Anal. Chem.

1980, 52,741-746

74 1

Background Correction in Atomic Absorption Utilizing the Zeeman Effect

F. J. Fernandez,* S. A. Myers, and Walter Slavin
The Perkin-Elmer Gorp., Norwalk, Conn. 06856

Zeeman effect atomic absorption was compared to conventional atomic absorption with respect to sensitivity, analytical range, and background correction performance. Employing a versatile Zeeman test Instrument, capable of both flame and furnace operation, sensitivities were compared for a total of 44 elements. Only four elements showed a loss of sensitivity exceeding a factor of two for the Zeeman-effect instrument. The Influence of magnetic field strength on sensitivity and linear range was studied. Side by side comparisons of background correction performance were carried out with several atomic absorption spectrophotometers using continuum source correction. This comparison involved the analysis of a variety of difficult sample matrices using the graphite furnace.

T h e widespread use of furnace atomization systems has greatly increased the requirements for accurate background correction in atomic absorption measurements. While background absorption is observed only in limited cases for flame analyses, the situation is reversed when using furnace atomizers. For many applications, detection limits are restricted by the ability to correct accurately for background. Virtually a l l manufacturers of atomic absorption instrumentation offer simultaneous background correction systems based on the continuum lamp system described by Koirtyohann and Pickett (1). However, background correction using continuum sources has limitations. These include: the need for several continuum sources to cover both the UV and visible wavelength regions, some degradation in signal-to-noise due to the additional light sources, possibility of inaccurate correction if t h e background is structured, limitations in the ability to match the intensities of the source and continuum lamps, and t h e need to have very good alignment between the optical beams of the source and continuum lamps. The combined effect of these limitations is that in analytical situations involving high background levels, errors can occur due to underor overcompensation for the background signal. T h e need for improved background correction capability has generated interest in applying the Zeeman effect to atomic absorption spectrometry. With a Zeeman-effect instrument, background correction is performed at or very close to the analyte wavelength without the need for auxiliary light sources. An additional benefit is that effective double-beam operation is achieved with a very simple optical system. In 1971 Hadeishi and McLaughlin (2) first reported the application of the Zeeman effect to the determination of mercury. Since then numerous workers have investigated the technique utilizing designs in which the magnetic field is applied directly to the light source (3-9) o r t o the sample vapor itself (10-14). T o date, however, most of the published results have concentrated on theoretical aspects (15) or upon the determination of a rather limited number of elements. Little data exist indicating how implementation of t h e Zeeman effect influences such factors as sensitivity and analytical range for many of the commonly determined elements. Even with respect to the most important virtue, providing more accurate background
0003-2700/80/0352-074 1$01.OO/O

correction, the published data are often inconclusive. This is in part due to the fact that, when specially constructed sources or furnaces are employed, it is very difficult to relate performance to existing commercial instrumentation. Our investigation had several major goals. First was to survey a large number of elements and wavelengths to study the effect of magnetic field strength on factors such as sensitivity and linearity. Twenty-eight elements were selected because of their importance for furnace atomic absorption and another 16 were selected as important with the flame. Second was to perform direct comparison tests of background correction performance vs. current commercial atomic absorption instruments utilizing continuum source correction. In designing our Zeeman test instrument, certain ground rules were defined based on these goals. Obtaining accuratp comparisons of background correction performance is virtually impossible if the comparison involves furnaces of' different design. Thus, it was highly desirable to utilize a standard commercial furnace for which experience exists. I t was also desirable t h a t the instrument have the capability for both flame and furnace operation.

