to thank Alex Wilson (SURRC) for the neutron activation analysis. REFERENCES Selenium Environmental Health Criteria 58 World Health Organisation Geneva 1987. Thomson C. D. Rea H. M. Robinson M. F. and Chapman 0. W. Proc. Univ. Otago Med. Sch. 1977 55 18. Neve J. in Selenium in Medicine and Biology eds. Neve J. and Favier A. Walter de Gruyter Berlin 1988 p. 97. Cohen G. and Hochstein P. Biochemistry 1963 2 1420. Neve J. in Selenium in Medicine and Biology eds.Neve J. and Favier A. Walter de Gruyter Berlin 1988 p. 137. Ge K. and Yang G. Am. J. Clin. Nutr. 1993 57 259s. Longnecker M. P. Stampfer M. J. Morris J. S. Spate V. Baskett C. Mason M. and Willett W. C. Am. J. Clin. Nutr. 1993 57 408. 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 Yang G. Zhou R. Yin S. Gu L. Yan B. Liu Y. and Li X. J . Trace Elem. Electrolytes Health Dis. 1989 3 77. Ohta Y. Nakano A. Matsumoto M. and Hoshi M. J. Radioanal. Nucl. Chem. Art. 1987 114 75. Tsalev D. L. Atomic Absorption Spectrometry in Occupational and Environmental Health Practice CRC Press Boca Raton Bratter P. Negretti de Bratter V. E. Gawlik D. Oliver W. Alvarez N. and Jaffe W. J . Trace Elem. Electrolytes Health Dis. 1993 7 111. Tsalev D. L. Taerovski E.I. Raitcheva A. G. and Barzev A. I. Spectrosc. Lett. 1993 26 331. Morris J. S. Stampfer M. J. and Willett W. Biol. Trace Elem. Res. 1983 5 529. Hocquellet P. and Candillier M. P. Analyst 1991 116 505. Kingston H. M. and Jassie J. T. Anal. Chem. 1986 58 2534. Neve J. Hanocq M. and Molle L. in Trace Element-Analytical Chemistry in Medicine and Biology eds. Bratter P. and Schramel P. Walter de Gruyter Berlin 1983 vol. 2 p. 859. Okamoto K. Morita M. Quan H. Uehiro T. and Fuwa K. Clin. Chem. 1985 31 1592. Aller A. J. and Garcia-Olalla C. J. Anal. At. Spectrom. 1992 7 753. Saeed K. and Thomassen Y. Anal. Chim. Acta 1981 130 281. Coetzee P. P. and Pieterse H. S. Afr. J. Chem. 1985 39 85. Matusiewicz H. Suszka A. and Ciszewski A. Acta Chim. Hungarica 1991 128 849. Eckerlin R. H. Hoult D. W. and Carnick G. R. At. Spectrosc. 1987 8 64. Chittleborough G. Sci. Total Environ. 1980 14 53. Styris D. L. Prell L. J. Redfield D. A. Holcombe J. A. Bass D. A. and Majidi V. Anal. Chem. 1991 63 508. Jackson K. W. Spectrochim. Acta Part B 1991 46 1841. Welz B. Schlemmer G. and Voellkopf U. Spectrochim. Acta Part B 1984 39 501. Krivan V. and Kuckenwaitz M. Fresenius' J . Anal. Chem. 1992 342 692. 1984 V O ~ . 11 ch. 26 pp. 167-178. Paper 4/05406G Received September 5 1994 Accepted November 28 1994 Journal of Analytical Atomic Spectrometry March 1995 Vol. 10 219

ISSN:0267-9477    

DOI:10.1039/JA9951000215

出版商:RSC    

年代:1995

数据来源: RSC

摘要:

Comparison of Chemical Modifiers for the Determination of Gold in Biological Fluids by Electrothermal Atomic Absorption Spectrometry* I Journal of 1 Analytical Atomic Spectrometry NIKOLAOS S . THOMAIDIS EFROSINI A. PIPERAKIT AND CONSTANTINOS E. EFSTATHIOU Laboratory of Analytical Chemistry Chemistry Department University of Athens University Campus 157 71 Athens Greece The use of various isomorphous metals as chemical modifiers on the determination of gold has been investigated. The temperature programmes and the masses of modifiers were carefully optimized. Noble metals copper nickel and rhenium increased the atomic absorption signal of gold and raised the maximum pyrolysis temperature from 700 to 1000 "C. In the presence of 20 pg of ascorbic acid this temperature was 1000°C whereas in combination with 1 pg of Pd it increased to 1150°C.On the basis of platform atomization and integrated absorbance measurements the best sensitivity for aqueous solutions was obtained with the mixed modifiers rhodium-rhenium (mo = 7.3 pg) ascorbic acid-palladium (m = 7.1 pg) or ascorbic acid-rhodium (m = 6.8 pg) whereas the characteristic mass in the absence of modifiers was 12.2 pg. The limit of detection was 0.23 pg 1-l in the absence of modifier whereas it was 0.12 pg 1-l in the presence of the mixed modifier rhodium-rhenium. In the absence of chemical modifiers a fractional order of release was observed for gold whereas in the presence of modifiers an approximately first- order release was observed. The determination of gold in biological fluids was interference-free and complete recovery was obtained when stabilized temperature platform furnace conditions were fulfilled.Constant signals without double peaks and shoulders were obtained only in the presence of modifiers. It was found that gold solutions stored in the autosampler cups during measurements of the samples were stable only in the presence of 0.5 g 1-l NH4SCN or 1.0 g 1-I ascorbic acid. Keywords Gold; electrothermal atomic absorption spectrometry; chemical modijier; biological fluid Gold is usually determined by electrothermal (graphite furnace) atomic absorption spectrometry (ETAAS) with adequate sensi- tivity (typically mo= 10 pg). However gold is a relatively volatile metal and interferences are observed from the sample Hence chemical modification is recommended in order to increase the maximum permissible pyrolysis tempera- ture and minimize interferences.Slavin et aL3 reported that the addition of 50 pg of Ni on the platform stabilized gold to 1100°C and reduced the interferences from the matrix of silicate rock samples. Egila et a1.2 found that a mixture of Pd and Mo acted as an effective modifier by reducing the inter- ferences from common salts and this mixture was used in the determination of Au in whole blood and serum.4 Vanadium has been used for the direct determination of Au in ores.5 Proteins and amino acids have been used as chemical modifiers for the determination of Au in biological fluids and tissues.6 Gold-containing drugs are administered to patients suffering * Presented at the Seventh Biennial National Atomic Spectroscopy Symposium (BNAAS) Hull UK July 20-22 1994.7 To whom correspondence should be addressed. from rheumatoid arthritis. Determination of gold in serum whole blood4 and urine7 is often required for monitoring the concentration of Au in these fluids because of its toxicity. Direct methods have been developed for the determination of Au in biological fluids and tissues with4v6 or w i t h o ~ t ~ ? ~ chemi- cal modifiers. In the absence of chemical modifiers low pyrolysis temperatures must be used so matrix-matched calibration is required. In contrast when modifiers were used the pyrolysis temperature could be raised and the calibration could be performed using aqueous standard^.^ The aim of the present study was to optimize the conditions for the use of existing (Pd Ni Cu) and new (Rh Rh-Re) modifiers.The choice of potential chemical modifiers was based on the studies of Tsalev et a1." of the possible substi- tution of the modifier atoms by the isomorphous analyte. The direct determination of Au in serum and urine was investigated and the use of modifiers in combination with platform atomiz- ation proved to be essential for interference-free measurements. The stability of Au solutions in the presence of chemical modifiers was examined and it was found that long-term stability was achieved in the presence of 0.5 g I-' NH4SCN. EXPERIMENTAL Instrumentation A Perkin-Elmer Model 5000 atomic absorption spectrometer equipped with an HGA 400 graphite furnace was used for the atomic absorption measurements.Pyrolytically coated graphite tubes with pyrolytic L'vov platforms were used throughout the study. Sample solutions (20 pl) were dispensed into the graphite tubes with an AS-1 autosampler and the modifier solution (5 pl) was injected with an Eppendorf micropipette with dispos- able polypropylene tips. The instrumental settings and the graphite furnace programmes are summarized in Table 1. The time-resolved atomic absorption pulses were obtained with an IBM compatible PC Quest 286/16 computer interfaced with the spectrometer through its RS232 port using a home- made control program written in Turbo Pascal 6.0. Data acquisition was performed at a maximum sampling rate of 50 absorbance indications (unprocessed and unfiltered) per second. The absorption peaks were filtered using the Savitzky-Golay algorithm and displayed on screen after appro- priate scaling.The peak height and integrated absorbance were calculated and simultaneously displayed along with the peak time. Hard copies of the absorption peaks were obtained by screen dump on a conventional dot-matrix printer. Furthermore common file operations allowed blank correc- tions signal averaging and multiple displaying of signals for comparison purposes. Reagents All chemicals used in this study were of analytical-reagent grade. All glass and polypropylene apparatus was kept in 10% Journal of Analytical Atomic Spectrometry March 1995 Vol. 10 221Table 1 Spectrometer- Instrumental operating conditions and temperature programmes for the determination of gold Wavelength 242.8 nm Bandpass 0.7 nm Lamp current 15 mA Purge gas Ar BG corrector Dz-ON Sample volume 20 p1 Modifier volume 5 P1 PIatjorm atornizatinn-- Ar flow Step Temperat ure/"C Ramp time/s Hold time/s ratelm1 min - ' Read Drying 180 10 20 300 Pyrolysis Va ri o us * 10 20 300 Cool-down 20 1 9 300 Atomization 1900 0 4 0 ON Cleaning 2650 1 2 300 * Tpyr are presented in Table 2.nitric acid at least overnight and then rinsed with 1% nitric acid followed by distilled water before use. The acids were of Suprapur grade (Merck Germany). Gold standards were pre- pared by diluting a 1 g 1-' Au (as AuC1,) stock solution (Titrisol Merck Germany) with de-ionized distilled water and acidified to a final HCl concentration of 0.5% v/v. Modifier stock solutions were prepared by dissolving appropriate amounts of their salts in acid followed by dilution with water.The modifiers studied were Ca Sr Y Cu Ni (as nitrates) W (as Na,WO,) Zr (as ZrOCI,) Re (as HReO prepared by dissolving Re powder in H,O,) and Ru Rh Pd Pt (as chlorides). All these metals are isomorphous with gold." A 1 yg 1-' NH,SCN solution was prepared from solid NH4SCN (Merck Germany). A 5 g I-' ascorbic acid solution was prepared from L-( +)-ascorbic acid powder (Merck Germany). Procedure Comparison of chemical modiJiers For the choice of the maximum pyrolysis temperature and the optimal mass of the modifiers the modifier solution (5 pl) was dispensed into the graphite followed by the gold solution (20 pl; 0.2 ng per injection). Calibration curves were con- structed by injecting 20 p1 samples of standard solutions con- taining 2.0 5.0 10.0 20.0 and 40.0 pg I-' Au on the platform; the linear range was up to 100 pg I - ' .The characteristic mass mo (pg) was calculated from the slope (b) of the standard curve using the equation m0=0.0044x 20/b for a sample volume of 20 p1. The limit of detection LOD (pg l-') was calculated from the equation LOD= 3 x S& where Sb is the standard deviation of ten blank firings. Deterrnination ofgold in serum and urine In order to determine the Au content in the serum of patients undergoing chrysotherapy the samples have to be diluted 100-400-fold. Nevertheless in order to assess the procedure developed recovery studies were performed both with spiked serum and urine samples from healthy subjects and patients after a 40-fold and 400-fold dilution respectively. A 0.5 g 1-' NH4SCN solution was used as diluent in all cases.In order to determine the concentration of the free gold ion in serum and subsequently the concentration of the bound gold 500 pl samples of serum were ultrafiltered using a disposable 13 mm syringe filter (0.45 pm) with tube tip PVDF (Whatman England). RESULTS AND DISCUSSION Stability of Gold Solutions The problem of the stability of Au solutions is well known.4 In the present study the Au standards were prepared in glass containers as indicated in ref. 4. However during the measure- ments the Au standards were contained in the plastic cups of the autosampler. It was observed that the atomic absorption signal of a 10 pg 1-' Au solution was reduced to 60% of its initial value after 2 h as illustrated in Fig.1A. The rate of Au adsorption was higher in the presence of chemical modifiers e.g. Rh (Fig. 1C). In contrast the Au solutions were stable for more than 2 h in the presence of NH,SCN (Fig. lB,D). A study of the effect of the NH4SCN concentration on the atomic absorption signal of Au showed it had no influence up to 1.0 g 1-'. Therefore all the Au solutions were prepared in a 0.5 g 1- ' NH4SCN solution. Comparison of Chemical Modifiers Optimization of modifier mass and temperature programme In the first part of this study we compared the maximum pyrolysis temperature (T,,,) and sensitivity obtained when all the metals mentioned under Experimental were used as chemi- 1 0.25 I--- I _ Q) 0.20 e 2 0.15 n 0 0.10 0 C m - m L a r Y m c - .- A &' 0.05 - 0' 0 20 I 40 I 60 I 80 I 100 I 120 1 Ti me/m in Fig.1 Stability of a 10 pg I - Au sotution stored in plastic cups of the autosampler during measurements at room temperature with different chemical modifiers A HNO 0.01% v/v; (B) 0.5 g I - ' NH,SCN; C 0.05 g 1 - ' Rh; and D 0.5 g 1 - ' NH,SCN + 0.05 g 1 - Rh. Similar trends were observed with integrated absorbance measurements 222 Journal of Analytical Atomic Spectrometry March f995 Vol. I0cal modifiers. It was found that Ca W and Zr influenced neither the Au sensitivity nor the pyrolysis temperature. Strontium decreased the Au signal whereas Y increased the background signal and caused overcorrection. All the other metals appeared to stabilize Au thermally and increased the Au sensitivity though to different extents.The influence of the mass of the modifier on the Au sensitivity and the pyrolysis temperature was investigated. It is known that the modifier mass significantly influences the sensitivity; a high modifier mass decreases the signal owing to secondary adsorption at the cooler ends of the graphite tube.12 On the other hand higher modifier masses stabilize the analyte to higher temperature^.'^ Therefore a careful optimization of the modifier mass was deemed necessary in order to achieve the maximum sensitivity and thermal stabilization. A Tpyr of at least 900-1OOO"C was essential in order to remove most of the matrix. The influence of increasing mass of Cu and Re on the Au signal at a Tpyr of 900°C is shown in Fig.2. The Au signal was recovered with as little as 0.5 pg of modifier. The same pattern was also observed with the other modifiers. However Cu (Fig. 2A) decreased the peak height signal of Au when a mass of > 2 pg was injected. The same phenomenon was apparent with all modifiers except Re. The integrated absorbance decreased gradually when a mass > 2 pg was used but this decrease was >lo% only in the presence of a mass of 2 5 pg. When Brajter et ~ 1 . ' ~ determined Au traces in Pt and Pd samples an increase of the Au peak height signal was observed with a slight excess of the matrix elements whereas a decrease was observed for a large excess of Pt or Pd. Similarly L'vov and Frech'' observed a decrease in the Au signal when Au was atomized in the presence of 20 pg of Ag.These phenomena were ascribed to trapping of analyte atoms in the presence of modifiers at the cooler tube This decrease was not observed when the mass of Re was optimized. The integrated absorbance and peak height were stable over a wide range (0.5-10 pg) of mass of Re. This can be easily explained since Re is a very refractory element and forms refractory compounds with carbon in the graphite tube; thus it could not be atomized at the atomization temperature of A u . ' ~ Also Re could not migrate towards the cooler ends of the tube whereas modifiers such as Pd could." The influence of the modifier mass on the maximum Tpyr is illustrated in Fig. 3 with Rh as an example. A near- stoichiometric amount of Rh raised the maximum Tpyr by 50°C but at least 1 pg of Rh was required to stabilize the Au signal at 1OOO"C.These results agree with the observations of Qiao and Jackson" who proposed that Pd stabilized volatile elements through a predominantly physical (dissolution of the analyte in molten Pd) rather than a chemical (formation of chemical compound) mechanism. The effect of different modi- Cu modifier mass/pg 0 0.5 1.0 1.5 2.0 2.5 3.0 0 2 4 6 8 10 Re modifier mass/pg Fig. 2 Influence of increasing amounts of Cu (A B) and Re (C D) on the peak height (A C) and integrated absorbance (B D) of 0.2 ng Au from platform atomization at 1900 "C pyrolysis temperature 900°C; ramp time 10 s; and hold time 20 s 1 1100 I 1050 1000 950 b 900 - EL 850 > bn 800 750 700 0.1 1 10 100 1000 10000 Mass Rhing Fig.3 Influence of modifier (Rh) mass on the maximum loss-free pyrolysis temperature ramp time 10 s; and hold time 20 s fier masses on the pre-atomization losses of Au at a pyrolysis temperature of 1OOO"C was examined; these losses were decreased even when 7.2 ng of Rh was present and Au could be pyrolysed for 10 s at 1000°C without significant losses. These experiments show that the mass of the modifier has to be carefully optimized when the temperature programme for an analysis is developed. Generally the optimum amount of the modifier tested was 1 pg. This amount of modifier could stabilize the Au signal at 10oO"C for 30 s with adequate sensitivity. In contrast > 5 pg of Re had to be used to obtain the same results. These amounts were used for the remaining study.Ascorbic acid was similarly tested as a potential chemical modifier for Au. Its mass was optimized as above and it was found that using platform atomization 10 pg of ascorbic acid were enough to recover the Au signal at a pyrolysis temperature of 900 "C (Fig. 4A and B). Therefore 20 pg of ascorbic acid were used with platform atomization for the remaining study. When ascorbic acid was contained in the gold solutions the latter were also stable during the measurements period. The pyrolysis ( Tpyr) and atomization temperatures (T,,,,) for Au determinations were optimized in the absence and in the presence of different chemical modifiers. The maximum Tpyr values are presented in Table 2. When no chemical modifier was present the maximum Tpyr was 700°C.In the presence of 1 pg of modifier (Cu Pd Pt Rh or Ru) the maximum Tpyr was 1000 'C. In contrast 5 pg of Re could stabilize Au up to 900 "C. One possible reason for this difference in behaviour between Re and the other modifiers is that the former could probably be driven from the tube during the pyrolysis step because of the formation of volatile oxides (mainly Re20,) in the tempera- ture range 400-900"C.18 In the presence of 20 pg of ascorbic acid the Tpyr was also 1OOO "C and it could be increased further to 1150 C when the mixed modifier ascorbic acid-Pd or ascorbic acid-Rh-Re was used. This increase in Tpyr in the presence of ascorbic acid could be ascribed to the formation 0 c (D 53 0.1 % n a n 1 1 I " 0 20 40 60 80 Amount of ascorbic acid/pg Fig.4 Influence of increasing amounts of ascorbic acid on A peak height and B integrated absorbance of 0.2 ng Au from platform atomization at 1900 T pyrolysis temperature 900 'C; ramp time 10 s; and hold time 20 s Jouriial of Analyticul Atomic Spectrometry March 1995 Vol. 10 223Table 2 Maximum pyrolysis temperatures (T,,,,) for the determination of 0.2 ng of gold in the presence of different chemical modifiers With 20 pg NH4SCN 700 lo00 8 50 lo00 lo00 lo00 loo0 900 lo00 With 20 pg ascorbic acid lo00 1 100 ND* 1150 ND 1050 ND lo00 1150 0 1 2 3 4 Atom i zat io n ti me/s * ND Not determined. of a carbon residue6.l9 during the thermal decomposition of ascorbic acid in the pyrolysis step. The optimum atomization temperature (Tatom) when the peak height was measured was 1900 "C.However when the integrated absorbance was meas- ured in the absence of chemical modifiers the best sensitivity was obtained at a T,, of 1500"C but the signal lasted for more than 5 s. When the T,, value was raised to 1900"C lower sensitivity was obtained owing to increased diffusional losses. Similar results were observed in previous studies.6*20 In the presence of chemical modifiers broad peaks were observed at a T,, value of 1500°C. Particularly in the presence of the mixed modifier ascorbic acid-Pd a very small signal was obtained. Therefore in the presence of chemical modifiers a T,, value of at least 1800°C had to be used in order to obtain a sharp absorption peak. The atomization temperature used in the rest of the study was 1900°C. Peak profiles The atomization signals for 0.2 ng of Au in the presence of different chemical modifiers are shown in Figs.5 and 6. It is apparent that the atomization efficiency is increased in the presence of the mixed modifiers Rh-Re ascorbic acid-Rh and ascorbic acid-Pd (higher integrated absorbance). It is also 0.4 1- I 0.3 0 m 0 + 0.2 a 0.1 0 I B 0 1 2 3 4 At om iza t ion ti me/s Fig. 5 Atomization signals for 0.2 ng of Au in the presence of 10 pg NHJCN A no metallic modifier Tp,,=700"C peak height 0.170 integrated absorbance 0.072 s; B 1 pg Cu TPF= 900 "C peak height 0.31 3 integrated absorbance 0.095 s; and C 1 pg Rh + 5 pg Re Tpyr = SOOT peak height 0.217 integrated absorbance 0.127 s. In all instances platform atomization was used at 1900 OC Fig. 6 Atomization signals for 0.2 ng of Au in the presence of 20 pg ascorbic acid A no metallic modifier Tpyr=900 "C peak height 0.210 integrated absorbance 0.086 s; B 1 pg Rh T,,,=900 "C peak height 0.240 integrated absorbance 0.131 s; C 1 pg Pd Tp,,=900"C peak height 0.262 integrated absorbance 0.120 s.In all instances platform atomization was used at 1900 "C obvious that in the presence of ascorbic acid the appearance of Au atoms was delayed. This could be attributed to the formation of a carbon residue" during the thermal decompo- sition of ascorbic acid and diffusion of gold atoms from this residue at the atomization step." Matthews and McGahan6 ascribed the enhancement in the sensitivity in the presence of proteins and amino acids to the carbon produced by their decomposition.that gold atoms in the gas phase are produced by evaporation of the pure metal. McNally and H ~ l c o m b e ~ ~ and Fonseca et found that Au shows a fractional order of release and this was attributed to the formation of microdroplets and evaporation from their edges. However a first-order process was obtained when the forma- tion of adatoms was expected e.g. when the number of active sites on the graphite surface was decreased.24 Qiao and Jackson" found that in the presence of the mixed modifier Pd-Mg the atomization of gold is a clearly first-order process and proposed that atomization occurs through diffusion out of molten Pd. Aller2' recently found an order of release of 0.8 when small amounts of Au (0.01-5 ng) were atomized either from platform or wall. The order was 0.94 when 0.1% of vanadium was pre~ent.~' The results obtained in this study agree with the reported findings.Table 3 shows that the peak time increases with increasing amounts of gold and so the order of release is fractional. In the presence of 1 pg of Cu 5 pg of Re or 20 pg of ascorbic acid the atomization process is approximately of first order since the peak time remains constant with increasing amounts of gold (Table 3). Similar results were obtained with the other modifiers (Rh Pd). In order to investigate the dependence of order of release on the amount of the modifier added the effect on peak times of increasing the amount of Au with different amounts of the same modifier was examined. The results in Table 4 show that 0.1 pg of Rh was sufficient to convert the fractional order of release to a first-order process.It is likely that modifiers such as Rh and Pd which are present as metals in the pyrolysis step," obstruct the formation of gold microdroplets and a solid solution could be formed therefore the atomization process occurs from the bulk of the modifier which is a first- It is Table 3 Peak times for increasing amounts of gold in the presence of different chemical modifiers Au mass/ ng 0.1 0.2 0.4 0.8 Peak time/s (n =4) 20 pg NH4SCN 1 pg c:u 5 Pg Re 0.58 -+ 0.02 0.70 k 0.04 0.84 & 0.04 0.88 * 0.01 0.88 f0.02 0.86 f 0.04 0.88 & 0.06 0.86 f 0.04 1.04 0.04 1 .OO k 0.04 1 .00 f 0.06 1.04 & 0.06 20 pg ascorbic acid 1.06 0.02 1.10 f 0.05 1.09 k 0.06 1.04 & 0.04 224 Journal of Analytical Atomic Spectrometry March 1995 Vol.10Table 4 Peak times for increasing amounts of gold in the presence of different masses of chemical modifiers Au mass/ ng 0.1 pg Rh 0.04 0.84 f 0.06 0.1 0.85 f 0.02 0.2 0.86 f 0.04 0.4 0.88 f 0.01 1 .o 0.88 k0.06 2.0 0.86 k 0.04 Peak time/s ( n = 3) 1 pg Rh 0 . 5 ~ ~ Re 1 pg Re 1.01 f 0.04 0.44 f 0.06 0.58 f 0.02 1.06 f 0.02 0.63 f 0.04 0.66 f 0.04 1.02 f 0.04 0.68 f 0.02 0.70 f 0.04 1.06 k 0.08 0.74 f 0.02 0.72 f 0.06 1.02 f 0.04 0.72 f 0.04 0.80 k0.02 1.08 f 0.06 0.74 & 0.06 0.84 f 0.04 order process." On the other hand it is likely that Re reacts with carbon at the active sites on graphite and decreases the number of active sites; thus the formation of adsorbed atoms rather than microdroplets is expected. However in the presence of a mass of Re of < 5 pg the atomization process was more complicated.It is possible that owing to Re loss during the pyrolysis step," not enough Re remains to convert the atomiz- ation process to a first-order release. Therefore higher masses of Re seem to be necessary to obtain a first-order rate of release. Table5 Sensitivity and Limits of Detection (LOD) for the determi- nation of gold LODIpg I - ' Modifier SCN- SCN-+Cu SCN- + Pd SCN- + Rh+ Re Ascorbic acid Ascorbic acid + Pd Ascorbic acid + Rh PH * I A* 0.26 0.23 0.23 0.23 0.2 1 0.23 0.10 0.12 0.14 0.19 0.20 0.20 0.1 1 0.16 PH 4.6 3.3 3.5 3.8 4.2 3.5 3.7 IA 12.2 9.1 9.2 7.3 9.8 7.1 6.8 *IA integrated absorbance; PH peak height; m, characteristic mass. Table6 Recovery of gold from spiked samples of serum and urine.The samples were subjected to a 40-fold dilution with a 0.5 g I-' NH4SCN solution. The gold content in all samples was below the limit of detection Au added/ P8 1 - ' Serum 1 5 10 20 40 Serum 2 10 Urine 1 5 10 20 40 Urine 2 10 Recovery (YO) No modifier 92.2 103.6 100.7 99.7 103.6 103.9 97.0 100.4 97.5 96.5 cu 96.7 102.0 101.0 100.5 101.0 103.5 97.4 100.5 99.8 99.6 Rh-Re 93.6 104.0 99.5 99.2 99.0 99.2 96.8 100.2 98.8 97.5 Analyticul .figures of merit Calibration curves were constructed following the procedure described under Experimental and applying the temperature programme given in Table 1. The results are summarized in Table 5. The characteristic masses derived from integrated absorbance measurements and in the presence of the mixed modifiers Rh-Re ascorbic acid-Pd and ascorbic acid-Rh were in agreement with those previously reported.6*20,26 The relative standard deviations of nine replicate injections of a solution containing 5 pg 1-' Au in the presence of 0.5 g 1-' NH,SCN were 3.2% and 3.6% for measurements of peak height and integrated absorbance respectively with platform atomization. The corresponding values for other modifiers (in parentheses) are 4.4% and 4.0% (Cu) 4.5% and 4.9% (Rh-Re) 1.8% and 3.6% (ascorbic acid) 1.4% and 1.5% (ascorbic acid-Pd) and 1.4% and 0.9% (ascorbic acid-Rh).Probably the presence of ascorbic acid contributes to an early formation of noble metals which in turn increases the overall reproducibility. Determination of gold in biological fluids When platform atomization was used the slopes of the cali- bration graphs in 40-fold diluted serum samples were higher than in aqueous gold solution because of the presence of serum proteins which was found to increase the Au signal.6 When matrix-matched curves were constructed from 400-fold diluted serum samples the slopes were similar to those obtained with aqueous standards.However double peaks or peaks with shoulders were obtained in the absence of modifier especially when old tubes were used. Double peaks were also obtained when the samples were diluted with 0.2 YO v/v Triton X-100. These peaks were not observed when the pyrolysis temperature was raised to 900°C. Peaks with shoulders were occasionally observed in the presence of ascorbic acid. This did not occur in the presence of chemical modifiers where constant results were obtained during the life of the tube.In order to test the accuracy of the method developed with or without modifiers recovery experiments were carried out with 40-fold diluted spiked samples (serum and urine) contain- ing no detectable Au and with serum from a patient undergoing chrysotherapy. The results are presented in Tables 6 and 7 respectively. The recoveries were almost quantitative when stabilized temperature platform furnace conditions were ful- filled. It is evident that platform atomization without modifiers is adequate for good recoveries to be obtained. Nevertheless for the reasons mentioned above the procedure seems to work better in the presence of modifiers Total and free Au content of samples obtained from two patients are presented in Table 8.Generally good precision was obtained (relative standard deviation 1-4%). CONCLUSIONS In the determination of gold in serum and urine complete recovery and adequate sensitivity were obtained only when Table 7 Recovery of gold added to serum of a patient with and without chemical modifiers -~ ~~~ ~ Without chemical modification c u modification Rh-Re modification Au added/ Au found/ Recovery Clg 1-' clg 1-' ( O/O) 0.0 8.49 5.0 14.04 111.0 10.0 19.47 109.8 20.0 29.03 102.7 40.0 48.05 98.9 Mean recovery (YO) 105.6 Au found/ Recovery clg 1-' ( Olo 1 8.87 14.06 103.7 19.02 101.5 29.72 103.4 49.35 100.7 102.3 Au found/ Recovery Pi3 1-' (O/O) 8.68 13.79 102.1 18.80 101.2 28.58 99.5 48.08 98.5 100.3 Journal of Anulyticul Atomic Spectrometry March 1995 Vol.10 225Table8 going chrysotherapy Determination of gold in serum samples of patients under- Modifier None c u Rh-Re Total Au*/mg 1-' Sample 1 3.57 f 0.13 3.47 f 0.04 3.44 rt 0.03 Sample 2 4.20 f 0.03 4.07 & 0.07 4.0 1 f 0.04 Sample 1 1.98f0.11 1.88 * 0.04 1.77 f 0.05 Sample 2 2.10 f 0.04 2.01 0.03 2.04 f 0.06 Free Au*/mg 1 - * Mean value of three determinations f 1 sb. platform atomization was used. Especially in the presence of chemical modifiers constant well-shaped signals were obtained throughout the measurements. Gold solutions stored in the autosampler cups during the measurements were stable only in the presence of 0.5 g 1 - * NH4SCN or 1.0 g 1-' ascorbic acid. In the presence of NH,SCN gold solutions with concen- trations of a few pg 1-' stored in plastic containers were stable for a week. N.T. thanks the Organizing Committee of BNAAS for a student bursary. The authors would also like to thank Prof. W. Frech for constructive comments during the symposium. REFERENCES 1 Slavin W. Graphite Furnace AAS. A Source Book Perkin-Elmer 2 Egila J. Littlejohn D. Ottaway J. M. and Xiao-quan S. J. Anal. At. Spectrorn. 1987 2 293. 1984 pp. 104-106. 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 Slavin W. Carnrick G. R. Manning D. C. and Pruszkowska E. At. Spectrosc 1983 4 69. Xiao-quan S. Egila J. Littlejohn D. and Ottaway J. M. J. Anal. At. Spectrom. 1987 2 299. Garcia-Olalla C. and Aller A. J. Anal. Chim. Acta 1991 252 97. Matthews D. O. and McGahan M. C. Spectrochim. Acta Part B 1987 42 909.Zhao Z. Jones W. B. Tepperman K. Dorsey J. G. and Elder R. C. J. Pharm. Biomed. Anal. 1992 10 279. Kamel H. Brown D. H. Ottaway J. M. and Smith W. E. Analyst 1976 101 790. Turkall R. M. and Bianchine J. R. Analyst 1981 106 1096. Tsalev D. L. Slaveykova V. I. and Mandjukov P. B. Spectrochim. Acta Rev. 1990 13 225. Tsalev D. L. Slaveykova V. I. and Mandjukov P. B. Fifth Colloquium Atomspektrometrische Spurenanalytik ed. Welz B. Bodenseewerk Perkin-Elmer Uberlingen 1989 pp. 177-205. Frech W. Li K. Berglung M. and Baxter D. C. J. Anal. Ar. Spectrom. 1992 7 141. Mandjukov P. B. Vassileva E. T. and Simeonov V. D. Anal. Chem. 1992,64 2596. Brajter K. and Slonawska K. J. Anal. At. Spectrom. 1987,2 167. L'vov B. V. and Frech W. Spectrochim. Acta Part B 1993 48,425. Koide M. Hodge V. Yang J. S. and Goldberg E. D. Anal. Chem. 1987 59 1802. Qiao H. and Jackson K. W. Spectrochim. Acta Part B 1991 46 1841. Haug H. O. J. Anal. At. Spectrom. 1992 7 451. Volynsky A. B. Tikhomirov S. V. Senin V. G. and Kashin A. N. Anal. Chim. Acta 1993 284 367. Welz B. Schlemmer G. and Mudakavi J. R. J. Anal. At. Spectrom. 1992 7 1257. Rowston W. B. and Ottaway J. M. Analyst 1979 104 645. Smets B. Spectrochim. Acra Part B 1980 35 33. McNally J. and Holcombe J. A. Anal. Chew. 1987 59 1105. Fonseca R. W. McNally J. and Holcombe J. A. Spectrochim. Acta Part B 1993 48 79. Aller A. J. Anal. Chim. Acta 1994 292 317. L'vov B. V. Spectrochim. Acra Part B 1990 45 633. Puper 4/04 928 D Received August 10 1994 Accepted November 14 1994 226 Journal of Anulyticul Atomic Spectrometry March 1995 Vol. 10

