The determination of thallium with the stabilized temperature platform furnace and Zeeman background correction

Spectrochrm~o

Act..

Vol. 438, Nos V-1 1, pp. 1157-l 165. 1988 c‘

Prmted m Great Bntam

058~8547/88 SO3 OO+.oO 1988 Pcr~amon Press pk

The determination of thallium with the stabilized temperature platform furnace and Zeeman background correction* D. C. MANNING and W. SLAVIN+ Perkin-Elmer Corporation, 901 Ethan Allen Hwy, Ridgefield, CT 06877, U.S.A. (Received 31 March 1988; in revisedform. 28 April 1988) Abstract-Palladium, about 5-1Opg as Pd (NO,),, is confirmed as an effective matrix modifier for TI. A pyrolysis temperature of 1000°C can be used, which is a much higher temperature than could be used with the previously proposed matrix modifier, H,SO,. The recovery of Tl was quite good in the presence of large amounts of NaCl but was inadequate in the presence of KCI. Abandoning the pyrolysis step and the matrix modifier proved to be convenient for determining Tl in both an NaCl and a KCI matrix. Thallium was fully recovered in these matrices in the presence of about 3Opg of either salt, or about 0.15%. assuming a 20~1 sample. INTRODUCTION

is widely determined in the graphite furnace in biological and environmental samples [l-7]. It has been a troublesome determination in the furnace because of a combination of properties. It is volatile and in many matrices it vaporizes below 500 “C. At the same time the vapor phase halides of Tl are very stable [8]. Thus when Tl and the halides are present together in the vapor phase, the Tl absorbance is reduced by molecular bonding to the halogen. FULLER [9] early observed that the presence of HCl or HClO,, even at concentrations less than 0.01 %, produced significant interference while 1% concentrations of HNO, or H,S04 had little effect. In fact, he found that the addition of H,SO, to chloride solutions reduced the interference. This, along with work by KUJIRAI [lo], led to the recommendation of H,SO, as a matrix modifier for Tl [ 111. Nevertheless, smaller quantities of chloride provided problems for Tl as compared to most other metals determined with STPF conditions. The stimulus for this project came from work reported by LETOURNEAU et al. [12] who determined five elements in eight matrices that simulated environmental waste waters. They used stabilized temperature platform furnace, STPF [ 133, methods and compared continuum and Zeeman background correction. For most analytes and matrices both systems gave reliable results. There were about 15 situations where continuum correction provided errors. In almost all of these situations Zeeman correction was error free. However there was interference for Tl in the presence of 0.1% mg/l NaCl or KCl, both with Zeeman and continuum correction. In this paper we duplicated their problems when their experimental conditions were used but conditions were found which avoided these errors. SCHLEMMER and WELZ [14] studied the use of a mixture of Pd and Mg nitrates as a universal matrix modifier. Their study, which included Tl, showed that this modifier permitted the use of higher pyrolysis temperatures than could be used with H,SO,. Prior to their work, SHAN et al. [ 151 found Pd to be a preferable matrix modifier for Tl. We studied the effect of the chlorides of Na and K on Tl when Pd was used as a modifier. The STPF conditions include atomization of the sample from a platform within the furnace tube, fast heating of the tube to provide a stable gas temperature when the analyte is vaporized and integration of the fast furnace signals. When this concept is used, the matrix, THALLIUM

*This paper is dedicated to Professor C. TH. J. ALKEMADE. One of the authors (W.S.) owes to him a great debt of gratitude for havmg suggested the concept of atomic absorption spectroscopy in his short note in J. Opt. Sot. Am. in 1955. Spectrochimica Acta, in the same year, carried the paper by Walsh which started many others in AAS. But Spectrochimm Acta was not widely read by chemical spectroscopists in the Umted States in those days. Professor Alkemade’s continued encouragement over the intervening years has been a source of pleasure as well as an important help to the work that we have done. +Author to whom correspondence should be addressed. SAL.3)43:9/11-L

