Atomization interferences in ICP atomic absorption spectrometry† Carl E. Hensman‡ and Gary D. Rayson* Department of Chemistry and Biochemistry, Box 30001 MSC 3C, New Mexico State University, Las Cruces, NM 88003, USA Received 6th February 1999, Accepted 12th May 1999 ICP atomic absorption spectrometry (ICP-AAS) has been presented as a possible solution for the analysis of complex samples. Unfortunately, the performance of an earlier configuration was limited as to the elements that could be determined with favorable figures of merit.The present study was undertaken to ascertain those fundamental parameters responsible for these limitations. The central channel through the plasma using a torch with an enlarged sample introduction tube (6.25 mm id) was found to exhibit lower Ar and Fe excitation temperatures than predicted. The presence of water vapor was also determined to have a significant impact on analytical signals from within the central channel-viewing region.The production of atomic species for detection by absorption was found to be strongly influenced by both analyte molecular bond strengths and the first ionization potential. for inductively coupled plasma atomic absorption spec- Introduction trometry (ICP-AAS) measurements.23–27 However, the con- An inductively coupled plasma atomic absorption (ICP-AA) clusion of these early studies was that, although ICP-AAS is technique has recently been developed utilizing a novel optical feasible, the energetic plasma was generally more suitable configuration allowing spatial dispersion of the discharge for AES.image.1 The results have been very promising with figures of The ICP torch presented in this ICP-AA study has a sample merit at least rivaling those of conventional lateral viewing introduction channel of 6.25 mm id. As previously mentioned, ICP atomic emission spectrometry (ICP-AES). Unfortunately, this may exaggerate cooling eVects within the plasma discharge this level of performance was not observed for all metals.This resulting in poor response reported for some analytes. To lack of response has been proposed to be associated with a understand the possible eVects of this potential cooling probcooling of the plasma central channel, thus inhibiting the lem, physical characterization of this unique torch is required. formation of the free atoms.1 The possible cooling eVect may This paper describes the results of those investigations.also be exaggerated by the large sample introduction tube (6.25 mm id) utilized for the ICP-AA technique. This study adopts a mechanistic approach to investigate the reasons for the lack of response from certain analytes to the ICP-AA Experimental technique. Such an understanding of the possible mechanisms Calculations involved may then enhance the utility of this technique. Conventional ICP torches have been studied for many years. Atomic iron was selected as a thermometric species because it These studies have resulted in an understanding of many of demonstrated desirable characteristics.17 Although there is a the processes occurring within the discharge.Unfortunately, a significant diVerence in the number of lines used for such complete understanding of this dynamic source remains elus- studies in the literature, six lines of atomic iron were selected. ive.2–14 However, many of the physical characteristics of the It was found that the standard error encountered using these discharge (both axially and radially) have been meas- Fe emission lines for the temperature determinations became ured.2–6,15,16 The acquisition of spatial profiles of the plasma unacceptable in areas of high background emission, such as emission has enabled the assessment of optimum conditions the toroidal ring of the plasma.To allow the satisfactory for the plasma discharge as well as insight into the atomization, calculation of excitation temperature within these regions, excitation and inter-element interference mechan- argon was also chosen as a thermometric species.The paramisms. 