ISSN:0267-9477    

DOI:10.1039/JA9951000221

出版商:RSC    

年代:1995

数据来源: RSC

摘要:

Studies on Solvent Extraction to Determine Iodide Indirectly by Electrothermal Atomic Absorption Spectrometry* Journal of Analytical 1 Atomic Spectrometry PILAR BERMEJO-BARRERA MANUEL ABOAL-SOMOZA ANTONIO MOREDA-PIREIRO AND ADELA BERMEJO-BARRERA University of Santiago de Compostela Department of Analytical Chemistry Nutrition and Bromatology Faculty of Chemistry E-15706-Santiago de Compostela (La Corun'u) Spain Preliminary studies addressed to the determination of iodine as iodide by indirect ETAAS and applying solvent extraction are described. The work was carried out to achieve a selective quantification of the iodide contained in the complex formed with mercury(11) in acid medium but selectivity problems were found. The formation of an ion association complex between mercury(rI) iodide and 2,2'-dipyridyl was studied.Several parameters were optimized related to the extraction process the optimal pH interval 6.8-7.6; volume of extractant 2 ml; and shaking time 30 s are proposed. Regarding the performance of the measurements the optimized graphite furnace temperature programme was also studied yielding 200 and 1OOO"C as the best temperatures for the mineralization and atomization steps respectively. The standard calibration and the standard additions methods were performed. The RSDs of the procedure were between 4.9 and 8.0% for the standard calibration and from 3.2 to 15.6% for the standard additions method. Keywords Iodide; solcent extraction; indirect method; electrothermal utornic ubsorption spectrometry The direct determination of non-metals presents problems owing to the fact that these species exhibit their main and most sensitive resonance lines in the vacuum ultra violet (VUV) spectral region below 190-200 nm.To overcome this difficulty the analyst can try to work at less sensitive secondary resonance lines that lie inside the conventionally attainable spectral range. But the sensitivity would be worse and spectral and chemical interferences could be important. Also work can be aimed at performing the measurements at the VUV wave- lengths that would give direct atomic absorption spectrometry (AAS) methods (although at the cost of increased instrument complexity'). The literature shows that this is a real possibility; for example Adams et aL2 and Lowe et aL3 have reported the determination of iodine by direct AAS procedures.However the necessary modifications to the instruments required for these direct methods cannot always be carried out. Both direct and indirect AAS methods for the quantification of iodine are described in the literature e.g. L'vov and Khartsy~ov~.~ and Kirkbright and Wilson6 reported the direct determination of iodine by electrothermal AAS (ETAAS) some years ago. In recent years the main efforts have been in the development of indirect AAS procedures rather than the optimization of direct methods. Some of the indirect methods proposed to determine iodine are based on the reducing effect of iodide in acid medium e.g. on chromium(v~)~ or selenium(~v).~,~ Iodide can also be precipi- * Presented at the Seventh Biennial National Atomic Spectroscopy Symposium (BNASS) Hull UK July 20-22 1994.tated with a silver(]) solution," but chloride and bromide interfere seriously. Methods based on the formation of com- plexes between iodide and metals such as mercury(II) have been described."-15 In these cases the determination of iodide is carried out by means of the measurement of the atomic absorption of the metal combined with it. Iodide also forms chelate ions with 1,lO-phenanthroline and metals such as cadmium(r1) or iron(II) the chelate ions can be extracted into an organic solvent to enable iodine determinati~n.'~.'~ Chakraborty and Das proposed recently the determination of iodideI8 and iodate" through the formation of ion association complexes between mercury(n) iodide (or iodate) and 2,2'- dipyridyl subsequent extraction of these complexes into an organic solvent and final quantification of the mercury con- tained in the complexes by cold vapour AAS.The present work is a series of preliminary studies whose main objective is the development of a method to determine iodine indirectly as iodide by ETAAS preceeded by a solvent extraction procedure. Two parts will be described in the following experimental work. The selective extraction into an organic solvent of the complex that mercury(i1) and iodide form in acid and the preparation and extraction of the ion association complex that mercury@) and iodide form with 2,2'-dipyridyl are described." In both cases mercury is the element whose atomic absorption is measured. EXPERIMENTAL Apparatus A Perkin-Elmer Model 1 lOOB atomic absorption spectrometer was used equipped with an HGA-700 graphite furnace atom- izer and an AS-70 autosampler.A mercury electrodeless dis- charge lamp connected to its power supply and operated at 4 W provided the 253.7 nm line at which the measurements were performed. A deuterium lamp was the background correc- tion system employed and the spectral bandwidth of the monochromator was 0.7 nm. Pyrolytic graphite coated graph- ite tubes with pyrolytic graphite (L'vov) platforms (both sup- plied by Perkin-Elmer GmbH Uberlingen Germany) were also used. Reagents Scharlau Sentmenat Barcelona Spain. grade Merck Poole UK. the atomizer and to purge internally. grade Scharlau. Acetic acid glacial (AcOH ). HPLC reagent grade 99.8% Ammonium dihydrogen orthophosphate.Analytical-reagent Argon NSO (99.9990%) purity. Used as sheathing gas for Diammonium hydrogen orthophosphate. Analytical-reagent Journal of Analytical Atomic Spectrometry March 1995 Vol. 10 2272,2'-dipyridyl. Analytical-reagent grade Merck Darmstadt Germany. lsobutyl methyl ketone (IBMK). Analytical-reagent grade Merck Poole UK. Mercury( 11,) nitrate stock standard solution. l.OOO+ 0.002 g 1- '. Prepared in HN03 1 mol 1-' Panreac- Montplet & Esteban S.A. Barcelona Spain. Nitric acid. Suprapur (69-70.5 YO) maximum mercury content 0.001 pg ml-' Merck Poole UK. Palladium powder 99.999%. Aldrich Chemical Milwaukee WI USA. Aqueous stock solution. Prepared2' by dissolving 300 mg of palladium powder in 1 ml of nitric acid and diluting to 100 ml with ultrapure water.Other solutions were prepared by suitable dilution of this one with the required palladium concentration. If dissolution was incomplete 10 pl of hydrochloric acid (Suprapur 35.0% maximum mercury content 0.001 pg ml-' Merck) were added to the cold nitric acid and heated to gentle boiling in order to volatilize the excess of chloride. Palladium Stock Solution in IBMK. To 1 ml of palladium aqueous solution (prepared as above and containing a given concentration of palladium) placed in a glass extraction tube was added 2ml of 2% m/v NaDDC solution 2ml of NaAcO/AcOH buffer solution (pH 4.6) and 4 ml of ultrapure water. After adding 3 ml of IBMK the content of the tube was shaken for 1 min and allowed to stand for the phases to separate (few minutes).Then the organic (IBMK) layer con- taining the palladium as a diethyldithiocarbamate complex (PdDDC) was collected to be used as chemical modifier and the aqueous phase was discarded. Potassium iodide. ACS reagent grade Sigma Chemical St. Louis MO USA. Sodium acetate (NaAcO). Analytical-reagent grade Merck Germany. Sodium diethyldithiocarbamate (NaDDC ). Analytical reagent grade Merck. Ultrapure water resistivity 18 MR cm. Obtained by means of a Milli-Q Water Purification System Millipore Bedford MA USA. All material (glassware plasticware and PTFE-ware) was washed and kept for at least 48 h in 10% v/v nitric acid and then rinsed several times with ultrapure water before use. It was necessary to use autosampler cups made of PTFE since the IBMK dissolves the usual plastic autosampler cups.Procedure Preparation and extraction of mercury(rr)-iodide complex Mercury(II) iodide and nitric acid solutions were dispensed into a glass extraction tube to give concentrations of about 300pgl-' 4Opg1-' and lop6 mol 1-' respectively. The volume was then adjusted to 5 ml with ultrapure water. Then 3 ml of IBMK were added to the tube and the mixture was shaken for 2 min and set aside for a few minutes to allow the phases to separate. Finally an aliquot of the upper organic (IBMK) layer (that contained the complex) was transferred to an autosampler cup to be analysed by ETAAS. Preparation and extraction of mercury( zr)-iodide-2,2'-dipyridyl ion association complex Aliquots of the following reagent solutions were dispensed into a separatory funnel in the order given:I8 mercury(I1) solution 0.06% m/v 2,2'-dipyridyl solution iodide solution and NH,H,PO,-( NH,),HPO buffer solution (pH between 6.8 and 7.6) t o give a volume of 5 ml.After that 2 ml of IBMK were added and the mixture was shaken for 30 s. Subsequently the funnel was set aside for a few minutes for the phases to separate. Finally a portion of the upper organic (IBMK) layer that contained the ion association complex was added to an aliquot of palladium solution in IBMK and analysed by ETAAS. RESULTS AND DISCUSSION Mercury (11)-Iodide Complex In 1981 Nomura and Karasawa" reported the indirect deter- mination of iodide by ETAAS through the formation of a complex between mercury(I1) and iodide in acid medium (stoichiometry proposed HgI,) and subsequent selective measurement of the absorption of radiation by that mercury linked to iodide.This work was the basis upon which we developed a method to determine iodide in tap water.20 As a continuation of that work the following research applies solvent extraction processes in order to separate and quantify only the mercury contained in the complex. By doing this an improvement both in selectivity and sensitivity is at least theoretically expected. Optimization of the graphite furnace temperature programme To obtain the best conditions to perform the measurements several proofs were carried out. The results are shown in Fig. 1 where the mineralization and atomization curves as well as the optimal temperatures in each case are shown.First an attempt was made to measure the extract for mercury without any chemical modifier. This gave 150 and 1200°C for the mineralization and atomization steps respectively (curves A Fig. 1). The possibility of using a chemical modifier was considered by adding an aliquot of palladium aqueous solution. This implied an enlargement of the time spent on each measurement because the extract and the modifier solution cannot be mixed. They had to be injected separately into the atomizer and an additional drying step was required between the injections of extract and modifier. These measurements yielded curves B and C (Fig. l) depending on the solution first injected. A different way of adding the chemical modifier was studied by using an IBMK solution containing palladium as PdDDC. The procedure to prepare this solution is described under experimental. The extraction process into IBMK is based on the work of Takeuchi et a1.22 and Tserovsky et al.23 Although this extractic 0.800 v) 0.600 .0 m e 2 0.400 m C I procedure is being improved (ie. re-optimized) B n A C D I 3 \ A C & Bl i I 1 I I 1 I 0 400 800 1200 1600 2000 Temperature/ C Fig. 1 Curves of mineralization and atomization for the complex at different conditions A without chemical modifier; B injecting the chemical modifier drying and injecting the extract; C injecting the extract drying and injecting the chemical modifier; and D using the modifier IBMK solution 228 Journal of Analytical Atomic Spectrometry March 1995 Vol. 10it is supposed to yield an extraction close to 100%. As illustrated in Fig.1 (curves D) with the use of this IBMK palladium solution as chemical modifier the optimal mineraliz- ation and atomization temperatures obtained were 200 and 900 "C respectively. From the mineralization and atomization curves displayed in Fig. 1 it is difficult to select the optimal conditions of measurement. However if the shape of the peaks obtained is also taken into account it can be concluded that the best performance is achieved when the modifier (palladium) is added and when it is added as an IBMK solution. All these considerations led us to use the optimized graphite furnace temperature programme described in Table 1. An additional cleaning step at 1000°C is also included to remove any contamination on the atomizer before the beginning of the next measurement.Composition of mercury( II)-iodide complex The molar ratio method was used to discover the composition of the complex. A series of extracts were prepared (following the optimized procedure described above) in which the concen- tration of mercury(I1) was constant (9.97 x mol I-' in the initial aqueous phases) and that of iodide varied between 1 x mol 1-' (in the initial aqueous phases) giving a molar ratio of iodide mercury of 0.1-5.0. The extracts were analysed by molecular absorption spectrophotometry using a Hewlett-Packard 8452A Spectrophotometer. The results obtained (Fig. 2) led us to conclude that the complex has a molecular formula that can be represented as [Hg13]- . This disagrees with Nomura and Karasawa" who reported a composition of HgI for the complex.In any case these molecular formulas refer to complex equilibria in solution and not to what is happening inside the graphite furnace and are of little help in predicting behaviour in AAS.24 and 5 x Table 1 Graphite furnace temperature programmes; values in par- entheses correspond to the programme optimized for the ion association complex T Ramp Hold Ar Flow rate Step /"C /s I S /ml min-' Dry 1 100 5 30(15) 300 Dry 2 110 5 30(15) 300 Mineralization 200 10 15 300 Atomization 900(1000) 0 7 0 (Read) Clean lo00 (1 100) 1 3 300 Operating Power Hg Electrodeless discharge lamp/W WavelengthInm Spectral handwidth/nm Background corrector Integration times Furnace Volume injected/pI 4 253.7 0.7 Deuterium lamp 7 Pyrolytic graphite coated graphite tubes with pyrolytic graphite (L'vov) platforms.20 0.800 v) 0 \ 0.600 2 $ a 0.400 a 4- ! CI ; 0.200 - n 0 1 2 3 4 5 Iodide Mercury (11) molar ratio Fig. 2 Application of the molar ratio method to find the composition of the complex Selectivity One of the main difficulties of indirect methods is to achieve an adequate selectivity. In our case the method would be selective enough if a clear difference could be established between the extraction with and without iodide present in the initial aqueous phase i.e. if the extraction only took place when iodide was present. To clarify this point two series of extractions were performed. In the first series the initial aqueous phases contained mercury and iodide (also nitric acid see procedure) in molar ratios of mercury iodide varying from 1 to 6.The mercury was added in excess of iodide in spite of the formula of the complex obtained to ensure that all the iodide present becomes involved in the formation of the complex. In the other series the starting aqueous solutions contained mercury (in the same concentrations as the corre- sponding solutions of the first series) and no iodide. Both series of extracts were then analysed by ETAAS giving the results in Table 2 where the concentrations of mercury(r1) and iodide in the initial aqueous solutions are also indicated. In view of those data it can he said that there is indeed a variation between both situations (i.e. mercury with and without iodide) but such a difference seems not to be enough to achieve a quantitative as well as selective determination of iodide by the described method or at least in its present form.Mercury(rr)-iodide-2,2'dipyridyl ion association complex In order to improve the selectivity a clear differentiation between performing the extraction when iodide was present (in addition to mercury) and not present was needed. This led to an adaptation of the procedures described by Chakraborty and Das1'*l9 to determine iodide by ETAAS. Firstly to discover an approximate idea of the selectivity of the method for iodide extractions were carried out (following the procedure described for iodide") on two aqueous solutions which contained besides the reagents needed mercury(I1) and iodide and only mercury. The organic solvent used was IBMK instead of ethyl acetate." Both extracts were subjected to Table 2 Comparison of absorbance when different amounts of mercury (11) are extracted with and without iodide; concentration of iodide in the initial aqueous phases was 1.5 pmol I - ' Concentration of Hg" in initial aqueous 1.5 3.0 4.5 6.0 7.5 9.0 Molar ratio Hg":I- 1 2 3 4 5 6 Absorbance solution/pmol I - Mercury (11) and iodide 0.2 15 0.367 0.498 0.8 10 0.695 1.144 Mercury (11) 0.124 0.370 0.385 0.435 0.356 0.744 Journal of Analytical Atomic Spectrometry March 1995 Vof.I0 229ETAAS using the temperature programme optimized for the mercury-iodide complex (see Table 1 ) and the palladium IBMK solution as chemical modifier. Analytical blanks were also prepared and measured. The data showed that the absorbance signal obtained in the absence of iodide (k when only mercury was present) was increased > 5 times when iodide was present (at the same mercury concentration); a quantitative difference between both situations.The results recorded for the blanks were acceptable. the absorbance signal. In view of this the minimal amount of 2,2'-dipyridyl was considered to be optimal; about 65 pl of a 0.06% m/v 2,2'-dipyridyl solution. Eflect of pH The pH usually seriously affects the extraction and complex formation processes. To find the optimal pH interval a new series of extracts was prepared from aqueous solutions to which aliquots of NH4H2P04-( NH4)2HP04 buffer solution were added to give a pH between 6.4 and 8.0. The extracts were measured as usual by ETAAS and the data produced (Fig.4) indicate that 6.8-7.6 is the optimal pH interval since its variation does not alter the absorbance signal significantly. To one of the separatory funnels ultrapure water (pH 6.0) was added instead of buffer solution giving an integrated absorbance of 0.161 (see Fig. 4). If this value is compared with the absorbances recorded when buffer solution was added it can be concluded that the extraction is much more significant with buffer than with water. Optimizution of the gruphite furnuce temperature programme The programme used for the mercury-iodide complex (Table 1) was taken as a basis. According to the procedure proposed," an IBMK extract containing the ion association complex formed with mercury(II) iodide and 2,2'-dipyridyl was prepared and after adding an amount of palladium modifier (prepared in IBMK) was analysed by ETAAS to find the optimal temperature and time for each step of the programme (Fig. 3).It can be seen that as for the mercury-iodide complex 200°C was the temperature considered as optimal for the mineraliz- ation step and 1000°C for the atomization step. Again the choice of this latter temperature was made after considering not only the highest signal but also the shape of the peaks obtained. Therefore although at temperatures of 850 and 900 "C greater signals were obtained than at 1000 "C the peaks were wider and lower when atomizing at those temperatures than at 1OOO"C. This was taken into account for proposing the atomization temperature. The problem encountered with the shape of the peaks is discussed later.The optimized temperature programme is also shown in Table 1. As the one optimized for the mercury(r1)-iodide complex the drying step has been divided into two and a final cleaning step (at 1100 T) has been included. EfSect oj'uolume qfextructant For a given concentration of analyte the smaller the volume of solvent the higher the concentration of analyte in the extract and accordingly the higher the absorbance signal recorded. A series of aqueous solutions that contained fixed concentrations of mercury(r1) (4.14 x mol l-') iodide (4.35 x lop6 rnol 1-') and 2,2'-dipyridyl (4.99 x lo-' rnol I - ' ) were prepared and extracted into different amounts of IBMK (1-5 ml). The results (Fig. 5) show an important decrease in v) 5 0.250 C m 2 v) n m Eflect of amount of 2,2'-dipyridyl A series of extracts were prepared from aqueous solutions containing fixed concentrations of mercury(r1) (4.15 x lo-' rnol 1-') and iodide (4.36 x low4 mol 1-') and concentrations of 2,2'-dipyridyl varying between 8.34 x and 8.33 x mol 1-'.The results obtained when these extracts were subjected to ETAAS indicate that the absorbance remains constant when the concentration of 2,2'-dipyridyl varies inside the interval studied (and for the concentrations of mercury(r1) and iodide studied). This not only agrees well with the results reported previously," but enlarges the interval within which the amount of 2,2'-dipyridyl can be varied without affecting 0 1 0.150 5.5 6.0 6.5 7.0 7.5 8.0 8 PH Fig. 4 Influence of pH of the buffer solution on absorbance signal 0.600 0.500 0.280 v) 0 C m -.$ 0.210 % d m -0 c. Q 0.140 0 C - 0.070 L I 0 p 0.400 0.300 4 -0 c. 0.200 I - 0.100 1 I I I I 1 1 3 4 0 2 5 Volume of IBMK/ml 0 250 500 750 1000 1250 1500 TemReratUfd C Fig. 3 Curves of mineralization and atomization using palladium in IBMK as chemical modifier for the ion association complex Fig. 5 Effect of amount of extracting solvent IBMK on absorbance signal 230 Journul of' Analytical Atomic Spectrometry March 1995 V d . 10the absorbance signal when the volume of IBMK is > 2 ml. Since a gradual decrease in the signal as the volume of extracting solvent increases can be expected the sudden decrease observed is unusual. The reason for this behaviour is still being investigated but the extraction process itself may have something to do with it.In spite of this and as indicated in Fig. 5 a 2 ml volume of extractant is recommended. Effect of shaking time The optimal time of agitation was studied by preparing five extracts following the usual procedure. Each extraction was performed with stirring times from 15 s to 2 min. Results indicate that the absorbance signal remains constant when the funnel is shaken for more than 30 s. If the funnels are stirred for a shorter time the signal obtained is slightly smaller. Therefore 30s was considered to be a shaking time long enough for the extraction to take place quantitatively although perhaps 15 s could be acceptable too provided that the change in the signal between that stirring time and 30 s is small. Stundard calihrution und stundurd additions curljes To obtain calibration and standard additions graphs extracts were prepared following the optimized procedure starting from aqueous solutions that contained concentrations of iodide between 0 and 75 pg 1-'.The standard additions method was applied to a sample of tap water. Although extraction processes usually give poor precisions in the measurements as well as linearity in the calibration graphs it was found that the graphs obtained are linear up to iodide concentrations (always referred to the initial aqueous solution) of about 75 pg 1-'. The slopes of calibration and standard additions graphs were 1.00 x lop3 and 8.84 x respectively giving a difference of about 13.1%. In this way the curves obtained also suggest that the measurements should be done by means of standard additions.Precision A brief study of the repeatability of the measurements was carried out using the data recorded to obtain the calibration curves. The number of data recorded as well as the mean values with their SDs and RSDs for each level of iodide concentration are shown in Table 3. The blank was 0.041 0.002 s. In the case of the standard additions method the usual trend in RSD and SD (k decrease as the signal increases) is observed while it is not in the standard calibration method. The SDs obtained from both methods are the same order of magnitude and the KSDs although showing values that are a little high can be considered as acceptable given the complexity of the extraction procedure followed. Peuk shape Closely related to the poor precision observed in the measure- ments is the shape of the peaks obtained.An example of a Table 3 iodide in starting aqueous solutions Repeatability of measurements at different concentration of Calibration Standard additions [Iodide] SD SD /pg 1 - 1 n ( x 10-3) RSD (yo) n ( x 1 0 - 3 ) RSD 0 10 2.27 5.5 10 1.77 15.6 25 10 5.40 8.0 11 4.41 8.9 50 10 5.20 4.9 9 2.59 3.5 75 10 6.51 5.8 10 2.41 3.2 analyte background . ..... .. ..... "=b':...::..; . 0 ~ 7 Time/s Fig. 6 Peaks obtained for analyte mercury and background peak recorded for the ion association complex is displayed in Fig. 6. The long tail exhibited by the peak could be the main cause of the bad precision achieved. As the peak does not reach the baseline a slight change in the height of the tail produces (or can produce) a significant increase or decrease in the value of the signal recorded provided that the measurements are carried out by integration absorbance.To overcome this difficulty firstly the integration time was increased with no success since the tail went on and on. Another possibility to achieve a non-tailed peak is to 'group' the signal because of the probable cause of the tail is slow atomization of all the mercury present something could be added (a suitable modifier for example) or done to make all the mercury atomize at the same time. Peak height measure- ments could be used but in view of the shape of the peaks it is not advisable since they are too wide for that mode of measurement . This problem of the shape of the peaks was observed when preparing the mercury(1r)-iodide complex also.CONCLUSIONS The formation of the ion association complex is a better choice than the mercury(I1)-iodide complex to determine iodide by indirect ETAAS combined with solvent extraction. Therefore further work will be devoted to the improvement of the latter method by studying parameters such as accuracy sensitivity and selectivity. The applicability to tap water samples or to more complicated matrices will be also an interesting aim of study. Efforts will first be made in overcoming the difficulty of the peak shapes. REFERENCES 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Dittrich K. Radziuk B. and Welz B. J. Anal. At. Spectrom. 1991 6,465 Adams M. J. Kirkbright G. F. and West T. S. Talanta 1974 21.573 Lowe M.D. Sutton M. M. and Clinton 0. E. Appl. Spectrosc. 1982 36 22 L'vov B. V. and Khartsyzov A. D. Zh. Anal. Khim. 1969,24 799 L'vov B. V. and Khartsyzov A. D. Zh. Prikl. Spectrosk. 1969 10 413 Kirkbright G. F. and Wilson P. J. Ar. Absorption Newsl. 1974 13 140 Christian G. D. and Feldman F. J. Anal. Chirn. Acra 1968 40 173 Christian G. D. Knoblock E. C. and Purdy W. C. J. Assoc. Of. Agric. Chem. 1965 48 877 Newberry C. L. and Christian G. D. J. Assoc. Off. Agric. Chem. 1965 48 322 de Almeida M. A. T. M. de Moraes S. and Barberio J. C. Report 1973 IEA-285 10 pp. Nomurd T. and Karasawa I. Anal. Chim. Acta 1981 126 241 Chuchalina L. S. Yudelevich I. G. and Chinenkova A. A Zh. Anal. Khirn. 1981 36 920 Kuldvere A. Analyst 1982 107 1343 Sun F.-S. and Julshamn K. Spectrochim. Acta Part B 1987 42 889 Journal of Analyticul Atomic Spectrometry March 1995 Vol. 10 23115 Wifladt A.-M. Lund W. and Bye R. Talanta 1989 36 395 16 Yamamoto Y. and Kinuwaki S. Bull. Chem. SOC. Jpn. 1964 37,434 17 Kumamaru T. Bull. Chem. SOC. Jpn. 1969,42 956 18 Chakraborty D. and Das A. At. Spectrosc. 1988 9 189 19 Chakraborty D. and Das A. Talanta 1989 36 669 20 Bermejo-Barrera P. Moreda-Piiieiro A. Aboal-Somoza M. Moreda-Pifieiro J. and Bermejo-Barrera A. J . Anal. At. Spectrom. 1994 9 483 21 Welz B. Schlemmer G. and Mudakavi J. R. J. Anal. At. Spectrom. 1988 3 695 22 Takeuchi T. Suzuki M. and Yanagisawa M. Anal. Chim. Acta 1966,36 258 23 Tserovsky E. Arpadjan S. and Karadjova I. Spectrochim. Acta Part B 1992 47 959 24 Komarek J. and Sommer L. Talanra 1982 29 159 Paper 4/05679E Received September 19 1994 Accepted Nouember 14 1994 232 Journal of Analytical Atomic Spectrometry March 1995 Vol. 10

