vaporizat ion interferences and the chemical composi t ion of desolvated and vaporizing solute part icles could be achieved th rough x-ray analysis techniques. Unfor tu- nately, these techniques were not available locally a t the t ime of this research.

ACKNOWLEDGMENTS This research was supported in part by the National Science Foun-

dation through Grants CHE 82-14121 and CHE 83-20053 and by the Office of Naval Research.

1. C. Th. J. Alkemade and R. Herrmann, Fundamentals o/Analyt- ical Flame Spectroscopy (John Wiley and Sons, New York, 1979), Chap. 4.

2. R. Kelly and P. J. Padley, Nature 216, 258 (1967). 3. I. Rubeska and B. Moldan, Anal. Chim. Acta 37, 421 (1967). 4. J. A. Holcombe, R. H. Eklund, and K. E. Grice, Anal. Chem. 50,

2097 (1978). 5. R. K. Skogerboe and S. J. Freeland, 3rd Annual FACSS Meeting,

Philadelphia, Pennsylvania, Nov. 1976, Federation of Analytical

Chemistry and Spectroscopic Societies, Pearl River, New York, Abstract 352.

6. R. K. Skogerboe, personal communication, 1982. 7. G. J. Bastiaans and G. M. Hieftje, Anal. Chem. 46, 901 (1974). 8. G. M. Hieftje and H. V. Malmstadt, Anal. Chem. 40, 1860 (1968). 9. N. C. Clampitt and G. M. Hieftje, Anal. Chem. 44, 1211 (1972).

10. N. C. Clampitt and G. M. Hieftje, Anal. Chem. 46, 382 (1974). 11. K. R. May, J. Sci. Instrum. 27, 128 (1950). 12. A. G. Childers and G. M. Hieftje, Appl. Spectrosc. 40, 688 (1986). 13. T. Hollander, Ph.D. Thesis, University of Utrecht, The Nether-

lands (1964). 14. C. B. Boss and G. M. Hieftje, Appl. Spectrosc. 32, 377 (1978). 15. R. C. Weast, Handbook of Chemistry and Physics (Chemical Rub-

ber Company, Ohio, 1976). 16. R. D. Cadle, Particle Size (Reinhold, New York, 1965), Chap. 2,

p. 91. 17. R. E. Russo, Ph.D. Thesis, Indiana University (1981). 18. D. A. McQuarrie, Statistical Mechanics (Harper and Row, New

York, 1979), Chap. 16. 19. S. P. Belyaev and L. M. Levin, Aerosol Science 5, 325 (1974). 20. M. B. Colket, L. Chiappetta, R. N. Guile, M. F. Zabielski, and D.

J. Seery, Combustion and Flame 44, 3 (1982).

Detection of Spectral Overlap Interference in ICP-AES with an Empirical Linewidth Ratio Technique

S. O. F A R W E L L * and C. T. K A G E L Department of Chemistry, University of Idaho, Moscow, Idaho 83843

ICP-AES line intensity data were fitted to a minimum-curvature smooth curve generated with the aid of a cubic semispline algorithm. These spline-fitted curves were used to obtain linewidth ratios which can pro- vide accurate indication of spectral overlap with the use of only sample and standard spectra.

Index Headings: Emission spectroscopy.

I N T R O D U C T I O N

Spectra l interferences place the major restr ict ion on cer tain appl icat ions of ICP-AES, par t icular ly in transi- t ion meta l matr ices, because of the richness of emi t ted spec t ra and line broadening processes in the plasma. 1,2 De te rmina t ion of t race meta ls in geological samples by ICP-AES presents spectral overlap problems of this type. For example , two s trong spectra l lines of gold are typi- cally used for analyt ical de terminat ions , 242.8 nm and 267.6 nm, bo th of which are subject to some degree of in terference by high levels of iron2 -5

In this work pre l iminary invest igations were directed toward character iz ing interference f rom a general sam- ple p repa ra t ion technique represent ing a high degree of spectra l interference. The ex ten t of interference on each emission line of in teres t was quant if ied with the use of convent ional overlap correct ion techniques. A new tech- nique based upon empir ical l inewidth rat ios was subse- quent ly developed and used to detec t potent ia l overlap

Received 27 September 1985; revision received 6 March 1986. * Author to whom correspondence should be sent.

interference. The predict ive capabi l i ty of this spectral overlap detect ion technique was then compared with the results f rom a convent ional correction method.