8,13,14,17–21 Among several fundamental ICP properties eters of the corresponding lines for Fe and Ar used for these that have been studied is the spatially resolved excitation calculations are listed in Table 1. The excitation (spectroscopic) temperature of the plasma. temperatures were determined from a logarithmic form of the Historically, Wendt and Fassel first suggested the use of the Boltzman equation.29 ICP as an atom reservoir for AAS.22 Successful absorption measurements were accomplished in that study by using multipass optics perpendicular to the plasma discharge. Several Ln A Ilki ( gkAki)B=A Ek (kTEB (1) other instrumental configurations have since been investigated where I is the relative intensity of a specific line, lki is the †Presented at the 1998 Winter Conference on Plasma wavelength, gk is the statistical weight of the upper energy Spectrochemistry, Scottsdale, AZ, USA, January 5–10, 1998.level for that transition, Aki is the transition probability, k is ‡Present address: Department of Geological Sciences, 275 Mendenhall the Boltzman constant (0.694 cm-1), Ek is the energy of the Lab, 125 South Oval Mall, Ohio State University, Columbus, OH 43210, USA. excited state, and TE is the excitation temperature. A plot of J. Anal. At.Spectrom., 1999, 14, 1025–1031 1025Table 1 Parameters used for excitation temperature calculations, in study. The output of the hollow cathode lamp (HCL) is first conjunction with eqn. (1) collimated by a lens, L1 (focal length 150 mm). This collimated radiation is directed through the center of a modified ICP l/nm E/cm-1 gA19,28 torch with a 6.25 mm id sample introduction tube. The second lens, housed in a cooling mount, L2 (focal length 300 mm), Iron 371.99 26 875 1.793 373.49 33 695 4.430 then focuses the image of the hollow cathode onto the entrance 381.58 38 175 6.636 slits of a 0.3 m focal length monochromator (Varian, Techtron) 382.44 26 140 0.2210 with a bandpass of 0.5 nm.The plasma discharge was pos- 360.89 35 856 3.985 itioned at a distance from L2 much less than its focal length 358.21 34 844 13.39 (approximately 125 mm). This results in the dispersion of the Argon 425.1 116 660 0.0243 light from the plasma discharge, thus minimizing the amount 425.9 118 871 0.3540 of light from the ICP illuminating the photomultiplier tube 426.6 117 184 0.1070 (PMT) while maximizing the throughput of light from the 430.0 116 999 0.1675 HCL. 433.3 118 419 0.2515 Sample introduction into the ICP torch was achieved using a Meinhard-type concentric glass nebulizer with a concentric Scott-type spray chamber. The sample solution was introduced ln A Ilki ( gkAki)B against Ek allows the application of a best-fit at 1 mL min-1 using a peristaltic pump (Model Rabbit, Rainin Instrument Co.).straight line with a slope inversely proportional to TE. Absorbance measurements were calculated from four separate measurements. A blank sample was introduced into Materials the plasma and plasma emission (Ice) and plasma emission Stock solutions of each metal, 1000 mg L-1, were prepared plus lamp emission (Icl) were initially recorded. Each sample by dissolution of the reagent grade nitrate salt in distilled, was then introduced into the plasma and similar measurements de-ionized water with 1% trace pure nitric acid.All standard were made yielding the values with (Icls) and without (Ices) the solutions were prepared daily by serial dilution. HCL signal. Subtraction of the measurements Icl-Ice yielded the intensity of the incident radiation (I0) and Icls-Ices results Instrumentation in the calculation of the transmitted radiation intensity (I ). Applying the Beer–Lambert law, The ICP-AAS configuration has been discussed in detail elsewhere.1 Table 2 summarizes the parameters used for the Abs=-log AIcls-Ices Icl-Ice B (2) ICP.Fig. 1 depicts the instrumental configuration used in this the absorbance (Abs) by ground state analyte atoms was thus Table 2 Excitation temperature calculated using iron as a thermocalculated. metric species Emission measurements were conducted on the same Applied power/W Excitation temperature/K instrumental set-up as the absorption measurements with three modifications.The HCL and collimating lens (L1) were 500 3833 removed. An imaging lens and an adjustable iris replaced the 600 4019 focusing lens (L2) and cooling mount. Finally, the imaging 700 4176 lens (focal length 98 mm) was positioned 289 mm from the 800 4312 900 4432 top of the load coil and 149 mm from the monochromator 1000 4540 entrance slits. Magnification of the image on the entrance slit 1100 4637 was 151.94. Thus, the portion of the image passing through 1200 4726 the 100 mm entrance slit of the monochromator corresponded 1300 4807 to a region of the discharge with a width of 52 mm. All 1400 4883 emission measurements (except for those using argon as the 1500 4953 1600 5019 thermometric species) were background-corrected with the 1750 5111 introduction of a blank into the plasma.Results and discussion Viewing regions