ISSN:0267-9477    

DOI:10.1039/JA9951000227

出版商:RSC    

年代:1995

数据来源: RSC

摘要:

Determination and Speciation of Heavy Metals in Sediments from the Cumbrian Coast NW England UK* Journal of Analytical Atomic Spectrometry ABD ULHAFID BELAZI CHRISTINE M DAVIDSON GILLIAN E KEATING AND DAVID LITTLEJOHN Department of Pure and Applied Chemistry University of Strathclyde 295 Cathedral Street Glasgow U K GI 1 X L MARTIN McCARTNEY Scottish Universities Research and Reuctor Centre East Kilbride Glasgow U K G75 OQU Chromium manganese nickel vanadium and uranium have been determined in inter-tidal sediments collected from locations along the Cumbrian coast. Total concentrations of chromium nickel and vanadium were determined by electrothermal atomic absorption spectrometry (ETAAS). Manganese was determined by flame AAS and uranium was quantified by alpha particle spectrometry with a surface barrier detector.Two alternative approaches to calibration were investigated in an attempt to obviate the need for routine use of the full standard additions procedure. It was found that the response following a single addition of standard solution at a concentration 50% of the initial estimate could be used to correct for the minor interference effects observed. Elevated levels of chromium (39.5 rfi 0.9 pg g- ’) vanadium (33.0k0.6 pg g-’) and uranium (39.0f 1.2 E!q kg-’) were observed at Whitehaven whilst concentrations of manganese were highest in samples from more northerly locations. The uranium enhancement is due to the extraction of phosphates from ore naturally rich in radionuclides at the nearby Albright and Wilson ‘Marchon’ chemical manufacturing plant.The chromium contamination may also arise from chemical manufacturing whilst the vanadium is thought to originate from oil spillage. Interferences associated with use of the European Communities Bureau of Reference ( BCR) sequential extraction protocol were investigated and the operationally- defined speciation of chromium manganese nickel and vanadium was then determined. Chromium nickel and vanadium were found mainly in association with the residual sediment phase except at Whitehaven and Maryport where increases in the oxidizable fractions were apparent. A large proportion of the manganese at all sites was present as exchangeable species Le. soluble in 0.11 mol I - ’ acetic acid and this was not affected by sample drying (at 60°C) nor by storage (for 6 months) prior to extraction.Keywords Speciation; sequential extraction ; heavy metals; sediment; electrothermal atomic absorption spectrometry Investigation of the distribution of heavy metals in sediments is important because under different environmental conditions these materials can act either as an efficient sink for metals released to the marine environment or as a source thereof. The proportion of a particular metal which will be released if a contaminated sediment is chemically perturbed will depend critically on the metal species present.’ A variety of methods have been proposed for speciation in solid samples. Direct methods of speciation such as nuclear magnetic resonance (NMR) and infrared (IR) spectroscopies * Presented at the Seventh Biennial National Atomic Spectroscopy Symposium (BNASS) Hull UK July 20-22 1994.are usually only applicable where analyte concentrations are high. More usually indirect methods must be used. Amongst the most popular of these are sequential extraction procedures in which samples are treated with a succession of reagents to liberate metals with different affinities for the matrix. The most widely applied procedure of this type is that recommended by Tessier et al. in 1979,2 many variants of which have since been reported (eg. refs. 3 and 4). Sequential extraction is an operationally defined’ form of speciation and the results obtained will depend on the extrac- tion reagents selected and the order in which the sediment is exposed to them. To overcome some of the difficulties in comparing results obtained by different workers the European Communities Bureau of Reference (BCR) recently proposed a sequential extraction protocol,6 based on the work of Salomons and Forstner,’ for use throughout the Community.This proto- col is currently being used in intercomparison studies and to produce reference materials certified for extractable as opposed to total metal contents.8 Although the BCR procedure has been applied to a number of river sediment^,^.^ the application of the procedure to marine sediments has not previously been reported. The coast of Cumbria in the north west of England is an interesting location for speciation studies since there are a number of potential sources of heavy metal release in the area both anthropogenic and natural.One of the most notable of the man-made sources is the Albright and Wilson ‘Marchon’ chemical manufacturing site located at Whitehaven. Amongst other processes the plant has a long history of the manufacture of phosphate products. Phosphate ores generally contain high concentrations of uranium and phosphate production has been associated with the technological enhancement of natural radioactivity at a number of locations,”*” including W hi tehaven.I2 This paper describes the development of methods for the determination of chromium manganese nickel and vanadium in aqua regia digests and acetic acid hydroxylammonium chloride and ammonium acetate extracts of intertidal sediments by flame (FAAS) and electrothermal atomic absorption spec- trometry (ETAAS).Interferences due to the presence of the aqua regia digest matrices the reagents used in the BCR sequential extraction protocol and species co-extracted from the sediment matrix by these reagents were investigated. A simple method was developed to correct for the interference effects observed involving use of single point standard additions. The methods developed were then applied in the determi- nation and speciation of the analytes in sediments from nine locations along the coast of Cumbria NW England UK. Uranium was also determined in the digests by alpha spec- trometry. The results obtained are discussed with respect to natural and anthropogenic inputs to the local marine environment. Journal of Anulytical Atomic Spectrometry March 1995 Vol. 10 233EXPERIMENTAL Instrumentation Chromium nickel and vanadium were determined in the sediment digests by ETAAS.The instrument used was a Perkin-Elmer PE 5000 spectrometer with Zeeman-effect back- ground correction and an HGA400 atomizer and AS40 autosampler. Samples were atomized from the walls of pyro- lytic graphite coated cuvettes. A matrix modifier (10 p1 of 1% m/v NH,H,PO,) was found to be useful in the determination of nickel and vanadium. However the modifier could not be used in the determination of chromium because it was contami- nated with this element. The furnace programmes used are given in Table 1. Manganese was determined by FAAS using a Unicam PU 9100 instrument with a fuel-lean air-acetylene flame. The wavelengths selected were Cr 357.9 nm; Mn 279.5 nm; Ni 232.0 nm; and V 318.4 nm.Uranium was determined by alpha particle spectrometry. An EG & G Ortec 576A dual alpha spectrometer with rug- gedized surface barrier detectors was used. Spectra were col- lected and processed with an EG & G Ortec Model 920 spectrum master multichannel buffer and analysed using Maestro I1 MCA emulator software. Reagents Nitric acid ammonium dihydrogenphosphate and acetic acid ( AnalaR grade) ammonium acetate (GPR) and Spectrosol solutions (1000 pg ml- ') of chromium (as Cr"') manganese nickel and vanadium were obtained from Merck (formerly BDH) Poole UK. Hydrochloric acid (analytical reagent) and hydrogen peroxide were obtained from Fisons Loughborough UK and hydroxylammonium chloride from Aldrich Gillingham UK.Radiochemical yield tracer (232U certified specific activity 0.994 Bq ml- ' on 12th June 1981) was obtained from AEA Technology Harwell UK. All dilutions were made with distilled water. Procedures Sampling A sediment sample was collected from Rockliffe on the Solway Firth SW Scotland on 20th October 1992. Eight sites along the Cumbrian coast were sampled on 2-3 November 1992. (Fig. 1). Intertidal sediment was collected with an acid-washed plastic scoop and returned to the laboratory in polyethylene bags. The sediment was oven-dried at 60°C and large aggre- gates were broken up. Stones and large shell fragments were discarded and the material that passed through a 2 mm mesh nylon sieve was retained for analysis. Table I ETAAS furnace programmes sample injection volume = 10 pl; and 10 ptl of 170 m/v NH,H,PO modifier was added for the determination of Ni and V Step TemperaturerC Ramp time/s Hold time/s Dry 130 5 30 Pyrolysis Cr 1400 Ni 1200 8 30 v 1200 Atomize* Cr 2500 Ni 2500 0 4 V 2700 Clean 2800 0 3 Pre-pyrolysis 400 30 5 * Peak area signal recorded.b IRISH SEA Fig. 1 Sample collection sites 1 Rockliffe; 2 Maryport; 3 Workington; 4 Harrington; 5 Whitehaven; 6 St Bees; 7 Nethertown; 8 Seascale; and 9 Ravenglass Sample preparation (aqua regia extruction) For determination of Cr Mn Ni and V 0.1 g of dried sediment was weighed into a PTFE bomb and 1 ml HNO plus 3 ml HCl added. The bomb was sealed and heated in an 850 W domestic microwave oven for 4 x 1 min at full power. The bomb was then cooled and the sample filtered (Whatman type 542 filter paper) into a calibrated flask and made up to volume (10 or 25 ml) with distilled water.The digests were analysed with respect to standard solutions prepared in 1% v/v nitric acid as described in the text. For determination of U 5 g of dried sediment were weighed into a silica crucible and ashed overnight in a muffle furnace (550 "C). The ashed sample was quantitatively transferred into a beaker and 0.1 ml of 232U tracer added. The sample was digested on a hot plate. Mineral acid (10 ml HNO plus 10 ml HC1) was added and the solution boiled to dryness. The residue was leached with 6 mol I-' HCI centrifuged or filtered and the leachate decanted into a separate beaker. The acid dissolution and leaching procedure was repeated twice. The combined leachate was then repeatedly extracted with di-isopropyl ether to remove iron taken to near dryness and rediluted in 9 moll-' HCl.Uranium was separated from interferents (other alpha-emitters and heavy metals which cause spectral interference and inefficient electroplating of the source respectively) by passing the solution through two columns ( 1 cm x 8 cm) of anion-exchange resin ( BioRad AGl-X8 100-200 mesh chloride form obtained from BioRad Laboratories Hemel Hempstead UK). As the sample is passed down the first column uranium plutonium and some heavy metals are retained whilst thorium and other interferents e g . aluminium are not. Uranium can then be eluted in 8 moll-' HNO leaving plutonium on the resin. The uranium solution is then taken to dryness and the residue dissolved in 9 moll-' HCl.Further contaminants are removed by passing the solu- tion through the second column and eluting with 1.2 mol I - ' HCl. For the Whitehaven sediments which were heavily con- taminated it was necessary to repeat the first column treatment. Finally the uranium was electroplated from solution (current 1 A; plating time 1 h) onto a clean stainless steel planchette (Nuclear Supplies Mirfield UK). Sequential extruction The reagents used in the BCR sequential extraction protocol together with the nominal sediment phases dissolved at each step are given in Table 2. The experimental procedure is summarized below and has been described in detail in a previous publication.' 234 Journal of Analytical Atomic Spectrometry March 1995 Vol.10Table 2 BCR sequential extraction procedure Step Reagent (s) Operationally-defined speciation 1 CH,COOH Exchangeable water and acid soluble 2 NH,OH.HCl at pH 2 Reducible 3 H 2 0 2 (8.8 mol 1- l ) ; Oxidizable (0.1 1 rnol I - . ' ) (0.1 rnol 1 - ' ) followed by CH3COONH (1.0 mol 1-' ) at pH 2 Step one. A 40 ml volume of 0.1 1 mol 1 - ' acetic acid was added to 1 g dried sediment in a polypropylene centrifuge tube and shaken overnight at ambient temperature on a mechan- ical end-over-end shaker. The mixture was centrifuged and the supernatant retained for analysis. Step two. A 40 ml volume of 0.1 moll-' hydroxylam- monium chloride (adjusted to pH 2 with nitric acid) was added to the residue from step one and extraction performed as described above. Step three.A 10 ml volume of 8.8 moll-' hydrogen peroxide was added to the residue from step two. The centrifuge tube was covered and the contents digested for 1 h at ambient temperature then for 1 h at 85 "C. The sample was evaporated to near dryness and the hydrogen peroxide treatment repeated. After taking to near dryness a second time 50 ml of 1 moll-' ammonium acetate (adjusted to pH 2 with nitric acid) was added and the extraction performed as described above. The material remaining after sequential extraction was digested with aqua regia as described above to give an indication of the amounts of metals associated with the more resistant (residual) components of the sediment matrix. Up to eight samples could be extracted simultaneously and the total time required to process a batch of samples (including overnight shaking and analysis) was of the order of five days.RESULTS AND DISCUSSION Development of Methods for the Analysis of Marine Sediment Digests by ETAAS Matrix interference effects due to the presence of a sediment digest were investigated. For manganese nickel and vanadium the slopes of response curves obtained in dilute acid (1 YO v/v HNO,) were similar to those of standard additions graphs obtained from a digest of sediment collected from Site 7 (Nethertown). A 27% suppression was observed for chromium (Table 3). Pyrolysis curves obtained indicated no losses of chromium in either standard solutions or sediment digests up to 1400 T although at higher pyrolysis temperatures the decrease in sensitivity for the digest was greater than for the standard solution.Arpadjan and KrivanI3 used radiotracers to study the behaviour of chromium in a graphite furnace and reported losses at pyrolysis temperatures of 1300 O C and above which varied for different types of sample and between samples and acid standard solutions. Unfortunately geological samples were not amongst the matrices investigated. Further investi- Table 3 digests; the errors quoted are Comparison of response curves in nitric acid and in sediment Is for thirteen degrees of freedom lo2 x Slope in lo2 x Slope in Difference in Element 1 '/o HNO sediment digest slope (%) Chromium 0.684 k0.017 0.500 + 0.01 6 - 27 Manganese 13.0 + 0.2 12.1 k0.3 -7 Nickel 0.193 0.006 0.185 k 0.006 - 4 Vanadium 0.146 f 0.003 0. I48 0.004 + 1 gation is therefore required to determine the cause of the interference effect observed.The interference on chromium worsened as the condition of the tube deteriorated. As up to 40% v/v aqua regia solutions were injected and high atomization temperatures used tube lifetimes were comparatively short (< 100 atomizations). Despite the satisfactory nature of most of the results in Table 3 it is possible that different matrix interference effects will be caused by different sediments. This implies that reliable sediment analysis may require the use of standard additions methods. Routine use of such methods however is not accept- able because of the length of time required to analyse many samples for each of the analytes. In this study two alternative approaches to calibration were investigated as part of a more extensive assessment of sediment interference effects.In the first method a single addition of standard solution (approximately equivalent to half the initial estimate of the intrinsic concentration) was made to each sample. The concentrations obtained by direct calibration with HNO standards were then corrected on the basis of the recoveries of the analytes added (Method A). The second procedure involved use of the standard additions graphs obtained with one sediment (from Site 7 Nethertown) for quantification of the analytes in the other samples (Method B). The results obtained by the two methods shown in Table 4 are generally comparable and of similar precision. Analyte recoveries are almost all within 84- 1 13% for manganese nickel and vanadium but significantly poorer and more variable for chromium.This is expected since chromium is the only analyte for which significant interference effects were observed. Distributions of Aqua Regia Extractable Metals The concentrations of chromium manganese nickel and vanadium in sediments collected from sampling sites along the Cumbrian coast are shown in Fig. 2. Results were obtained using calibration Method A the single point standard addition. Enhanced concentrations of chromium (39.5 0.9 pg g- '; n = 3 ) and vanadium (33.0k0.6 pg g-'; n=3) were observed at Whitehaven whilst manganese levels were highest (around 1025 pg g- ') at the more northerly sites of Maryport Workington and Harrington. Nickel concentrations were in the range 4.8-17.6 pg g-I and displayed no clear trend along the sampling transect.The concentrations determined at Site 1 (Rockliffe) are within the range of those measured some time previously in the Solway Firth,I4 indicating no increase in pollutant levels of chromium manganese or nickel in this area. Significant quantities of chromium were discharged under authorization by the Marchon chemical plant in 1992,'' and it is possible that this is the source of the chromium contami- nation in Whitehaven harbour. The vanadium enhancement may be due to spillage of fuel in the harbour. Crude oil and its derivatives often contain high levels of vanadium.I6 The manganese distribution is more difficult to explain as a previous study" revealed that concentrations in freshwater sediments are higher to the south of Whitehaven rather than to the north. The distribution of the most abundant naturally occurring uranium isotope 238U is shown in Fig.3. There is a large enhancement at Whitehaven which can be attributed to either effluent discharge or spillage of raw materials associated with phosphate production at the Marchon chemical plant. The activity of 234U the first a-emitting decay product of 238U in the uranium natural decay series is 39.2 & 1.2 Bq kg- ' and the 234U 238U activity ratio is 1.01 k0.04. This supports the con- clusion that the uranium is derived from a natural until recently undisturbed source such as phosphate ore. In a geochemical system which is unperturbed for a long period Jouriiul of Analytical Atomic Spectrometry March 1995 Vol.10 235Table 4 Comparison of results obtained by different calibration methods; errors quoted are f Is n=3 Element Site Chromium Manganese Nickel Vanadium 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 Concentration Recovery Concentration/pg g- 1% HNO standard s/ 13.1 f0.4 13.6 f 0.4 12.0 k 0.4 14.4 f 0.4 30.4 f 0.7 7.4 f 0.2 10.3 f 0.4 7.5 * 0.2 7.6 * 0.4 488 f 20 988 f 23 881 f 2 2 998 f 23 349 f 20 164f21 340 f 20 93 f 20 134f21 8.6 f 0.5 16.3 L- 0.6 8.6 f 0.5 11.0f0.5 16.2f0.5 4.1 f 0.4 5.0 f 0.5 14.4 f 0.6 6.0 f 0.5 8.0 f 1 .O 7.0 f 0.3 9.3 f 0.4 11.0f0.8 34.3 f 0.6 4.3 k 0.3 5.1 f0.3 2.4 f 0.5 4.6 f 0.7 vg g-I W) 71.3 73.1 75.0 90.0 77.2 68.0 72.3 82.0 70.2 89.0 91.0 90.2 93.0 87.0 84.3 88.0 77.0 79.3 91.4 92.5 89.2 88.0 93.0 84.5 88.5 93.5 95.1 105.0 100.0 103.2 101.3 104.0 108.1 112.3 106.1 105.4 Method A* Method B t 18.4 f 0.6 17.7 f 0.6 18.6 f 0.5 18.4 k 0.6 16.0f0.5 16.1 f0.6 16.0 f 0.4 14.0 k 0.6 39.5 f0.9 41.5 & 1.3 1 1 .O & 0.3 10.0 k 0.3 14.2f0.5 14.0f0.6 9.1 f 0.2 8.6 f 0.3 10.8 f 0.6 10.2 f0.5 549f23 547f 17 1086f26 1075f23 976 f 25 963 f 21 1074f25 1088f23 401f23 402k16 194f25 202& 17 387f23 391 -t 16 121 f 2 6 130+ 18 170f27 174f 17 9.4 f 0.5 9.2 f 0.6 17.6 f 0.7 17.3 & 0.7 9.6 f 0.6 9.2 k 0.5 12.5 f 0.6 12.0 & 0.6 17.4f0.5 17.1 f0.7 4.8 f 0.5 4.4 f 0.4 5.6 f 0.6 5.3 f0.5 15.4f 0.6 15.3 f 0.6 6.3 f 0.5 6.5 f 0.5 7.6 f 0.9 7.6 f 1 .O 7.0 f 0.3 6.6 f 0.3 9.0 f 0.4 9.2 f 0.4 10.9 L- 0.8 10.8 + 0.8 33.0f0.6 33.8k0.5 4.0 0.3 4.2 f 0.5 4.5 f 0.3 5.0 10.3 2.3 k0.5 2.3 f0.4 4.4 f 0.7 4.5 & 0.7 * Method A values obtained by calibration with acid standards corrected for analyte recoveries as determined by single point standard addition.t Method B calibration with respect to full standard additions for Nethertown sediment digest. the nuclides in a particular natural decay series will achieve secular equilibrium i.e. all members of the series have the same activity. Modern seawater however typically has a 234U 238U activity ratio of 1.14 due to preferential leaching of 234U from crustal rocks by groundwater." Table 6 ETAAS detection limits in sequential extraction reagents (pg.g-l sediment dry mass); detection limits were calculated on the basis of 3s (blank solution) Hydroxylammonium Ammonium Element Acetic acid chloride acetate Chromium 0.04 0.03 0.06 Manganese* 0.4 0.44 0.6 Nickel 0.05 0.04 0.07 Vanadium 0.2 0.3 0.4 * FAAS.Development of Methods for the Analysis of Marine Sediment Extracts by ETAAS Interference effects associated with the determination of chro- mium nickel and vanadium in marine sediment extracts obtained by the BCR procedure were investigated. To deter- mine the effects of the extraction reagents used slopes of response curves obtained from standard solutions prepared in 1% v/v HNO were compared with those obtained from solutions prepared in each of the extractants (Table 5). None of the reagents had any significant (ix. > f 10%) effect on the analytical response for manganese nickel or vanadium but the presence of ammonium acetate caused a suppression of the chromium signal.This effect was found to be reproducible. In a replicate experiment a 29% suppression was observed. In previous work slight enhancements relative to uquu regiu standard solutions were observed for chromium in acetic acid and ammonium acetate whilst hydroxylammonium chloride was found to have no effect on the slope of the response curve.' Detection limits for each analyte in each of the extractants are shown in Table 6. Slightly higher detection limits were obtained using ammonium acetate for all the analytes owing to poorer precision of the AAS signals. The response curves for the analytes in the reagent matrices were also compared with standard additions graphs for extracts of Nethertown sediment. Species co-extracted by acetic acid caused slight (around 10%) suppression for all analytes in the Step 1 extract whilst no significant interference effects were observed in Step 2 and Step 3 extracts.Quantification of manganese nickel and vanadium in extracts of this sediment was therefore possible by direct calibration with standard solutions in 1% HN03. This procedure was also adequate for the determination of chromium in Step 1 and Step 2 extracts but the use of a reagent-matched standard was required for analysis of the Step 3 extract. Since different interference effects may be caused by different sediment matrices the calibration method based on a single standard addition described above was also used in the analysis of the sediment extracts. Table 5 Comparison of response curves in nitric acid and in sequential extraction reagents; errors quoted are k Is for thirteen degrees of freedom Element Chromium Manganese Nickel Vanadium 10' x Slope in 1% HNO 1.20f0.02 8.85 f 0.08 0.39 fO.O1 0.160 f 0.005 Reagent CH,COOH NH *O H.HC1 CH,COONH CH,COOH N H ,O H.l i C1 CH,COONH CH,COOH NH,OH.HCl CH,COONH CH,COOH N H,OH.I-ICl CH3COOUH4 10' x Slope in reagent 1.10 f 0.02 1.1Of 0.03 0.90 f 0.03 8.25 f0.08 8.45 f 0.08 8.40 * 0.1 1 0.38 f 0.01 0.37+0.01 0.36 f 0.0 1 0.170 f 0.005 0.175 f 0.006 0.145 f 0.003 Difference in slope (%) -8 -8 - 25 -7 - 5 - 5 -3 - 5 8 +6 +9 -9 236 Journal of Analytical Atomic Spectrometry March 1995 Vol. 100 -40 -30 -20 -10 0 10 20 30 N 200 4001 h 0 - 40 - 20 0 20 40 0 ' I 1 I I I I I -40 -30 -20 -10 0 10 20 30 S Distance from Whitehaven/km Fig.2 Concentrations of aqua-regia extractable heavy metals in sediments from sampling sites north and south of Whitehaven (a) chromium (b) manganese (c) nickel and ( d ) vanadium 40 r f 35 30 lo 25 r Y 5 20 > Y .- 15 10 5 2 0 -40 -30 -20 -10 0 10 20 30 N S Distance from Whitehaven/km Fig. 3 Activities of acid extractable uranium in sediments from sampling sites north and south of Whitehaven Metal Speciation by Sequential Extraction The concentrations of metals extracted from the sediment samples at each stage of the sequential extraction procedure are given in Table 7. The agreement between the total amount of analytes extracted in the sequential procedure and the results obtained by aqua regia dissolution is generally good although some large discrepancies do occur e.g.for nickel in the sediment from Site 6 and for vanadium at Site 7. The operationally-defined speciations of chromium nickel and vanadium are similar (Fig. 4) but differ markedly from that of manganese. At most sites the majority of the Cr Ni and V is associated with the residual sediment fraction but there is an increase in more labile forms of these elements particularly oxidizable species at Site 2 (Maryport) and Site 5 (Whitehaven). For vanadium the concentration in the residual sediment fraction at Site 5 (8.2 r_t 0.2 pg g- ') is within the range (2.3-10.9 pg g-') of total vanadium concentrations at other sites. This indicates that the enhancement level of vanadium in Whitehaven harbour is due almost entirely to species not present at the other locations sampled; perhaps as suggested earlier derived from oil residues.The manganese speciation at Sites 2 and 3 where enhance- ment in the total concentrations was observed does not differ markedly from that at the other sites. In all the sediments a large proportion of the manganese was extracted in Step 1 of the procedure. This would seem to suggest that the element is present in exchangeable forms rather than as the expected hydrated manganese oxides. Many workers have discussed the possibility that reagents used in sequential extractions may prematurely 'attack' non- target phases i.e. phases not intended to be dissolved until later in the procedure. Rauret et aL4 investigated removal of manganese and iron from river sediment by 1 mol 1-1 sodium acetate adjusted to pH 5 with acetic acid (the Step 2 reagent of Tessier7s2 procedure) and concluded that up to 20% prema- ture dissolution of the 'reducible' phases could take place if reagent volume sample mass ratios were high (50 ml g-').In contrast Robinsonlg found that a more vigorous treatment with 1 mol 1-' acetic acid (the Step 1 reagent of a sequential extraction scheme proposed by Filipek et d2*) removed only 2% of the manganese from a composite sample known to contain 25% (by mass) manganese oxide ore. In the present work the sediment was extracted with 40 ml g-' 0.11 moll-' acetic acid at pH 2.7. Handling and pretreatment of samples also strongly influ- ences the results obtained by sequential extraction procedures with drying and storing sometimes having a profound effect on the lability of manganese.2' When the sediments used in this study were originally sampled (in November 1992) no speciation studies were envisaged and no special precautions taken to minimize post-sampling species transformations.It is therefore possible that oven drying followed by prolonged storage at room temperature has significantly altered the intrinsic manganese species distribution. Journal of Analytical Atomic Spectrometry March 1995 Vol. 10 237Manganese Nickel Vanadium Table 7 quoted are Is n= 3 Comparison of amounts of metals extracted by the BCR procedure with those obtained by aqua regia dissolution (pg g-'); errors Sum* Aqua regia Element Site Step 1 Step 2 Step 3 Residual C 1+2+3+R dissolution Chromium 1 0.30 0.67 0.90 16.0 17.9 f0.2 18.4f0.6 2 0.63 1.04 2.50 9.62 13.8k0.1 18.6k0.5 3 0.64 0.56 1 .oo 11.4 14.0 f 0.1 16.0 k 0.5 5 1.60 2.03 8.80 31.0 43.4 f0.7 39.5 k 0.9 6 < 0.04 0.1 0.47 6.8 7.4f0.1 11.0 f0.3 7 0.40 0.74 0.77 7.40 9.3 f 0.2 14.2 f0.5 8 < 0.04 0.10 0.30 8.20 8.6k0.1 9.1 f 0.2 9 0.07 0.60 0.36 7.80 8.8 k0.1 10.8 f.0.6 1 321 47.0 10.4 60.0 439 f 10 549 f 17 2 598 172 28.1 203 1002 f 16 1086 f. 23 3 423 150 51.4 394 1018 k 16 976 k 21 5 143 31.1 45.0 120 339 k 12 401 16 6 60.9 4.2 < 0.6 97.0 162+9 194+ 17 7 280 51.2 1.2 84.4 417 f 10 387 f 16 8 101 5.3 < 0.6 72.0 178 f 8 148f13 9 91.5 6.4 < 0.6 57.0 155f7 170f 17 1 0.3 < 0.04 < 0.07 14.0 14.3 f 0.1 12.2 k 0.6 17.6 k 0.7 2 4.0 3 0.4 0.1 4.0 9.0 13.5 f 0.1 10.0 k 0.5 5 2.1 0.6 5.3 11.8 19.8 f 0.2 17.4k 0.5 6 0.2 0.2 < 0.07 12.2 12.6 k 0.2 8.6 k0.5 5.6 f 0.6 7 0.6 < 0.04 < 0.07 6.1 6.7 k 0.2 8 0.1 < 0.04 < 0.07 15.5 15.6f0.2 15.4 k0.6 9 0.3 < 0.04 < 0.07 6.6 6.9f0.1 6.3 f0.5 1 <0.1 < 0.3 1.3 7.2 8.5 k0.5 7.6 f 0.9 2 <0.1 1.2 1.9 7.1 10.2 f 0.8 7.5 k 0.6 3 <0.1 < 0.3 1.4 11.1 12.5 k 0.3 11.2 k0.7 5 3.8 8.4 9.0 8.2 29.4 f 0.2 33.0 f 0.6 6 <0.1 < 0.3 < 0.6 8.0 8.0 f 0.2 6.5 f 0.5 7 0.9 < 0.3 0.8 4.6 6.3 f 0.3 4.5 f 0.5 8 <0.1 < 0.3 < 0.6 2.0 2.0 k 0.2 2.3 f 0.5 9 <0.1 < 0.3 < 0.6 5.0 5.0 k0.3 4.4 f 0.7 2.0 4.4 6.6 17.0 f 0.4 * Errors quoted are f Is calculated from the standard deviation values of individual concentrations (in each case n = 3).Table8 Effects of drying and storage on the amount of manganese extracted from sediment by the initial step(s) of the BCR and Tessier sequential extraction procedures Drying temperature/ Site "C 5 60 60 Ambient 7 60 60 Ambient Storage time/ months 0 6 6 0 6 6 Procedure* BCR step 17 BCR step 1 Tessier step 1% Tessier step 25 BCR step 1 Tessier step 1 Tessier step 2 BCR step 1 BCR step 1 Tessier step 1 Tessier step 2 BCR step 1 Tessier step 1 Tessier step 2 Result/ 159 171 2 60 176 2 56 163 168 5 87 160 5 85 I % g-' * Sediment mass= 1 g (dry mass).7 BCR step 1 40 ml of 0.11 mol 1-' acetic acid pH 2.7. $ Tessier step 1 8 ml of 1.0 mol 1-l sodium acetate pH 8.2. 3 Tessier step 2 8 ml of 1.0 mol 1-' sodium acetate adjusted to pH 5.0 with acetic acid. Further experiments were performed to investigate this and determine whether premature dissolution of manganese- containing species occurs during Step 1 of the BCR procedure.Fresh sediment samples were collected from Whitehaven (Site 5) and Nethertown (Site 7) in February 1994. Sediment subsamples were dried at 60 "C and at room temperature. The subsample which had been dried at 60 "C was extracted immediately and again six months later whilst the subsample which had been dried at room temperature was extracted only after storage. Neither drying nor storage was found to have a significant influence on the amount of manganese extracted by 0.11 moll-' acetic acid (Table 8). A group of scientist involved in the BCR Programme reported no instability in the speciation (as determined by the BCR procedure) of Cd Cr Cu Ni Pb and Zn when an air-dried river sediment was stored for up to eight months but manganese was not amongst the analytes investigated.22 The oven-dried and stored sediment was also treated with the first two reagents of the Tessier sequential extraction procedure2 (1.0 mol I-' sodium acetate at pH 8.2 followed by 1.0 moll-' sodium acetate adjusted to pH 5.0 with acetic acid) since these reagents would together be expected to dissolve similar phases of the sediment to Step 1 of the BCR procedure.Significantly smaller amounts of manganese however were liberated by the Tessier scheme (Table 8) which suggests that either the low pH of the BCR reagent or the high reagent volume sample mass ratio used may indeed cause premature dissolution of manganese-containing species.CONCLUSIONS Observed enhancements in concentration of chromium vanadium and uranium in sediments from coastal sites in Cumbria can be correlated with anthropogenic activity in the area. The BCR sequential extraction scheme has been applied successfully in the speciation of metals in marine sediments. Few interferences were encountered in the analysis of extracts of these sediment by ETAAS and most could be overcome by use of matrix modifier or reagent-matched standard solutions. However because interference effects are likely to vary for sediments of differing composition it is recommended that a standard additions method be used wherever possible. In this study a single standard addition was made at a concentration 238 Journal of Analytical Atomic Spectrometry March 1995 Vol.10100% 80% 60% 40% 20% 0 % 100% 80% 60% 40% 20% 0% Bm exchangeable reducible 0 oxidisable residual 1 2 3 5 6 7 8 9 1 2 3 5 6 7 8 9 Sampling site Fig. 4 Operationally-defined speciation of heavy metals in sediments from sampling sites around Whitehaven (a) chromium (b) manganese (c) nickel and ( d ) vanadium about half of the estimated intrinsic level to assess interferences in a reasonable time. Both the distribution and speciation of manganese were difficult to explain and further work is necessary to establish (a) the source(s) of manganese in the area and (b) the most appropriate method to determine manganese speciation in marine sediments. Within the past two years phosphate processing at the Albright and Wilson Marchon plant has altered significantly with the cessation of on-site ore processing and the introduc- tion of effluent treatment.The current work provides an indication of the levels of contamination present in the marine environment at the time when shipment of uranium-rich ore to Whitehaven ceased. Further sampling and analysis is now being undertaken which will allow trends in inventories of uranium its decay products and other heavy metals associated with phosphates production to be assessed. REFERENCES 1 Forstner U. Contaminated Sediment-Lecture Notes in Earth Science 22 eds. Bhattacharji S. Friedman G. M. Neugebauer H. J. and Seilacher A. Springer-Verlag Berlin 1989. 2 Tessier A. Campbell P. G. C. and Bisson M. Anal. Chem. 1979 51 844. 3 4 5 6 I 12 13 14 15 16 1 ’I 18 Rauret G.Rubio R. Lopez-Sanchez J. F. and Casassas E. Int. J. Environ. Anal. Chem. 1989 35 89. Rauret G. Rubio R. and Lopez-Sanchez J. F. Int. J. Environ. Anal. Chem. 1989 36 69. Ure A. M. Fresenius’ 2. Anal. Chem. 1990 337 577. Ure A. M. Quevauviller Ph. Muntau H. and Griepink B. EUR Report No. 14763 CEC Brussels 1992. Salomons W. and Forstner U. Metals in the Hydrocycle Springer New York 1984. Ure A. M. Quevauviller Ph. Muntau H. and Griepink B. Int. J. Environ. Anal. Chem. 1993 51 135. Davidson C. M. Thomas R. P. McVey S. E. Perala R. Littlejohn D. and Ure A. M. And. Chim. Acta 1994 291 277. Kobal I. Brajnik D. Kaluza F. and Vengust M. Health Phys. 1990 58 81. Van der Heijde H. B. Klijn P. J. Duursma K. Eisma D. De Groot A. J. Hagel P. Koster H. W. and Nooyen J. L. Sci. Total Environ. 1990 90 203. McCartney M. Kershaw P. J. and Allington D. J. J. Environ. Radioact. 1990 12 243. Arpadjan S. and Krivan V. Fresenius’ Z . Anal. Chem. 1988 329 145. Halcrow W. Mackay D. W. and Thornton I. J. Mar. Biol. Ass. UK 1973 53,721. Donaldson B. R. personal communication. Jones K. C. Lepp N. W. and Obbard J. P. in Heavy Metals in Soils ed. Alloway B. J. Blackie Glasgow 1990 p. 316. Regional Geochemistry of the Lake District and Surrounding Areas British Geological Survey Keyworth 1992. Anderson R. F. LeHuray A. P. Fleisher M. Q. and Murray J. W. Geochim. et Cosmochim. Acta 1989 53 2205. Journal of Analytical Atomic Spectrometry March 1995 VoE. 10 23919 Robinson G. D. Chem. Geol. 1984/1985 47 97. 20 Filipek L. H. Chao T. T. -and Carpenter R. H. Chem. Geol. 1981 33 45. 21 Kersten M. and Forstner U. in Chemical Speciation in the Environment eds. Ure A. M. and Davidson C. M. Blackie Glasgow 1995. 22 Fiedler H. D. Lopez-Sanchez J. F. Rubio R. Rauret G. Quevauviller Ph. Ure A. M. and Muntau H. Analyst 1994 119 1109. Paper 4/05 I I ID Received August 22 1994 Accepted November 8 1994 240 Journal of Analytical Atomic Spectrometry March 1995 Vol. 10