E X P E R I M E N T A L

M a t e r i a l . A l l glassware was prec leaned and condi- t ioned by overnight contac t with a 10% aqua regia so- lution and subsequen t rinsing with 18 M-ohm water. Commerc ia l ly available 1000 gg /mL meta l s tandards f rom Anderson Labora tor ies were used. The gold s tan- dard was verified against s t andard solutions p repa red bo th f rom J. T. Baker te t rachloroaur ic acid and 99.99 % pure gold foil f rom Alfa. Higher concentra t ions of iron in solution were p repared with the use of J. T. Baker reagent -grade n i t ra te or chloride salts.

M e t h o d s . Digestions. One assay- ton (29.17 g) of dried and ground samples (200 mesh) was weighed into 500- m L Florence flasks, 50 m L conc. HC1 were added, and the covered solution was boiled for 0.5 h. Then, 15 m L conc. HN03 were added, and digestion cont inued for 1 h. Next , 20 m L 6 N HC1 were added, and digestion con- t inued for ano ther hour. Solutions were subsequent ly cooled, fil tered th rough W h a t m a n #4 filter pape r into 100-mL volumetr ic flasks, and made to volume with deionized water.

S p e c t r o m e t e r Sys t em . Emission studies were per- fo rmed on an Applied Research Labora tor ies Model 35000 C ICP emission spec t romete r with a H a m m a m a t - su R955 p h o t o m u l t i p l i e r . T h e 1-m C z e r n y - T u r n e r monochroma to r has a resolut ion of 0.02 n m with fixed

944 Volume 40, Number 7, 1986 0003-7028/86/4007-094452.00/0 APPLIED SPECTROSCOPY ,c~ 1986 Society for Applied Spectroscopy

20-#m slits. A stepper motor drives the grating (1200 lines/mm) in 0.0056-nm steps to the desired nominal wavelength. Software peak-searching algorithms then scan a fixed number of steps (7 in calibration, 4 in anal- ysis) on either side of the nominal wavelength, integrat- ing the intensity at each step for a specified interval. Peaks are located when one point has two neighbors of decreasing intensity on either side and has a net inten- sity greater than 20% of background. If no "peak" is located according to these criteria, then the average of the two points around the nominal wavelength is re- ported as "peak" intensity. When peaks are located, the maximum point and its two nearest neighbors are used to compute a Gaussian profile, the maximum intensity of which is the reported peak value. A scan mode may be used to scan a specified window centered on a nom- inal wavelength value. In this mode, no peak searching is performed and net intensity values at each wavelength step are reported.

Data Analysis and Computer Simulations. Computer Facilities. The VM-CMS time-sharing facility of an IBM 4341 mainframe was used to analyze digitized spectra acquired on the slew-scan ICP system. A Waterloo BA- SIC Ver. 36 programming environment was used with a virtual machine size of 950K.

Corrections for Spectral Overlap. For routine spectral overlap corrections in ICP-AES, Boumans' technique 7 is often employed Total line intensity (Xv) at an analytical wavelength, X., is given by:

XT(xo) = X.(Xo) + XAx°) + ~ X~(X~) (1) i - 1

where XB is background intensity, Xa is intensity con- tributed by analyte, and X~ is intensity contributed by interferent "i" of "n" interferents. The calibration func- tion, in the absence of interferences, for an analyte to be determined with an analysis line X. can be expressed by:

XAxo) - xAx. ) C~ = (2)