Three diVerent viewing methods of the plasma discharge were employed during the study for the calculation of excitation temperatures: radially resolved emission (temperature) measurements, integrated emission measurements, and integrated absorbance measurements.Fig. 2(a) shows a typical temperature profile of the plasma as a function of radial position. For these measurements, the torch was aligned concentrically along the optical axis and then moved in increments of 0.27 mm (significantly greater than the calculated spatial resolution of the system) across the optical axis for each set of operating conditions. Fig. 2(b) shows a typical intensity profile from integrating the central viewing region of the plasma as a function of the applied forward rf power. The Fig. 1 Schematic representation of optical configuration for solid vertical line in Fig. 2(a) indicates the viewing region of absorption measurements: HCL, hollow cathode lamp; L1, quartz the discharge. This method was also used for all analyte collimating lens (focal length 150 mm); L2, quartz camera lens (focal length 300 mm); FS, field stop; MC, monochromator.emission measurements. Fig. 2(c) shows a typical absorbance 1026 J. Anal. At. Spectrom., 1999, 14, 1025–1031Fig. 3 Calculated excitation temperatures as a function of radial position from the torch center. (a) Argon as the thermometric species. (b) Iron as the thermometric species. ($) 600 W, (,) 1000 W and Fig. 2 (a) Radial profile of argon excitation temperatures; 0 mm (&) 1400W applied rf power.indicates the center of the axially oriented torch. (b) Mg atomic emission (285.2 nm) as a function of Fe excitation temperature integrated throughout the region indicated by the solid line in (a). (c) Ag atomic absorbance (328.1 nm) as a function of Fe excitation metric species. Again, it can be seen that the measured temperature integrated throughout the region indicated by the dotted excitation temperatures with 600 and 1000 W applied power line in (a). exhibited very similar radial profiles.At 1400 W, an excitation temperature increase was observed much closer to the center profile also from the central viewing region of the plasma as of the torch than was indicated using Ar as the thermometric a function of applied power. Incident radiation from the species. If the measured temperatures within the emission HCL is passed through the area indicated by the dotted line viewing region (i.e., the center of the discharge) for the 600 in Fig. 2(a) and then focused onto the slits of the and 1400 W power conditions were integrated, a relative monochromator. excitation temperature increase of 978 K was calculated. It should be noted that the relative excitation temperature Temperature determinations increase between 600 and 1400 W for the fitted data (Table 2) was 864 K. The relationship between calculated excitation temperature A region of constant argon excitation temperatures as a and applied power using iron as the thermometric species was function of viewing position was observed in the center of the found to follow a curve-fit logarithmic relationship: plasma.This was approximately the same radius as the sample TE=1020ln (P)-2506.2 (3) introduction channel. This suggests that the sample introduction channel has created a cooler region within the plasma where P is the indicated applied rf power in Watts. It should discharge. However, the iron excitation temperature radial be emphasized that this is an empirical relationship and profiles indicated that this cooling eVect was limited.This facilitated the application of temperatures instead of applied limited cooling eVect hypothesis is also supported by the powers in subsequent discussions. The calculated excitation distribution of magnesium ion to atom emission ratios (Mg temperatures are presented in Table 2. II/I ) across the plasma discharge. Fig. 4 shows these Mg I, Mg II and Mg II/I distributions. Radial profiles As the applied power was increased, magnesium ion production within the plasma increased [Fig. 4(a)]. The mag- Fig. 3(a) shows the radial temperature profile of the discharge using argon as the thermometric species for three diVerent nesium ion intensity was observed to be highest near the inside skin of the toroidal ring and to decrease towards the center power levels (600, 1000 and 1400 W). The excitation temperature of the central portion of the discharge was found to be of the torch.As expected this decrease in Mg II was mirrored by an increase in Mg I [Fig. 