ISSN:0267-9477    

DOI:10.1039/JA9951000233

出版商:RSC    

年代:1995

数据来源: RSC

摘要:

On = I i ne Preco ncen t rat ion of C h ro m i u m ( I I I) and Speciation of Chromium in Waters by Flame Atomic Absorption Spectrometry* Journal of Analytical Atomic Spectrometry BENYAMIN PASULLEAN CHRISTINE M DAVIDSON AND DAVID LITTLEJOHN? Department of Pure and Applied Chemistry University of Strathclyde Cathedral Street Glasgow UK GI 1 X L The chemistry of complexation has been investigated and conditions optimized for on-line preconcentration of Cr"' by resins with immobilized quinolin-8-01 or iminodiacetate functional groups. When solutions were buffered to 0.1 rnol 1-' acetate at pH 9 the detection limit of Cr"' was 6 ng ml-' for preconcentration on the quinolin-8-01 resin at 6 ml min-' for 3 min. With the iminodiacetate resin (Muromac A-1) the optimum buffer conditions were 0.1 rnol I-' acetate at pH 4 which gave a Cr"' detection limit of 2 ng ml-' (3 min preconcentration at 6 ml min-').The major ions in sea-water did not interfere with preconcentration of 100 ng ml-' Cr"' by Muromac A-1. However suppressive interferences were caused by 100 pg ml-' of Ca2+ or Fe2+ with the quinolin-8-01 resin. Despite having similar log stability constants of complexation with either resin the interference caused by Ca2+ was much greater than that by Mg2+. Muromac A-1 was used for preconcentration of Cr"' in estuarine- and sea-waters prior to determination by flame atomic absorption spectrometry (FAAS). By reduction of CrV' to Cr"' it was also possible to determine the total concentration of ionic Cr in the waters which allowed calculation of the CrV' concentration by difference.The Cr"' concentration in the samples from the Clyde estuary were 3-8 ng ml-' with the CrV' concentration about 0.7 ng ml-'. Only Cr"' (3 ng ml-') was found in the sea-water samples from the Cumbrian Coast. Keywords On-line preconcentration; quinolin-8-01 and iminodiacetate resins; chromium speciation; atomic absorption spectrometry On-line preconcentration has been used with different atomic spectrometry techniques to reduce or remove interference effects and improve detection limits.'-8 Many reports concern the preconcentration of transition metal ions and the most commonly used resins are those with immobilized quinolin-8- 0 1 ~ 9 ~ 9 ~ or i m i n o d i a ~ e t a t e ' * ~ ~ ~ ~ * ~ functional groups. Problems have often been encountered in the on-line preconcentration of Cr"' with these and other resins and this has been attributed6 to the inertness of Cr"' aquo-comple~es.~ Prakash et a1.' reported that neither ethylenediaminetriacetic acid nor sulfon- ated quinolin-8-01 immobilized on cellulose gave better than 10-40% recovery of Cr"' at pH 5.5.Porta et aL3 precomplexed Cr"' with 8-hydroxy-7-iodoquinoline-5-sulfonic acid and achieved 63-73% recovery at pH9.2 by extraction on an anion-exchange resin. Ebdon et aL8 found that an iminodiacet- ate-based resin could not be used successfully for on-line preconcentration of Cr"' at pH 5.3 using 1.5 mol 1-' ammonium acetate as a buffer. In contrast Hirata et al.' had no apparent difficulties in preconcentrating Cr"' from solutions containing 0.5 mol 1-' ammonium acetate with Muromac A-1 * Presented at the Seventh Biennial National Atomic Spectroscopy Symposium (BNASS) Hull UK July 20-22 1994.t To whom correspondence should be addressed. an iminodiacetate-based resin when the pH was adjusted to 3-4. It has been reported that control of pH and the use of a suitable auxiliary complexing reagent is essential for on-line preconcentration of ions of the Group 13 elements by immobil- ized quinolin-8-01.'~-~~ This study shows that by optimizing the chemistry of complexation efficient preconcentration of Cr"' can also be achieved with quinolin-8-01 or iminodiacetate- based resins. A comparison of the two resins has been made and interference effects from concomitant ions studied. A number of methods have been developed for selective determination of Cr"' and/or Cr" as reviewed recently by Sperling et a l l 3 Some procedures like that reported by Jen et a/.' involve complexation of Cr"' by e.g.EDTA and then separation by reversed-phase liquid chromatography on c8 or c18 columns to obtain individual detection of Cr"' and Cr". A similar approach was adopted by Milacic and Stupar," who achieved separation of CrV1 and anionic complexes of Cr"' by fast protein anion-exchange liquid chromatography with off- line detection by flame AAS (FAAS) Sperling et a1.13 used activated alumina in the acid form for sequential species- selective sorption of Cr"' at pH 7 and CrV' at pH 2. The pH conditions were selected to retain one or the other of the chromium ions. By careful control of flow rate the non-sorbed species could be made to elute in front of the sorbed ion so that both species could be determined sequentially in one analysis.Consequently the sensitivity of determination for the non-sorbed species was poorer than that of the sorbed ion. In situations where there was a large difference in the concen- tration of the ions addition of a wash cycle before elution was beneficial. For a 3ml volume loaded onto the column over 35 s a sensitivity enhancement of 25 was achieved and the detection limit by FAAS was about 1 ng ml-' for each sorbed ion. Elmahadi and Greenway16 achieved speciation of Cr"' and CrV1 by passing solutions through columns containing immobilized algae. The Cr"' was eluted with thiourea in 0.1 mol 1-' HClO and CrV' by 0.1 moll-' NaOH.For 5 ml of sample the best detection limits reported were 20 and 30ngml-' of Cr"' and Crvl respectively for elution into an air-acetylene flame. In the present study CrV' was converted to Cr"' and the iminodiacetate-based resin was used with FAAS to determine concentrations of Cr'" and total dissolved ionic Cr (i.e. Cr"' plus CrV1) in coastal sea-water samples and estuarine waters. EXPERIMENTAL Apparatus A Unicam PU9400 spectrometer was operated with a nitrous oxide-acetylene flame for atomic absorption measurements of Cr at 357.9 nm. The bandpass was 0.5 nm the fuel flow rate 4.0 1 min-' and burner height 6 mm. Peak height absorbance signals were recorded on a chart recorder. A single-line continuous flow preconcentration system'0." was used.The Omnifit glass column (3 mm id. and either 100 Journal of Analytical Atomic Spectrometry March 1995 Vol. 10 241or 150mm in length) was fitted with gripfit connectors and was modified by removing the microporous plastic end-discs and replacing them with discs of Whatman GF/A glass fibre filter. The 100mm column was filled with 0.5046g of con- trolled-pore glass beads with immobilized quinolin-8-01 (125-177 pm particle size; Pearce Chemical Company UK) and the 150mm column was packed at different times with 0.41-0.66 g of Muromac A-1 (250-500 pm particle size; Muromachi Kagaku Kogyo Kaisha Japan). Muromac A-1 has the same iminodiacetate functional groups as Chelex- 100 but is free of the swelling properties that are encountered with Chelex-100.17 It was supplied as the sodium form but was washed with 2 mol 1-' HC1 for at least 1 h before use.Hitherto Muromac A-1 has been used mainly by Japanese worker^.^,'^,^^ Measurements of pH were made with a Whatman PHA 3030 pH meter. Reagents A SpectrosoL solution of 1000 pgml-' Cr"' (BDH now Merck) was used to prepare calibrant solutions by serial dilution with distilled water. Analytical grade hydrochloric acid and nitric acid (May and Baker) were used to prepare the eluent. Buffer solutions were prepared from analytical grade tartaric acid (Hopkins and Williams) glacial acetic acid oxalic acid malonic acid and citric acid (Merck). All the buffer solutions except acetate were prepared by dissolving the acids in distilled water to prepare a 2 mol 1-' stock solution.The acetate buffer was prepared by mixing 2 mol 1-' acetic acid with 2 moll-' ammonium acetate. Hydrochloric acid and ammonia (Merck) were used to adjust the pH of the buffer solutions to the required value in the range pH 1-12. On-line Preconcentration Procedure The system used has been described previously in detail."-12 Samples and standard solutions containing 0.1 mol 1-' buffer adjusted to the required pH were passed through the column at 6 ml min-' for 3 or 6 min. The Cr"' and any retained concomitant ions were eluted at 6 ml min-' with a mixture of 2 mol 1-' HC1 plus 2 mol 1-1 HN03 into the nebulizer of the nitrous oxide-acetylene flame. During the preconcentration period the effluent from the column was pumped to waste. Reduction of Cr"' to CrU1 The procedure was based on that reported by Isshiki et aL2' Normally 2 ml of 2 moll - ' HCl was added to 80 ml of sample or standard solution followed by addition of 2 ml of 1 mol 1-' hydroxylamine.The solution was left at room temperature for at least 20 min before 10 ml of 1 moll-' acetate was added the pH adjusted to 4 and the volume made up to 100ml with distilled water. The solution was then preconcentrated within 10min or after 60min to ensure complete collection of Cr"' by the Muromac A-1 resin. Samples River water samples were collected from five locations in the Clyde estuary (Yoker Old Kirkpatrick Dumbarton Castle Cardross and Ardmore) and two locations off the Cumbrian coast (Ravenglass and Whitehaven). The samples were filtered through Acro-disc 0.45 pm filters (Gelman UK) acidified with HCl to 1% v/v and stored in plastic containers at 4°C before analysis.RESULTS AND DISCUSSION Effects of pH and Buffers on Complexation of Cr"' Most complexes of Cr"' are hexa-coordinated and in aqueous solution are characterized by their kinetic inertness.' At low pH Cr"' exists as [Cr(H20)J3' and at high pH Cr"' is hydrolysed and eventually precipitated as Cr(OH) .' The pK of hydrolysis of Cr"' is 4.01 and the pH at which precipitation starts to occur calculated as described in ref. 11 is 4.05 for a concentration of 0.1 pgml-l Cr"'. Hence at pH values above 4 loss of Cr"' from solution is expected unless an auxiliary complexing agent is used to solubilize the Cr. Previous experi- ence'0-12 has shown that acetate citrate malonate oxalate and tartrate can be useful buffering reagents for on-line precon- centration of metal ions likely to precipitate at neutral and alkaline pH.The effects of these complexing reagents on preconcentration of 300 ng ml-' Cr"' by immobilized quinolin- 8-01 are shown in Fig. 1. Quinolin-8-01 is amphoteric as the molecule has basic nitrogen and phenolic hydroxyl groups. At low pH the quinolin-8-01 is protonated and so reaction with Cr"' is unfavourable. With acetate and oxalate buffers Cr absorbance signals were obtained between pH 7-12. The opti- mum pH with oxalate was pH 11 but the sensitivity was not as good as that with acetate at pH 9. No signals were obtained at any pH with citrate malonate or tartrate buffers. The log stability constants (I1) of complexes of Cr"' with acetate and malonate are 4.63 and 8.26 respectively.21 It is possible therefore that the malonate complex is too strong and prevents reaction of Cr"' with the immobilized quinolinate.Although the hydroxide complex of Cr'" has log stability constants of 10.01-28.6 (p1-p4),21 the relative concentrations of acetate and hydroxide in solution will influence complex formation and apparently favour reaction with acetate. As stability constants were not available for the other organic buffers further expla- nation of the results is difficult but it can be surmised that citrate and tartrate also form strong complexes with Cr"' and have stability constants that are greater than that of Cr"' with quinolinate at any pH. When no organic buffers were used no Cr AAS signals were obtained confirming Stary's opinion22 that reaction of Cr"' with quinolin-8-01 does not occur at room temperature.However the presence of acetate at pH 9 seems to have a substantial effect on the reaction improving the efficiency of complexation to a level suitable for use in on-line preconcentration. The optimum pH range is narrow which implies that careful adjustment of the pH of solution is necessary. At present no hypothesis can be offered to explain the influence of acetate on the reaction of Cr"' with quinolinate. The effects of pH and buffer on the preconcentration of Cr'" by Muromac A-1 are shown in Figs. 2 and 3. Atomic absorption signals of Cr were obtained for acetate citrate malonate and 0.25 0.20 V 8 0.15 z 2 0.10 0.05 0 2 4 6 8 10 12 14 PH Fig.1 Effect of pH and buffer on the on-line preconcentration of 300 ng ml-' Cr'" with immobilized quinolin-8-01 a 0.1 moll-' acet- ate; 0 0.1 moll-' oxalate and * 0.1 moll-' citrate malonate or tartrate 242 Journal of Analytical Atomic Spectrometry March 1995 Vol. 100.05 1 0 2 4 6 8 10 12 PH Fig.2 Effect of pH and buffer on the on-line preconcentration of 100ngml-' Cr"' with Muromac A-1 H 0.1 moll-' acetate; 3 0.1 moll-' tartrate and A no buffer 0'25 r---l 0.20 1 I I T 8 0.15 t I \ 0 2 4 6 8 10 PH Fig.3 Effect of pH and buffer on the on-line preconcentration of 100 ngml-' Cr"' with Muromac A-1 L 0.1 mol I - ' citrate and V 0.1 moll - ' malonate tartrate over the range pH 2-5. The optimum pH with acetate malonate and tartrate was pH 4 and for citrate it was pH 3.The optimum pH with acetate is similar to that obtained with Muromac A-1 by Hirata et ul.,' who used ammonium acetate as the buffer. However Hirata et al. found that there was not much difference in the Cr"' response over the pH range 2.5-5. The sharp optimum for Cr"' with acetate indicated in Fig. 2 is more in line with other results,2*8 which suggests that efficient on-line extraction of Cr"' with iminodiacetate-based resins is not possible unless there is strict control of pH. Hirata et u1.l also reported that citrate was not considered a useful buffer because low recoveries were obtained for the elements con- sidered in their study (Cr"' Fe"' Ti V). However Fig. 3 indicates that citrate is almost as good a buffer as acetate for on-line preconcentration of Cr"' with Muromac A-1. Without a buffer the preconcentration of Cr"' by Muromac A-1 was much poorer and was dependent on the flow rate through the column.At flow rates of 2.8 and 6 ml min-' the absorbance signals obtained after 3 min preconcentration of 100 ng ml-' Cr"' were 0.09 and 0.04 respectively. In contrast similar absorbance signals were obtained for Cr following preconcentration at these flow rates when the modifiers were used (e.g. absorbances of 0.29 and 0.28 at 2.8 and 6.0 ml rnin - ' respectively for 0.1 mol I - ' acetate). This implies that the kinetics of the reaction with Muromac A-1 were improved when the acetate complex of Cr"' was formed. Further evidence of the difference in reaction kinetics was obtained by passing a 10 pg ml-' Cr"' solution at pH 4 through the column with the effluent flowing directly into the nebulizer of the AA spectrometer. When the solution contained 0.1 mol I-' acetate all of the Cr was retained by the column for 21 s (equivalent to 2.1 ml of solution) and the resin became saturated in 60 min.However without acetate an AAS signal of Cr was measured immediately the solution was passed through the column and it was 180 rnin before resin saturation occurred. Over the pH range 1-5 the predominant forms of Cr"' are Cr3+ and CrOH2+ according to Sperling et al.13 Although the log stability constant of the hydroxide complexes of Cr"' are higher than the p1 value of the acetate complex the concentration of acetate is much higher than that of hydroxide in this pH range and so dominates complex formation.With immobilized quinolin-8-01 the complexes formed by Cr"' and malonate citrate or tartrate were apparently stronger than the quinolinate complex and so prevented efficient on-line preconcentration of Cr"'. The situation was different with Muromac A-1 as the log stability constants of iminodiacetate complexes of Cr"' (pl = 10.9; pz=21.9)21 are higher than the PI of the malonate-Cr"' complex (8.26) and apparently also greater than those of the citrate and tartrate complexes although the values were not available. Hence reaction of Cr with the iminodiacetate groups was possible with these buffers at pH 2-5. Analytical Characteristics The enrichment factors detection limits and characteristic concentrations of Cr"' obtained by on-line preconcentration with immobilized quinolin-8-01 and Muromac A-1 are given in Table 1.The results are for a preconcentration time of 3 rnin at 6 ml min-'. When longer times of up to 18 min were used with Muromac A-1 the enrichment factor increased and the detection limit and characteristic concentration decreased in proportion. The enrichment factor obtained with Muromac A-1 was greater than that with quinolin-8-01 probably because the Cr"' complex formed with iminodiacetate is stronger than the quinolinate complex. Also the competing effects of hydrox- ide are smaller at pH4 than at pH 9. The Cr calibration graphs obtained by AAS with both types of resin were linear to at least an absorbance of 0.8 which corresponded to concentrations of 100-fold greater than the Cr"' detection limits.Impurity levels of Cr in the reagents were low and did not cause a problem in this work. Blank signals were negligible Table 1 Comparison of analytical characteristics for on-line precon- centration of Cr"' with resins based on immobilized quinolin-8-01 and iminodiacetate Resin* Characteristic Particle size/pm Column length/mm Optimum pH and buffer Enrichment factort Detection limitf/ng ml- ' Characteristic concentration$/ng ml- ' Efficiency of extraction7 (YO) Breakthrough capacity" /meq uiv g - ' Muromac A-1 Quinolin-8-01 (iminodiacetate) 125.- 175 250-500 100 150 9; acetate 4; acetate 23 61 6 2 6 2 - 78 0.14 0.26 * For 3 min at 6 ml min-'. t Ratio of calibration graph slopes obtained with and without Based on 3s of the signal of the lowest calibrant concentration on-line preconcentration.(absorbance d0.05; n = 5 ) . 5 Equivalent to an absorbance of 0.0044. 7 Same value calculated for 0.1 0.2 and 0.5 pgml-' Cr"' (ref. 11). II See ref. 11 for method of calculation. Jotrrnul of Analytical Atomic Spectrometry March 1995 Vol. 10 243Table 2 A-1 for 3 min at 6 ml min-'; acetate buffer at pH of 9 and 4 for quinolin-8-01 and Muromac A-1 respectively Effect of extraneous ions on the absorbance signals of Cr after on-line preconcentration by immobilized quinolin-8-01 and Muromac Extraneous ion Na+ (as NaCl or NaNO Ca2+ (as C~CI,) Mg2+ (as MgCl,) Fe3+ (as FeCI,) Fe2+ (as FeSO,) Cu2+ [as Cu(NO,),] Cr"' (as Na,Cr04) C1- (as NH4CI) SO,,- [as (NH,),SO,] NO3- (as NH,N03) Concentration/ pg ml-' 1000 5000 loo00 10 20 100 100 500 50 100 50 100 50 100 1 2 loo00 20000 500 2000 100 500 Recovery (YO)* Quinolin-8-01 100 97 89 105 94 47 101 97 96 94 100 27 100 92 100 100 98 96 101 100 100 100 Muromac A - l 97 95 89 101 100 89 98 97 100 100 99 95 98 97 102 96 95 100 101 100 99 * Percentage ratio of absorbance signal of Cr with and without extraneous ion; 400 and 100 ng ml-' Cr"' for quinolin-8-01 and Muromac A-1 respectively.(absorbance <0.005) for preconcentration times of up to at least 30 min. Table 3 Mass of extraneous ions extracted by Muromac A-1 during on-line preconcentration of Cr"' for 3 min at 6 ml min-I; 100 ng ml - ' Cr"' Interferences As it was intended to analyse sea estuarine and river waters for Cr the effects of various potential interferents were studied. Different concentrations of the individual ions were added to solutions of Cr"' containing 0.1 moll-' acetate with the pH adjusted to 9 or 4 for preconcentration of 400 or 100 ng ml-' Cr"' by immobilized quinolin-8-01 or Muromac A- 1 respect- ively.The absorbance values obtained with and without the extraneous ions were compared in order to obtain a percentage recovery of the signal. The results are given in Table 2. With quinolin-8-01 suppressive interferences of > 20% were caused by 100 pg ml-' concentrations of Caz+ and Fe". A small effect by loo00 pgml-' Na was also observed. The difference in the interference effects caused by Caz+ and Mg2+ was unexpected as the log stability constants of the quinolinate complexes are similar 3.27 and 3.66 respectively.21 Also the interference caused by Ca2+ was greater than that of Cu2+ or Fe3+ which form stronger complexes with quinolinate (log stability constants of 12.2 and 12.3 respectively).Few interferences were observed with Muromac A-1 with only 10000pgml-' Na+ and 100pgml-' Ca2+ causing noticeable effects. The mass of Ca2+ Cu2+ Fe2+ Fe3+ and Mg2+ extracted by Muromac A-1 from individual 100 pg ml-' solutions of the ions at pH 4 were calculated from FAAS analyses of eluate solutions. The results in Table 3 indicate that considerable amounts of the extraneous ions can be retained by the resin without affecting the preconcentration of 100 pg ml-' Crrrr (Table 2). The breakthrough capacity of Muromac A-1 (Table 1) corre- sponds to an uptake of 1900 pg Cr3+ so it is not surprising that negligible interferences (Table 2) were caused by 360-175Opg of Ca2+ Cu2+ Fez+ Fe3+ or Mg2+ (Table3).However the efficient uptake of Ca2+ is more difficult to explain. The log stability constants of iminodiacetate com- plexes of Ca2+ and Mg2+ are similar (3.6 and 3.7 respectively) and much lower than those of Cu2+ and Fe3+ which are Ion* Amount retained/pg Extraction (YO) Ca2 + 1750 97 c u 2 + 1650 92 Fez' 1060 59 Fe3+ 465 26 Mg2+ 3 60 20 * For 100 pg ml-' of each ion; total mass loaded 1800 pg. similar to that of Cr3+ 10.9. It seems that the literature values of stability complexes only have limited value in explaining the results obtained during on-line preconcentration with quinolin-8-01 and Muromac A-1. Other factors such as steric effects particle surface morphology and the role of competing ligands (acetate and hydroxide) need to be considered more fully.Table 4 Effect of hydroxylamine and acetate buffer on the reduction of Cr"' and subsequent collection of Cr"' by Muromac A-1 Volume of 1 mol 1-' NH,OH*/ml 0.5 1 .o 2.0 1 .o 1 .o 2.0 2.0 2.0 2.0 Volume of 1 mol I - ' acetatet/ml 1 1 1 5 10 5 10 15 20 Recovery 70 70 75 90 85 100 100 95 90 ( O/O ) * Solution acidified with 2 ml of 2 mol 1-' HCl before NH,OH i After addition of acetate and pH adjustment volume is made up $ Based on the ratio of the absorbance signals obtained for reduced solution added. to 100 ml with distilled water. CrV' and Cr"' (100 ng ml-') after preconcentrations. 244 Journal of Analytical Atomic Spectrometry March 1995 Vol.10Analysis of Estuarine Water and Sea-water Muromac A-1 was used for on-line preconcentration of Cr"' in the analysis of waters because it has a greater enrichment factor than quinolin-8-01. As CrV1 was not extracted by Muromac A-1 and did not interfere with on-line preconcen- tration of Cr"' speciation measurements were possible through separate determination of the concentrations of Cr"' and total dissolved ionic Cr. Chromium is known to exist as both Cr"' and CrV' in sea- water,23 with the concentration of Cr"' generally much lower 0.25 I 0.20 8 0.15 C a +? 0.10 t 0.05 1 0 10 20 30 40 50 60 70 80 Tim e/m i n Fig. 4 Effect of the time delay between addition of acetate to reduced solutions of CrV1 and the start of on-line preconcentration of Cr"' by Murornac A-1 (see Table 4 for details) than that of Cr".Hence procedures for the speciation of Cr are required to investigate the distribution of Cr"' and Crvl in estuarine coastal and open sea-waters. In this study the detection of elevated concentrations of Cr was of interest as part of a pollution monitoring programme. When conversion of one oxidation state to the other is attempted in the speciation of Cr to obtain an estimate of the total concentration of dissolved ionic Cr it is usually Cr"' that is oxidized to CrV'.13 This generally requires a period of boiling with an oxidant such as CerV (ref. 24) or peroxydisulfate with a silver catalyst,25 for quantitative conversation of C P to CrV1 in sea-water. However the efficient collection of Cr"' by Muromac A-1 permits selective preconcentration of this ion and calculation of the total Cr concentration after reduction of CrV1 to Cr"' with hydroxylamine as proposed by Isshiki et Some modification to their reduction procedure was required in this study to obtain quantitative reduction of Cr" as the amount of hydroxylamine and acetate buffer added to the solutions affected the recovery of Cr by Muromac A-1 as shown in Table 4.It was necessary to add 2 ml of 1 moll-' hydroxylamine to the CrV1 solution (which was eventually made up to 100ml) to achieve complete reduction to Cr'". This is twice the amount recommended by Isshiki et aLZ0 The concentration of acetate in the final solution has to be optim- ized to achieve efficient collection of Cr"' by Muromac A-1. The optimum concentration was 0.05-0.1 moll- ' acetate obtained by adding 5-10 ml of 1 mol I-' acetate to the Cr solution (final volume 100 ml). It is not clear why addition of higher concentrations of acetate reduced the collection efficiency (see Table 4).It was also discovered that the time delay between the addition of acetate buffer and the start of preconcentration Table 5 Concentrations of Cr"' total ionic Cr and CrV1 (ng ml-I) in estuarine and coastal waters obtained by on-line preconcentration on Muromac A-1 and analysis by FAAS and ETAAS; mean and standard deviation of triplicate analyses Sample location* Yoker Old Kirkpatrick Dumbarton Cardross Ardmore Ravenglass Whitehaven Techniquet F E F E F E F E F E F E F E Cr"' 8.1 & 0.3 9.8 f 1.4 4.5 k0.3 5.2f 1.3 4.6 f 0.2 4.5 f 1.3 3.8 f 0.2 3.8 f0.7 3.1 f0.2 3.8 f 1.3 2.8 f 0.2 2.7f0.6 3.1 f 0.2 2.8 0.6 Total Cr 8.7 f 0.3 10.2 f 1.8 5.3 f 0.4 6.3 f 1.3 5.3 k 0.3 5.2 f 1.2 4.5 k 0.4 4.2 f 0.7 3.8 f 0.3 4.9 & 0.6 3.1 k 0.4 3.0 k 0.6 3.4 f 0.3 3.2 f 0.6 CrV'$ 0.6 f 0.4 0.8 k 0.5 0.7 & 0.4 0.7 f 0.5 0.7 f 0.4 - - - - - * Ravenglass and Whitehaven are Cumbrian coastal waters; the rest are from the Clyde estuary with Yoker farthest Ardmore closest to the t F = FAAS (present method); E = ETAAS.3 Obtained by subtraction of Cr"' from total Cr. mouth of the estuary. Table 6 Recovery of added Cr"' and CrV' to water samples from the Clyde Estuary Sample Cr added*/pg Yoker 0 2 Old Kirkpatrick 0 2 Dumbarton 0 2 Cardross 0 2 Ardmore 0 2 Cr'" found/pg 0.67 2.66 0.34 1.32 0.36 2.36 0.31 2.30 0.25 2.24 Recovery (%) Total Cr found/pg Recovery (YO) 99 98 100 99 99 0.72 2.72 NDt 0.40 2.37 0.37 2.35 0.28 2.25 100 ND 98 99 98 * Separate additions of Cr"' or Cr"' to 80 ml volumes of water samples.t ND = not determined due to lack of sample. Journal of Analytical Atomic Spectrometry March 1995 VoZ. 10 245influenced the efficiency of the concentration of Cr"' produced by reduction of Crv' as indicated in Fig. 4. Provided the solution was preconcentrated within 10 min of the addition of buffer the expected amount of Cr"' was collected by the Muromac A-1 resin. For delay times between 10-60minl a lower recovery was obtained. Thereafter the amount of Cr retained increased again to the expected level. Although a rigorous explanation of this unusual occurrence is not possible at present the effect is caused by the hydroxylamine as similar trends were observed when the reducing agent was added to solutions of Cr"' before addition of acetate and preconcentration.The concentrations of Cr"' and total dissolved ionic Cr obtained for each of the samples collected from the Clyde estuary and the Cumbrian coast are given in Table 5. The solutions were pumped through the Muromac A-1 column for 6 min at 6 ml min-'. For the Clyde samples a difference due to more than chance was demonstrated for the Cr"' and total Cr results by the t-test (P=0.05) so concentrations are also given for Cr" obtained by subtracting the Cr"' concentration from the total Cr Concentration. No statistical difference was demonstrated for the Cr"' and total Cr concentrations in the Ravenglass and Whitehaven samples.The results in Table 5 suggest that the source of Cr in the Clyde is upstream of Yoker as the concentration of Cr"' decreases as the river moves downstream to Ardmore. In contrast the concentration of CrV' does not change significantly. The solutions prepared for on-line preconcentration were also analysed by electrother- mal AAS (ETAAS) and similar concentrations were obtained although the precision was poorer (Table 5). When the samples from the Clyde estuary were spiked individually with 1 or 2 pg of Cr"' or Cr" to increase the concentration of each ion to 13-27 ng rnl-' recoveries of 98-100% were achieved by the preconcentration FAAS method as shown in Table 6. The repeatability of the method was also acceptable.When the samples from Cardross and Yoker were analysed on three separate days the Cr"' concen- trations obtained were 3.8 k0.2 3.8 f0.2 and 3.8 f0.2 ng ml-' for Cardross and 8.1 f0.6 8.1 f0.3 and 8.3 f0.3 ng ml-' for Yoker. The corresponding total Cr concentrations were 4.4 k 0.3 4.5 k 0.4 and 4.4 f 0.2 ng ml - ' for Cardross and 8.8k0.7 8.7k0.3 and 9.0k0.2 ng ml-' for Yoker. In each case the mean and standard deviation of triplicate results are given. CONCLUSION The study has shown that on-line preconcentration of Cr"' can be achieved with resins based on quinolin-8-01 or iminodiacet- ate functional groups provided the pH is optimized and acetate is used as an auxiliary complexing reagent. Muromac A-1 an iminodiacetate-based resin was preferred for 'complexation of Cr"' as it gives a larger enrichment factor and exhibits lower interferences from concomitant ions.The possibility of Cr"'-Crv' speciation has been demonstrated using a con- venient method of reducing CrV' to Cr"' to allow determination of the total dissolved ionic Cr concentration. Although the application described in the work concerned analysis of estuarine and sea-waters by FAAS the methodology for preconcentration can be used with ICP optical emission spectro- metry or mass spectrometry. In future studies either of these techniques may be more appropriate considering the concentrations of Cr"' and Crv' in the samples. Abd Ulhafid Belazi is thanked for providing the results obtained by ETAAS. REFERENCES 1 7 - 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 Hirata S.Umezaki Y. and Ikeda M. Anal. Chem. 1986,58,2602. Prakash N. Csanady G. Michaelis M. R. A and Knapp G. Mikrochim. Acta 1989 111 257. Porta V. Sarzanini C. and Mentasti E. Mikrochim. Acta 1989 111 247. Beauchemin D. and Berman S . S. Anal. Chem. 1989 61 1857. Heithmar E. M. Hinners T. A. Rowan J. T. and Riviello J. M. Anal. Chem. 1990 62 857. Alexandrova A. and Arpadjan S. Analyst 1993 118 1309. McLaren J. W. Lam J. W. H. Berman S. S. Akatsuka K. and Azerado M. A. J. Anal. At. Spectrom. 1993 8 279. Ebdon L Fisher A. S. Worsfold P. J. Crews H. and Baxter M. J . Anal. At. Spectrom. 1993 8 691. Cotton A. A. and Wilkinson G. Advanced Inorganic Chemistry 4th edn John Wiley and Sons New York 1980. Mohammad B. Ure A. M. and Littlejohn D. J . Anal. At. Spectrom. 1992 7 695. Mohammad B. Ure A. M. and Littlejohn D. J. Anal. Ar. Spectrom. 1993 8 325. Mohammad B. Ure A. M. and Littlejohn D. Mikrochim. Actu 1994 113 325. Sperling M. Xu S. and Welz B. Anal. Chim. Acta 1992,64,3101. Jen J.-Y. Ou-Yang G.-L. Chen C.-S. and Yang S.-M. Anulysr 1993 118 1281. Milacic R. and Stupar J. Analyst 1994 119 627. Elmahadi H. A. M. and Greenway G. M. J. Anal. At. Spectrom. 1994 9 547. Fang Z. L. Ruzicka J. and Hansen E. H. Anal. Chim. Acta 1984 164 23. Kumamaru T. Matsuo H. Okamoto Y. and Ideda M. Anal. Chim. Acta 1986 180 171. Hirata S. Hondu K. and Kumamaru T. Anal. Chim. Actu 1989 221 65. Isshiki K. Sohrin Y. Karatani H. and Nakayama E. Anal. Chim. Acta 1989 224 55. Kotrly S. and Sucha L. Handbook of Chemicul Equilibria in Analytical Chemistry Ellis Horwood Chichester 1985. Stary J. Anal. Chim. Acta 1963 28 132. Rubio R. Sahoquillo A. Rauret G. and Quevauviller Ph. lnt. J . Environ. Anal. Chem. 1992 47 99. de Andrade J. C. Rocha J. C. and Baccan W. Analyst 1984 109 645. Welz B. Microchem. J. 1992 45 163. Paper 4/04558K Received July 25 1994 Accepted September 27 1994 246 Journal of Analytical Atomic Spectrometry March 1995 Vol. 10