EXPERIMENTAL
The optical configuration of the test instrument is shown schematically in Figure 1. The design utilizes a dc-operated electromagnet, positioned around the atomizer, in conjunction with a rotating polarizer to switch between T and u polarization states. This approach, which is similar to that reported by others (12,13),was chosen since it permitted the use of standard light sources and atomization systems. A cJarrell-Ash 0.5-m monochromator, a rotating Rochon polarizer, a Perkin-Elmer Model HGA-2100 graphite furnace, and a Varian Model V-4005 electromagnet were used. Polarizer6 constructed of both magnesium fluoride and calcite were used. As the polarizing prism rotates (50 Hz), the signal varies sinusoidally between a peak amplitude corresponding to T transmittance (incident plane of polarization perpendicular to the magnetic field) and a minimum amplitude corresponding to the smaller u transmittance (incident plane of polarization parallel to the magnetic field). Standard signal processing techniques were employed to create two dc reference voltages proportional to these x and u transmittances. The presence of a background signal attenuates these signals proportionally so that the ratio of the transmittances is unaffected. After log amplification,separate T and u absorbances are obtained, and electronic subtraction yields a background corrected x u signal. During the experiments, all three signals were simultaneously recorded. To achieve high magnetic fields, it was necessary to minimize the gap size of the electromagnet. The graphite contact rings of the furnace were extended 50 mm and contoured to match the shape of the magnet pole caps. With this modification, a maximum field strength of 18 kG was obtained using an 18-mm air gap. Extending the contact rings required no change to the graphite furnace tube or purge gas flows and had 110 measurable effect on performance. By changing from 25- to 100-mm diameter pole caps, a standard Perkin-Elmer burner system could be installed. With an air gap of 19 mm, a maximum field strength of 11.6 kG could be achieved. For operation with a N20/C2H2 flame, the pole caps were water cooled. Comparison results were obtained using several Perkin-Elmer atomic absorption spectrophotometers (Models 460,603, and 5000),a l l equipped with a Model HGA-2100
~

0 1980 American Chemical Society

742

ANALYTICAL CHEMISTRY, VOL. 52, NO. 4, APRIL 1980

-,
l 8 O O H r CHOPPER

'
L

MAGNET COIL

3
Figure 1. Optical schematic of the Zeeman test instrument

-~
?

MONOCHROMATOR

Table I. Comparison of Relative Absorbance, Zeeman AA vs. Standard AA Using the Furnace ele me n t Ag A1
A , nm

Zeeman classa
A

3 0.55
0.41

6 0.55 0.81 0.67 0.65 0.53 0.95 0.26 0.26 0.93 0.79 0.71 0.61 0.81 0.19 0.29 0.75
0.19

field strength, kG 9 12 0.32 0.95 0.59 0.49 0.61 0.98 0.51 0.35 0.94 0.90 0.85 0.46 0.65 0.42 0.30 0.74 0.13 0.62 0.62 0.31 0.87 0.83 0.7 3 0.90 0.7 6 0.72 0.47 0.93 0.95 0.76 0.81 0.96 0.47 0.72 0.87 0.18 0.92 0.44 0.33 0.46 0.98 0.76 0.37 0.94 0.91 0.73 0.36 0.49 0.58 0.27 0.68 0.33

15
0.11

18

As Au Ba Be Bi
Cd co Cr cu
Fe

In

Mn Mo Ni Pb Pd Pt Sb Se
Si

Sn Te Ti T1
V

Zn
a

328.1 309.3 396.2 193.7 242.8 553.6 234.9 223.1 222.8 228.8 240.7 357.9 359.3 324.8 327.4 248.3 303.9 325.6 279.5 313.3 232.0 283.3 217.0 247.6 265.9 217.6 196.0 251.6 286.3 224.6 214.3 365.4 276.8 318.4 213.8

A A A A
N

N A A N A A A A A A A A A A A
N N
N -b

0.35 0.54 0.30 0.81 0.04 0.10 0.50 0.36 0.24 0.69 0.87 0.41 0.18 0.42
0.10

0.82 0.31 0.21

0.31

0.07 0.70
0.21 0.11 0.18 1.00

0.28 0.24 0.61 0.23 0.32

0.11

A A N
N N

A A A A N
=

0.42 0.35 0.56 0.36 0.48 0.51 0.16 0.22 0.56 0.13 0.43 0.25 normal.

