JOURNAL O F ANALYTICAL ATOMIC SPECTROMETRY JULY 1994 VOL. 9 759 Furnace Atomization Plasma Emission Spectrometry at Controlled Pressures Shoji lmai Department of Chemistry Joetsu University of Education Joetsu Niigata 943 Japan Ralph E. Sturgeon* and S. N. Willie National Research Council of Canada Institute for Environmental Chemistry Ottawa Ontario Canada KIA OR9 A furnace atomization plasma emission source was operated in a controlled pressure He environment (200-2000 Torr) (1 Torr=133.322 Pa). The response from several elements (Cd Pb Ag Mn Cu Fe and Co) as well as emission from He I at 388.8 nm and the temperature of the centre electrode (CE) were measured. Below 800Torr the temperature of the CE increases with increasing pressure as does emission from the analytes and He I.This can be attributed to increases in the gas density and collision frequency. Response reached an optimum slightly above atmospheric pressure and declined thereafter. At high pressures (> 1200 Torr) the emission from Cd Pb and Ag again increased probably as a result of an increase in the efficiency of deposition of these analytes onto a cooler CE (secondary site). The effect of pressure on detection limit and linear range of calibration for Cd and Mn was also examined. Keywords Furnace atomization plasma emission spectrometry; centre electrode; deposition efficiency; pressure effect Furnace atomization plasma emission spectrometry (FAPES) is a relatively new atomic emission technique for ultra-trace analysis based on a combined source that features an atmos- pheric pressure r.f.plasma as an excitation medium sustained within a graphite furnace acting as a vaporizer/atomizer.1'2 Furnace atomization non-thermal excitation spectrometry (FANES) operates at reduced pressures [ < 200 Torr ( 1 Torr = 133.322 Pa)]. The effect of this parameter on the excitation characteristics of the plasma have been in~estigated.~ With increasing pressure the mean free path of collision partners decreases resulting in more efficient heating and a loss of highly energetic particles. Several studies of the effect of pressure on response in electrothermal atomic absorption spectrometry (ETAAS) have also been Reduced pressure atomization decreases the mean residence time of the analyte thereby reducing sensitivity. Elevated pressure induces line broadening and lowers the diffusive loss rate.In FAPES atomic emission probably arises via collisions of analyte species with energetic particles such as electrons and metastable He atoms. The effect of pressure on response in FAPES would thus reflect its net effect on both plasma characteristics and on the transport of analyte species in the gas phase. Increasing pressure depletes a fraction of the highly energetic plasma collision partners and increases the frequency of energy exchange. However transport of gaseous analyte in the FAPES workhead is more complex than in ETAAS atomizers due to the presence of the centre electrode (CE). Double analyte emission peaks are frequently observed in FAPES.12-'5 Analyte initially deposited on the furnace wall (primary site) condenses on the CE (secondary site) following thermal desorption from the wall and subsequently re-atomizes from this radiationally- heated s~rface.'~*'~ The largest response is due to the latter process.With volatile elements such as Cd and Pb deposition on this secondary site is limited by its temperature which rises in the presence of the plasma during the pyrolysis stage.I5 This work was undertaken in an effort to study the effect of pressure on response in FAPES. Experiment a1 Apparatus All studies were conducted with a water-cooled integrated contact cuvette (ICC) pyrolytic graphite coated graphite * To whom correspondence should be addressed. NRCC No. 37558 furnace housed within a 10cm vacuum 6-way cross fitted with a feed-through for r.f.power as described previously.16 A coaxial pyrolytic graphite coated 1 rmn diameter graphite CE supported in a Ta holder was used to deliver power from a crystal controlled 13.56 MHz 1500 W r.f. Dionex generator (Model PM 112-1 500). Impedance matching was achieved with a manually adjusted Heathkit antenna tuner (Model SA-2060A Benton Harbor MI USA). The furnace was pow- ered by a Perkin-Elmer Model 2200 supply and fitted for maximum power heating via an optical feedback circuit. The chamber could be evacuated to 25 Torr pressure with a rotary pump and backfilled with high-purity He gas. The FAPES workhead was interfaced to a Spectrometrics Model SMI I11 echelle grating 0.75 m polychromator (Spectrometrics Andover MA USA). The transfer optics for this spectrometer have been described previ~usly.'