4(b)]. At 600 W, the Mg I was similar in magnitude for all three powers. Unexpectedly, as the high-energy plasma of the toroidal ring is approached, the reasonably constant across the center portion of the discharge, only decreasing in the area of higher magnesium ion emission. response of 600 and 1000 W applied power continued to be similar. At 1400 W, a relative increase in the excitation tem- The application of 1000 Wpower showed a significant increase in Mg II intensity towards the center of the discharge while perature was observed closer to the plasma center.The radial profile at 1400 W also showed an overall increase in the maintaining the same Mg I intensity towards the toroidal ring inner skin as for 600 W. Applying 1400 W of power yielded excitation temperature in the region of the toroidal ring. Fig. 3(b) shows the radial temperature profile of the discharge relatively consistent atom emission intensity across the discharge center.This was similar to the atom emission intensities for each of these three power levels using iron as the thermo- J. Anal. At. Spectrom., 1999, 14, 1025–1031 1027sion intensity decreased significantly at about 3 mm from the plasma center. The Ba I and Fe I signals exhibited emission maxima at that same location. It is suggested, from the above discussion, that the sample introduction channel creates an environment of low energy at the center of the discharge.At the same time a high-energy environment is created at the inner skin of the toroidal ring. This results in a non-uniform energy environment across the central channel region, resulting in analyte-dependent regions of maximum excitation. An alternative explanation for variations in emission intensities across the torch would involve a variation in the number densities of species within this region.However, such a mass transportation explanation would predict similar behavior for diVerent analytes. This was not observed (Fig. 5). Even so, the impact of such number density variations should be considered in the interpretation of these data. Mercury emission Fig. 6(a) demonstrates relative emission intensities for Hg I using wet aerosol introduction. As the applied power is increased the emission becomes stronger and moves towards the center of the torch. The profile shape, although similar to Fe I and Ba I, is much closer to the inner skin of the toroidal ring. This suggests that a higher energy environment is required to excite mercury than with any of the previously discussed analytes.Introducing atomic mercury into the plasma as a dry aerosol removes the solvent eVects on the analyte.30–32 Clearly, Fig. 4 (a) Mg II (279.5 nm); (b) Mg I (285.2 nm); (c) Mg II/Mg I as the highest emission is seen to be at the center of the plasma a function of radial position from the torch center.($) 600 W, (,) torch [Fig. 6(b)]. Because the presence of water vapor would 1000 W and (&) 1400W applied rf power. not be predicted to impact the distribution of the independently formed Hg vapor within the discharge, this suggests that the solvent is suppressing the production of excited analyte atoms at the center at 1000 W. Combining the information from ion at the center of the torch. In other words, the solvent may be and atom emission results in the magnesium ion-to-atom ratio the primary source of the previously discussed reduced energy [Fig. 4(c)] the conditions of 600 and 1400 W applied power at the center of the plasma torch. However, this is not seen in had nearly non-variant profiles across the discharge center. An applied power of 1000 W resulted in a similar profile, as seen in Fig. 3(b). The center of the discharge had a similar ion and atom emission intensity as a lower power plasma. As the inner skin of the plasma was approached, ion and atom emission intensities increased until a value similar to that experienced with a higher power plasma was realized.Fig. 5 clearly indicates regions of emission across the plasma discharge for the radial profiles of Ag I, Cd I, Ba I, Fe I and Ar I. The Ag I and Cd I emission signals are at a maximum at the center of the channel. This indicates that a suYcient amount of energy is available for atom excitation. This emis- Fig. 6 Relative emission intensities of Hg I (253.7 nm) as a function Fig. 5 Normalized emission intensity for (,) Ag I (328.1 nm), (&) of radial position from the torch center. (a) Hg introduced as a wet aerosol. (b) Hg introduced as a dry aerosol. ($) 600 W, (,) 1000W Cd I (228.8 nm), ($) Ba I (553.5 nm), (2) Fe I (344.1 nm) and (+) Ar I (545.1 nm) as a function of radial position from the torch center. and (&) 1400W applied RF power. 