0.59 0.65 0.43 0.72 0.69 0.46 0.82 0.61 0.81 0.51 0.88 0.84 0.54 0.76
0.91

___

0.55 0.23 0.88 0.90 0.87 0.93 0.79 0.58 0.43 0.93 0.98 0.90 0.72
0.91

0.44 0.76 0.69 Line not classified.

0.37 0.68 0.94

0.99 0.95 0.36 0.94 0.89 0.70 0.30 0.35 0.57 0.19 0.62 0.48 0.71 0.48 0.16 0.82 0.97 0.99 0.94 0.77 0.44 0.38 0.93 0.98 0.97 0.64 0.86 0.32 0.64 0.97

0.98 0.33 0.94 0.85 0.70 0.28 0.26 0.50 0.09 0.54 0.50
--_

0.41 0.12 0.78

1.00 1.00

0.94 0.73 0.32 0.34 0.93 0.98 0.99 0.58 0.80 0.34
0.60

0.97

Zeeman pattern; A = anomalous, N

or HGA-2200 graphite furnace. Subsequent experiments have shown that the Btendue of the spectrophotometer may be related to the degree of upset of the backgroundcorrected signals. This was not investigated in the experiments reported here. The light sources employed were Perkin-Elmer Intensitron hollow cathode lamps or electrodeless discharge lamps (As, P, and Se) operated at the manufacturer's recommended current or power levels.

RESULTS AND DISCUSSION Sensitivity. The initial focus of this investigation was to study the effect of magnetic field strength on analytical sensitivity. Tables I and I1 summarize furnace and flame results obtained at various field strengths. A total of 44 elements was surveyed, using the primary resonance lines normally used

for atomic absorption (AA) analysis. The relative absorbance data were calculated by defining the absorbance obtained in the standard AA mode as unity. The standard AA absorbance was obtained as either P or 0 absorbance measured a t zero field. Several elements were determined both in the furnace and the flame. The relative absorbances measured were the same. In this paper we use the terms normal and anomalous Zeeman patterns as used by de Loos-Vollebregt and de Galan (15)and Brown (16). Koizumi and Yasuda (8) identified the pattern classifications for most of the lines used in this study. In general, it was found that transitions with similar Zeeman patterns yielded similar performance. For normal class Zeeman transitions, very little sensitivity loss should occur

ANALYTICAL CHEMISTRY, VOL. 52, NO. 4, APRIL 1980

743

Table 11. Comparison of Relative Absorbance, Zeeman AA vs. Standard AA Using the Flame element Bb
ca
A,

lor

nm

Zeeman
classa A

Ga Geb
Li Mg Na

Pb
Rh

Ru
Sr Ub
Wb

Zrb

249.7 422.7 287.4 265.1 253.7 766.5 670.8 285.2 589.0 213.6 343.5 349.9 460.7 358.5 255.1 360.1

N A N
N A A N A A

A A
N -C A A

field strength, kG 3 6 9 0.57 0.18 0.42 0.92 0.60 0.96 0.59 0.20 0.75 0.59 0.06 0.82 0.38 0.12 0.66 0.54 0.61 0.33 0.08 0.18 0.30 0.45 0.74 0.07 0.41 0.53 0.60 0.58 0.75 0.18 0.81 0.72 0.68 0.86 0.80 0.86 0.69 0.74 0.51 0.52 0.40 0.65 0.35 0.77 0.90 0.93 0.90 0.80
___

,011

1 -

IO

IO

IO0

1000

CU CONCENTRATION - flg

Figure 2. Copper calibration curves (324.8 nm) measured in the furnace at a field strength of 12 kG

a Zeeman pattern; A = anomalous, N = normal. C,H, flame. Line not classified.