ISSN:0267-9477    

DOI:10.1039/JA9951000241

出版商:RSC    

年代:1995

数据来源: RSC

摘要:

Speciation of Arsenic by the Determination of Total Arsenic and Arsenic(iii) in Marine Sediment Samples by Electrothermal Atomic Absorption Spectrometry* P. BERMEJO-BARRERA M. c. BARCIELA-ALONSO M. FERRON-NOVAIS AND A. BERMEJO-BARRERA Department of Anulytical Chemistry Nutrition und Bromatology Faculty of Chemistry University of Santiago de Compostela 15706 Santiago de Journal of Analytical Atomic Spectrometry Compostelu Spain A method for the determination of arsenic(II1) and total arsenic was studied. Arsenic(II1) was chelated with 2% sodium diethyldithiocarbamate in water extracted with isobutyl methyl ketone and determined by electrothermal atomic absorption spectrometry using palladium as chemical modifier. The effects of pH extraction time and amounts of reagents required for the extraction were studied.The detection limit of this method was 25 pg kg-'. Different modifiers [ LaCI,-HN03 Pd-Mg(N03)J for the total arsenic determination in slurries of marine sediment samples were also investigated; LaCI,-HN03 was avoided because a double peak was observed. The detection limit for the determination of total arsenic using Pd-Mg(N03)* as modifier was 44 pg kg- '. The precisions obtained for the different amounts of arsenic(ix1) and total arsenic were 0.57-2.99 and 0.83-0.45% respectively. The graphite furnace programmes accuracy and interierences for both procedures were also studied. The method has been applied to the determination of arsenic(rri) and total arsenic in sediment samples from Galicia (North West of Spain). Keywords Arsenic species sediment slurry sampling electrothermal atomic absorption Spectrometry Arsenic is a ubiquitous trace element and may be found in water food and sediments.Although the extensive use of arsenic compounds in insecticides herbicides feed additives and wood preservatives has been decreasing during the last years,' the emission of arsenic compounds from industrial discharges waste incineration etc is still significant. Nowadays there is considerable evidence to suggest that the toxicity and physiological behaviour of arsenic are dependent on its chemi- cal form the arsenic(v) species being less toxic than arsenic(Ir1). It is necessary therefore to develop sensitive methods for the determination of arsenic in its different forms. Arsenic is usually determined by hydride generation atomic absorption spectrometry.In the last years some problems due to the use of sodium tetrahydroborate have been avoided' with the intqDduction of flow injection. Furthermore since trivalent and' pentavalent arsenic show different behaviour in the generation process some papers have been published on the selective determination of As"' in presence of AS".^-^ The use of a flow injection on-line separation of As"' with sodium diethyldithiocarbamate ( NaDDC) reduction of AsV and pre- concentration for electrothermal atomic absorption spec- trometry (ETAAS) allows the determination of As"' and total arsenic in a short time.' Recently Chen et aL8 determined * Presented at the Seventh Biennial National Atomic Spectroscopy Symposium (BNASS) Hull UK July 20-22 1994 arsenic(v) and arsenic(II1) species in environmental samples by coprecipitation with zirconium hydroxide and were able to differentiate between As"' and AsV by using different temperature programmes in the graphite furnace. Metal determination in sediment samples has usually been performed by flame AAS or ETAAS with a prior sample preparation step.In order to avoid problems related to conven- tional wet-oxidation and dry-ashing procedures the use of slurry sampling was proposed to determine lead in marine sediment samples.' Slurry sampling has also been applied to the determination of As in mussel samples using Pd-Mg( NO,) as chemical modifier." In this work a method for the determination of As"' and total arsenic in marine sediment samples is developed.Arsenic(rI1) is separated by solid-liquid extraction with NaDDC and extracted into isobutyl methyl ketone (IBMK). Aliquots of the organic phase are determined by ETAAS using palladium in IBMK as chemical modifier according to Tserovsky et a/." Total arsenic is determined by ETAAS using slurry sampling and Pd-Mg(NO,) and LaC1,-HNO are studied as chemical modifiers to stabilize the As. EX PER I M ENTA L Apparatus Measurements were performed using a Perkin-Elmer Model 1 l00B atomic absorption spectrometer equipped with an HGA-400 graphite furnace atomizer and an AS-40 auto- sampler. An electrodeless discharge lamp operated from an external power supply at SmA which provided a 193.7nm line with a spectral bandwidth of 0.7 nm was used. Deuterium background correction and pyrolytic graphite coated graphite tubes with L'vov platforms were used.All measurements made during this study used integrated absorbance with an inte- gration time of 3 s. The temperature and time programmes for the atomizer are shown in Tables 1 and 2. The volume injected was 20 pl. Agitator (Vibromatic) and magnetic agitator (Agimatic) from Selecta (Spain) were used in the slurry preparation. Table 1 ETAAS programme for total As determination Temperature/ Ramp time/ Step "C S Dry 1 100 10 Dry 2 200 25 Pyrolysis 480 10 Pyrolysis 1200 10 Atomize 2500 0 Clean 2600 1 Hold time/ 20 25 10 15 3 3 S Gas flow/ ml min-' 300 300 300 300 stop 300 Journal of Analytical Atomic Spectrometry March 1995 Vol. 10 247Table 2 ETAAS programme for As"' determination Temperature/ Ramp time/ Hold time/ Gas flow/ Step "C S S ml min-' Dry 1 50 2 8 300 Dry 2 120 10 15 300 Pyrolysis 1100 15 30 300 Atomize 2500 0 3 stop Clean 2600 1 3 300 Reagents Arsenic trichloride stock standard solution 1 mg ml-' of As (BDH Poole UK).Prepare every test solution with ultrapure water immediately before use. Triton X-100 Polyethylene glycol mono (p-l,1 ,3,3-tetra- methylbutylphenyl). Ether for gas chromatography obtained from Merck (Darmstadt Germany). Palladium aqueous solution. Prepare by dissolving 300 mg of palladium (99.999%) (Aldrich Chemical Milwaukee WI USA) in 1 ml of concentrated nitric acid and diluting to 100 ml with ultrapure water. If dissolution is incomplete add 10 p1 of hydrochloric acid (Suprapur 35% with a maximun arsenic content of mg 1-' BDH) to the cold nitric acid and heated to gentle boiling in order to volatilize the excess of chloride.Palladium IBMK solution. Transfer 1 ml of aqueous palladium solution containing 180 mgl-I of Pd with 4ml of water into a test-tube. Add 2ml of acetic acid buffer (pH =2.63) 2 rnl of NaDDC and 3 ml of IBMK and shake for 1 min. Allow the two phases to separate and then remove the organic phase containing the palladium. Magnesium nitrare Suprapur Merck (Darmstadt Germany). Nitric Acid Suprapur (69.0-70.5 YO). Maximum arsenic content 5 pg I-' Merck (BDH) Poole UK. Lanthanum trichloride stock standnrd solution Suprapur 99.99% Aldrich Chemical Milwaukee WI USA 100 mg 1-'. Prepare by dissolving 26.7340 g of lanthanum trichloride in 100 ml of ultrapure water.Reference material PACS- 1 Harbour Marine Sediment. Obtained from the National Research Council of Canada (NRCC) Ottawa Ontario Canada. Sodium diethyldithiocarbamute Merck Poole UK. Prepare by dissolving 2 g of NaDDC in 100 ml of ultrapure water. Acetic acid glacial HPLC grade Scharlau. Sodium acetate Krist Suprapur Merck Darmstadt Zsobutyl methyl ketone (IBMK) AnalaR Merck Poole UK Argon N50 purity. Use as a sheath gas for the atomizer and to purge internally. Ultrapure water resistivity 18 Mi2 cm-'. Obtained using a Milli-Q water purification system (Millipore). All glassware material was kept in 10% nitric acid for at least 48 hours and washed three times with ultrapure water before use. Germany. Procedure for the Slurry Preparation Lyophilized marine sediment samples were ground to reduce them to a particle size of less than 250pm.A portion of the sample 0.25 g was weighed and placed in a polyethylene vial and zirconia beads (5 g) and 3 ml of water were added. The vial was agitated in a flask shaker (Vibromatic) for 60 min and the beads were separated using a sieve funnel (Haldenwanger Technische Keramic Dusseldorf Germany). Finally the slurry was adjusted to its final volume (10 ml) by addition of water and the amount of Triton X-100 necessary to obtain a concen- tration of 0.1% Triton X-100. A portion of the slurry with an appropriate amount of chemical modifier was transferred into the autosample cups and stirred magnetically before being measured. The procedure for the slurry preparation was reported in a previous work.' Extraction Procedure Lyophilized marine sediment sample (0.1 g) with 5.2 ml of water were transferred into a test-tube. A 2 ml volume of acetic acid buffer (pH = 4.63) 1.2 ml of NaDDC and 2.0 ml of IBMK were added and the solution was shaken vigorously for 4 min.The two phases were allowed to separate and then the organic phase was removed for ETAAS analysis using the HGA-400 programme in Table 2. RESULTS AND DISCUSSION Total Arsenic Determination Optimization of the furnace programme Experiments were carried out to optimize the temperatures and times for the drying pyrolysis and atomization steps for both palladium-magnesium nitrate and lanthanum trichloride as chemical modifiers. These conditions were optimized by means of several measurements for a slurry sediment sample and aqueous standards.With both modifiers two drying steps were used to dry the slurry completely 100 "C for 10 s followed by 200°C for 25 s (Table 1). Two pyrolysis steps were used; the first at 480°C and the second at a higher temperature. In previous work on lead determination in sediment slurries,' the incomplete ashing of the organic matrix of the slurry presented problems with the background signal and the use of air or oxygen as an auxiliary gas in the pyrolysis step was proposed. In the arsenic determination in sediment slurries there were no problems with the background signals and improvement was observed with the use of air or oxygen in the first pyrolysis step. To optimize the second pyrolysis temperature the pyroly- sis curve was performed for both modifiers.When lanthanum- nitric acid was used the optimum pyrolysis temperature was 1200"C but a broad signal with a double peak was observed (Fig. 1). Different amounts of lanthanum and nitric acid were used but no improvement was obtained. This behaviour has been observed previously with the use of this chemical modifier in the determination of arsenic in mussel slurries." The pyrolysis curves for palladium-magnesium nitrate as modifier are shown in Fig. 2 and it can be seen that in both cases (slurry sediment sample and aqueous standard) 1200 "C can be used as optimum pyrolysis temperature. The determination of the optimum atomization temperature was carried out by studying different atomization temperatures between 1400 and 2700 "C.The optimum atomization tempera- tures were 2100°C for a slurry sediment sample and 2000°C for arsenic aqueous standards. We selected 2100 "C as optimum because at this temperature there are no problems with the aqueous standards. To avoid memory effects the graphite tube was cleaned at 2300°C for 3 s with an argon flow of 300mlmin-'. The optimized temperature programme for the arsenic determi- -. . . - __ 0.500 I-- - I 0 1 2 Time/s 3 Fig. 1 as chemical modifier in total arsenic determination Double peak observed with the use of lanthanum-nitric acid 248 Journal of Analytical Atomic Spectrometry March 1995 VoI. 100.15 0.12 0 m 2 0.09 < a 2 0.06 0.03 0 500 800 1100 1400 1700 2000 Temperature/ C Fig. 2 Pyrolysis curves when palladium-magnesium nitrate was used as chemical modifier in total arsenic determination A slurry sediment sample; B aqueous standard nation in marine sediment slurries with palladium-magnesium nitrate modifier is shown in Table 1.Amount of chemical modiJier A series of measurements were carried out to determine the optimum concentration of both palladium and magnesium nitrate to be used by adding different amounts of each to a series of slurries in the absence of the other. Variations in the integrated absorbance for different amounts of palladium and magnesium nitrate for sediment slurries are shown in Fig. 3. The optimum amounts chosen were 15 mg 1-' of palladium and 10 mg 1-' of magnesium nitrate. Calibration and standard additions graphs To obtain a calibration graph appropriate volumes of solutions of palladium magnesium nitrate and Triton X-100 were added to standard aqueous solutions containing arsenic concen- trations between 0 and 20pg1-' to give concentrations of 15 mg 1- ' 10 mg 1- ' and 0.1 % respectively.The standard additions method was used over the same range of concen- trations using a sediment sample. The obtained equations were as follows Calibration graph Qa= -2.2 x 1.1 x lO-'C r=0.999 Standard additions graph Qa=8.9 x 10-'+1.23 x lO-'c r=0.999 where Q is the integrated absorbance and c is the arsenic concentration in pg 1-I; both graphs are shown in Fig.4 where they exhibit similar slopes. This means that aqueous calibration is a real possibility and this type of calibration was therefore used. Sensitivity The limit of detection (LOD) the lowest concentration level that can be determined to be statistically different from a blank is defined as LOD = 3SD/m (where rn is the slope of the calibration graph) corresponding to a 99% confidence level.The limit of quantification (LOQ) is defined as the level above which quantitative results can be obtained with a specified degree of confidence. At the 99% confidence level the value recommended is LOQ = 10SD/m. In both cases the SD is the within-run standard deviation of a single blank determination. The values obtained were 0.22 pg 1-' for LOD and 0.73 pg I-' for LOQ based on ten replicate determinations of the blank. The LOD and LOQ for the sediment sample using 0.1 g of sample diluting a slurry to 10 ml and taking 500 p1 to prepare the final solution with the modifier (1000 pl) were 44 and 146 pg kg- ' respectively.The characteristic mass m is defined as the mass of analyte in picograms required to give a signal of 0.0044 s for integrated absorbance. The characteristic mass obtained was 6.8 f0.17 pg. Precision The within-run precision (RSD) of the method (instrumental and matrix factor) obtained for ten replicate analyses of a single sample during the same run was 2.5% for 4.8 pgl-' of arsenic. The within-batch precision of the method obtained for ten replicates of three samples with different concentrations of arsenic added was also investigated. To study the within-batch precision three samples with 5 10 and 20 pg 1-' of arsenic added were used and the results were 0.8 0.5 and 0.5% respectively.Accuracy To study the accuracy of the method the certified reference material PACS-1 with a certified arsenic content of 21 1 t 11 mg kg- ' was used. The reference material and the 0.24 0.20 0 t m 0.16 0 n a 0.12 0.08 / I I- .- - 17 .,L I 12 12 1' Fig.3 Effect of palladium and magnesium nitrate on the integrated absorbance signal for total arsenic determination 0.39 0.29 CL C ; 0.19 w 11 Q 0.09 - 0.01 I I I I 0 4 8 12 16 20 IAsJ/pg I - ' Fig. 4 Calibration graphs for total arsenic determination Journal of Analytical Atomic Spectrometry March 1995 Vol. 10 249Table 3 Interferences on the determination of total arsenic 0.12 8 : a m 0.10 Ll 0.08 Ion Ag + AP + Cd2+ co2 + Cr3 + c u 2 + Fe3 + Mg2' Mn2+ Ni2 + Pb2+ Si Zn2 + F - I - S2 - SCN - S203'- so42- - - - Concentration added/ 1000 15000 600 500 1600 3000 400 2500 5000 lo00 2000 250 800 500 400 1500 300 300 1500 pg 1-' Absorbance variation ( O/o 1 - 6.9 + 10.0 + 3.5 + 7.0 - 10.3 - 9.3 - 6.4 + 9.8 - 10.1 +7.1 + 5.0 - 4.3 + 2.0 - 8.2 - 9.8 + 6.7 - 4.4 - 5.2 + 7.3 Concentration equivalent in the sediment/mg kg-' 1 600 24000 960 800 2560 4800 640 4000 8000 1600 3 200 400 1280 8000 6400 2400 480 480 2400 blank were measured three times.The result obtained expressed as mean f SD was 21 8 0.06 mg kg- '. Interferences Experiments were performed to discover the extent to which the proposed method is susceptible to interferences from elements known to interfere in the determination of arsenic by AAS. An interference was defined as significant if a variation of more than &lo% in the measurement was observed.The obtained results are given in Table 3. The levels of these elements in marine sediment samples are usually lower than those necessary to produce interferences. Arsenic( 111) determination Extraction pH The effect of changes in pH at which arsenic(rr1) was complexed and extracted was studied using 0.1 g of sediment samples and carrying out the extraction procedure at different acidities. The results (Fig. 5 ) show that the optimum pH is 4.63. Amount of NaDDC Different amounts of NaDDC were added to a series of 0.1 g of sediment samples and the extraction procedures were carried 0.124 8 0.104 m + a d 0.084 0.064 0.044 3.7 4.0 4.3 4.6 4.9 5.2 5.5 PH Fig. 5 Effect of pH on arsenic(iI1) extraction with NaDDC out.It was found that 0.4% of NaDDC was sufficient for total complexation and that an excess did not have any adverse effect (Fig. 6). Amount of IBMK To study the effect of organic solvent on the extraction procedure four solutions containing 0.1 g of sediment sample 0.4% of NaDDC and different amounts of IBMK (1 2 3 and 4 ml) were prepared. After extraction all organic phases were diluted to 4 ml. No improvement in the absorbance signal was observed for the four volumes. A volume of 2 ml was selected as the minimum volume necessary to develop this work. Extraction time efeect Using 0.1 g of sediment sample the extraction procedure was carried out for different times (0.5 1 2 3 4 and 5 min). The results in Fig. 7 show that 4 min is sufficient for total extraction. Optimization of furnace programme Experiments were carried out to determine the best tempera- tures and times for the various steps of the graphite furnace programme.Two drying steps were used; the first at 50 "C and the second at 120°C. The temperatures for the pyrolysis and I I -.-I- 0.06 t - 1 0 0.1 0.2 0.3 0.4 0.5 0.6 [NaDDCI (YO) Fig. 6 Effect of NaDDC concentration on arsenic(II1) extraction 250 Journal of Analytical Atomic Spectrometry Murch 1995 Vol. 100.100 I 0 2 3 4 5 Ti me/mi n L l l l l 0.025 0.010 - Fig. 7 Effect of time on arsenic(tt1) extraction with NaDDC atomization steps were optimized by means of several measure- ments for the extraction of arsenic(rI1) in the presence and absence of palladium. In Fig. 8 the pyrolysis curves are shown with and without palladium added.The As is lost at pyrolysis temperatures greater than 1000 “C but when the palladium was added the As was stabilized to 11OO”C so the use of palladium as chemical modifier is recommended. The optimum atomization temperatures were 2400 and 2500 “C in the absence and presence of palladium. The internal gas-stop mode and maximum power were used for the atomization step. The furnace conditions are given in Table 2. 0.4 0.3 0 0 C m 2 0.2 0 D Q 0.1 0 300 800 1300 1800 Temperature/ C Fig. 8 B without palladium Pyrolysis curves for arsenic(ii1) A with palladium added; Table 4 Interferences on the determination of arsenic(I1t) Amount of chemical modijier A series of measurements were carried out to determine the optimum concentration of palladium by adding different amounts of palladium to a series of sediment extracts.The optimum palladium concentration was 5 mg 1- ’. Cdihrution and sensitivity To obtain a calibration graph several 1 ml volumes of standard arsenic solution containing 0-80 pg 1- ’ of arsenic(rr1) were subjected to the extraction procedure. Then aliquot(s) of these samples were placed in cups and the optimum amount of palladium was added. The final concentrations of arsenic(1lr) in the cup were 0 8 16 and 32 pg1-’. The standard additions method was used with the same range of concentrations of arsenic used in the calibration being added to a sediment sample. The equations obtained were as follows Calibration graph Qa= -3.2 x 10-3+3.1 x lOP3C r=0.999 Standard additions graph Q,=2.6 x 10-’+3.02 x 10-3C r=0.999 where Q is the integrated absorbance and c is the arsenic(1rr) concentration in pg 1-l.Both graphs had similar slopes so it is not necessary to apply the standard additions method for sample measurement. To study the sensitivity of the method the LOD LOQ and rn were calculated and the results obtained were 25 and 137 pg kg-’ for LOD and LOQ and 28.1 kO.93 pg for m,. Precision and accuracy The within-run precision (RSD) of the method (instrumental and matrix factors) obtained for ten replicate analyses of a single sample during the same run was 2.1 1 YO. The within-batch precision of the method obtained for ten replicates of four samples with different concentrations of arsenic(m) added was also investigated. To study the within- batch precision four samples with 8 16 and 32 pgl-’ of arsenic added were used and the results were 2.99 1.03 and 0.57% respectively.The accuracy of the method was investi- gated by determining the recovery of standard additions of arsenic(n1) to sediment samples. The recoveries were 93 102.8 and 99.9% for 8 16 and 32pg1-’ of arsenic added to a sediment sample respectively. Ion Ag + Cd2 + co2 + Cr3 + c u z + Fe3 + Mn” Ni2 + Pb2+ Si Zn2 + F - I - SCN - S203’- so,2- Concentration added/ 800 lo00 750 800 400 600 5000 1000 2000 loot) 800 300 400 1300 500 lo00 I % - ’ Absorbance variation ( n/o 1 - 1.2 - 8.9 - 5.8 + 7.6 + 1.1 + 4.3 0.0 - 1.2 - 2.7 + 9.7 - 8.2 + 9.5 - 2.4 - 3.5 + 9.5 - 9.4 Concentration equivalent in the sediment/mg kg-’ 32 40 30 32 16 24 200 40 80 40 32 12 16 52 20 60 Journal of Analytical Atomic Spectrometry March 1995 Vol.10 251Table 5 Concentration of arsenic(II1) and total arsenic in marine sediment samples from Galicia coast (Spain) n 1 2 3 4 5 6 7 8 9 10 Total As/mg kg- ' 41.1 49.0 44.0 59.4 55.5 25.4 90.7 59.4 67.2 76.4 As"'/mg kg- ' 0.06 0.09 0.06 0.05 0.05 0.04 0.03 0.06 0.10 0.07 Interferences To study the interferences in the arsenic(m) extraction method the experiments were carried out with suitable cations to form complexes in the same pH range with NaDDC and to produce absorption at a similar wavelength to arsenic. The results are given in Table4. The levels of these elements in marine sediment samples are less than those necessary to produce interferences. Applications The method was applied to the determination of arsenic in marine sediment samples from the Galicia coast (North West of Spain).Two sub-samples taken from each sediment sample were prepared in the form of slurries and for the extraction procedure and two sub-samples were subjected to AAS. The results obtained are shown in Table 5. The values obtained for total arsenic are 25.4-90.7 mg kg- and for arsenic(II1) 0.05-0.1 mg kg- REFERENCES 1 2 7 8 9 10 11 Dickerson 0. B. in Metals in the Environment ed. Waldron H. A. Academic Press London 1980. Lin Y.-h. Wang Xi.-r. Yuan D.-x. Yang P.-y. Huang B.-I. and Zhuang Z.-x. J. Anal. A t . Spectrom. 1992 7 287. Aggett J. and Aspell A. C. Analyst 1976 101 341. Howard A. G. and Arbab-Zavar M. H. Analyst 1981,106 213. Holak W. and Specchio J. J. At. Spectros. 1991 12 105. Lopez A. Torralba R. Palacios M. A. and Camara C. Talanta 1992,39 1343. Sperling M. Yiu X.-f. and Welz B. Spectrochim. Acta Part B 1991 46 1789. Chen Y.-k. Qi W.-g. Cao J.-s. and Chang M.-s. J. Anal. At. Spectrom. 1993 8 379. Bermejo-Barrera P. Barciela-Alonso C. Aboal Somoza M. and Bermejo-Barrera A. J. Anal. At. Spectrom. 1994 9 469. Bermejo-Barrera P. Lorenzo-Alonso M. J. Aboal-Somoza M. and Bermejo-Barrera A. Mikrochim. Acta in the press. Tserovsky E. Arpadjan S. and Karadjova I. Spectrochirn. Acta Part B 1992 47 959. Paper 41054856 Received September 9 1994 Accepted November 7 1994 252 Journal of Analytical Atomic Spectrometry March 1995 Vol. 10