1157

D. C. MANNING and W. SLAVIN

1158

up to some level, will have no influence upon the analytical result. Chloride matrices are typically troublesome because most metal chlorides are quite volatile, vaporizing at lower temperature than other salts of the metal. At the same time, the metal halide vapors are quite stable [8J. Therefore, in the presence of large amounts of chloride, it is difficult to convert all of the analyte to the atomic vapor, which is necessary for the STPF technique to work well. When the maximum matrix that can be tolerated for a particular analyte is lower than is analytically desirable, it stimulates a search among alternatives to extend this range. For chlorides, this search usually involves driving off as much of the chloride as possible prior to the vaporization of the analyte and preventing its recondensation in an environment where it will be revolatilized during the atomization step when the analyte element is in the vapor phase. There are several ways to do this but the use of matrix modifiers is the most frequently used. Instead of using an organized strategy to extend the STPF technique, most workers use some of the features and, if interference-free analyses do not immediately result, they resort to empirical tactics. The experimental work in this paper was completed about a year and a half ago but the manuscript was set aside in the hope that, with additiona work, still further quantities of chloride could be accommodated. In the meantime, WELZ et ai. [16] have successfully pushed the frontier still further. In addition, their paper discusses several recent references. We will discuss the Welz et al. work in the Discussion Section. EXPERIMENTAL We used the Perkin-Elmer Zeeman/5100 for most of this work, along with the AS-60 autosampler. The furnace program is shown in Table 1. All of this work was done with the 276.8-nm line of Tl and a spectral slit width of 0.7 nm. A Tl EDL was used at 7 W. It provided an energy signal on the Zeeman/5100 of 53 units. Typically, 20~1 samples were aliquoted onto the platform. The Zeeman/SlOO can be programmed to perform special sequences automatically. This operating mode is called “obey”. Several of the pyrolysis studies were done using this program. The ~ginning and final temperatures are specified, along with the incremental temperature steps (generally lO@‘C).The operator designates the sample cups to be used and whether a matrix modifier is to be added separately. When the sequence is begun, the system makes the requested number of firings automatically, sampling from the designated cups and increasing the pyrolysis temperature after each successive firing. The integrated absorbance data are collected and printed in a convenient format. The Zeeman/Sl~ has the capability to use the “max. power” operation for two temperatures, which in previous Perkin-Elmer systems was only available for a single temperature. This permitted the pyrolysis as well as the atomization temperature to be achieved rapidly. The minimum temperature that can be set with max power is about 950-1000°C. In the early part of this study we had difficulty delaying the analyte signal until the tube temperature was stable. Our previous experience {unpublished) had shown that this stemmed from the use of “grooved tubes” for the platforms. FALK and GLISMANN [17] had shown that electrical contact with the tube permitted current to flow through the platform during the atomization step thus providing active heating. To perform properly, the platform should be heated oniy by radiation from the tube wall. In order to obtain canditions as close as possible to theoretical, platforms were used in ungrooved pyrolytically coated tubes in this study. Care was taken to center the platform along the length of the tube and to be sure the platform lay in the horizontal plane. These platforms are not as convenient to use as captured platforms in grooved tubes. The Zeeman/Sl~ and the AS-60 have the capability, not previously available on Perkin-Elmer instruments, of taking up material from several sample cups in succession, the modifier first, and dispensing all the material in one insertion in the furnace. This makes the system more rapid, Table 1. Furnace program with pyrolysis Ramp(s)

Hold(s)

140

1

1000

0 I 0 1

45 45

Temp ( “C)

20 1800

2650

15 5 15

Ar flow (mi/min) 300 300 300 0 (READ) 300

Determination

of thallium

1159

particularly during methods development. This capability was used for most of this work. This arrangement is compared to the conventional procedure where each solution is taken separately into the furnace. We found a 10% gain in analytical sensitivity using the new procedure. This was puzzling for some time but experiments finally confirmed that it resulted from the washout of residual sample by the subsequently dispensed material in the pipet. This should be remembered by anyone using this convenient new capability. For essentially all of the experiments reported below, the experimental protocol was arranged so that the data were collected in random order. For instance, the recovery studies alternated randomly the various concentrations of matrix compounds. The single exception to this were the pyrolysis studies performed by the “obey” program, where the instrument controls the experimental protocol. Each point collected was plotted in the figures below, none were averaged or omitted. The Pd nitrate matrix modifier was prepared from Pd powder, Johnson Matthey Chemicals, Ltd, Royston, Hertfordshire, England, supplied by Alfa Products, Ventron, Danvers, MA 01923. The stock solution was prepared at 0.3% Pd and further diluted as needed. The Pd powder was dissolved in 5 ml of cont. HNO, and 5 ~1 of cont. HCL on a hot plate at low heat. When the reaction was completed the solution was washed into a calibrated loo-ml volumetric flask, cooled and brought to the mark with deionized distilled water.