N,O/ for Bi (223.1 nm) is probably also attributable to the presence of hyperfine structure. In considering sensitivity with a Zeeman-effect instrument, it is essential to define the capabilities of the magnet. While desirable for experimental studies, the unit we employed is far too large to be practical for use in a commercial instrument. A dc electromagnet providing a maximum field strength of 9 kG appears to be analytically satisfactory and is a realistic compromise with respect to size, weight, and power considerations. Using the data in Tables I and 11, and assuming a variable field limited to 9 kG,the sensitivity comparison considering only primary wavelengths is as follows: For 26 of the 44 elements studied, less than a 25% sensitivity loss is observed relative to standard AA sensitivity. Somewhat greater sensitivity loss, in the range of 25-50%, is evident for 14 additional elements. A loss greater than a factor of two is observed in only four cases; Bi at 223.1 nm, Cu a t 324.8 nm, In a t 303.9 nm, and Li a t 670.8 nm. I t is significant t h a t if a 9 kG permanent magnet were used, an additional 11 elements would fall into this category. Overall, the outlook considering analytical sensitivity appears to be reasonably favorable using a variable field limited to 9 kG. Relative to standard atomic absorption sensitivity, only a moderate loss in sensitivity is observed for all but a few elements. Even in these cases, it is doubtful that performance will be significantly affected considering the excellent sensitivity inherent in the furnace technique. Analytical Curves and Working Range. A more serious potential limitation in considering the Zeeman effect is the question of analytical working range. For transitions free of hyperfine structure and displaying a normal Zeeman pattern, x and u components will be completely separated a t sufficiently high magnetic fields. Consequently, unless extremely high analyte concentrations are analyzed, the analytical curve will be similar to that obtained by standard atomic absorption. However, for transitions displaying an anomalous Zeeman pattern, some overlap between displaced A and u components usually occurs. In these cases, curvature will generally be greater and may be so severe as to produce double-valiied analytical curves. Another factor contributing to doublevalued behavior is the level of stray light within the optical system (17). The Cu calibration curves shown in Figure 2 illustrate the reason for this double-valued behavior. As sample concentration increases, x absorbance begins to saturate as in standard AA. The u absorbance saturates a t higher concentrations because of the greater separation of the emission line from the displaced u components. When the increase in u

since x absorbance is concentrated in a single, undisplaced spectral component. This single A component will be well separated from wavelength displaced u components even a t relatively low field strengths. As expected, normal class transitions typically did yield good sensitivity a t low field strengths. However, one exception was Be (234.9 nm) for which good sensitivity was obtained only a t relatively high field strengths. One contributing factor for this behavior could be significant Doppler line broadening of the emission line, since Be has a very low atomic weight. This in turn may lead t o greater u absorbance and a lower A - u signal. While not a normal class transition, the poorer than expected performance for Li (670.7 nm) a t low field strengths may also be due in part to Doppler broadening. Of the elements surveyed, over half have primary wavelengths with anomalous Zeeman patterns. In these cases, somewhat poorer sensitivity than that obtained by standard AA would be expected since both A and u absorbance are split into several components. As the field strength is increased, these components will broaden out and shift away from the center frequency. Since the u components will shift more rapidly than x components, sensitivity initially increases. This is followed by a decrease in sensitivity as the x components shift away from the center frequency. The significance of this behavior is t h a t the field strength providing optimum sensitivity varies considerably depending upon the particular Zeeman pattern. For example, the field strengths optimum for the anomalous Zeeman patterns of the Ag (328.1 nm) or Mo (313.3 nm) lines are much different than the optimum field strengths for the normal Zeeman patterns of the P b (217.0 nm) or Sn (224.6 nm) lines. I t is thus obvious that the flexibility of a variable field electromagnet offers significant advantages over a permanent magnet with respect to minimizing sensitivity losses. The presence of hyperfine structure influences sensitivity. Significant hyperfine structure is observed a t the primary resonance lines for Cu, 324.8 nm, and for In, 303.9 nm. Consequently, the Zeeman patterns are no longer symmetric about the center frequency. As the magnetic field increases, there is greater overlap between the emission line broadened by hyperfine structure and the wings of the displaced u components. The resulting complex behavior is indicated by the relative absorbance data which show an initial local maximum with field strengths in the 3-6 kG range. Higher field strengths first produce a decrease, then a further increase in relative absorbance. The relatively poor performance evident

744

ANALYTICAL CHEMISTRY, VOL. 52, NO. 4, APRIL 1980

SAMPLE
006-

Table 111. Absorbance Level Where Calibration Curve Rollover Observed-Flame Results element
Bi Cd
h,