1028 J. Anal. At. Spectrom., 1999, 14, 1025–1031Fig. 7 Calculated excitation temperatures as a function of radial position from the torch center, using Ar as the thermometric species.With ($) and without (,) a water aerosol at 1000W applied power. the radial excitation temperature profiles of the plasma under dry and wet vapor conditions (Fig. 7). The dry metal vapor data show an increase in excitation temperature as compared with the wet atomic vapor. It also has a significantly broader region of high excitation temperature towards the center of the plasma torch.If Hg I was to agree with the argon excitation temperature studies for the dry vapor conditions, an emission maximum should be seen at 2 mm from the torch center. This was not observed [Fig. 6(b)]. It can therefore be concluded Fig. 8 Power profiles of atomic Hg emission and absorbance that the solvent aVects the plasma by more than a simple (253.7 nm) with (,) and without ($) a water aerosol as a function reduction of energy or ‘cooling’ across the radius of the sample of integrated Fe excitation temperatures (based on applied power introduction channel.Because of the method of introduction levels). of iron into the plasma (i.e., as a solution aerosol ), it was not possible with this sample introduction system to use this Table 3 Values for the dissociation of the metal oxide (MO) bond thermometric species to measure plasma temperatures for the and first ionization potential (IP) of the metal with the previously reported1 limits of detection using ICP-AAS dry vapor condition.Use of argon as the thermometric species did not suVer from this limitation. However, power-consistent Dissociation energy Limit of measurements of temperatures in the central channel of the Element of MO/eV33 IP/eV33 IP/MO detection1,a plasma were not obtained [Fig. 3(a)]. This complex energy reduction was further experienced Ag 1.4 7.57 5.4 1.84 when viewing the emission and absorption of mercury with Al 5.0 5.98 1.2 1120 Be 4.6 9.32 2.0 respect to excitation temperature (applied power) increase Cd 3.8 6.11 1.6 2.19 (Fig. 8). Emission intensity for the wet aerosol plasma initially Co 3.7 7.86 2.1 4.51 increased with plasma temperature similarly to the dry metal Cr 4.2 6.76 1.6 14.5 vapor condition. At about 4200 K (700–800W), the emission Cu 4.9 7.72 1.6 2.47 intensity with the wet aerosol increased more significantly than Fe 4.0 7.87 2.0 14.4 for the dry aerosol plasma condition until a plateau region Mn 4.0 7.43 1.9 1.71 Ni 4.3 7.63 1.8 7.50 was exhibited by each at approximately 5100 K (1600– Pb 4.1 7.4 1.8 14.7 1750 W).Interestingly, while the dry aerosol plasma showed Tl <3.9 6.11 >1.6 67.7 a strong increase in absorbance as the excitation temperature Zn 4.0 9.39 2.4 1.88 was increased, the wet aerosol plasma remained constant. The Hg 2.29 10.44 4.6 discrepancy between the wet aerosol plasma emission and Nd 7.4 5.45 0.74 N.D.b absorbance is suggested to result from the sensitivity of U 7.7 6.1 0.79 N.D.W 7.98 7.2 0.90 N.D. mercury emission as compared with that of absorption Y 7 6.51 0.93 N.D. measurements in the plasma. aAll detection limits are indicated in units of ng mL-1. bN.D. indicates that this metal was not detected by atomic absorption in the ICP. Temperature profiles Previously, it has been indicated that certain analytes exhibited poorer limits of detection by the ICP-AAS technique.1 Table 3 with a wet aerosol as applied power was increased.There is an initial increase in absorbance followed by a decrease in lists both the first ionization potentials and metal oxide bond strengths for a set of representative analytes and their pre- absorbance resulting in a characteristic curve. The turning point of the curve has been observed to be analyte-dependent. viously reported limits of detection.1 In particular cases the metal’s oxide bond strength is greater than its first ionization This behavior is strongly indicative of an initial low-energy limited process, i.e., the conditions for atomization of the potential (further indicated by a IP/MO less than one).Interestingly, analytes with a IP/MO value of less than one analyte from its metal oxide become more favorable with increasing applied power. A high-energy limited process then correlate well with those analytes that exhibit poorer figures of merit by the ICP-AAS technique. Fig. 9 shows typical follows this, i.e., a reduction of atomic analyte as the applied power increases due to the formation of the corresponding absorbance and emission profiles for an analyte introduced J. Anal. At. Spectrom., 1999, 14, 1025–1031 1029Fig. 9 Typical atomic absorbance (,) and emission ($) profiles for Fig. 11 Normalized Cd ion (,, 214.4 nm) and atom ($, 228.8 nm) Ag (328.1 nm) as a function of integrated Fe excitation temperatures. absorbances as functions of integrated Fe excitation temperatures. ion.