~ A rectangular vertical slice of the plasma was always viewed with one vertical edge bounded by the edge of the CE the other by the tube wall and both horizontal edges bounded by the upper and lower tube walls.Owing to the radial symmetry of the source viewing of this image permitted a representative fraction of the source to be studied. This could be quickly verified by imaging the entire cross-section of the source onto the large photodiode detector element and observing identical effects of system parameters on measured intensity as when PMT detection of the sub-region was used. All optical components and the FAPES source were aligned with the aid of an 8 mW He-Ne laser. Analyte resonance lines were isolated with the use of appropriate hollow cathode lamps operated in d.c.mode. Atomic absorption transients were measured in the absence of the usual r.f. CE and support and also with the CE coaxially suspended within the ICC via a short piece of 1 mm i.d. stainless-steel tubing in an arrangement that permitted the hollow cathode light beam to pass through and illuminate the interior of the ICC. Photocurrents were fed to a current-to- voltage amplifier having a gain of lo9 digitized with 12-bit resolution and stored to disc using an IBM AT processor. All data manipulations were performed using in-house software written in Turbo Pascal version 4 (Borland International). Temperature measurements of the graphite surfaces were made with both an Ircon Series 1100 (Niles IL USA) auto- matic optical pyrometer and a Thermodot Model TD-6BH (Infrared Industries Santa Barbara CA USA) optical pyrom-760 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JULY 1994 VOL.9 eter. The latter permits focusing to a viewed region < 1 mm in diameter and was calibrated to 1300°C by focusing onto a hole drilled into a graphite block heated in a muffle furnace and the recorded temperature compared with the output from a thermocouple. The lower temperature limit was 120°C. Temperature measurements of the CE were obtained by sight- ing the Thermodot pyrometer through the sample dosing hole of the ICC furnace during the time when the furnace was not being heated. Those of the furnace wall during the atomization cycle were taken with the Ircon pyrometer (blackbody assumed) in the absence of the CE.The temperature of the furnace wall during the pyrolysis stage was measured using a chromel-alumel thermocouple. Pressure measurements in the chamber were made with one or other of two pressure transducers an Ashcroft Transducer Model ASHKlGlOOD7M0242Jl (Cole Parmer Chicago IL USA) and a Setra Systems Model 204 pressure transducer (Acton MA USA). The former was used at high pressure and the latter for low pressure conditions. Reagents High-purity He (Matheson Whitby Ontario Canada) was used as the plasma gas and for purging of the source. Stock solutions of all of the elements were prepared by dissolution of the high-purity metals (Cd Pb Ag Mn Cu Fe and Co) in sub-boiling distilled HN03. Working standards were prepared by dilution of the stocks with deionized distilled water acidified to 1% v/v with HN03.Procedures Both FAPES and AAS measurements were made according to the following procedures. Volumes of sample (5 pl) were pip- etted by hand onto the interior wall of the ICC furnace using an Eppendorf pipette fitted with polypropylene tips. The sample was then dried for 30 s at 80 "C under reduced pressure. Helium was admitted to the chamber to adjust the pressure to the desired level and all gas inlet and outlet valves were closed. The r.f. power was then applied and the plasma ignited spontaneously. Following a further 5 s plasma stabilization period the atomization stage was activated the signal recorded and the r.f. power turned off. Data acquisition commenced with a trigger pulse to the computer commensurate with the beginning of the atomization stage. The pyrolysis and atomiz- ation conditions for each element are summarized in Table 1 along with the wavelengths and heating rates used.Maximum power heating mode was used during atomization. All tempera- tures refer to the pre-set values as read from the front meter panel of the HGA-2200 power supply. The actual measured temperatures of the pyrolysis stage were 86 240 and 330°C for pre-set values of 80 250 and 400"C respectively. Measurements of He I emission at 388.8 nm and the tempera- ture of the CE were made using the same procedure. Table 1 Experimental conditions for analyte atomization Results and Discussion When pressure increases collision frequency as well as plasma gas density increase.This results in a depletion of the fraction of high-energy species due to their more efficient thermal equilibration. The intensity of the triplet He I line