00 pq C"

nm

maximum absorbance level 6 kG 9 kG 11.6kG

0.450 0.910

005-

SAMPLE
0 04

80 D I

C"

SAMPLE
40

ll

003

cu In
Ni

Pb Se

223.1 228.8 324.8 303.9 232.0 217.0 196.0

0.770 0.450 0.280 0.470 0.630 0.700

0.970 1.25
0.110

0.775 1.05 0.975

1.20 1.05 1.60 0.360 0.890 1.20

1.05

SAMPLE

(I

& Cu

'1

02 BKG. COURECTION

absorbance exceeds the incremental change in T absorbance, the analytical curve turns downward, back toward the concentration axis. We call this effect rollover of the analytical curve. While primarily a problem with transitions having very complicated Zeeman patterns, this effect will eventually occur even with normal class transitions if analyte concentrations are carried to extremely high levels. Table I11 summarizes results illustrating this double-valued behavior for a number of elements. The absorbance values listed are the maximum values obtained before calibration curve rollover was observed. With only several exceptions, it was found that while this double-valued problem exists, the useful working range is not severely restricted, assuming that reasonable field strengths are used. Again, considering the situation using a variable field limited to 9 kG, with the single exception of the In 303.9-nm line, even those transitions with the most complicated Zeeman patterns yielded analytical curves exceeding a minimum of 0.7 absorbance. The secondary In line a t 325.6 nm (Table I) is preferable due to the significant hyperfine structure evident a t the primary 303.9nm line. For furnace applications at ultratrace concentration ievels, a working range that is in some instances restricted to this absorbance level should not be a serious limitation. Such a restriction would be of much more consequence for flame applications where variations in concentration ranges are typically much greater. It is important to consider several points regarding this double-valued behavior. First, such behavior introduces the possibility of ambiguous results. This problem does not exist with conventional AA systems where calibration curves eventually reach an asymptote a t very high concentration levels but do not curve downward. By monitoring the shape of the analyte peak, it should be possible to avoid the potential error associated with double-valued behavior. With analyte concentrations great enough to cause rollover, the signal first rises with increasing atom concentration, then falls with the onset of double-valued behavior. As the atom concentration declines towards the end of the atomization cycle, the signal will rise a second time to the initial maximum before falling to zero. The analyte signal thus consists of two distinct peaks. If the maximum peak height is monitored, the analytical curve will remain a t an asymptote once rollover occurs. The occurrence of double-valued behavior might also be detected by monitoring the derivative of absorbance vs. time which would pass through zero three times in the presence of this effect, but only once in its absence. Secondly, as noted by de Loos-Vollebregt and de Galan (19,a design utilizing an ac modulated magnetic field will improve performance regarding both sensitivity and working range. B a c k g r o u n d Correction. The reason for employing the Zeeman effect is the potential for improved background correction performance relative to continuum source correction. Zeeman background correction on the analyte corrects a t exactly the same analyte wavelength. Thus if the background is structured, correction is far more likely to be accurate than is continuum background correction ( 1 4 ) . A

Flgure 3. Comparison results for the determination of Cu (324.8 nm)

in an artificial seawater matrix. Furnace atomization temperature was 2500 O C
Pb 5 T D

03-

PO STO

1%

yc1

1 0 0 Pa

,
MG

1%

NoCI

1 0 3 pg Pb

ZEEMAN CORRECTION

I
c6pREcrm

Flgure 4. Comparison results for the determination of Pb (283.3nm) in a 1 % NaCl matrix. Furnace atomization temperature was 2300 OC