RESULTS We determined that Pd stabilized the Tl signal to about a pyrolysis temperature of 900-1000 “C, consistent with the findings of SHAN et al. [ 151 and SCHLEMMERand WELZ [ 141.

We then sought the smallest amount of Pd that would permit this pyrolysis temperature in the presence of a large amount (30 pg) of NaCl. The resulting experiment is shown in Fig. 1. Five to 1Opg of Pd was shown to provide the desired performance. The maximum signal from that experiment provided a characteristic mass of 21 pg, in adequate agreement with the expected value of 17 pg/0.0044A.s [13]. A pyrolysis curve was prepared for the combination of Tl, NaCl(40 pg), and Pd (6 pg). The result, in Fig. 2, showed that the combination could be pyrolyzed to 1050°C. Fig. 3, then, showed the amount of NaCl that the conditions of Table 1 (pyrolyzing at 1000°C in the presence of 6 pg of Pd) would accommodate. The 150 pg of NaCl that can be tolerated was equivalent to 0.75% NaCl, using 20~1 samples. Note that, in Fig. 3, the absence of NaCl provided the same signal for Tl found when increasing amounts of NaCl up to 150 pg were present. Our preliminary exploratory experiments had shown that the presence of Mg(NO,),, suggested by SCHLEMMERand WELZ [14], had provided a slightly smaller Tl signal, which is why we dropped the Mg salt in the matrix modifier. Figure 4 shows that the Tl signal is significantly different when increasing amounts of Mg are added in the presence or absence of NaCI. Using the conditions from Fig. 3, we were not able to recdver the Tl signal in the presence of quite small amounts of KCl, as shown in Fig. 5. The minimum at 5 pg was difficult to

Fig. 1. The amount of Pd necessary as a matrix modifier. The sample was 1 ng of TI in 3Opg of NaCI. The pyrolysis temperature was 900°C and the atomization temperature was 1800°C. The calculated characteristic mass was 21 pg/O.O044 A’s for S-10 pg of Pd.

D. C. MANNING and W. SLAVIN 0.24 -

Fig. 2. Pyrolysis

with 4Opg NaCl

curve for

0 24

Fig. 3. Recovery

I ng of Tl in 40 ng of NaCl and 6 peg of Pd

as Pd(NO&

r

of the signal for 1 ng of Tl in 6 fig of Pd and various amounts temperature was 1000°C.

of NaCI. The pyrolysis

Frg. 4. Recovery of the srgnal for 1 ng of TI m 6 p’g of Pd and various amounts of Mg(NO,),. upper curve no NaCl is present and, in the lower curve, 30 pg of NaCl is present

In the

explain so we performed a pyrolysis study in the presence of 2 pg of KCl, an amount close to the minimum from Fig. 5. The result was the even more surprising curve shown in Fig. 6. More Tl was lost in pyrolysis at 800°C than at a higher temperature. This situation was unusual but it had been observed previously for Se by CARNRICK et al. [18] who showed a very similar pyrolysis curve (their Fig. 9) for Se in the presence of 0.1% H,SO,. WELZ et al. [ 191 in their Fig. 1 showed that Se(IV) displayed the same pyrolysis anomaly. The problem for Tl reported in the LETOURNEAU et al. [ 121 paper, who could not recover Tl in the presence of 2Opg of KCl, was quantitatively consistent with the results in Figs 5 and 6. Thus the failure, up to this point in the experiment, of the present technology to provide Tl measurements free of interference from KC1 resulted from a pyrolysis loss as the sample was

Determination

oh Fig. 5. Recovery

010 ;

A pg

of thallium

n

lb

KC;

1161

;o

;o

of the signal for 1 ng of Tl in 6pg of Pd and various pyrolysis temperature of l@OO”C.

Id0 amounts

of KC1 using a

with 6pg Pd and 2pg KCI

Fig. 6. Pyrolysiscurve for 1 ng of T1 and 6pg of Pd in 2pg of KCI.