ISSN:0267-9477    

DOI:10.1039/JA9951000247

出版商:RSC    

年代:1995

数据来源: RSC

摘要:

Automatic Wavelength Calibration Procedure for Use with an Optical Spectrometer and Array Detector* Journal of Analytical Atomic Spectrometry DARAN A. SADLER AND DAVID LITTLEJOHN Department of Pure and Applied Chemistry The University of Strathclyde 295 Cathedral Street Glasgow U K GI 1 X L CHARLES V. PERKINS AT1 Unicam Limited York Street Cambridge U K CBl 2PX Echelle spectrometers with cross-dispersion are often used in emission spectroscopy owing to their high spectral resolution and good light throughput. The resultant two-dimensional dispersion plane is ideally suited to array detectors such as the charge-injection device (CID) or charge-coupled device (CCD). The successful coupling of an khelle spectrometer with an array detector permits true simultaneous spectroscopy to be performed over a large spectral range.In order to correctly identify spectral features it is necessary to have an accurate wavelength calibration function which maps the CCD/CID pixel co-ordinates to wavelength. A new approach to the wavelength calibration of optical spectrometers with array detection is proposed that does not involve a direct modelling of the spectrometer dispersion. Instead the difference between an ideal conceptual spectrometer and the physical instrument is modelled. The procedure is able to compensate for the effects of manufacturing tolerances and local temperature and pressure conditions. Preliminary results obtained by simulation with a computer-modelled Cchelle spectrometer has shown that a subpixel accuracy in the predicted position of spectral lines can be achieved over a temperature range 5-35 "C.Keywords Optical emission spectrometry; ichelle spectrometer; wavelength calibration; charge-injection device; array detector Emission spectrometers based on the echelle grating' are often used in analytical spectroscopy due to the high resolution and good light-throughput that can be achieved. An echelle grating obtains high resolution by using high orders instead of a high groove density. The dispersion of such a grating is also enhanced by the large blaze angle at which it is used typically > 45". Each diffraction order has a very short free-spectral range and hence to cover a wide spectral region it is necessary to utilize multiple orders. These orders need to be spatially separated from each other in the exit plane of the spectrometer.Order separation is achieved by an auxiliary dispersive element such as a prism or low-dispersion grating placed in the spectrometer so as to disperse the wavelengths in a direction perpendicular to the dispersion of the echelle. This arrange- ment is often described as cross-dispersion. An echelle spec- trometer with cross-dispersion results in a two-dimensional representation of the emission spectrum. By matching the physical dimensions of an array detector such as a charge injection device (CID) or a charge coupled device (CCD) to the extent of the spectrometer dispersion it is possible to record simultaneously the emission spectrum over a wide spectral range. This offers the analyst considerable * Presented at the Seventh Biennial National Atomic Spectroscopy Society (BNASS) Hull UK July 20-22 1994. advantages in comparison to a sequential single-channel instrument.Most elements emit at a number of wavelengths and the choice of spectral line for elemental analysis is deter- mined by line sensitivity and the effect of spectral interferences. As the array detector system will record multiple spectral lines from the same analyte the necessity for an a priori decision concerning line selection is removed which may simplify method development. In addition the wide wavelength cover- age offered by an array-detector-kchelle system allows true simultaneous multi-element spectroscopy to be performed. A number of articles detailing the application of CIDs and CCDs to analytical spectroscopy have been Various echelle spectrometers with array detection have been reported in the literature.Bilhorn and Denton6 designed a CID-based echelle system for elemental analysis with a three- electrode dc argon plasma. The sensitivity and dynamic range of the system were found to be comparable to an emission spectrometer with photo-multiplier tube detection.' A CCD- based echelle spectrometer has been designed by Scheeline et a[.* for use with a variety of emission sources. The system has been used to investigate the matrix dependence of analyte excitation in a high voltage spark discharge' and for the determination of electron temperature and density in a spark discharge." Barnard et al." describe a custom-designed array detector the segmented-array CCD detector (SCD) for use with a specific echelle grating spectrometer" designed for inductively coupled plasma optical emission spectroscopy (ICP-OES).This system is unique in that the SCD is actually a number of individual linear CCDs fabricated onto the same device. Whilst this does not give the full wavelength coverage offered by conventional array detectors this unique structure offers true random-access readout and a high UV quantum efficiency. WAVELENGTH CALIBRATION In order to perform either qualitative or quantitative analysis with an array-detector-kchelle system it is necessary to deter- mine the relationship between the spectral line wavelength and pixel position on the array detector. This mapping between wavelength and position is the wavelength calibration function and is dependent upon the optical design of the spectrometer.In order to take advantage of the high spectral resolution offered by the echelle grating and the stability of the array detector the wavelength calibration function should predict the position of a spectral line with sub-pixel accuracy. By regular re-calibration the wavelength calibration function will also compensate for drift in the position of a spectral line associated with temperature and (if using a prism for cross- dispersion) pressure changes in local ambient conditions. Therefore any automatic procedure for determining the wave- length calibration function must be flexible enough to compen- Journal of Analytical Atomic Spectrometry March 1995 Vol. 10 253sate for small changes in the function whilst still retaining a high positional accuracy.There are a number of alternative approaches to determining the wavelength calibration function of an optical spectrometer with array detection. The method chosen is determined by a number of factors including the required accuracy and the optical design. The CID-based spectrometer of Bilhorn and Denton‘ maintains calibration through reference to the pos- ition of the mercury 253.65 nm line emitted from a Pen-Ray mercury pen lamp. The co-ordinates of analyte lines are stored in computer memory relative to the pixel upon which the 253.65 nm line falls. By frequently determining the position of this reference point the wavelength calibration may be updated. This procedure assumes that any temperature or pressure related effects result in a linear shift in the wavelength cali- bration function.As will be shown later this assumption is incorrect if a prism is used for cross-dispersion and hence can result in inaccuracies in the predicted positions of spectral lines. The approach taken to wavelength calibration by Miller and Scheeline13 is to model the dispersion of the spectrometer from a knowledge of the theoretical performance of the optical components. By adjusting various parameters in the mathemat- ical model it is possible to compensate for changes in the wavelength calibration function. This procedure is relatively straight-forward to implement for the case of an echelle spectrometer with grating cross-di~persion,’~ however when using a prism for cross-dispersion it becomes extremely difficult to implement.For the case of the grating cross-dispersion system the dispersion of the echelle and the cross-dispersor are perpendicular and non-interactive. However when using a prism there is a complex and non-linear interaction between the prism and grating which is difficult to model. In addition the refractive property of the prism is a complex and often unknown function of wavelength temperature and atmos- pheric pressure. For these reasons this approach is not suitable for spectrometers with prism cross-dispersion. A wavelength calibration procedure often implemented in spectrometers using only a single order is to model the dispersion with a low-order p~lynomial.‘~.’~ Moore and Furst16 applied this procedure to an echelle spectrometer with cross-dispersion provided by a separate quartz-prism spec- trometer.The exit slit of the echelle spectrometer was also the entrance slit for the quartz-prism spectrometer which used a photographic plate as the detector. The dispersion of each spectrometer can be expanded to a low-order polynomial and the dispersion of the entire system is described by two coupled polynomial equations. Unfortunately it was found that when using a system with an internal prism cross-dispersor the interaction between the echelle and the prism dispersion means that a simple polynomial equation does not describe the positions of the spectral lines on the array detector with sufficient accuracy. It is clear therefore that a procedure is required that can perform high accuracy wavelength calibration on an kchelle spectrometer with an internal prism cross-dispersor for use with array detection. METHODOLOGY The procedure for wavelength calibration described here does not model the spectrometer dispersion with an explicit func- tional form.Instead the difference between an ideal conceptual spectrometer and the physical instrument is modelled. In this context ‘ideal spectrometer’ is taken to mean an instrument in which the positions and performance of all components optical or otherwise are free from manufacturing variability and that the spectrometer behaves identically to its ray-trace model. For any spectrometer operated at room temperature there exists a mapping between the x and y co-ordinates of the array 254 Journal of Anulvticul Atomic Soectrometrv.March 1995. detector pixels and the wavelength j of a spectral line in diffraction order m. This mapping for the ideal spectrometer may be written as (x y)=am j . ) x*O,(m E;) y*O,(m i.) These mappings between the x y co-ordinates and the m E space may not be known explicitly. However for any given values of m and i. the corresponding x y co-ordinate may be found from a computer ray-trace model of the spectrometer. The wavelength calibration function of a physical spec- trometer is distorted from the ideal mappings due to manufac- turing tolerances and local temperature and pressure conditions. For any co-ordinate (x y ) this distortion may be represented by a vector with magnitude (Ax Ay) such that Ax = x’ - O,(m i ) which may be broken down into two separate mappings and and Ay=y‘-O,(rn i.) Where x’ and y’ are the actual co-ordinates of a spectral line of wavelength i in order m.Owing to the physical causes underlying the distortion of the ideal mappings it is expected that Ax and Ay will vary in magnitude across the plane of the array detector. It is possible to define a pair of difleerencefunctions representing the depen- dence of Ax and Ay upon the array detector pixel co-ordinates 1.e. and These difference functions are unique to each spectrometer and respond to variations in operating temperature and atmos- pheric pressure. Once the difference functions have been found the predicted position (x’ y’) of a spectral line of wavelength i. in order rn is given by x’ = 0,(113 i ) +f,(.U y) and A mathematically equivalent approach using one-dimensional difference functions is described by Grosser and Collins” to correct for positioning errors in the grating drive mechanism of a scanning single-channel monochromator. The difference functions may be determined by utilising a calibration lamp that emits a number of spectral lines at known wavelengths.A comparison of the actual array detector co-ordinates of the calibration lamp lines with the positions predicted by the ideal spectrometer mappings allows the difference functions to be determined. This procedure can be automated and performed at regular intervals to allow changes in local atmospheric conditions to be compensated. EXPERIMENTAL The procedure to determine the wavelength calibration func- tion by the method of difference functions has been tested using computer simulation.A computer ray-trace model of an echelle spectrometer with prism cross-dispersion has been simulated with the BEAM4 Optical Ray Tracer program (Stellar Software Berkeley CA USA). For the model a Czerny-Turner’* configuration for the Vol. 1 0collimator and camera mirrors was chosen with a CaF prism for cross-dispersion positioned after the echelle grating in the light path. The spectrometer was designed such that the spectral range from 167 to 460 nm would fit onto a dispersion plane equal in size to the commercially available Kodak KAF-1300L CCD. Table 1 lists the specifications of the spec- trometer and CCD system. The longest dimension of the CCD (the x co-ordinate) was aligned parallel to the dispersion of the prism the inter-order direction.Each diffraction order although subject to a slight tilt then runs in the general direction of the y co-ordinate or intra-order direction. Three other spectrometers based upon the same design as in Table 1 were simulated with the BEAM4 program. Each of these spectrometers designated as SP1 SP2 and SP3 had slight differences representative of likely manufacturing toler- ances in the position and properties of some of the optical components. Table 2 shows the differences between SP1 SP2 SP3 and the ideal spectrometer of Table 1. To model tempera- ture effects the refractive index of the prism material was varied. No other temperature-related effects were simulated as it is expected that the refractive index will be the most temperature sensitive parameter in a well-designed spec- trometer.The refractive index p of CaF was calculated using the Sellmeir formula determined by Ma1its0n.l~ The variation of the refractive index with temperature was modelled using the empirical equation 0.0425 . ( pm) - 0.0 1 37 * = (-1.169+ dT ) x 10-5 peroc The equation was derived from data obtained from Merck2' and is accurate to within 3% over the wavelength range used. The performances of each of the spectrometers SPI SP2 and SP3 were simulated at three temperatures 5 24 and 35 "C. The ideal spectrometer model was simulated at 24 "C. In order to calculate the difference functions the emission from a Hg discharge lamp was simulated upon the entrance aperture of each spectrometer. The wavelengths of the spectral lines emitted from the lamp are shown in Table 3 along with the order in which they appear.The position of the 20 lines in the spectral region 167-460nm was determined for SP1 SP2 and SP3 at each of the three temperatures and for the ideal conceptual spectrometer From these positions Ax and Ay were calculated for each line. A pair of difference functions for SP1 SP2 and SP3 at each temperature modelled was then deter- Table 1 Details of spectrometer design Focal length of collimator mirror Focal length of camera mirror Groove density of echelle Blaze angle of echelle Prism material Included angle of prism CCD pixel dimensions CCD dimensions Order for h= 167 nm Order for h = 460 nm Dispersion in order 135 Dispersion in order 49 500 mm 235 mm 79 lines per mm 63.4" 42.8" 16 pm x 16 pm 20.48 mm x 16.38 mm 135 49 5.6 nm mm- 2.0 nm mm-' CaF2 Table3 Spectral lines from an Hg discharge lamp in the spectral region 167-460 nm Wavelength/nm 184.9 1 194.17 226.22 237.83 248.20 253.65 265.20 280.35 289.36 296.73 302.15 3 12.57 313.17 318.77 334.15 365.02 365.44 404.66 407.78 435.84 Order 122 116 100 95 91 89 85 80 78 76 75 72 72 71 67 62 62 56 55 52 mined using the general linear least squares method.21 The equations used to describe the difference functions were f x ( x y ) = + + (uJx2) -k (a4y) + (u5-x-Y) + (a,y2) + (a7xy2) and Table 2 Differences between SP 1 SP2 SP3 and ideal spectrometer designs with the u and b parameters as variables. A number of different functions some including cubic terms in x and y were evalu- ated.However the above equations were found to provide statistically satisfactory fits to the calibration data. The x2 statistic for each of the difference functions was calculated. In all cases the value of the x2 per degree-of-freedom was less than 0.6 indicating a statistically good fit to the available data. Table 4 lists the a and b coefficients for SPl at 5 and 24°C. The coefficients were calculated with the dimensions of the x and y positions in mm. RESULTS AND DISCUSSION The applicability of this procedure for wavelength calibration was tested by a comparison of the expected position of a number of spectral lines to the position predicted by the use of the difference functions. For each of the spectrometers SP1 SP2 and SP3 at the three temperatures modelled six spectral lines evenly distributed in each of orders 49 69 89 108 122 and 135 were simulated incident upon the spectrometer entrance aperture.The positions of these 36 lines upon the CCD are the 'expected' positions of the spectral lines. The same 36 spectral lines were also simulated incident upon the ideal spectrometer model and their ( x y) position on the CCD were found. The difference functions for a particular spec- trometer derived from the calibration lines were used to determine the magnitude of Ax and Ay at any given ( x y ) co-ordinate of a spectral line. By adding Ax and Ay to the Rotation of collimator mirror about Rotation of echelle about grating normal Groove density of echelle Included angle of prism Focal length of camera mirror/mm vertical axis Ideal spectrometer SP1 SP2 SP3 - 5" - 5.2" -4.8' - 5.2" 0" 79 42.8" 235 0.5" - 0.5" - 0.5" 79.0006 78.9994 79.0006 42.86" 42.56" 42.74.234 236 236 Journul of Analytical Atomic Spectrometry March 1995 Vol. 10 255Table 4 Difference function coefficients for SP1 at 5 and 24°C 5 "C 24 "C Subscript a and b 1 2 3 4 5 6 7 a - 0.0001955 0.010387 16 - O.oooO732 1 0.0030077 1 -0.00027590 - 0.00048686 - 0.00000203 b 0.00496609 - 0.0220283 - 0.00012422 - 0.00025018 -0.00011167 -0.00016583 - 0.000000422 a - 0.00002 16 0.01040342 0.00294804 - O.oooO7264 - 0.00028008 - 0.00047859 0.00000266 b 0.00715383 - 0.0002570 -0.00017825 - 0.00026209 - O.ooO10852 - 0.OOO 19545 0.00000084 ideal spectrometer position of the spectral lines the 'predicted' positions of the 36 lines were deduced for each spectrometer.The effect of both manufacturing tolerances and changes in ambient conditions is to shift the position of all spectral lines upon the array detector. The magnitude of this shift may be large compared to the resolution of the spectrometer and errors in the assignment of spectral lines may occur. Table 5 shows the apparent wavelengths of six spectral lines evenly distributed across the operating spectral range for spec- trometers SP1 SP2 and SP3 at 24°C. The differences in the apparent wavelengths of the six spectral lines are due to the differences in the position and performance of the optical components that define SP1 SP2 and SP3 i.e. the manufactur- ing tolerances. After calibration by the use of the difference functions all the spectral lines shown in Table 5 are predicted to within 0.0004 nm of the actual wavelength. The most significant effect of changes in ambient temperature and pressure is on the refractive properties of the prism.This will cause a change in the separation of the orders on the array detector. A secondary and much smaller effect is to alter the dispersion within each order. The effect of a tempera- ture change on the x co-ordinate of a spectral line is shown in Table 6. The shifts in the x co-ordinate of six spectral lines from their nominal positions are shown as a fraction of the slit-image height for SP1 at 5 24 and 35 "C. The slit-image height used was 72 pm. At 24°C the shift in the x co-ordinate is due to manufacturing tolerances alone (the nominal design of the spectrometer was simulated at 24°C).However it can be seen that significant changes in the order separation occur with a change in temperature. The largest temperature related effects occur at the higher wavelengths this is due to the wavelength dependence of dp/dT. Table 7 shows the error in the prediction of the centre of the slit-image also as a fraction of slit height after calibration by the difference function procedure. The position of all the spectral lines is correct to Table 5 spectrometers at 24 "C Apparent wavelengths of six spectral lines for un-calibrated Wavelength h/nm SP 1 SP2 SP3 167.165 167.146 167.181 167.148 184.691 184.68 1 184.698 184.685 207.845 207.84 1 207.844 207.852 253.096 253.108 253.085 253.108 327.29 1 327.3 18 327.272 327.308 462.979 463.02 1 462.961 462.992 Table6 Shift in the x direction (inter-order) of a spectral line as a fraction of slit-height for SPl before calibration; slit-height = 72 pm Wavelength hlnm 5 "C 24 "C 35 "C 167.165 - 1.07 - 1.19 - 1.26 1 84.69 1 - 0.64 - 0.49 - 0.40 207.845 - 0.34 - 0.04 0.13 253.096 - 0.03 0.4 0.63 327.29 1 0.10 0.59 0.88 462.979 0.04 0.58 0.89 Table7 slit-height for SP1 after calibration; slit-height = 72 pm Error in the predicted centre of slit image as a fraction of Wavelength h/nm 5 "C 24 "C 35 "C 167.165 0.06 0.00 0.04 184.69 1 0.00 0.00 0.02 207.845 -0.01 0.00 0.02 253.096 0.00 0.00 - 0.04 327.29 1 0.00 0.00 - 0.05 462.979 0.00 0.00 - 0.05 within 6% of the height of the slit image.This corresponds to a maximum error in the prediction of approximately 0.25 pixels. The mean error calculated over all 36 test spectral lines and the range of the errors in the predicted position of all 36 test spectral lines are shown in Table 8 for the y co-ordinate direction. All values given are the absolute errors in pm and should be compared with the dimension of a Kodak CCD pixel 16 pm. The equivalent data for the error in the predicted position for the x co-ordinate direction are shown in Table 9. From Tables 8 and 9 it is evident that the difference function approach to wavelength calibration is able to correctly predict the position of all spectral lines with sub-pixel accuracy. For the intra-order direction the absolute error in prediction corresponds to a spectral error very much less than the limiting spectral resolution of the instrument defined as the spectral width of a pixel and hence accurate assignment of spectral features can be made.The largest error in the predicted position of a spectral line corresponds to just 0.0004 nm which Table 8 (intra-order); dimensions in pm Mean and range of error in predicted position for y direction 5 "C 24 "C 35 "C SP 1 Mean 0.2 1 0.02 0.02 Range -0.47 to 2.0 -0.38 to 0.49 -0.28 to 0.32 Mean - 0.04 0.02 - 0.08 Range -0.49 to 0.36 -0.47 to 0.27 -0.67 to 0.15 Mean - 0.28 - 0.04 -0.01 Range -0.26 to 0.52 -0.57 to 0.24 -0.56 to 0.81 SP2 SP3 Table 9 (inter-order); dimensions in pm Mean and range of error in predicted position for x direction 5 "C 24 "C 35 "C SP 1 Mean 0.65 - 0.06 2.0 Mean 0.68 0.00 2.1 Range -0.74 to 4.7 - 0.60 to 0.58 - 3.3 to 4.6 Mean 0.69 - 0.04 2.1 Range -0.77 to 5.2 -0.88 to 0.33 -2.2 to 4.3 Range -0.6 to -5.0 -0.81 to 0.71 -3.2 to 4.3 SP2 SP3 256 Journal of Analytical Atomic Spectrometry March 1995 Vol. 10occurs at ~ 1 6 7 n m .For the y co-ordinate direction there appears to be no correlation between temperature and the error in prediction of the position of a spectral line. This indicates that the form of the difference function used is able to fully compensate for temperature effects. However in the x co-ordinate direction the mean error in the predicted position is almost identical for each spectrometer at a given temperature. This indicates there is a direct corre- lation between temperature and the residual error in prediction. The form of the difference function used in this direction is not fully compensating for the non-linear change in order separation due to temperature changes. It is possible that a different form of the x co-ordinate difference function will more accurately compensate for temperature effects and this is under investigation at present.Even though the difference functions are not fully compensating for temperature related effects in the x co-ordinate direction the predictions produced by this procedure are still accurate to within temperature range of 5-35 "C. CONCLUSION A new approach to wavelength calibration spectrometer with array detection has been tested. The procedure works by calibrating 3 pixel over a for an optical proposed and the difference between the actual spectrometer and an idealized conceptual model.Testing of the difference function procedure has been performed by computer modelling of three spectrometers based upon the same nominal design with simulation of typical manufacturing tolerances built in to the assessment. Each of the spectrometers has been simulated at three temperatures over the operating range 5-35°C. It has been demonstrated that without adequate wavelength calibration serious errors in the assignment of spectral features can occur. By calibrating the difference functions with a standard mercury lamp the positions of all spectral lines in the wavelength range from 167 to 460 nm were predicted with sub-pixel accuracy. REFERENCES 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 Keliher P. N.and Wohlers C. C. Anal. Chem. 1976 48 333A Bilhorn R. B. Sweedler J. V. Epperson P. M. and Denton M. B. Appl. Spectrosc. 1987 41 1114 Bilhorn R. B. Epperson P. M. Sweedler J. V. and Denton M. B. Appl. Spectrosc. 1987 41 1125 Epperson P. M. Sweedler J. V. Bilhorn R. B. Simms G. R. and Denton M. B. Anal. Chem. 1988,60 32712 Sweedler J. V. Jalkian R. D. Pomeroy R. S. and Denton M. B. Spectrochim. Acta Part B 1989 44 683 Bilhorn R. B. and Denton M. B. Appl. Spectrosc. 1989 43 1 Bilhorn R. B. and Denton M. B. Appl. Spectrosc. 1990 44 5381 Scheeline A. Bye C. A. Miller D. L. Rynders S. W. and Owen R. C. Appl. Spectrosc. 1991 45 334 Bye C. A. and Scheeline A. Appl. Spectrosc. 1993 47,2031 Bye C. A. and Scheeline A. Appl. Spectrosc. 1993 47 2022 Barnard T. W. Crockett M. I. Ivaldi J. C. Lundberg P. L. Yates D. A. Levine P. A. and Sauer D. J. Anal. Chem. 1993 65 1231 Barnard T. W. Crockett M. I. Ivaldi J. C. and Lundberg P. L. Anal. Chem. 1993 65 1225 Miller D. L. and Scheeline A. Spectrochim. Acta Part B 1993 48 El053 Wollman S. T. and Bohn P. W. Appl. Spectrosc. 1993 47 125 Tseng C.-h. Ford J. F. Mann C. K. and Vickers T. J. Appl. Spectrosc. 1993 47 1808 Moore F. L. Jr. and Furst B. J . Opt. Soc. Am. 1972 62 762 Grosser Z. A. and Collins J. B. Appl. Spectrosc. 1991 45 993 The Design of Optical Spectrometers eds. James J. F. and Sternberg R. S. Chapman and Hall London 1969 Malitson I. H. Appl. Opt. 1963 2 1103 Crystan Handbook Merck Poole UK Numerical Recipes in Fortran Second Edition Press W. H. Teukolsky S. A. Vetterling W. T. and Flannery B. P. Cambridge University Press Cambridge 1992 Statistics for Technology Third Edition Chatfield C. Chapman and Hall London 1983. Paper 41051 441 Received August 22 1994 Accepted November 18 1994 Journal of Analytical Atomic Spectrometry March 1995 Vo!. 10 257