benefit of perhaps equal importance is that double-beam background correction can be realized for all elements without the alignment and intensity limitations encountered using continuum sources. To assess background correction performance, direct comparison tests were performed using the Zeeman instrument and several conventional atomic absorption instruments utilizing continuum source correction. A variety of sample matrices, which past experience has shown create problems with respect to accurate background correction, were chosen for this comparison. All comparisons were performed using the same sample solutions and furnace parameters. Figure 3 compares results for the direct determination of Cu in an artificial seawater matrix. Results with the Zeeman instrument were obtained operating a t a field strength of 2.5 kG. For this analysis, the matrix of approximately 4 % NaCl creates considerable problems. Even after ashing a t 900 "C for 40 s, remaining matrix material generates a background level ranging from 1.5 to 2 absorbance. The results obtained using D2correction (Model 5000) show considerable base-line upset due to under- or overcompensation for the very high background. This is particularly evident in the signal for the unspiked sample where the multiple peaks make interpretation difficult. With Zeeman correction, the high background has little effect on the base line as is evident by the well defined Cu peaks. Figure 4 shows similar comparison results for Pb, determined in a 1%NaCl matrix (18). Because of the relatively low ashing temperature of 500 O C required when determining Pb, the NaCl matrix is not removed prior t o atomization. Consequently, the level of background absorption exceeds 1.5 absorbance. This high background level again clearly influences the base-line stability obtained using Dz correction (Model 460). There is substantial improvement evident in the results obtained with the Zeeman instrument which was operated a t a field strength of 9 kG. Results for the determination of Se in a calcium phosphate matrix are compared in Figure 5. Using an ashing temperature of 400 OC, the background level ranged from 0.7 to 1.2 absorbance. The large negative signals obtained using Dz correction (Model 603) indicate overcompensation for the high background signal. In this case, the negative signals are so large that it is difficult t o interpret accurately the Se peaks

ANALYTICAL CHEMISTRY, VOL. 52, NO. 4, APRIL 1980

745

ZEEMAN CORRECTION

D2 BKG CORRECTION

0.5% r n H m

1%

3 nq Se

Flgure 5. Comparison results for the determination of Se (196.0 nrn) in a calcium phosphate matrix. Furnace atomization temperature was 2600 O C

in the spiked solutions. Results using the Zeeman instrument operated at a field strength of 6 kG show that the high background has little effect on the base line. As a result, the Se signals in the spiked solutions are clearly defined and easy t o quantitate. Another type of interference to consider is spectral interference (14). With continuum source correction, spectral interference can occur because all spectral features of the background within the monochromator bandpass are sampled, producing a n average background measurement which may differ from the background a t the analyte wavelength. The determination of Se in a n iron matrix (Figure 6) is an example of such a spectral interference (19). Using D2 correction (Model 603),several iron lines close to the primary S e line at 196.0 nm fall within the spectral bandpass. When solutions containing high iron concentrations are analyzed, absorbed radiation from t h e D2 lamp produces negative absorption signals. No iron spectral interference was observed operating the Zeeman instrument at the primary 196.0-nm Se line (6 kG field strength) since background correction is made a t exactly the same wavelength. Yamada et al. (20) have shown that Zeeman correction eliminates similar spectral interference when determining S b in the presence of Pb, Cu, and Fe using D, correction.

CONCLUSIONS
T h e comparison data confirm that background correction based on t h e Zeeman effect provides significant advantages over continuum source correction. This improvement was clearly evident in several cases involving very high background levels where performance using continuum sources was de-