carried through the temperature range from 500 to about 1000 “C. The same effect, though of smaller degree, was found for NaCl and Mg(NO,),, consistent with the LETOURNEAUet al. [ 123 results and those in Fig. 4. The recovery of Tl in the presence of both KC1 and NaCl was decreased both when the pyrolysis ramp was slowed and when the pyrolysis time was lengthened. Since the Zeeman/5100 can use max power heating during the pyrolysis step, this should reduce the time to a minimum when the system is at the intermediate temperature where losses occur. This strategy worked and the interferences were distinctly smaller when the max power feature was used for the pyrolysis step. This arrangement also reduced the length of time necessary for the pyrolysis step. There was still residual interference, especially for KCl. A typical pyrolysis study for these experiments is shown in Table 2, an example of the “Obey” program which automatically performed the study using 1 ng of Tl, 6 pg of Pd and 20 pg of KCl. This is to be compared with the data in Fig. 6 where only 2 pg of KC1 was used. Finally the experiment was tried which, in hindsight, seemed obvious. An important advantage of Zeeman background correction is that it tolerates much larger backgrounds without producing erroneous data. In this experiment Tl was being lost at temperatures below those that drove off the matrix. So the experiment was performed without a pyrolysis step and Tl was successfully recovered. Since the purpose of the matrix modifier was to make it possible to use a higher pyrolysis temperature, the modifier might not be necessary if the sample was not pyrolyzed. In fact the Tl signal was fully recovered in the presence of both KC1 and NaCl without a pyrolysis step and without matrix modifier. We did use 10% HNO, as the diluent for the salt and this had the practical effect of reducing the size of the background. Under these (no pyrolysis) conditions, the background absorbance peaks for the high concentrations (500 pg) of KC1 and NaCl were less than 0.4 absorbance units, which was easily handled by the Zeeman background correction system. Figure 7 shows the

1162

D. C MANNINGand W. SLAVIN Table 2. Pyrolysis study Temp.

Atomic signal (A. s)

400 500 600 700 800 900 1000 1100 1200 1300

0 18

0.13 0.12 0.11 0.10 0.10 0.10 0.10 0.08 0.01

Background signal (A. s) 064 0.60 0.57 0.55 0.39 0.06 0.04 0.04 0.03 -0.01

Fig. 7. Recovery of the stgnal for 1ng of Tl m 10% HNO, and vartous amounts of NaCl with no pyrolysis and no matrix modifier and an atomization temperature of 1800°C. The platform was used with ungrooved tubes.

recovery of Tl in the presence of NaCl using no pyrolysis step and no matrix modifier. The KC1 data were virtually identical to the NaCl data of Fig. 7. The profiles found for 50 pg of KC1 and no KC1 are compared in Fig. 8. The Tl profile changed considerably, but the integrated absorbance signal changed less than 4%. The data in Fig. 7 can be compared with that in Fig. 3. Characteristic mass For any study using STPF conditions it is important to consider and report the characteristic mass, m,,, that is found. This provides an indication of the quality control of the determination and it permits the operators to know that their equipment and technique is working as it should. The several figures of this report provide a Zeeman signal between 0.19 and 0.25 A*s for 1 ng of Tl. The average of these values converted to characteristic mass yields an m, of 19.5 pg/O.O044 As. We used the NBS SRM 1643b, Trace Elements in Water, to test the sensitivity. Tl is present at 8.0 + 0.2 pg/l which, with the 40 ~1 sample that was used, provided 320 pg of Tl. We used the conditions of Table 1 but without the pyrolysis step and without a matrix modifier and obtained an m, of 21.0 pg, in adequate agreement with our standards. The sensitivity loss by Zeeman AAS was calculated from several determinations with standards only and no matrix modifier and no pyrolysis. The sensitivity loss is the difference between the integrated signal for simple AAS and Zeeman AAS divided by the AAS signal and expressed as a percent. The value was very stable at 35%. FERNANDEZ et al. [20] reported 34% for a similar experimental configuration. Table 3 lists m. values from a number of other papers. Zeeman data have been converted to non-Zeeman values using the 34% sensitivity loss from FERNANDEZ et al. [20]. The values range from 9 to almost 13 pg, about f 18% from the average of 11 pg. We did not investigate linearity or analytical range in this study because we had recently done this [22] for Tl. The peak A signals for Tl at this wavelength become asymptotic to an

Determination

1163

of thallium

Time (s) Fig. 8. Absorbance profiles for 1 ng of Tl in 10% HNO,. In the upper curve, without additional matrix, the peak absorbance was 0.262 A and the peak area was 0.193 As. In the lower curve with 5Opg of KCI (0.25% KC]) the peak absorbance was 0.2OOA and the integrated absorbance was 0.185 As. The background (Bg) and Tl Zeeman (Tl) profiles are labelled.