(It should be noted that although this discussion uses discussed assumptions, that as the applied power is increased analyte oxides as a starting component, the authors acknowl- there should also be a marked increase in ion absorption. edge that the analyte may be initially present as other molecular Fig. 11 demonstrates this increase in ion absorption (even species.) Those target analytes whose ionization potentials are though the instrument is not presently optimized for ion less than the metal oxide bond strengths (i.e., IP/MO is less absorption measurements).than one) would then be predicted to experience conditions not favorable for atomization as the applied power increases. At the point where there would be suYcient energy to ionize Summary the analyte, the metal oxide bond will not have been broken. Under applied power conditions where the metal can be An instrumental configuration for ICP-AAS has been eYciently atomized, there would then be more than suYcient described.This configuration has been noted to be unsatisfacenergy for the eYcient ionization of the analyte. In this tory for specific metals. The present study has addressed the manner, it is conceivable that there would never be a suYcient problem of these poorly responding metals, particularly in number density of analyte atoms present to yield a detectable relation to the unique sample introduction diameter of the absorption signal.[ This may account for the lack of a signal plasma torch (6.25 mm id). when viewing mercury in a wet aerosol, Fig. 7(b).] Previous Initial studies were made associating the applied power to eVorts1 to measure the atomic absorption signal for each of the spectroscopic excitation temperature for the plasma disseveral elements that are in this category (i.e., Nd, U, W and charge. These resulted in a logarithmic fit of the applied power Y) yielded no discernible signal, as would be predicted from to the temperature. Radial profile studies of the thermometric their respective IP/MO ratios of less than 1.0 (i.e., 0.74, 0.79, species (iron and argon) within the plasma discharge in 0.90 and 0.93, respectively).combination with radial studies of magnesium atom and ion Aluminum has an IP/MO ratio closer to one than most emission suggest a non-uniform energy environment within other detectable analytes (Table 3). It also demonstrates a the viewing region utilized for the atomic absorption studies.poor response to the ICP-AAS technique.1 As the applied This environment extends to an approximate 3 mm radius power was increased while monitoring aluminum, it can be from the torch/discharge center, with the center being of lowest seen (Fig. 10) that there is no significant signal response until energy and that energy increasing as the inner skin of the about 4700 K (1200 W). This response is supportive of the toroidal ring is approached.It was demonstrated that the lack above explanation of the poor sensitivity for aluminum. The of analyte atom number density in the absorption-viewing refractory nature of aluminum oxide means that a relatively region was predominant due to the analyte solvent. However, higher energy environment is required to dissociate the it is suggested that solvent is not solely responsible for the metal–oxygen bond. It also would be expected, under the ineYcient atomization; rather, the problem lies with a combination of eVects caused by the diameter of the sample introduction channel.These eVects may also be due to increased gas expansion in the plasma, ineYcient rf coupling and ineYciency of excitation species mass transport from the plasma. Further studies are required to characterize the contributions mentioned and will be the subject of future papers. It is suggested that atomic absorption measurements are limited by atom formation at lower applied powers due to the presence of molecular analyte species and ion formation limits the atom number density at higher applied powers.This would result in the inability to observe those metals