graded by extraneous signals due to under or over compensation. With the Zeeman instrument, base-line stability was maintained at substantially higher background levels. However, the attractiveness of the Zeeman technique for improved background correction must be weighed against the disadvantage of seme sensitivity loss plus the more important problems of reduced linearity and double-valued analytical curves. An overall assessment of the Zeeman technique is complicated by the fact that the particular design approach employed is an important factor. Depending upon the magnet location, the Zeeman effect may be produced in the atomic vapor or in the light source. The magnetic field may be aligned in an orientation longitudinal or transverse to the optical path. Various polarization systems (rotating polarizers, fixed polarizers, variable phase retardation plates, etc.) may be used, and the Zeeman effect may also be produced by modulating the magnetic field. Regardless of which configuration is used, the fact that different transitions provide optimum performance a t widely varying field strengths greatly favors the use of a variable-field electromagnet over a permanent magnet. Stephens (22) has reported a source-split configuration using a longitudinal field. However, a longitudinal alignment appears to offer no significant advantage over t he much more commonly used transverse orientation. The obvious advantage of a source-split configuration is the ease with which any atomization system can be used. Competing against this advantage is the difficulty encountered operating standard hollow cathode lamps in a strong magnetic field. Constructing numerow light sources capable of providing adequate stability in high magnetic fields is a project of considerable scope. In considering background correction performance, there is a slight disadvantage using a source-split configuration since the background is measured slightly off the center frequency. However, the wavelength interval is so small that in practice it would probably be very rare to experience measurable error from structured background. While background correction performance will probably not be significantly influenced by the position or type of magnet used, this is not the case regarding sensitivity and analytical range. A configuration using a modulated field (ac electromagnet) and a polarizer to eliminate the x signal could lead to improved sensitivity and analytical range ( I 7 ) . The sensitivity loss observed when the x absorption line is split into several components is overcome by making the signal plus background measurement with the field off, followed by a background correction measurement ( u absorbance only) at maximum field strength. If the maximum field used is sufficiently high that u absorbance is extremely small, then it should always be possible to obtain sensitivities very close to
6 "9 Se STD

0.4 -

a3-

z a

6 "9 Se STD.

a20.1 -

2 a

0--

ZEEMAN- CORRECTION
was 2600 O C

0 2 BACKGROUND CORRECTION

Flgure 6. Comparison results for the analysis of solutions containing high iron concentrations at the 196.0-nm Se line. Furnace atomization temperature

746

Anal. Chem. 1980, 52, 746-753 (8) Koizumi, H.; Yasuda, K. Spectrochim. Acta. Part6 1978, 37, 237-255. (9) Koizumi, H.; Yasuda, K. Anal. Chem. 1976, 48, 1178-1182. (10) Dawson, J. B.; Grassam, E.; Ellis, D. J.; Keir. M. J. Analyst(London) 1976, 707, 315-316. (11) Grassam, E.; Dawson, J. B.; Ellis, D. J. Analyst (London) 1977, 102, 804-818. (12) Koizumi, H.; Yasuda, K. Spectrochim. Acta, far2 6 1978, 3 1 , 523-535. (13) Koizumi, H.; Yasuda, K.; Katayama, M. Anal. Chem. 1977, 4 9 , 1106-1 112. (14) Koizumi, H. Anal. Chem. 1978, 5 0 , 1101-05. (15) de Loos-Vollebregt, M. T. C . ; de Galan, L. Spectrochim. Acta, Part 6 1970, 33, 495-511. (16) Brown, S. D. Anal. Chem. 1977, 4 9 , 1269A-1281A. (17) 1979. 33. . . de Loos-Vollebreat, - M. T. C.: de Galan. L. ADD/. . . Soectrosc. . 616-626. (18) Manning, D. C.; Slavin, W . Anal. Chem. 1978, 50, 1234-1238. (19) Manning, D. C. At. Absorpt. News/. 1978, 77, 107-108. (20) Yamada, H.; Uchino, K.; Koizumi, H.; Noda, T.; Yasuda, K. Anal. Lett. 1978, 10, 855-867. (21) Stephens, R. Talanta 1979, 2 6 , 57-59.

standard atomic absorption sensitivities. Even more impor- . .

tant, the linearity of the analytical curves should be improved. Apart from the technical problems associated with modulating
the field, it must also be shown that in such a system the background does not change when the field is modulated. In considering overall performance for a Zeeman-effect atomic absorption spectrophotometer, it would appear that this approach may retain the improved background correction capability with minimum sacrifice in sensitivity and analytical range.