at 388.8 nm corresponding to the 3p3P-2s3Sl transition'* is presented in Fig. 1 as a function of pressure at various plasma forward powers. The ICC was maintained at a constant 330°C. As the power increases 4-fold from 20 to 80 W the He I emission intensity increases 60-fold at atmospheric pressure. When the total source intensity is integrated by focusing the image onto a large (5.8 x 5.8 mm) wide-band response (320-1 100 nm) pho- todiode detector (Model S1336-8BKY Hamamatsu Photonics Japan) the output was found to change 138-fold. In a low pressure r.f. plasma electron temperatures are generally inde- pendent of power whereas electron densities increase with power.lg Sturgeon et reported that the He excitation temperature increased marginally from 3000 to 3260 K as the forward power increased from 20 to 80 W in an atmospheric pressure FAPES source.Within the precision of the meas- ured excitation temperatures (i.e. k 180 K) the increase in output intensity could be accounted for by assuming the usual exponential relationship between the excitation temperature and energy i.e. the expected change in the relative emission intensity at 80 and 20 W follows from the estimated change in excitation temperature (260 K) coupled to the energy level involved (23 eV) in an exponential relationship. This is accompanied by a volume expansion of the plasma both radially and longitudinally within the ICC.At low power (< 40 W) emission intensity decreases continuously with increased pressure. This occurs as a result of decreasing excitation energy of collision electrons due to a decrease in mean free path coupled with a shrinkage in plasma volume which eventually leads to extinction of the plasma at pressures 0 400 800 1200 1600 2000 Pressu reflorr Fig. 1 Emission intensity of He I 388.8 nm as a function of pressure at a tube wall temperature of 330°C and plasma forward powers of A 20; B 30; C 40; D 50; E 60; F 70; and G 80 W Element Cd Pb Ag Mn cu Fe c o A/nm 228.8 283.3 328.1 279.5 324.8 248-3 242.5 Pyrolysis temperature*/oC 80 250 400 400 400 400 400 Atomization temperaturet/'C 1100 1 700 2000 2400 2400 2400 2400 Heating rate/"C s-l 1610 1600 1530 1530 1530 1530 1530 * Time.80 s for all elements studied; plasma ignition and stabilization occurs during final 5 s of pyrolysis stage. i Time 4 s for all elements studied.JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JULY 1994 VOL. 9 761 of 1000 and 1200 Torr for powers of 20 and 30 W respectively. At higher powers (> 50 W) intensity initially increases with pressure reaches a maximum and then declines. At 50 W a maximum occurs at 800 Torr and the emission declines rapidly above 1000 Torr. Several competitive factors contribute to these observations. Increased pressure increases He plasma gas (emitter) density but decreases the mean free path for electron excitation thereby reducing the high-energy tail of the electron energy distribution function.Additionally as noted above the plasma volume shrinks at high pressure. Even at 80 W power the plasma is eventually extinguished at pressures > 2000 Torr. Under such conditions electron energies become too small to sustain ionization. Alternatively this may be due to a decrease in the efficiency of coupling of r.f. power into the source. Fig. 2 displays the He I line intensity at 388.8 nm as a function of pressure for a 50 W input power (plasma continuously re-tuned as the pressure is changed). Imaging the source with a 10 ym widex25 pm high entrance slit or a 200x300 pm slit shows the same dramatic decrease in intensity with pressure as that observed when the entire source output intensity is integrated by focusing its image onto the large photodiode detector.Unfortunately this measurement cannot account for the losses of power in the source in the form of simple heating of the He gas and the ICC assembly. Although the CE can be seen to cool as the pressure is increased the net IR radiation emitted from the source may be increasing and since He is such an efficient conductor of heat especially at higher pressure the 50 W delivered to the FAPES source may simply be radiated away under conditions that are not energetic enough to support ionization. The temperature of the CE is an important factor in influencing the efficiency of condensation of gaseous molecular and atomic analyte species (such as those of Cd and Pb) onto this ~urface.'