MA(X,)

where CA is analyte concentration and MA is analyte sensitivity (intensity/concentration). Likewise, the cali- bration function of an interferent at its analytical wave- length X~ is:

XT(X3 -- X~(X~) C~ = (3)

M~(X,)

Note that in each case of calibration (analyte and inter- ferent), total line intensity is considered the sum of cal- ibrant and background intensities. The interelement calibration quantifying the effects of an interferent on an analyte line is generally expressed in terms of "ap- parent analyte concentration" (Cs) measured for rele- vant interferent levels. Interferent concentrations (C~) are measured at an analytical line of the interfering ele- ment and applied in the following correction equation:

CiMi(X,) (4) c~ MAx°) "

Contribution of multiple linear spectral interferences on an analytical line can be expressed by:

"' C~Mk(Xo) c~ = ,~ MAx°) (5)

i=1 i~1

If interelement effects are nonlinear, then the apparent analyte concentration contributed by an interferent will be given by:

C~ = ~ k~C/ (6) j = l

in which an appropriate interpolating polynomial is de-. fined to approximate the nonlinear effect. First-order correction (j = 1) and second-order correction (j = 2) are applied routinely, 5 but higher order polynomials are not as frequently used. Multiple nonlinear corrections result in the following formula for the total interferent contributions, C,j

i ~ l i= l i - 1

The background-corrected apparent concentration, CT, is derived from the net line intensity and the analyte calibration function:

X~.(Xo) - X~(Xo) Cr = = CA + Cs. (8)

MAx°)

For analyte/interferent pairs, critical concentration ra- tios (CCR) may be computed on the basis of intensity2 The concentration ratio at which the contributed inten- sities from interferent and analyte arE., equal is termed CCR~0. With the use of the formula for single linear interference and appropriate substitution, CCR~o results in an apparent analyte concentration equal to the true analyte concentration (100% error if uncorrected). CCRw is defined as the concentration ratio when XJXo = 0.1, and results in an error of 10% if uncorrected. Each in- tensity measurement results from peak intensity at a single wavelength location, with only one wavelength necessary per analyte or interferent. Once correction factors have been generated, measured analyte concen- trations can be corrected if interferents are quantified at other wavelengths and their contribution to apparent analyte concentration is subtracted.

Whereas this preceding technique provides good cor- rection for spectral overlap, it does not flag the overlap, i.e., except by measurement of intensity at an interfer- ent wavelength. Thus, the analyst will be unaware of an overlap unless the interference has been previously identified and measured. The line-shape-based algo- rithm employed in this paper can be used to detect the occurrence of spectral overlap and then the resultant predictions of overlap interference can be compared to the overlap as quantified by Boumans' CCR~0.

Line Shapes. Kawaguchi et al2 reported the calcula- tion of experimental profiles for ICP emission lines as Voigt profiles which result from the convolution of the actual line profile (assumed to be purely Gaussian) and the instrumental broadening (assumed to be primarily Lorentzian). 1° Posener 11 described Voigt profiles by the expression:

H(a ,v )=a- I '~ e - y 2 dy

7r ~,-oo a 2 + (v ----y2) (9)

APPLIED SPECTROSCOPY 945

Cr 205.552

>- l--

Z LU

Z

Ni 227.021 Zn 213.856 Au 267.595

"CCR~o is the concentration ratio where the contributed emission in- tensities from the interferent and analyte are equal.

b CR is the actual sample ratio of interferent to analyte.

where , is a reduced frequency dependent on the width of the Gaussian broadening and the distance from the line center, and the a-parameter expresses the Lo- rentzian/Gaussian character of the profile. For a tran- sition of known shape (i.e., a-parameter), H(~) describes the intensity in units of reduced frequency. Kawaguchi et a lp estimated that most ICP emission lines have a-parameter values from 0.2 to 0.7.

The difficulties of fitting data to this relatively com- plex line shape combined with the variability of line- width with concentration 12 suggest that empirical spec- tra of pure analytes in relevant concentration ranges be used to define the case for zero spectral overlap. This approach also allows the instrument-independent use of the overlap detection algorithm.