LITERATURE CITED
Koirtyohann, S. R.; Pickett, E. E. Anal. Chem. 1966, 3 8 , 585-587. Hadeishi, T.; McLaughlin, R. D. Science 1971, 174, 404-407. Hadeishi, T.; Church, D. A.; McLaughlin, R. D.; Zak. B. D.; Nakamura, M.: Chana. B. Science 1975. 787. 348-349. Koizumi, h.; Yasuda, K. Anal. Chem. 1975, 4 7 , 1679-1682 Stephens, R.; Ryan, D. E. Talanta 1975, 2 2 , 655-658 Stephens, R.; Ryan, D E. Talanta 1975, 2 2 , 659-662. Veinot, D. E.; Stephens, R. Talanta 1976, 2 3 , 849-850

RECEIVED for review November 30, 1979. Accepted January

15, 1980.

High Frequency Excitation of Spark-Sampled Metal Vapor

David M. Coleman,* Mario A. Sainz, and Hal T. Butler
Deparfment of Chemistry, Wayne State University, Detroit, Michigan 48202

First results are presented of a new analytical discharge system wherein the normal processes of samp/ingand spectral excifaflon are temporally and spatially segregated into Independently optlmlzable events. Specifically, a high voltage spark Is first used as a sampling device to create a toroidally-shaped reservoir of species sampled from a metallic electrode surface. Later in time, after the spark sampling current has ceased, thls material is inductively re-excited, at high frequency, utilizing critical coupling Conditions. This approach results in dmpliflcatlon of observed emission spectra with most energy seemingly concentrated in the atom and ion resonance lines (speclation). Further, one observes relatively sharp spectral emitting lines and nearly total elimination of spectral background continuum. Because of the optical thinness of the re-exclted torus, Increased freedom from physlcal matrix effects and Increased dynamic range are anticipated, but have not been experimentally verified.

coupling of the sample vapor and of the secondary magnetically induced field. Some similarity to a conventional high duty cycle-mode (CW-mode) inductively-coupled plasma is evident.

The major goal in this laboratory has centered around the development and understanding of new luminous electrical discharges which allow separation of sampling and of spectral excitation phenomena which are normally integrated in direct spectrochemical analysis of metal surfaces. The long term goal is to provide control and optimization of each step to analytical advantage. We report here first results of a new approach to direct spectrochemical analysis of a metallic electrode surface. This approach involves direct metallic vapor introduction, with spatial integrity, into an observation region via controlled high voltage spark sampling of discrete electrode surfaces. The properties of the spark discharge process serve to locate this vaporized metal (ions and atoms) in space with well-defined toroidal geometry. Time-delayed re-excitation of this resident vapor, with a triggered high frequency pulsed energy train, is then caused t o occur a t a time corresponding to critical
0003-2700/80/0352-0746$01.00/0

BACKGROUND Analytical discharges such as the spark, the arc, or the inductively-coupled plasma, all require that the same plasma function to both sample and/or prepare the sample and then to electronically excite the spectral levels of interest. In the cases of the spark and arc, the first steps involve direct sampling, vaporization, propagation, atomization, and ionization of material from an electrode surface. For the inductively coupled plasma, analysis involves sample preparation (acidic dissolution, dilution, pH adjustments, and the addition of spectroscopic anti-matrix buffers), presentation (ultrasonic or mechanical nebulization), energy-consuming desolvation of a cold sample in the plasma, and somewhat inefficient excitation. I t is not unreasonable to assume that selection of analysis conditions which may indeed provide for most efficient sampling are, in fact, frequently inconsistent with energy levels necessary to also produce optimum spectral excitation. The work reported in this paper supports that view. Recent experimental studies ( I ) on the atmospheric pressure spark discharge revealed presence of a hithertofore undetected toroidally-shaped structure, wrapped about the interelectrode axis, which is long-lived and which forms only after the spark sampling current had decayed to low values. This torus was originally observed by Hosch ( 2 ) and further characterized by Klueppel, Coleman, et al. (3) in time- and spatially-resolved schlieren studies. Coleman and Walters ( 4 ) subsequently presented first experimental evidence substantiating a postulate that this torus is composed of significant (but as yet unquantified) numbers of metal atoms and ions of material originally sampled from the cathodic electrode. Very recent work by Araki and Walters (5) studying absorption in a 323-MHz quarter-wavelength stable spark disQ 1980 American Chemical Society