~ The CE is heated by both radiation from the wall and collision of excited species from the plasma.The negative bias potential which develops on the electrode in a free running system2' induces further heating by He ion bombardment. The temperature of the CE is presented in Fig. 3 as a function of the pressure at various ICC surface temperatures and a plasma power of 50 W. For any given pyrolysis temperature the electrode temperature initially increases as the pressure rises. Increased frequency of collision of He ions with the electrode surface at high gas density elevates its temperature. After the electrode temperature reaches a maximum in the 800-1000Torr region for a tube temperature of 86 "C 800 Torr for 240 "C and 1000-1200 Torr for 330°C it then declines. At higher tube temperatures the negative bias potential on the CE increases.21 This induces increased He ion bombardment and further heating which is / +4+ B \ 0 400 800 1200 1600 2000 Pressu reflo r r Fig.2 He I intensity at 388.8 nm measured using A 1 0 x 2 5 pm entrance slit; B 200 x 300 pm entrance slit; and C 5.8 x 5.8 mm photo- diode (320-1 100 nm spectral bandwidth) 700 1 1 600 i C I 300 ' I I 1 J 200 600 1000 1400 Pressureflorr Fig.3 Temperature characteristics of the centre electrode in response to changes in pressure for a 50 W plasma. Pyrolysis temperature A 86; B 240; and C 330 "C otherwise masked by the collapse of the plasma at the higher pressures. Fig. 4 illustrates some typical emission transients at various pressures. Signals shown for Fe are representative of those obtained for Mn Cu and Co. Consequently only Cd Pb Ag and Fe were selected for display purposes along with those from the He I 388.8 nm line.When the primary site for deposition of analyte is the tube wall double peaks frequently occur as a result of re-distribution of analyte from the wall to the cooler CE.I5 Because the temperature of the CE is low enough for condensation of molecular oxides of Cd and Pb to occur (cJ Fig. 3) double peaks (resolved and unresolved) are obtained for these elements. In the case of Ag and Fe transfer of atomic vapour from the wall to the CE with their subsequent desorption as this surface heats is responsible for the signal shapes observedi5 for elements such as these. Fig. 5 illustrates the effect of pressure on the peak height and area intensities for these elements.As pressure increases there is an initial increase in both intensities. Whereas one maximum occurs for Fe at 1000 Torr Mn at 1060 Torr Cu at 900Torr and Co at 1000Torr two maxima are obtained for Cd (at 880 and 1340 Torr) Pb (at 930 and 1350 Torr) and Ag (at 760 and 1550 Torr). Under the operating conditions used the temperature of the CE begins to decrease at 1200 Torr for Cd and Pb and at 1100 Torr for Ag (cJ Fig. 3). Thus the observed intensity increase may arise as a consequence of the more efficient deposition of analyte onto the CE as the pressure rises. It is likely that excitation mechanisms involving electron collisions and re-combination dominate in the plasma. The influence of metastable He atoms is probably negligible and their concentration would decrease with pressure as a conse- quence of collisional deactivation. Moreover energy transfer by metastable species exhibits a sharp resonance behaviour restricting the process to a few energy levels and hence elements.Predicting the effect of ambient pressure on response from the analyte is difficult. Measured intensity is proportional to the analyte number density and the frequency of collision with the excitation partner. As pressure increases analyte residence time increases owing to reduced diffusional loss. Concurrently the plasma volume can be seen to shrink thereby increasing r.f. power density (for a constant forward power of 50 W) and hence collision partner (electron) density. To a first approximation these factors lead to a quadratic depen- dence of the analyte emission intensity on applied pressure.However as pressure increases the mean free path decreases and there is a loss in the high-energy tail of the electron energy distribution. This may lead to a decrease in the excitation capabilities of the plasma but in the absence of a defined distribution and detailed rate information one can speculate that this heating process has no impact on the excitation of the relatively low energy levels associated with the transitions762 JOURNAL OF ANAL,YTICAL ATOMIC SPECTROMETRY JULY 1994 VOL. 9 (d) 0 0.5 1 .o 1.5 2.0 D . * 0 0.5 1 .o 1.5 2.0 0 0.5 'I .o 1.5 2.0 Ti rn e/s Fig. 4 Typical signals at various pressures with a 50 W plasma. (a) 1 ng Cd A 300; B 510; C 760; and D 1400 Torr. (b) 1 ng Pb A 300; B 540; C 930; and D 1400Torr.