Overlap Detection. Digitized line intensity data from this investigation were fitted to a minimum curvature smooth curve that was generated through the use of a modified version of a published cubic semispline algo- rithm. 13 This procedure yields smooth curves based sole- ly on experimental data and does not require a line-

14

18

12 #

10

8

6

4

2

0

J

I

-10 0 10

STEP NUMBER FIG. 2. Interference effects of geological samples on Cr 205.552 nm. Cr (2.5 mg/L) = ' + ' ; Cu (1000 rag/L) = '# ' ; Fe (1000 mg/L) = '$'; and a typical sample digest (MA1) = '* '

shape model. Furthermore, the continuous curves gen- erated by this spline algorithm permit exact experimen- tal linewidths to be computed from these digital data via conventional baseline determinations and the defi- nition of full width at half maximum (FWHM). The spline-fitted curves were subsequently used to obtain FWHM values for standard and sample spectra. The resu l tant w i d t h rat ios (FWHMs,mple/FWHMst,nd,,d) were e m p l o y e d to s ignal the p r e s e n c e of p o t e n t i a l spec tra l over lap in ter ference in u n k n o w n samples . In t h o s e cases

0

15

12

g co z W

z 8

Ob

I

TABLE I. Spectral overlap interferences in typical geological samples.

Analyte Peak Sample/ wavelength Inter- separation analyte Width

(nm) ferent (urn) CCR~o ° CR b ratio

Ag 328.068 Fe -0 .042 5000 3000 1.1 M n +0.008 2500 200

Cu -0 .033 7000 80 1.2 Fe -0 .015 10,000 5000

M n - 0 . 0 3 9 20,000 50 1.0 Cu ±0.005 500 4 1.0 Fe +0.016 1500 5000 1.5 M n -0 .045 1500 250

- l g 0 lg

STEP NUMBER FIG. 1. Interference effects of geological samples on Ag 328.068 nm. Ag (2.5 mg/L) = '+ ' ; M n (1000 mg/L) = '$'; Fe (1000 mg/L) = '*'; and a typical digest (MA1) = '# ' .

946 Volume 40, Number 7, 1986

6

, i - - - , i

4 > - t--.

z LiJ F - z 2

I !

-10 0 10

STEP NUMBER FIG. 3. Interference effects of geological samples on Ni 227.021 nm. Ni (2.5 rag/L) = ' + ' ; M n (1000 mg/L) = '*'; A1 (1000 mg/L) = '$'; and a typical sample digest (SF1) = '# ' .

60

40

20

0

-IB

I I I I I

t r r r r r r r - - • " . . . . . . . . .

80

STEP NUMBER Fro. 4. In ter ference effects of geological samples on Zn 213.856 nm. Zn (100 mg/L) = ' + ' ; Cu (1000 mg/L) = '*'; and a typical sample (MA1) = ' # ' .

of no significant spectral interference, this ratio should approximately equal one if the standard and sample spectra are obtained at similar analyte concentrations. Ratios greater than one indicate line broadening in the sample, which serves as an indicator of significant spec- tral overlap from an adjoining interferent emission line.

RESULTS AND DISCUSSION

Six different analytical lines (i.e., Ag 328.068 nm, Cr 205.552 nm, Cu 324.754 nm, Ni 227.021 nm, Zn 213.856 nm, and Au 267.595 nm) that exhibit potential spectral

>-

LO Z F-- Z

B

0 - - I " - - - - I

-10 fl 10

STEP NUMBER FIG. 5. In ter ference effects of geological samples on Au 267.595 nm. Au (2.5 mg/L) = ' + ' ; sample MA1 = '*'; s ample SF1 = '$'; and sample SF4 = ' # ' .