Table

Zeeman

Temperature

3. Characteristic

Calculated non-Zeeman

m,(pg)

(“C)

m,(pg)

19.5 17

1900 1400 1800 1400 1500 1600 1500 1800

12.9 11.2

16 16.3 15 19.

mass for Tl

Measured non-Zeeman m,(pg)

9.6 10.6 10.7 10. 9.0 12.5

Reference This study SLAVIN and CARNRICK [13] L’vov et al. [21] BARNETT et nl. [22] SLAVIN et al. [23] SCHLEMMER and Web [14] SLAVIN et al. [24] WELZ et al. [16]

absorbance of about 0.8 A after which they rollover. In contrast, the integrated absorbance signals continue to increase to a signal of about 2 A.s (20 ng) after which the curve flattens. This extended range of linearity is another advantage of using A.s signals.

CONCLUSIONSAND DISCUSSION This paper shows that there are situations where the pyrolysis step can be abandoned, and in many of these cases the matrix modifier can be abandoned also. This alternative will not work in every situation because backgrounds can be too large even for Zeeman correction to handle. The same alternative should be tried when matrix modifiers are troublesome, for instance when the contamination of the modifier is too large to be corrected by a blank measurement. There is, of course, nothing novel in avoiding the use of the pyrolysis step and the matrix modifier. In our laboratory we establish accurate characteristic masses by using a simple standard or the NBS SRM 1643b, Trace Metals in Water [25]. L’vov et al. [21] compared experimental m, values with values from theoretical calculations. They avoided altogether the pyrolysis step and a matrix modifier. It is routinely advantageous to compare the m, that is found without a pyrolysis step or matrix modifier when a method is developed that uses a

1164

D. C. MANNINGand W. SLAVIN

modifier and a pyrolysis step. Differences between the values found with and without a pyrolysis step will provide insight into potential weaknesses of the method. Many graphite furnace papers report interference studies similar to those we have shown here for the effect of NaCl and KC1 on Tl. These studies are often difficult to reproduce in detail. For example, SHAN et al. [15] had shown that 100,ug of NaCl had no effect on the absorbance of Tl in the presence of Pd using similar instrumentation but different conditions. Our own paper [23] on the effect of halides on several metals including Tl showed no effect of 200 pg of NaCl on Tl using STPF conditions including 1% H,SO, as a matrix modifier. Yet LETOURNEAU et al. [12] found a significant interference (more than 10%) from as little as 20 pg of NaCl. Indeed, we show in Fig. 3 no effect on Tl absorption from as much as 15Opg of NaCl using the pyrolysis step and the Pd matrix modifier. We were not always able to reproduce that result. Partly this results from the occasional failure to mix the matrix modifier and the sample fully when they are dispensed separately onto the platform. The interference occurs as a result of the trapping of some chloride either on the cool ends of the furnace or by binding on or intercalation within the graphite. Thus subtle changes in the experimental conditions or the gradual degradation of the tube can alter the freedom from interference. Increased analytical throughput is an important potential improvement for furnace instrumentation. There are many applications which, on a Zeeman corrected instrument, could be performed much more quickly by avoiding both the pyrolysis step and the matrix modifier. If a 5 or lo-p1 sample is used to reduce the drying time to about 10 s, the entire determination can be completed in less than 1 min. Another approach to fast furnace analyses is simultaneous multielement instrumentation. For such an approach to be effective, common analytical conditions must be found for all elements. The pyrolysis temperature is the most important non-common parameter for furnace analyses. If the pyrolysis step can be avoided it will increase the attractiveness of multielement furnace AAS. The work by WELZ et al. [16] sought to determine Tl near the furnace detection limit in seawater. Using a 10~1 sample of a salt solution containing about 4% NaCl meant that the conditions had to accommodate about 4OOpg of NaCl without interference. Welz et al. confirmed our experience that avoidance of the modifier and the pyrolysis step permitted full recovery of Tl in the presence of about 3Opg of NaCl. That procedure did not work at the much higher concentrations of NaCl present in their samples. They also confirmed that the use of the Pd + Mg matrix modifier and a pyrolysis step at about 1000 “C usually permitted recovery of Tl in 400 ,ng of NaCl; but not always. The source of the variability was not found. After considerable experimental effort directed towards trying to approach more closely the STPF conditions, they found that they could consistently recover Tl from 4OOpg of NaCl if they pyrolyzed the matrix modifer at 1000 “C prior to depositing the sample on the platform and adding about 5% H, in the argon stream within the furnace tube. Raising the temperature of the matrix modifier solution to 1000°C insured that the Pd was reduced to the metal prior to depositing the sample and, thus, was in a form that could more readily react with the Tl. The use of hydrogen in the argon stream for the determination of Tl had been suggested much earlier by L’vov et al. [8] to reduce the effect of vapor phase chlorides because the bond strength between hydrogen and chlorine is large, and the hydrogen reduces the amount of free chlorine that can react with the Tl in the vapor phase. While Welz et al. used Pd + Mg as the modifier, they reported that, for the Tl determination, virtually the same performance was found using Pd alone if the modifier was pyrolyzed at 1000°C prior to deposition of the sample. Thus, on a practical basis, Tl may now be determined in complex matrices just as well as other analytes for which problems have not been found. The most convenient way is to use no modifier, no pyrolysis step and simple aqueous standards. If the analyst knows that the salt content of the sample is greater than about 30 pg (20~1 of an 0.15% solution) or if a simple recovery experiment shows this to be the case, then the more time-consuming procedure using Pd, reduced at 1000°C on the platform, and an internal argon stream containing 5% H,, should be used.