whose molecular bond strengths are greater than the first ionization potential. The correlation between an ionization potential-to-metal oxide bond strength ratio less than one and the non-detectable metals was demonstrated. This was reinforced by the applied power profile of aluminum not showing a relatively reasonable signal for both emission and absorption until 4700 K (1200 W), probably due to the aluminum refractory species.Fig. 10 Al atomic absorbance (,) and emission ($, 308.2 nm) as a function of integrated Fe excitation temperatures. This is consistent with the need for the higher temperatures 1030 J. Anal. At. Spectrom., 1999, 14, 1025–103112 J. A. M. van der Mullen, A. C. A. P. van Lammeren, attainable with a N2O–C2H2 flame as compared with the more D.C. Schram, B. van der Sijde and H. J. W. Schenkelaars, common air–C2H2 flame-gas mixture. Spectrochim. Acta, Part B, 1987, 42, 1039. Finally, it is suggested that to increase the analyte response 13 S. Nowark, J. A. M. van der Mullen and D. C. Schram, either dry aerosol introduction should be used to increase Spectrochim. Acta, Part B, 1988, 43, 1235. atomization eYciency (i.e., analyte desolvation or electrother- 14 M. Huang, D. S. Hanselman, P.Yang and G. M. Hieftje, Spectrochim. Acta, Part B, 1992, 47, 765. mal vaporization), or ion absorption measurement should 15 N. N. Sesi, D. S. Hanselman, P. Galley, J. Horner, M. Huang and also be employed in conjunction with atomic absorption G. M. Hieftje, Spectrochim. Acta, Part B, 1997, 52, 83. measurements within the plasma. 16 J. M. de Regt, F. P. J. de Groote, J. A. M. van der Mullen and D. C. Schram, Spectrochim. Acta, Part B, 1996, 51, 1371. 17 D. J. Kalnicky, R.N. Kniseley and V. A. Fassel, Spectrochim. Acknowledgements Acta, Part B, 1975, 30, 511. 18 D. J. Kalnicky, V. A. Fassel and R. N. Kniseley, Appl. Spectrosc., The authors thank Sandia National Laboratories for the 1977, 31, 137. donation of the inductively coupled plasma. Also, financial 19 H. Kawaguchi, T. Ito and A. Mizuike, Spectrochim. Acta, Part B, support from the National Science Foundation (grant 1981, 36, 615. #CHE-9312219) and the New Mexico Water Resources 20 L. M.Faires, B. A. Palmer, R. Engleman and T. M. Niemczyk, Research Institute (grant USGS #14–08–0001-G2035) is grate- Spectrochim. Acta, Part B, 1984, 39, 819. 21 G. R. Kornblum and L. de Galan, Spectrochim. Acta, Part B, fully acknowledged. 1974, 29, 249. 22 R. H. Wendt and V. A. Fassel, Anal. Chem., 1966, 38, 337. 23 C. Veillon and M. Margoshes, Spectrochim. Acta, Part B, 1968, References 23, 503. 24 S. Greenfield, P. B. Smith, A. E. Breeze and N. M. D. Chilton, 1 C. E. Hensman, J. F. Mihalic and G. D. Rayson, Anal. Commun., Anal. Chim. Acta, 1968, 41, 385. 1997, 34, 355. 25 J. M. Mermet and C. Trassy, Appl. Spectrosc., 1984, 38, 876. 2 J. W. Olesik and K. R. Bradley, Spectrochim. Acta, Part B, 1987, 26 G. D. Rayson and D. Y. Shen, Spectrochim. Acta, Part B, 1991, 42, 377. 46, 1237. 3 P. B. Farnsworth, Spectrochim. Acta Rev., 1991, 14, 447. 27 M. A. Mignardi, B. T. Jones, B. W. Smith and J. D. Winefordner, 4 P. B. Farnsworth, D. A. Rodham and D. W. Ririe, Spectrochim. Anal. Chim. Acta, 1989, 227, 331. Acta, Part B, 1987, 42, 393. 28 I. Rief, V. A. Fassel and R. N. Kniseley, Spectrochim. Acta, Part 5 A. F. Parisi, G. D. Rayson, G. M. Hieftje and J. W. Olesik, B, 1973, 28, 105. Spectrochim. Acta, Part B, 1987, 42, 361. 29 S. Nakamura, J. Anal. At. Spectrom., 1995, 10, 467. 6 J. W. Olesik, Spectrochim. Acta, Part B, 1990, 45, 975. 30 I. Novotny, J. C. Farinas, W. Jia-liang, E. Poussel and 7 J. M. de Regt, F. P. J. de Groote, J. A. M. van der Mullen and J. M. Mermet, Spectrochim. Acta, Part B, 1996, 51, 1517. D. C. Schram, Spectrochim. Acta, Part B, 1996, 51, 1527. 31 J. W. Olesik, J. A. Kinzer and B. Harkleroad, Anal. Chem., 1994, 8 D. S. Hanselman, N. N. Sesi, M. Huang and G. M. Hieftje, 66, 2022. Spectrochim. Acta, Part B, 1994, 49, 495. 32 D. G. J. Weir and M. W. Blades, Spectrochim. Acta, Part B, 1994, 9 F. H. A. G. Fey, D. A. Benoy, M. E. H. van Dongen and 49, 1231. J. A. M. van der Mullen, Spectrochim. Acta, Part B, 1995, 50, 51. 33 Analytical Chemistry Handbook, ed. J. A. Dean, McGraw-Hill, 10 H. U. Eckert, Spectrochim. Acta, Part B, 1985, 40, 145. New York, 1995, pp. 7.13–7.14. 11 J. A. M. van der Mullen, I. J. M. Raaijmakers, A. C. A. P. van Lammeren, D. C. Schram and B. van der Sijde, Spectrochim. Acta, Part B, 1988, 43, 317. Paper 9/01071H J. Anal. At. Spectrom., 1999, 14, 1025–1031 1031