>-

(.Y) z l..U l-- Z

>- I-- H

U3 Z LLI l--- Z

I 'I

-10 0 10

STEP NUMBER FIG. 6. In ter ference effects of base meta l s on An 267.595 nm. Au (2.5 mg/L) = ' + ' ; Mn (1000 mg/L) = '*'; and Fe (13,100 mg/L) = ' # ' .

overlap interference were studied. Pure analyte and in.- terferent solutions were analyzed to establish interfer- ence levels (CCR~0) for each interferent/analyte pair. Then several geological samples were analyzed for ana.- lytes and interferents via both the linewidth ratio tech- nique and Boumans' procedure. The resultant data from this investigation are summarized in Table I.

The degree of interference is indicated by comparing the CCR~0 value in Table I with the corresponding sam- ple/analyte concentration ratio (CR). As shown, the width ratio increases as the sample/analyte CR value approaches the CCRso value. The two emission lines for Ni and Zn in Table I do not show significant spectral interference, as evidenced by their low CR values relatiw) to their CCRso values. A similar comparison of CR with CCR~0 data for the Ag, Cr, and Au lines in Table I in- dicates an interference due to Fe. The width ratio of 1.5 for Au in Table I is the largest, as predicted from the fact that the corresponding CR value is much greater than the CCR~0 value. Width ratios of 1.1 for Ag and 1.2 for Cr in Table I are less than the Au width ratio. Con- sequently, one would predict a lesser degree of interfer- ence for Ag and Cr than for Au, but would still expect greater interference problems than when the width ratio is 1.0.

The ICP-AES spectral data, which were used to derive the information in Table I for each of the six analytical emission lines, are discussed in the following sections.

Ag 328.068 nm. Typical sample spectra shown in Fig. 1 indicated two neighboring background emissions bracketing this Ag emission line. Scans of interferents identified iron lines at 328.026 nm and 328.13 nm, con- stituting a mild spectral overlap in the former case. Scans of other potential interferents (Pb, Mn, Cu, Na, Mg, K, A1, and Ca) revealed a spectral overlap interference from Mn 328.076 nm, as noted in Table I.

APPLIED SPECTROSCOPY 947

Cr 205.552 nm. Typical sample spectra shown in Fig. 2 indicated a spectral overlap and a high spectral back- ground. Scans of interferents showed a nearby Cu(II) line (205.497 nm) which contributes a wing overlap, and a weak Fe(II) emission at 205.527 nm documented by Michaud and Mermet, 1 but not by Parsons et al.~4 The relative intensities of Cu and Fe emissions to that of Cr in the 0.1-nm window were such that the CCRso for both interferences (Table I) were well above expected sample ratios.

Cu 324.754 nm. Copper levels in the samples exam- ined were relatively high and this line showed excellent freedom from overlap. Off-peak background correction compensated for the small, level increases in baseline due to a weak nearby Fe emission.

Ni 227.021 n m . The relatively weak Ni signal in Fig. 3 is located on a sloping high background, between the wings of neighboring emissions. The wing overlap is at- tributed to Fe and A1 (Fe 227.086 nm and 226.910 nm and A1 226.922 nm) on the basis of interferent scans and comparison with the tables of Michaud and Mermet. 1 Mn was the only overlap observed in which the inter- ferent peak was within the 0.1-nm scan window. Because the observed scans for 1000 mg/L Fe and Cu showed direct overlap of low intensity (2-3 mg/L apparent Ni) and because such emissions are not reported by other workers, ~4,~ it was concluded that this observed "inter- ference" was due to nickel impurities in the Fe and Cu standard solutions.

Zn 213.856 nm. High concentrations of zinc relative to neighboring emissions were observed for all samples studied. As shown in Fig. 4, a direct overlap of Cu (213.851 nm) was observed. This interference has also been reported by other workers. TM While the effect was small (3 mg/L apparent Zn/1000 mg/L Cu) relative to typical Zn/Cu ratios in sample solutions, significant errors could result in trace Zn determinations for certain matrices (e.g., copper ores and certain plating baths).