ISSN:0267-9477    

DOI:10.1039/JA9951000253

出版商:RSC    

年代:1995

数据来源: RSC

摘要:

Optimal Accuracy Precision and Sensitivity of Inductively Coupled Plasma Optical Emission Spectrometry Bioanalysis of Aluminium* TREVOR J . BURDEN J . J . POWELL AND R. P. H. THOMPSON Gastrointestinal Luhorcitorj The Rayne Institute St Thomas' Hospital London U K SEl 7EH P. D. TAYLOR Department of Chemistry Unicersity College London London U K WC2H OAI The accurate and precise analysis of trace elements in biological samples by inductively coupled plasma optical emission spectrometry (ICP-OES) has not been established particularly for elements such as Al Ni and Pb for which the literature is contradictory and methodology and sample preparations are poorly reported. We outline important features for the successful analysis of aluminium by ICP-OES. Three analytical lines for aluminium were considered; 167.020 nm is the most sensitive but an unresolvable iron interference (z 1 500 Fe Al) means that this line cannot easily be used for some biological samples particularly blood.The line at 308.215 nm is least sensitive. The line at 396.152 nm is the most suitable but requires baseline correction for a broad hydroxy peak at 397.000 nm and a calcium peak at 396.847 nm. There is also potential interference from a previously unreported iron peak at 396.121 nm (z 1 2500 Fe Al). Sheath gas height of the plasma viewing zone and sample diluent were most critical for optimum analytical conditions and these parameters may be interdependent. For a sheath gas flow of 1.05 I min-' urine diluted 1 + 4 with a 100 g 1- ' potassium chloride solution yielded a significantly enhanced net aluminium signal compared with urine diluted 1 +4 with water whereas reducing the sheath gas flow to 0.4 I min-' gave a relatively greater signal for the sample diluted with water.An analytical profile with a window size of 0.1 nm containing 54 wavelength increments each of 5 s integration time further maximized sensitivity and precision. Spectra required dynamic background correction and deconvolution. Thus sub pg I-' detection of aluminium is possible with 0.9 pg 1 - achievable in diluted clinical samples. These approaches are widely applicable for other elements by ICP-OES. Keywords Aluminium ; hiologicd samples ; inductively coupled plusma optical emission spectrometry Aluminium is an established toxin that can for example cause encephalopathy or bone pathology' in adults with renal dis- ease,' and in pre-term infants particularly those also with renal failure.2 More recently an aluminium associated anaemia has been reported in haemodialysis patients with elevated body levels of the metaL3 At these high levels aluminium is easily determined but such gross toxicity is now rare.Other disorders may be caused by aluminium at lower levels; for example the relationship between Alzheimer's disease and aluminium' remains unresolved. Understanding the potential long term effects of this modern toxin on the population will therefore - * Presented at the Seventh Biennial National Atomic Spectroscopy Symposium (BNASS) Hull UK July 20-22 1994. Journal of Analytical Atomic Spectrometry require great analytical precision but unfortunately many studies attempting to demonstrate small but possibly import- ant increases in the levels of aluminium in the body have been hampered by two problems.First aluminium is normally only present in body fluids at very low levels while it is environmen- tally ubiquitious' and thus is a major contaminant during sample collection preparation and analysis. Secondly alu- minium in biological samples is difficult to determine owing to the complex matrix nature of the analyte and low levels present. Electrothermal atomic absorption spectrometry (ETAAS) has therefore generally been preferred and methods have been developed to overcome the matrix of clinical samples and the high atomization temperatures required for alu- mini~m.~,' Nevertheless both time and particularly skill are required to undertake such analyses by ETAAS and the linear range for calibration is limited.Inductively coupled plasma optical emission spectrometry (ICP-OES) is a suitable technique for the determination of many metals of biological interest but the results for aluminium have been contradictory.* The high temperature of the atomiz- ation/ionization plasma of ICP-OES suits refractory elements such as silicon and aluminium and also quickly destroys the matrix thus minimizing this analytical interference. The con- tinuous integrated signal gives superior precision compared with ETAAS but the reported detection limits vary consider- ably (0.3-10 pg Furthermore there is little information on the requirements of sample volume sample dilution prep- aration and analyte interferences or even agreement on the choice of analytical line.We have therefore examined these questions and assessed the suitability of a bench-top monochro- matic ICP-OES instrument for the analysis of aluminium in biological samples. EXPERIMENTAL Instrumentation All analyses were carried out using a Jobin-Yvon JY 24 sequential ICP-OES instrument (Instruments SA Longjumeau France) using a standard Meinhardt nebulizer and a conventional Scott-type double-pass spray chamber. Sample Preparation Ultrapure water (Elga UHP High Wycombe UK) was used throughout and all aluminium standards were prepared from a 1000 mg 1-' aluminium stock standard solution (Spectrosol Merck Poole UK). Potential interference solutions were pre- pared accordingly from sodium potassium phosphorus cal- cium magnesium and iron salts (ARISTAR grade Merck or Gold Label quality Aldrich Chemicals Gillingham Dorset.UK). Aluminium contaminations were established as described Journal of Analytical Atomic Spectrometry March 1995 Vol. 10 259below using ICP-OES at 396.152 nm with background correction. Pooled samples of serum blood and urine from healthy volunteers were diluted in various ways spiked with alu- minium and used to assess the best working parameters and sample preparation. Precision was assessed using these samples at differing spiked aluminium concentrations and accuracy checked using biological standards (Seronorm trace elements urine batch no. 108 and Seronorm trace elements serum batch no.116 Nycomed AS Oslo Norway). Dilutions were carried out as shown under Results using either 0.10 moll-' nitric acid (Merck 'Aristar' grade prepared in ultrapure water) a solution of 100 g 1-' potassium chloride (Merck 'Analar' grade prepared in ultrapure water) a solution of 1OOgl-' sodium chloride (Merck 'Analar' grade prepared in ultrapure water) or a solution of 3.1 g 1-' caesium chloride (Aldrich Chemicals; prepared in ultrapure water). The salt-containing solutions were rendered aluminium-free using a solid-phase chelator which is based on a hydroxy-pyridone and is presently pa tent pending. Polypropylene laboratory ware was used throughout (Nalgene Merck) having been acid washed for 24 h in 2.7 moll - ' hydrochloric acid (Analar grade Merck) then 24 h in 0.32 mol 1-' nitric acid (Aristar grade Merck) and finally twice for 12 h in ultrapure water.This gave consistently undetectable aluminium contamination in ultrapure water. Emission Lines Sensitive aluminium lines are at 167.020 237.132 308.21 5 309.271 309.284 394.401 and 396.152 nm. No obvious spectral interferences are found in standard tables of spectral lines," but in body fluids the major inorganic biological elements namely calcium iron magnesium phosphorus potassium and sodium occur in such great excess of aluminium that minor spectral lines from these elements become apparent or spectral interferences extend over considerable wavelengths. These elements were therefore investigated for their effects at up to 1000 mg 1- ' with aluminium levels up to 50 pg 1- '.Limits of detection were calculated by examining peak profiles and represent the minimum resolvable analytical peak above the background noise. The net peak was obtained from the raw data points by non-linear least squares fit of a Gaussian peak superimposed on a quadratic background. RESULTS AND DISCUSSION Initial Assessment Analytical line 167.020 nm. Despite being the most sensitive line for alu- minium (detection limit less than 0.9 pg1-') it suffers a partially resolvable interference from iron. This iron inter- ference has been reported12 at 0.010nm distance from the aluminium peak but the two elemental lines actually are closer at 4 x nm from each other (Fig. 1). The aluminium peak is approximately 500 times more sensitive than the iron peak and so it may be possible to use this line if iron levels in the sample are low.237.312 nm. This was the least sensitive line for aluminium and was shifted from the baseline by iron and surrounded by iron peaks. Phosphorus also interfered. 308.215 nm. A signal for aluminium was given that was clear of immediately surrounding interferences but there was a hydroxy peak exactly at this wavelength that would affect the precision of any aluminium measurements made using this line. Phosphorus (1000mg1-') gave a small peak at this wavelength (Fig. 2) equivalent to an apparent signal of 10 pg 1 aluminium which decreased with decreased phos- t - m 0) a A . .... . ' .. . . . . .. .... 166.990 167.020 167.050 Mnm Fig. 1 Overlap of aluminium and iron lines at 167.020 nm in aqueous solution; A profile of aluminium (100 pgl-'); B profile of iron (50 mg I - ' ) 308.185 308.215 308.245 Mnm Fig.2 Aluminium line centred at 308.215 nm may suffer minor interferences from hydroxy iron and phosphorus species A analysis of ultra high purity water (no detectable aluminium present) with the hydroxy peak centred at 308.215 nm; and B an aqueous solution containing phosphorus (loo0 mg 1-') and iron ( lo00 mg I-'). Interference at 308.21 5 nm was further increased owing to the presence of phosphorus and the background marginally raised owing to the presence of iron phorus concentrations until this interference became negligible at 100 mg I - ' phosphorus. Iron (lo00 mg 1-') increased the apparent aluminium concentration by about 2 pg 1-I and this became negligible at less than 500 mg I-' iron.Care is therefore needed with samples that could have high concentrations of either of these two elements such as urine bone digest samples and blood which even when diluted for ICP-OES may have phosphorus levels exceeding 100 mg 1-'. However this line gave a perfectly linear response for aluminium standards of between 0 and 10mg1-' prepared in both ultrapure water and saline. The effect of dissolved solids for example in the case of saline and other experimental fluids could be allowed for by matrix-matching the standards. However the matrix within biological samples is always unknown and so compari- son with spiked samples to obtain a true (sample based) standard curve would be required although most biofluids would not contain sufficient aluminium to be detected at this wavelength.260 Journal of Analytical Atomic Spectrometry March 1995 Vol. 10309.271 and 309.284 nm. Interference from only supra- physiological levels of magnesium took place although these two aluminium peaks themselves were not fully resolvable making quantitative analysis difficult. A hydroxy peak occurred at 309.282 nm and caused further interference. 394.401 nm. This line was only half as sensitive as the line at 396.152 nm (below) in agreement with standard texts of emission lines." and also like the line at 396.152 nm was shifted from the baseline by calcium as previously reported.13 396.152 nm. This was the most sensitive line for aluminium in the visible region. Its peak was shifted from the baseline by a hydroxy peak centred at 397.000 nm and spanning 2.400 nm (Fig.3). The shape and height of this hydroxy peak was unchanged by acid concentrations of 0-10% m/v nitric acid. Calcium further increased both the baseline shift and the slope of the aluminium peak (Fig. 3) but did not alter the net signal. Hence the true aluminium signal could be computed (see under Emission Lines) from this sloping background as for example in Fig. 4 for the analysis of aluminium in urine. A previously unreported iron peak centred at 396.121 nm could also interfere (Fig. 5). Although this became insignificant at levels below 15 mg 1-I iron whole blood even when diluted 1 +9 for analysis contained 40-60 mg 1-l iron,14 which interfered with the analysis of aluminium at 396.152 nm.This iron line (396.121 nm) was 2500 times less sensitive than the aluminium line (396.152nm) and thus presented a five-fold lower inter- ference at this wavelength than at 167.020 nm. Furthermore the aluminium and iron lines were better resolved at 396.152 nm (3.1 x lop2 nm apart) than at 167.020 nm (4 x nm apart). Hence the contribution from iron could be computed as described above (see under Emission Lines) using a two Gaussian fit as for example in Fig. 6. The 396.152nm line was therefore the most suitable for further assessment with biological samples. All of the above lines were investigated with plasma gas flow maintained at 121min-I and sample pump speed at 1.0 ml min-l. Sample Analysis with Acidified Aqueous Diluent Analyses of urine blood and serum were carried out by diluting the samples with 0.10 mol 1-' nitric acid.The integration times that were chosen to determine the length of analysis also affected the amount of sample uptake precision and detection limits. The detection limit for aluminium in diluted clinical samples with a 60 point 5 s integration was 1.5 pg l-l but required 6ml of sample whereas a 60 point 1 s integration Ca line Al line Ca line Ar or OH line (397.970 nrn) (395.705 nrn) (396.1 52 nrn) (396.847 nm) 396.200 397.000 397.800 Unrn Fig.3 A broad hydroxy peak from analysis of ultra high purity water (no detectable aluminium present j centred at 397.000 nm. The aluminium line at 396.152nm is shifted from the baseline by the hydroxy peak. The small peak at 396.847 nm is due to some low level calcium contamination. B aqueous solution of calcium (100 mg 1-I) increases the baseline slope of the aluminium line at 396.152nm.A small peak centred at 395.705 nm is also from a calcium signal. At 397.970nm there is an argon or hydroxy peak which is common to both profiles 798.0 789.0 780.0 771 .O A c. u) t - x In c al c w .- c - 20.0 15.0 10.0 5.0 A - h - h - Fig. 4 (a) Urine diluted (1 + 2) with a solution of potassium chloride shows the aluminium peak centred at 396.152 nm and raised from the background by calcium with net slope increasing from left to right. (b) The analytical peak at 396.152 nm is resolved from the raw data by a non-linear least squares fit of a Gaussian peak superimposed on a quadratic background. (c) Gaussian peak is extracted from the raw data by subtraction of the background to yield the resolved analyt- ical peak had a detection limit of 3 yg 1-I and required only 2 ml of sample.The following studies were made with a 1 s integration time and dilutions were 1 + 9; (sample + diluent) for whole blood 1 + 7 for serum and 1 + 2 for urine. Dilutions were based on our previous results15 for efficient nebulization of these samples where the slope of the diluted sample-based standard curve approximates (& 15%) to that of the pure diluent. Urine is often supersaturated with salts at room temperature and was therefore heated to 40°C and allowed to cool before analysis since we have shown that this avoids sample precipi- tation.16 Diluted sample-based standard curves were used.The net aluminium signal was calculated as before (see under Analytical Line) using a single or double Gaussian approxi- Journal of Analytical Atomic Spectrometry March 1995 VoE. 10 261396.121 396.152 Unm Fig. 5 Profile from an aqueous solution of iron (50 mg I-') showing a peak at 396.121 nm that interferes with the aluminium peak at 396.152 nm mation with dynamic background correction such as in Figs. 4 and 6. The pooled samples after dilution had baseline aluminium concentrations of less than the ICP-OES detection limits and were therefore checked using ETAAS (Table 1). Precision was assessed in these diluted samples after spiking with aluminium (Table 1); with aluminium concentrations of 10 pg 1-' the standard deviation was generally c 5%. Standard deviations of measurements of diluted blood samples were slightly higher than those for urine and serum perhaps owing to interference of the aluminium line by the high levels of iron in blood.The accuracies of the reference materials (Table 2) were statistically indistinguishable from the certified values. The standard Meinhardt nebulizer remained largely patent although occasional blockages with diluted blood samples were experienced. Further work could consider alternative nebulizers such as the PTFE high-solids (Ebdon or V-groove) or direct injection. Under certain circumstances the presence of easily ionized elements"*'* may enhance the net analyte signal and this was further investigated. Sample Analysis with Electrolyte Containing Diluent Diluent and instrumental conditions It has been reported that diluting serum for analysis in an aqueous solution of potassium chloride" enhances the net alu- minium signal at 396.152 nm.It has also been suggested" that a combination of electrolyte (particularly sodium chloride) concentration of the diluent and the rate of sheath gas flow critically affects the signal-to-background noise ratio of this line. Although detectable aluminium contamination was removed from these diluents prior to acidification using a novel solid-phase chelator (see under Sample Preparation) any solid-phase chelator with a high affinity for aluminium could be used such as Chelex 100 [Dionex (UK) Farnborough UK]. Peak profiles at 396.152 nm are shown (Fig. 7) for urine diluted (1 +4)19 with either acidified ultrapure water or an acidified potassium chloride (100 g 1 - I ) solution. At a sheath gas flow of 0.4 1 min-' the urine diluted with acidified water showed the greater net signal.However by increasing the sheath gas flow to 1.05 I min-' the potassium chloride diluted sample showed the greater signal-to-noise ratio compared with the aqueous signal at the lower sheath gas flow rate. Reasons for this are not clear although examination of the shape of 36.6 36.0 35.4 34.8 c. .- C 3 Fig. 6 (a) Blood diluted (1 +9) with a solution of potassium chloride showing interference of aluminium (approximately 10 pg I - ' ) at 396.152 nm by iron (approximately 50 mg I-') at 396.121 nm. (b) Peaks at 396.121 nm for iron and 396.152 nm for aluminium are resolved from the raw data by a non-linear least squares fit of two Gaussian peaks superimposed on a quadratic background.(c) Gaussian peak for aluminium at 396.152 nm is extracted from the raw data the plasma (Fig. 8) showed that increasing the sheath gas flow raised the height of the central annulus of the plasma. Thus at the lower sheath gas flow of 0.4 1 min-' which favoured the aqueous diluent most aluminium is presumably atomized in the tail flame of the plasma while at the higher sheath gas flow of 1.05 1 min-' which favoured the potassium chloride diluent most aluminium is atomized in the annulus region of the plasma. Peak profiles at 396.152 nm are also shown in Fig. 9 for the same urine sample with differing diluents (1 +4) (a) acidified ultrapure water (b) caesium chloride (3.1 g 1-')9 (c) sodium chloride (100 g l-'),'' and ( d ) potassium chloride (100 g l-').I9 Sheath gas flow was adjusted for each diluent in order to maintain the plasma shape shown in Fig.8(c). In agreement with Chappuis et al.19 potassium chloride produced the greatest enhancement (Fig. 9) in signal intensity and was therefore further investigated. In contrast to Mauras and Allaing we could not reproduce the increase in signal intensity at 262 Journal of Analytical Atomic Spectrometry March 1995 Vol. I0Table 1 Precision of analyses in biological samples Diluted sample* Serum Serum Serum Serum Serum Serum Urine Urine Urine Urine Urine Urine Whole blood Whole blood Whole blood Whole blood Whole blood Whole blood n 2 6 6 6 6 6 3 6 6 6 6 6 2 6 6 6 6 6 Expected value/ 0.6@ 3.60 5.60 7.60 10.60 40.60 5.00 7.00 9.00 12.00 42.00 1.6@ 4.60 6.60 8.60 11.60 41.60 Yg 1-lt 2.0q Value found/ < 3.0 P8 1 - 9 3.05 4.66 7.41 10.20 41.3 < 3.0 5.29 7.08 8.82 12.20 43.0 < 3.0 4.11 5.97 7.96 11.33 44.9 Expected value ("/.I 84.7 83.2 97.5 96.2 101.7 105.8 101.1 98.0 101.7 102.4 89.3 90.5 92.4 97.7 107.9 Standard deviation W) 39 34 7.1 4.5 1.9 38 15 5.7 2.9 2.6 22 28 17 6.1 2.5 * Serum diluted 1 + 7 whole blood diluted 1 + 9 urine diluted 1 + 2.t First value of each sample type shows the baseline level of aluminium in the diluted sample and successive values show the level of baseline $ Results obtained with ICP-OES at 396.152 nm (see text) using 2 ml diluted sample and 60 point 1 s integration times. 5 Results obtained using ETAAS.plus spike. Table 2 Accuracy of analyses in biological reference materials Seronorm Certified value/ Mean value sample Batch No. Dilution n Pg 1-' obtained/pg 1-l Serum Serum Urine Urine 116 116 108 108 1+7 1+9 1+1 1+2 106k6 106+6 161* 161* 108.4 108.7 158.0 159.8 * Preliminary recommended value with independent analytical values of 160 and 162 pg 1-I. Fig. 7 Peak profiles of aluminium at 396.152 nm for a urine sample diluted (1 +4) with either A acidified ultrapure water or B potassium chloride (100 g 1-l) (a) sheath gas flow of 0.40 1 min-l (the aqueous diluted sample shows greater net aluminium signal); and (b) sheath gas flow of 1.05 1 min-l (the potassium chloride diluted sample shows greater enhancement of the aluminium signal and a signal to back- ground ratio superior to both peaks in (a) 396.152 nm using caesium chloride (2.5 g 1 final concentration). Plasma power was kept at 1.2 kW plasma gas flow at 12.0 1 min-' and sample pump speed at 1.0ml min-'.Photomultiplier voltage and nebulizer gas pressure both had minimal effects on signal-to-noise ratio for aluminium at 396.152 nm. Calibration and recovery experiments Previous worklg with a potassium chloride based diluent has analysed serum using a 1 +4 dilution and so this was further investigated here as was urine diluted 1 +4. Blood samples were not included because of viscosity and likely problems with blockage of the nebulizer. Potassium chloride (100 g 1-I) diluent was acidified (0.22 moll-' nitric acid) for urinary analysis but not for serum analysis.Acid concentrations greater than 0.3 mol l-' will start to precipitate out proteins from serum but this effect is considerably enhanced in the presence of potassium chloride (100 g 1-I) and thus only acid (up to 0.3 moll-') or electrolyte but not both may be used as the diluent for such samples. Results of serum and urine analysed in this manner are shown in Fig. 10. Slight deviation from the standard curves are shown (f7.6%; SD of the aqueous standard curves) and both enhancement and suppression were experienced. One reason for this may be that since the instrument set up relies partly on visual assessment of the portion of the annulus in the plasma (Fig. S) minor variations may marginally alter the relative ratios of emission signals for standard and diluted sample.Nevertheless for plasma diluted 1 +4 with the potass- ium chloride solution deviation from the standard curve was Journal of Analytical Atomic Spectrometry March 1995 Vol. 10 263Fig. 8 Changes in appearance of the plasma observed with increasng sheath gas flow rates when using a solution of yttrium nitrate as a marker (a) 0.40 ( h ) 0.80; (c) 1.05; and ( d ) 1.20 1 min-I. Increasing sheath gas flow raised the height of the central annulus [pink zone in (c.) and ( 4 3 of the plasma. A sheath gas flow of 0.40 1 min-' (a) was the optimum flow rate for aqueous diluted biological samples and a sheath gas flow of 1.05 1 min-' (c) was the optimum flow rate for potassium chloride diluted biological samples. The viewing tube of the spectrometer is outlined by a dotted white line in (c) less than for plasma added even at greater dilutions to nitric acid-acidified ultrapure water (Fig. 1 1 ).Thus the potassium chloride diluent not only enhanced signal-to-background ratio but also reduced the differences in emission intensities between standard and diluted sample of the same analyte concentration. It is still necessary however to use a sample-based standard curve for accurate analyses. Pooled or typical samples may be diluted and used to produce a standard curve yielding accurate data. Again further dilution experiments could be investigated using alternative nebulizers. Limits qf detection From this work the limit of detection for aluminium in diluted biological samples was 0.9 pg I-'. However the limits of detection for each sample type were dictated by the necessity 264 Journal of Analvtical Atomic Spectrometrjl Murch 1995 Vol.10t - (d tl) 3 Fig. 9 7 I -CSCI h- Peak profiles for the same urine sample diluted ( 1 +4) with acidified ultrapure water potassium chloride (100 g l-l) caesium chloride (3.1 g 1-') and sodium chloride (100 g 1-'). Potassium chlor- ide caused the greatest enhancement in signal intensity to dilute and was e.g. for serum 4.5 pg 1-1 with an analysis time of 6 min per sample. However precision was poor at this low level which is still outside the range of the very low levels of aluminium in normal serum and blood. For such samples ETAAS is therefore still recommended. In urine the limit of detection for aluminium by ICP-OES was also 4.5 pg 1-1 (dilution 1+4) and this is at the low end of the range of normal aluminium levels in urine.Indeed from our previous work,16 urine when correctly prepared may be diluted 1+1 for analysis giving a potential limit of detection of 1.8 pg 1-l. This low dilution is not so applicable to serum which is more viscous and less easily nebulized. CONCLUSIONS The emission line at 167.020 nm gave good detection limits (c 1.0 pg 1-l) for aluminium in aqueous solutions but iron interference means that this line cannot easily be used for all biological samples in particular blood. Nevertheless for plasma and serum which contain approximately 1000 pg 1-1 iron the added contribution to the levels of aluminium is about 2.0 pg 1-l. This could be subtracted or for most pur- poses ignored since most samples measured by ICP-OES would have aluminium levels in excess of 10 pgl-'.Two further lines could be used for bioanalysis of aluminium at 308.215 and 396.152 nm. The line at 308.215 nm was not as sensitive as that at 396.152 nm and had small hydroxy and phosphorus inter- ferences. However this line does not require background correction and would be particularly useful for rapid analysis such as with a polychromator based spectrometer. The 396.152 nm line may be used to analyse all diluted clinical samples but requires computer aided signal correction of peak profiles since there was a small iron interference at 396.121 nm and a hydroxy/calcium interference yielding a sloping background. The importance of these two interferences will vary between spectrometers.The sheath gas flow rate needs to be optimized and either acidified de-ionized water (0.22 mol 1-1 nitric acid) or potassium chloride (100 g 1-l) solution used as the diluent. The latter has the advantage of better signal noise ratio on analysis and closer approximation to the diluent-based standard curve as well as negating inter- sample matrix effects. However care is needed to produce an analyte free solution of potassium chloride and samples con- taining undigested protein (blood serum etc.) cannot be diluted with acidified potassium chloride since a precipitate is observed. Whether samples were diluted with acid or potassium 1000 (a 1 800 - 600 - 400 - 0 20 40 60 80 100 120 0 20 40 60 800 1 600 I A 0 25 50 75 100 [Alurninium]/pg I-' Fig. 10 Plots of urine and serum samples diluted with a solution of potassium chloride and spiked with various concentrations of alu- minium compared with similarly spiked acidified potassium chloride solutions.(a) Potassium chloride solution (100 g 1-l) acidified to 0.10 mol I-' nitric acid (A) and urine diluted 1 +4 with potassium chloride solution (100 g 1-I) and acidified to 0.10 moll-' nitric acid B; deviation of urine from potassium chloride solution was + 6.9%. (b) A different potassium chloride solution (100 g 1-I) acidified to 0.10 moll-' nitric acid (A) and a different urine sample diluted 1 +4 with potassium chloride solution (100 g 1-l) and acidified to 0.10 mol I-' nitric acid (B); deviation of urine from potassium chloride solution was -6.8%. (c) Potassium chloride solution (100 g I-') acidified to 0.10 moll-' with nitric acid (A) and serum diluted 1 +4 with potassium chloride (100 g 1-I) solution (B); deviation of serum from potassium chloride solution was - 6.8% chloride solution they still required sample-based calibrations for accurate results.The detection limit was 0.9 pg 1-1 for aluminium in biological samples which had been appropriately diluted with potassium chloride solutions and required a 6 ml final volume and analysis time of 6 min; or the detection limit was 1.8 pg 1-1 requiring a 2 ml volume and analysis time of 2 min. With a diluent of acidified ultrapure water these detec- tion limits were 1.5 pg 1-l and 3 pg l-' respectively. The degree of dilution depends on the sample type and the diluent used and in this work for example was 1 + 2 for urine 1 + 7 for plasma/serum and 1+9 for blood (sample+acid) or was 1 + 4 for serum and urine (sample + potassium chloride solu- tion). These dilutions approximated (2 15%) to the normal Journal of Analytical Atomic Spectrometry March 1995 Vol.10 2650 50 100 150 200 [Aluminiumypg r' Fig. 11 A serum sample diluted (1 + 7) with ultrapure water acidified to 0.10 mol I - ' nitric acid and spiked with various concentrations of aluminium and B spiked ultrapure water sample acidified to 0.10mol I-' with nitric acid; deviation of serum from water was - 14.3% standard curves (non-sample based) but reduced the effective detection limit of the technique for blood to 15.0 pg I-' and serum and urine to 4.5 pg 1-'. Optimal dilutions cannot be recommended as these will depend on the sample type diluent nebulizer and the agreement sought between diluted-sample based standard curves and diluent standard curves. Work with more efficient nebulization should further improve sensitivity and make possible the analysis of normal levels of aluminium in blood.We are grateful to the Wellcome Trust and the Jean Shanks Foundation for their support and for the continuing support of the Special Trustees of St Thomas' Hospital. REFERENCES 1 Stewart W. K. in Aluminium in Food and the Environment eds. Massey R. C. and Taylor D. Royal Society of Chemistry Cambridge 1989 pp. 6-19. 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Freundlich M. Zillerrelo G. Alikbol C. Strauss J. Faugere M. C. and Malluche H.H. Lancet 1985 ii 527. Mc.Gongle R. J. S. and Parsons V. Nephron 1985 39 1. Wisniewski H. M. and Sturman J. A. in Aluminium and Health- a critical review ed. Gitelman H. J. Marcel Dekker New York Zief M. and Mitchell J. W. in Contamination Control in Trace Element Analysis Chemical analysis series Wiley New York 1976 vol. 47. Leung F. Y. and Henderson A. R. At. Spectrosc. 1983 4 1. Bertholf R. L. Brown S. Renoe B. W. Wills M. R. and Savory J. Clin. Chem. 1983 29 1087. Savory J. Brown S. Bertholf R. L. Mendoza N. and Wills M. R. in Methods of Enzymology 158 Academic Press. 1988 Mauras Y. and Allain P. Anal. Chem. 1985 57 1706. Schramel P. Wolf A. and Klose B. J. J . Clin. Chem. Biochem. 1980 18 591. Meggers W. F. Corliss C. H. and Scribier B. F. Tables of Spectral Line Intensities Part 2 National Bureau of Standards Monographs 145 2nd edn. US. Government Printing Office Washington 1975. Nygaard D. D. Leighty D. A. Appl. Spectrosc. 1985 39 363. Lichte F. E. and Hopper S. Anal. Chem. 1980 52 120. Geigy Scientific Tables ed. Lentner C. Ciba-Geigy Basle Switzerland 1981 vol. 3 p. 58. Powell J. J. Ph. D. Thesis University of London London 1994. Burden T. J. Powell J. J. and Thompson R. P. H. Anal. Proc. 1994 31 153. Rybarczyk J. P. Jester C. P. Yates D. A. and Koirtyohann S. R. Anal. Chem. 1982 54 2162. Hieftje G. M. and Wu M. Spectrochim. Acta Part B 1994 49 149. Chappuis P. Poupon J. and Rousselet F. Clin. Chim. Acta 1992 206 55. Brenner I. G. Geological Survey of Israel personal communi- cation 1993. 1989 pp. 125-165. pp. 289-301. Paper 4/05460A Received September 7 1994 Accepted November 30 1994 266 Journal of Analytical Atomic Spectrometry March 1995 Vol. 10