(c) 1 ng Ag A 300; B 570; C 760; and D 1340Torr. ( d ) 5 ng Fe A 300; B 500 C 760; and D 1400Torr. (e) He I A 300; B 760; and C 1400 Torr. Atomization conditions as specified in Table 1; He transient obtained using conditions for Fe t A 200 600 1000 1400 1 1 I I I 200 600 1000 1400 Pressu renorr Fig. 5 Effect of pressure on emission response with a 50 W plasma for (a) Cd 1 ng; (b) Pb 1 ng; (c) Ag 1 ng; and (d) Fe 5 ng. A peak height and B areaJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JULY 1994 VOL. 9 763 monitored for these elements. It is clear from Fig. 5 that no linear dependence of the intensity on pressure exists. A quad- ratic dependence on applied pressure is implied by the data presented in Fig. 6 wherein the response for each element has been normalized to the maximum obtained in the pressure range investigated.This may arise as a fortuitous trade-off of numerous factors which could lead to the proposed relation- ship. It is evident from the data in Fig. 4(e) that the intensity of the He I line at 388.8 nm generally decreases as the pressure increases. Since the ‘residence’ t h e of this specie3 is not altered by pressure changes (only emitter density) excitation con- ditions must be changing presumably via the change in the electron energy distribution function. Helium i s more sensitive to changes in this function than the other analytes. Above 800-1000 Torr the data presented in Fig. 6 begin to deviate from linearity. The slopes of the lines for Cd Pb and Ag are higher than that f6r Fe (as well as for Mn Co and Cu which are not displayed).The lower slope for Fe Mn Cu and Co may arise because there is a greater density of excited electronic states available for these elements over which the collisional excitation energy can be distributed. An extensive study of the parameters affecting the rate of loss of atomic vapour from the ICC operating in both the AAS and FAPES modes was undertaken. Rates were calculated from a plot of log(response) uersus time based on data from the decay side of the transient. Linear relationships were obtained and the loss rate data were reproducible within +_ 15% (relative standard deviation from several determi- nations). When the CE is present during atomic absorption measurements no plasma is present. Double peaks which arise due to analyte desorption from both the tube wall and the CE were also observed in AAS when the electrode was present.This has earlier been reported by Hettipathirana and Blades.12 In such case the late peak was used to estimate the loss rate. The following general observations could be drawn. At atmospheric pressure the presence of the CE in the AAS mode decreases the loss rate from the observation volume by 3-6-fold for the volatile elements Cd Pb and Ag and 1.5-2-fold for Mn Cu Fe and Co. The CE serves as a condensation site for atomic vapour and acts as a second surface source for re-desorption similar in operation to that of a L‘vov platform. As the ambient pressure increases the influence of the CE decreases because it is more efficiently heated by conduction and the platform effect diminishes.The rate of loss from the ICC is 2-3-fold greater in the FAPES mode with the plasma present than in the AAS mode with only the CE present. This is probably a consequence of the greater temperature of the diffusion medium (He) in the presence of the plasma as well as possible ionization of the analytes. As the pressure is increased this disparity is reduced because the plasma volume 1.20 W a 0.80 2 t; 0.60 + .- W 0.40 c g 0.20 I I . I 1 0 300 600 900 1200 1500 Pz/lo9 Torr2 Fig. 6 Normalized integrated emission intensity uersus the square of the pressure for A Cd 1 ng; B Pb 1 ng; C Ag 1 ng; and D Fe 5 ng shrinks and the gas cools. When no CE is present for the AAS measurements vapour loss rates for Cd Pb and Ag are less in the FAPES system because of the platform effect offered by the cool CE.For the less volatile elements loss rates in the FAPES system are larger than i s GAS without the CE because the temperature of the dihsion medium (He) is higher when a plasma is present. As the applied p u r e is increased the loss rate for volatile elcmtnts in $’APES tends to increase relative to AAS because t k tm~mfure of the CE increases and condensation efficiency dccmwes. Table2 summarizes data far the dative limit of detection (LOD) and linear range of calkation for Cd and Mn at various pressures. At reduced pressure a decrease in sensitivity and an increase in the LOD are reported for ETAAS systems and the upper limit of the calibration curve is extended.’T6 This occurs as a result of a decrease in the analyte number density coupled with inamwed diffusive f a s ~ mtt.At akvated pressure decreased sensitivity is also noted resulting from absorption line broadening due to the Lorentz effect. However the linear range of the calibration curve is extended 2-6-f0ld.