Au 267.595 n m . Study of this analytical line provided considerable insight into the nature of background cor- rection in slew-scan ICP-AES. The typical sample scans in Fig. 5 showed broad, distorted peaks on a high spec- tral background. Spiking a diluted digest solution with 2.5 mg Au/L increased the intensity of emission at 267.595 nm, but determined the presence of a "shoul- der" at 267.611 nm due to a relatively weak Fe emission. The nature of the background emissions in the region of this line is more fully evident from Fig. 6. Both Mn and Fe produced interference effects, as listed in Table I, although the effect of Fe was more serious due to its relatively high concentration.

CONCLUSIONS

The detection of spectral interferences in complex samples analyzed by ICP-AES can be achieved through application of the described width ratio method. This

relatively simple width ratio method is an alternative to the more time-consuming, conventional procedure of comparing sample spectra with reference information from standard spectra and wavelength tables. Typically, the conventional procedure is performed on "represen- tative" samples as a part of method development and correction factors are determined for those interferences discovered during this developmental phase. Two dis- advantages are inherent in this conventional approach: (1) all those identified interferents must be accurately quantified in all samples, and (2) individual samples are not corrected for interferents not found in the "repre- sentative" samples.

The width ratio method on the other hand provides the analyst with a qualitative indication of potential spectral overlap on individual samples with the use of only the curve-fitting algorithm in combination with the standard and sample spectra that are routinely acquired during calibration and actual sample analysis. Thus, with only a minimal increase in analysis time, the analyst can use the width ratio method to assess the spectral purity of each elemental emission line that is employed for quantification. When used in conjunction with conven- tional overlap correction techniques, the reliability of the corresponding ICP-AES data from complex sample matrices will increase.

The major limitation of the width ratio technique is its inability in practice to detect direct spectral overlap, which is also not detected by visual comparison and must be deduced from emission data at alternate wavelengths. Conceptually, such techniques based upon linewidth comparisons should be able to detect most cases of di- rect overlap since the intrinsic widths will be different; however, an extremely high resolution spectrometer would be required.

1. E. Michaud and J. M. Mermet, Spectrochim. Acta 37B, 145 (1982). 2. J. Xu, H. Kawaguchi, and A. Mizuike, Appl. Spectrosc. 37, 123

(1983). 3. A. M. Harris, J. B. Lengton, and F. Farrell, Talanta 25, 257 (1978). 4. R. B. Wemyss and R. H. Scott, Anal. Chem. 50, 1694 (1978). 5. R. K. Brown in Precious Metals, Proc. Int. Prec. Met. Inst. Conf.

6th, M. I. EI-Guindy, Ed. (Pergamon Press, New York, 1983), pp. 393-408.

6. J. W. Graham, J. W. Welch, and K. I. McPhee, Waterloo BASIC Primer and Reference Manual (WATCOM Publications, Water- loo, Ontario, 1983).

7. P. W. J. M. Boumans, Spectrochim. Acta 31B, 147 (1976). 8. P. W. J. M. Boumans, Spectrochim. Acta 35B, 57 (1980). 9. H. Kawaguchi, Y. Oshio, and A. Mizuike, Spectrochim. Acta 37B,

809 (1982). 10. C. F. Bruce and P. Hannaford, Spectrochim. Acta 26B, 207 (1971). 11. D. W. Posener, Aust. J. Phys. 12, 184 (1959). 12. R.J. Lovett and M. L. Parsons, Spectrochim. Acta 35B, 615 (1980). 13. F. R. Ruckdeshel, BASIC Scientific Subroutines, Vol. II, BYTE

(McGraw-Hill, Peterborough, New Hampshire, 1980), pp. 305-310. 14. M. L. Parsons, A. Forster, and D. Anderson, An Atlas of Spectral

Interferences in ICP Spectroscopy (Plenum Press, New York, 1980).

15. P. W. J. M. Boumans, Line Coincidence Tables for ICP-AES (Ple- num Press, New York, 1980).

948 Volume 40, Number 7, 1986