ISSN:0267-9477    

DOI:10.1039/JA9951000259

出版商:RSC    

年代:1995

数据来源: RSC

摘要:

Simultaneous Measurement of Isotope Ratios in Solids by Laser Ablation with a Twin Quadrupole Inductively Coupled Plasma Mass Spectrometer* LLOYD A. ALLEN HO-MING PANG ARNOLD R. WARREN AND R. S. HOUKT Ames Laboratory- U S . Department of Energy Department of Chemistry Iowa State University Ames IA 5001 1 USA A twin quadrupole inductively coupled plasma mass spectrometer is used with laser ablation sample introduction in both the steady state (10 Hz) and single pulse modes. The twin quadrupole instrument allows the measurement of ions at two m/z values simultaneously. A 10 Hz laser produces a noisy steady-state ion signal in both channels. The signals are highly correlated and the majority of flicker noise cancels when the ratio is calculated. Isotope ratios for copper are measured with an RSD of 0.26%.Transient signals for single laser pulses are also obtained. The integrated peak area of the transient signals is correlated even when flicker noise on the transient is present. Copper isotope ratio measurements for several laser pulses are measured with an RSD of 0.85%. For a single pulse the RSD is 1.6%. Keywords Laser ablation; inductively coupled plasma mass spectrometry; isotope ratio; solids analysis Since the original work of Gray in 1985,' laser ablation (LA) inductively coupled plasma mass spectrometry (ICP-MS) has been used for the direct analysis of a wide variety of solids including p o ~ d e r s ~ ' ~ glas~es,~*~ and single mineral grains.' This technique combines the attractive features of ICP-MS and LA. ICP-MS offers excellent detection limits and the ability to give isotopic information while LA requires little or no sample preparation does not require the sample to be conductive and can provide spatial resolution.However LA is an erratic process. Using a short pulsed laser (pulse width GlOns) the sample material is vaporized or ejected to the surrounding gas environment and a microplasma is formed above the ablation site." This process creates particles of various sizes which may or may not leave the ablation cell. The material leaving the cell is carried to the plasma via a transfer tube. In the transfer tube the particles can undergo several processes including deposition collisions and coalesc- ence into larger particles." These ablation and transport processes yield solid particles with a substantial size distri- bution which causes flicker noise in the ICP. At higher signal levels plasma flicker noise usually limits the precision obtain- able for isotope ratios or internal standard ratios in commercial quadrupole ICP-MS instruments due to the time lag required for peak hopping or scanning. For this reason LA-ICP-MS with conventional quadrupole devices is seldom if ever used for isotope ratio determinations.In principle simultaneous measurement of the various ions of interest should eliminate flicker noise from the resulting ratio. Ion ratios can be measured simultaneously using a magnetic sector equipped with multiple Faraday cup collectors.8~12~13 This instrument employs either solution neb~lization'~ or LA8 with RSD values better than 0.03% and 0.05% respectively.* Presented at the Seventh Biennial National Atomic Spectroscopy Symposium (BNASS) Hull UK 20-22 July 1994. t To whom correspondence should be addressed. Journal of Analytical Atomic Spectrometry Recent work with a double focusing device of Mattauch- Herzog ge~rnetry,'~ and with time-~f-flight'~.'~ and ion-trap analy~ers,'~~'~ also shows substantial promise for providing simultaneous or sequential detection on a very rapid time scale. Recently we described an ICP-MS instrument in which the ion beam from the ICP is split into two parts.lg Each part is then sent to its own quadrupole mass analyser and detector. Results for solution nebulization demonstrate that signals for two isotopes of the same element and internal standard analyte isotopes are highly correlated.Precision is improved signifi- cantly by taking the ratio of the individual signals. This paper presents results for laser ablation of a copper sample with the twin quadrupole ICP-MS. A repetition rate of 10 Hz at an energy level of 6 mJ pulse-1 is used to yield a steady state Cu' signal. Transient signals are obtained with a single laser pulse at both 11 and 20 mJ pulse-'. Significant noise on the transient signal is obtained at the lower laser energy. However the noise is highly correlated in either the steady state or transient case and precision can be improved significantly by taking the ratio of the two isotope signals. EXPERIMENTAL The ICP-MS system used in this work has been described previo~sly.~' Briefly the ion beam from a typical ICP is split into two parts by an ion beam splitter placed in the third vacuum stage. The splitter is essentially two electrostatic analysers back to back.Once split each portion of the beam is sent to its own quadrupole mass analyser and electron multiplier. The ICP-MS operating conditions are listed in Table 1. The voltages applied to the ion lenses for both the steady-state and single-pulse experiments were optimized to yield a maximum Cu' signal in both channels using 10Hz laser pulses. A Nd:YAG laser (Model NY 82-30 Continuum) was fre- quency doubled to yield a beam at 532nm with a maximum Q-switched energy of 400mJ pulse-'. The pulse width was 8ns. Decreasing the Q-switch delay attenuated the beam to approximately 6 mJ pulse-' in the 10 Hz experiment and 11 and 20 mJ pulse-' for the two transient experiments.As shown in Fig. 1 a quartz lens (f= 10 cm) was used to focus the laser beam onto a rotated copper sample. The beam was directed to the sample slightly off centre in order to yield fresh ablation sites during each ablation sequence. Laser energy was measured in real time with a 50 mm pyroelectric energy detector (Model Rm-3700) and universal radiometer (Model RjP-736) from Laser Precision (now Laser Probe). A copper plate was cut to fit the dimensions of the cell placed on a stepper motor (AMSI Corp. model 301SM) and rotated at 30 rpm. No other sample preparation was required. Copper was selected for two reasons (a) the isotope ratio is -2 1 and is therefore measured relatively easily and (b) the previous paperlg describing this device also concentrated on Cu isotope ratios.The ablation cell (Fig. 1) consisted of a glass pipe with an Journal of Analytical Atomic Spectrometry March 1995 Vol. 10 267Table 1 Instrumental components and operating conditions Component Operating conditions ICP Plasma Therm generator (now RF Plasma Products) Model HFP-2000D RF Plasma Products torchbox (modified in-house for horizontal operation with home-made copper shielding box) Ames Laboratory construction Ion extraction interface Vacuum system'' Three stages differentially pumped Welded stainless steel Ames Laboratory construction Mass Analysers From VG Plasma Quad Model SPX 300 with RF-only pre-filters; Model SXP 603 Controllers and RF Generators Galileo Model 4870 pulse counting mode Counting Electronics EG&G ORTEC Model 660 dual 5 kV bias supply Model 9302 amplifier/discriminator Model 994 dual counter/timer Electron Multiplier Forward power/kW 1.25 Reflected power/W 40 Aerosol gas flow/l min - 0.8 Outer gas flow/l rn-l 16 Intermediate gas flow/l min - ' 0.8 Sampler orifice/mm diam.1 Skimmer orifice/mm diam. 1 Sampling distance from coil on centre/mm 6 Sampler-skimmer separation/mm 12 Differential pumping orifice/mm diam. 1.5 Operating pressures/Torr Expansion chamber 1.2 Second (ion lens) chamber 6 x Third (quadrupole) chamber 6 x Mean Rod Bias/V 0 Bias voltage/V -3OOO MS interface )[Ll+irgy and ICP i:L:i N- Doubled \ Fig. 1 to plasma Schematic diagram of ablation system and transport tube i.d. of 2.54 cm and a volume of approximately 40 cm3.The beam passed into the cell through a quartz window and was focused to maximize the intensity of the microplasma above the sample surface. The cell was held onto an O-ring in the stepper motor. The outer case of the stepper motor was sealed with epoxy cement in order to maintain consistent gas flow dynamics through the cell. The ablated particles were trans- ported to the ICP with argon at a flow rate of 0.8 lmin-' (velocity of -41 cm s-') through Tygon tubing (1 m long 6.4 mm i.d.). RESULTS AND DISCUSSION Steady State Signals A mass spectrum for copper is shown in Fig. 2 with a 10 Hz laser repetition rate to yield a steady state signal. The irregular peak shapes are similar in both channels and are due mainly to noise in the ablation process.Under normal conditions of solution nebulization the peak shapes are much smoother as usually obtained with a typical quadrupole device. The back- ground level is -50 counts s-'. A plot of count rate uersus time is given in Fig. 3 for a dwell time of 0.5 s. The absolute stability af each signal is rather 5 1 1 " 60 62 64 66 68 70 m/z Fig. 2 Mass spectra from both mass analysers using 10 Hz laser ablation of the Cu sample. The spectrum from quadrupole 2 is offset to the right for clarity poor at 8.5% RSD. However the precision improves to 1.0% RSD when a ratio of the two signals is taken. Visually it can be seen that the two signals fluctuate together across the 12 s data acquisition period. The dotted lines in Fig. 3 represent regression lines for each ion signal.These lines show that the sensitivity for both 63Cu+ and 65Cu+ drifts upward with time in this experiment. The ratio for the points making up the regression line was calculated and a slight drift ( - 1.8%) in the ratio was observed. This drift however was substantially less than the drift in the individual regression lines which was - 20% demonstrating the ability of the instrument to compen- sate for some drift in the sample introduction process. Mermet and Ivaldi2' used ICP atomic emission spectroscopy with an echelle spectrometer and array detector to obtain two different signals for the same emission line of an element. Correlation of signals was demonstrated by plotting the nor- malized intensity of one signal versus that of the other.Fig. 4 shows the correlation plot for the ICP-MS data in Fig. 3. A high degree of correlation was obtained ( r = 0.9978) even 268 Journal of Analytical Atomic Spectrometry March 1995 Vol. 10- 12 v) v) 3 c 2 1 I I I 1 I I I Time/s 0 2 4 6 8 10 12 14 Fig. 3 Plot of count rate uersus time during steady state laser ablation (10 Hz 6 mJ per pulse) of the Cu sample; dwell time =0.5 s. The RSD of each signal is 8.5% while the RSD of the ratio is 1 .O%. The dashed lines are regression lines for the signals. See text for explanation 100 + 3 u L $ 90 h c. C 3 8 .- R U - m 80 z 70 80 90 I00 Normalized count rate P3Cu + ) Fig. 4 Correlation plot for copper isotopes; dwell time = 0.5 s and correlation coefficient is 0.9978 though the signals fluctuate over a large range (70-100 units).A correlation coefficient close to unity indicates that flicker noise cancels when the ratio is measured.” Table 2 shows the effect of increasing dwell time on precision. The measured precision improves by increasing the number of counts accumulated as would be expected from counting statistics up to a dwell time of 2.0 s. However the precision deteriorates for dwell times longer than 2.0s as seen in the previous paper,” for reasons that are unclear at this time. Table 2 also lists the precision predicted for the ratio based on Table 2 Effect of dwell time on precision of Cu isotope ratio Mean ratio* 63Cu + ?’Cu + Dwell time/s 0.5 1.82 1 .o 2.04 1.5 2.43 2.0 2.43 2.5 2.43 3.0 2.39 %RSD of 63Cu+ :65Cu+ ratio Measured Counting statistics 1.04 0.2 1 1.17 0.14 0.55 0.17 0.39 0.14 0.41 0.13 1.03 0.13 * Accepted natural abundance ratio = 2.24. counting statistics the calculation for which was previously shown.’’ Clearly all the measured values of precision listed in Table 2 are poorer than the counting statistics limit.Perhaps the beam position at the entrance to the splitter the relative transmission of the two mass analysers and/or the relative response of the detectors drifts on a time scale of several seconds or longer. Note that the measured isotope ratio (2.4 in most cases) does not exactly equal the actual isotope ratio for Cu (2.24) due to mass bias through the splitter and different responses from the quadrupoles and detectors. The measured ratio can be changed manually by adjusting a set of ion lens voltages prior to the splitter.Previous results using solution nebuliz- ation indicate that the measured ratio can be related to the actual ratio by the use of a calibration curve.” The ratio values shown in Table 2 were determined in the order shown over a 1 h time period. For the earlier measure- ments (dwell time=0.5 and 1.0 s) the measured ratio drifts from 1.82 to 2.04. The ratio then stabilized at -2.4 for the subsequent measurements. Such drift in the measured ratio is common with this device particularly with measurements in the early phases of a day’s experiments. Fig. 5 shows a plot of count rate uersus time for an optimum dwell time (2.0s). The absolute stability of the signals is somewhat better at 4.5%. The precision improves to 0.39% when the ratio of the two signals is calculated. The correlation plot (not shown) has a correlation coefficient of 0.9986. As was the case in the initial study,” the precision of the ratio can be further improved by averaging 5 consecutive ratios.The RSD of 5 such averaged ratios is O.26% which is only -60% worse than the best RSD of 0.15% obtained for Cu isotope ratios during solution nebu1i~ation.l~ Transient Signals Transient signals were obtained by firing individual laser pulses. The sample was rotated as before and the dwell time for the counters (0.1 s) was selected to give an adequate number of points to define the shape of the ablation pulse. The ablation crater from an individual shot had a central pit that was 25 pm diameter and zz 10 pm deep. This pit was surrounded by an irregular distorted region of several concentric ridges similar to those shown by Denoyer et aL2’ for a Q-switched Nd:YAG laser at 1064 nm.This edge region was x 140 pm diameter. A plot of signal uersus time for 7 such laser pulses with an average energy of * 11 mJ pulse-’ is shown in Fig. 6. Several 10 1 I I I 1 I I 0 10 20 30 40 50 Time/s Fig. 5 Plot of count rate uersus time during steady state laser ablation of the Cu sample dwell time= 2.0 s. The RSD of each signal is 4.5% while the RSD of the isotope ratio is 0.39% Journal of Analytical Atomic Spectrometry March 1995 Vol. 10 2692.5 I I v) L $ 1.5 m $ \ 1.0 !! C 2 0.5 0 4- 0 25 50 75 100 125 150 Time/s Fig.6 Plot of count rate oersus time for seven laser pulses average laser energy = 11 mJ per pulse dwell time =0.1 s.The RSD of the integrated area for each isotope is 4% while the RSD of the isotope ratio is 0.85%. The plot for quadrupole 2 has been offset upward by 50000 counts s - ' for clarity. The inset is an expansion of the right- most peak of the peaks especially peaks 3 5 and 7 show significant fluctuations in signal during the peak. However this fluctuation is still correlated in both channels as indicated clearly by the inset in Fig. 6. The correlation plot (not shown) corresponding to the areas from Fig. 6 yields a correlation coefficient of 0.9796. The isotope ratio was obtained by integrating over the entire peak in both channels. The RSDs of the individual signals and the ratio are 4% and 0.85% respectively. The counting statistics limit for the RSD of this ratio is 0.33%.A plot of signal versus time for a higher laser energy ( ~ 2 0 mJ pulse-') is shown in Fig. 7. The peaks in this plot have a smoother more uniform shape than was the case at lower laser energy. At 20 mJ pulse-' the Q-switch delay is more nearly optimum which yields better beam quality and reproducibility. The Cu' counts in Fig. 7 are higher than those in Fig. 6 but do not increase linearly with energy because the latter part of the laser pulse is absorbed by the microplasma formed above the sample surface." At higher pulse energies a greater percentage of the pulse energy is absorbed by the microplasma and does not reach the sample. The RSD of the individual peaks is 5% while the ratio yields an RSD of 0.85%. Thus reasonable precision for an isotope ratio can be obtained for either an irregular or smooth ablation pulse.The counting statistics limit in this case is 0.27%. A correlation plot for this experiment is shown in Fig. 8. Again there is significant fluctuation in the integrated peak areas but these fluctuations are correlated in the two signals measured. Several applications of laser ablation require removal of a i 3*0 t 0 75 150 225 300 Time/s Fig. 7 Plot of count rate versus time for 11 laser pulses average laser energy=20 mJ per pulse dwell time=0.1 s. The RSD of the integrated area for each isotope is 5% while the RSD of the isotope ratio is 0.85%. The plot for quadrupole 2 has been offset upward by 50000 counts s - * for clarity 80 85 90 95 100 Normalized count rate ( ' Y u + ) Fig.8 Correlation plot for copper isotopes using individual laser shots; dwell time =0.1 s and correlation coefficient is 0.9888 very small amount of material and/or a high degree of spatial resolution. In order to evaluate the precision in the isotope ratio for only one laser pulse the ratio for seven consecutive points across the top of each individual peak in Fig. 7 was calculated. A summary of the results for these measurements is given in Table 3. The average RSD for these ratios is 1.6%. These measurements demonstrate the ability of this LA-ICP-MS system to give an isotope ratio from a single laser pulse with adequate precision for some applications. To our knowledge this has not been previously accomplished with ICP-MS. In Table 3 the mean ratio ( FZ 1.3) has substantial bias.This is not surprising because the bias is strongly affected by the lens voltages used at the entrance to the beam splitter so different isotope ratios are obtained on different days depending on the particular lens voltages used. CONCLUSION Results indicate that using a twin quadrupole TCP-MS pre- cision can be improved by simultaneous measurement of ion signals. Effective cancellation of plasma flicker noise was observed for both 10 Hz and single pulse laser ablation. The signals for the isotopes measured were highly correlated. Preliminary results indicate that a modest RSD (1-2%) can be obtained on an individual laser pulse. The precision obtained in all experiments is still poorer than that predicted by counting statistics for reasons that are unclear at this time.Also ion transmission must be improved in order to make the instrument an effective tool for trace elemental analysis. Table3 Precision of the Cu isotope ratio for the seven top points across each peak in Fig. 7 Peak 1* 2 3 4 5 6 7 8 9 10 l l t Mean ratio 63Cu+ 65Cu+ 1.34 1.34 1.31 1.33 1.31 1.32 1.34 1.32 1.32 1.30 1.30 %RSD of 63Cu+ 6 5 C ~ i 1.57 1.59 2.63 1 .so 1.21 I .37 1.37 1.51 1.06 2.46 1.19 * First peak in Fig. 7. t Last peak in Fig. 7. 270 Journal of Analytical Atomic Spectrometry March 1995 Vol. 10Possible ways to improve transmission are currently being studied. Ames Laboratory is operated for the US Department of Energy by Iowa State University under Contract No. W-7405-ENG-82. This work was supported by the US Department of Energy Environmental Remediation and Waste Management Office of Technology Development.REFERENCES 1 2 3 4 5 6 7 8 9 Gray A. L. Analyst 1985 110 551. Mochizuki T. Sakashita A. Tsuji T. Iwata H. Ishibashi Y. and Gunji N. Anal. Sci. 1991 7 479. Watling R. J. Herbert H. K. Delev D. and Abell I. D. Spectrochim. Acta Part B 1994 49 205. Kogan V. V. Hinds M. W. and Ramendik G. I Spectrochim. Acta Part B 1994 49 333. Arrowsmith P. Anal. Chem. 1987 59 1437. Denoyer E. R. J. Anal. At. Spectrom. 1992 7 1187. Thompson M. Goulter J. E. and Sieper F. Analyst 1981 106 32. Walder A. J. Abell I. D. Platzner I. and Freedman P. A. Spectrochim. Acta Part B 1993 48 397. Chenery S. and Cook J. M. J. Anal. At. Spectrom. 1993 8 299. 10 11 12 13 14 15 16 17 18 19 20 21 Moenke-Blankenburg L. in Laser Micro Analysis ed. Winefordner J. D. and Kolthoff I. M. Wiley New York Arrowsmith P. and Hughes S. K. Appl. Spectrosc. 1988,7,1231. Walder A. J. and Freedman P. A. J. Anal. At. Spectrom. 1992 7 571. Walder A. J. Platzner I. and Freedman P. A. J. Anal. At. Spectrom. 1993 8 19. Cromwell E. F. and Caley C. E. 42nd Conference on Mass Spectrometry and Allied Topics Chicago IL June 1994 Poster No. WP 272. Myers D. P. and Hieftje G. Microchem. J. 1993 48 259. Myers D. P. Li G. Yang P. and Hieftje G. M. J. Amer. SOC. Mass Spectrom. 1994 5 1008. Barinaga C. J. and Koppenaal D. W. Rapid Commun. Mass Spectrom. 1994 8 71. Koppenaal D. W. Barinaga C. J. and Smith M. R. J. Anal. At. Spectrom. 1994 9 1053. Warren A. R. Allen L. A. Pang H. and Houk R. S. Appl. Spectrosc. 1994 48 1360. Mermet J. M. and Ivaldi J. C. J. Anal. At. Spectrom. 1993,8,795. Denoyer E. R. Fredeen K. J. and Hager J. W. Anal. Chem. l991,8,445A. 1989 pp. 43-48. Paper 4/05398B Received September 5 1994 Accepted November 17 1994 Journal of Analytical Atomic Spectrometry March 1995 Vol. 10 271

ISSN:0267-9477    

DOI:10.1039/JA9951000267

出版商:RSC    

年代:1995

数据来源: RSC