Determination

of thalhum

1165

But, perhaps more importantly, this work with Tl will illustrate the importance of finding conditions which permit the STPF technique to work. We have no doubt that others will find simpler alternative procedures to protect the Tl determination from the effect of chloride, but this work indicates that proving the feasibility of STPF conditions stimulates the effort to find simpler and more rugged methods for furnace AAS. Acknowled~ement~The Obey program which saved us considerable thank GLEN CARNRICK for his help and suggestions.

time was written by RICHARD GIDDINGS. We

REFERENCES [l] [Z] [3] [4] [S] [6] [7] [8] [9] [lo] [ll] [12] [13] [14] 1151 [I61 [17] [18] 1193

F. Ah, H. Berndt and J. Messerschnndt, Artzl. Lab. 28, 243 (1982). J. F. Chapman and B. E. Leadbeatter, Anal. Left. 13, 439 (1980). M. Hoenig and R. de Borger, Spectrochim. Acta 38B, 873 (1980). H. A. Chandler and M. Scott, At. Spectrosc. 5, 230 (1984). C. D. Wall, Clin. Chim. Acta. 76, 259 (1977). Z. Grobenski, R. Lehmann, B. Radziuk and U. Viillkopf, At. Specrrosc. 7, 61 (1986). J. P. Riley and S. A. Siddiqui, Anal. Chim. Acta. 181, 117 (1986). B. V. L’vov, Spectrochim. Acta 338. 153 (1978). C. W. Fuller, Anal. Chim. Acta 81. 199 (1976). 0. Kujirai, T. Kobayashi and E. Sudo, Z. Anal. Chem. 297, 398 (1979). W. Slavin, G. R. Carnrick and D. C. Manning, Anal. Chim. Acta 138, 103 (1982). V. A. Letourneau, B. M. Joshi and L. C. Butler, At. Spectrosc. 8, 145 (1987). W. Slavin and G. R. Carnrick, At. Spectrosc. 6, 157 (1985). G. Schlemmer and B. Welz, Spectrochim. Acta 41B, 1157 (1986). Shan Xlao-quan, Ni Zhe-ming and Zhang Li, Talanta 3, 150 (1984). B. Welz. G. Schlemmer and J. R. Mudakavi, to be published. H. Falk and A. Ghsmann, Z. Anal. Chem. 323, 748 (1986). G. R. Carnick, D. C. Manning and W. Slavin, Analyst 108, 1297 (1983). B. Welz, G. Schlemmer and U. Vijllkopf, Spectrochim. Acta 39B, 501 (1984).

[20] 1211 [22] [23] [24] [25]

F. J. Fernandez, W. Bohler, M. M. Beaty and W. B. Barnett, At. Spectrosc. 2, 73 (1981). B. V. L’vov, V. G. Nikolaev, E. A. Norman, L. K. Polzik and M. Mojica, Spectrochim. Acta 41B, 1043 (1986). W. B. Barnett, W. Bohler, G. R. Carnrxk and W. Slavin, Spectrochim. Acta. 4OB, 1689 (1985). W. Slavm, G. R. Carnrick and D. C. Manning, Anal. Chem. 56, 163 (1984). W. Slavin, G. R. Carnrick and D. C. Manning, Anal. Chim. Acta 138, 103 (1982). W. Slavin. D. C. Manning and G. R. Carnrick, J. Anal. Atom. Spectrom. 3, 13 (1988).