**~ In reduced pressure FAPES there is a decrease in sensitivity and an increase in LOD but no extension to the upper limit of the calibration curve. Decreased sensitivity is likely to be due to decreased collision frequency (excitation rate) decreased analyte number density and redwed residence time. Although both of the latter parameters increase with pressure neither sensitivity and LOD nor linear range of calibration is im- proved. This arises as a result of a decrease in the excitation efficiency as a consequence of shorter mean free paths as well as enhanced self-absorption.Emission sources are typically characterized as having linear dynamic ranges of 4-6 decades. The FAPES source spans no more than 2-4 probably due to self-ab~orpti0n.l~ At high pressure the reduced efficiency of excitation leaves a greater population of ground state analyte atoms both in the ICC source as well as in a cloud between the source and the detector enhancing the self-absorption effect. At low pressure reduced excitation efficiency may occur at high analyte masses when the volume fraction of the analyte vapour approaches 1-2% of that of the He. At 940Torr it is evident that optimum analytical features are achieved reflecting the balance between the competitive factors of collision frequency analyte number density and excitation efficiency.Conclusion The complexity of effects due to pressure in FAPES occurs as a result of the influence of pressure on both the excitation characteristics and on the distribution of gaseous analyte Table 2 Normalized (to 760Torr) LOD and sensitivity and linear range for Cd and Mn Pressure/ Element Torr Cd 300 760 940 1160 1400 2000 Mn 300 760 940 1240 1400 2000 LOD*/pg Relative Log sensitivity? (linear range) 9-1 0 12 3.1 1-0 1 .o 3.7 1.6 1.1 3.9 1.1 0.83 3.9 2-5 082 3-8 1.0 0.99 3.3 1.4 1.1 1.9 1.0 1.0 2.4 0.84 2.2 2.2 0.9 1 1 *4 2.1 0.95 1.4 2.0 0.9 1 1.6 1.8 * Based on signal/s,=3. t Calculated from slope of calibration curve normalized to response at 760 Torr.764 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JULY 1994 VOL.9 species. An optimum pressure for analytical work may be unique for each element and is likely to be slightly higher than atmospheric pressure this being determined by the competition between increasing density longer residence time secondary site adsorption-desorption processes and decreasing excitation energy for collision. A more complete understanding of these observations will require Langmuir probe diagnostics of the plasma and two dimensional imaging of the source. The efficiency of deposition of analyte onto secondary sites is increased with increased pressure and the rate of re-desorption from such surfaces is limited by their decreased temperature at higher pressures. The effect of pressure on the interference by easily ionized elements and the influence of controlled bias potential on performance at various pressures warrant investigation.The authors thank G. Jolly Bell Northern Research Ottawa for the loan of the r.f. generator. S.I. thanks the NRCC for partial financial support while in Ottawa. References Liang D. C. and Blades M. W. Spectrochirn. Acta Part B 1989 44 1059. . Sturgeon R. E. Willie S. N. Luong V. T. Berman S. S. and Dunn J. G. J. Anal. At. Spectrom. 1989 4 669. Falk H. Hoffmann E. and Ludke Ch. Prog. Anal. Spectrosc. 1988 11 417. Donega H. M. and Burgess T. E. Anal. Chem. 1970,42 1521. Hassel D. C. Rettberg T. M. Fort F. A. and Holcombe J. A. Anal. Chem. 1988 60,2680. 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 Wang P. and Holcombe J. A. Spectrochim. Acta Part B 1992 47 1277. Sturgeon R. E. Chakrabarti C. L. and Bertels P. C. Spectrochim. Acta Part B 1977 49 1100. Sturgeon R. E. and Chakrabarti C. L. Anal. Chem. 1977 49 1100. Sturgeon R. E. and Chakrabarti C. L. Prog. Anal. At. Spectrosc. 1978 1 5. Fazakas J. Spectrochim. Acta Part B 1982 37 921 Fazakas J. and Zugravescu P. Gh. Appl. Spectrosc. 1988,42,521. Hettipathirana T. D. and Blades M. W. J. Anal. At. Spectrom. 1992 7 1039. Smith D. L. Liang D. C. and Blades M. W. Spectrochim. Acta Part B 1990 45,493. Sturgeon R. E. Willie S. N. Luong V. T. and Berman S. S. Anal. Chem. 1990 63 2370. Imai S. and Sturgeon R. E. J. Anal. At. Spectrom. 1994 9 493. Sturgeon R. E. Willie S. N. Luong V. T. and Dunn J. G. Appl. Spectrosc. 1991 45 1413. Berman S. S. and McLaren J. W. Appl. Spectrosc. 1978,32 372. Weise W. L. and Martin G. A. Wavelength and Transition Probabilities for Atoms and Atomic Ions US Department of Commerce NSRDS-NBS (US) No. 68 Washington D.C. 1980. Lin I. J. Appl. Phys. 1985 58 2981. Sturgeon R. E. Willie S. N. and Luong V. T. Spectrochim. Acta Part B 1991 46 1021. Sturgeon R. E. Luong V. T. Willie S. N. and Marcus R. K. Spectrochim. Acta Part B 1993 48 893. Paper 3/06496D Received November 1 1993 Accepted March 21 1994