1.00 indicates that the relationship between the emission signal and concentration is linear and passes through the origin. The detection limits for calcium, copper, and hafnium are shown in Table I. The detection limits are defined as the concentrations of the element in solution which when aspirated into the plasma gives emission signals equal to twice the standard deviation in the background measurements ( 5 ) . The standard deviation was obtained as Y5 of the peak-to-peak noise level determined over a period of time equal to many time constants. Flame atomic absorption and emission detection limits are included in Table I for rough comparison purposes. (5) T. J. Vickers and J . D . Wlnefordner, "Flame Spectrometry," in "Analytical Emission Spectroscopy," Part I / , E. L . Grove, Ed., Marcei Dekker, Inc , New York, N.Y , 1972. o, 333 (6) 0. Menis and T. C. Rains. in "Analytical Flame Spectroscopy," R. Mavrodineanu, Ed., Macmilian and Co. Ltd., London, New York, 1970, p 60. (7) R . L. Warren, Analyst (London), 90, 549 (1965). (8) G. D. Christian and F. J . Feldman, "Atomic Absorption Spectroscopy." Wiiey-Interscience, New York, N.Y.. 1970, p 174. (9) S. Slavin, W. B. Barnett, and H. L. Kahn, At. Absorption Newslett.. 11, (2) 38 (1972). ~~~
There are at least two important factors other than the intensity of a line which affect the detection limits. The emission from the center of the arc is very sensitive to minor current fluctuations. Therefore, as the position for observation of an analyte line moves toward the center of the arc, these fluctuations in the background emission can become a limiting factor. Thus, although Figure 6 shows the peak intensities for Ca(1) a t 3.8 mm and Ca(I1) at 1.9 mm to be about equal, the detection limit for the Ca(I1) line a t 1.9 mm is 50 times poorer than the detection limit for the Ca(1) line. A second important factor is the effect of the spectral lines neighboring the line of interest. Although, for a concentration of 1000 ppm, the hafnium ion line is over six times as intense as the neutral atom line, Figure 6, its detection limit is only four times better than the detection limit for the neutral atom line. The ion line is in the region of OH band emission while the neutral atom line is found in a "spectrally clean" region of the arc. Received for review June 18, 1973. Accepted December 18, 1973.
High Resolution Atomic Absorption Spectrometry Using an Echelle Grating Monochromator Peter N. Keliher and Charles C. Wohlers Chemistry Department, Villanova University, Villanova, Pa., 19085
A direct comparison is made, using a high resolution echelle spectrometer, between atomic absorption using a line source (hollow cathode lamps), and atomic absorption using a continuum source (150-W xenon lamp). Sensitivities using the continuum source tend to be less than those using a line source, but not to any great extent. The effective spectral bandwidth using a continuum source with the echelle spectrometer is wider than for a line source but much narrower than that used in most previous studies of atomic absorption with continuum sources. Linear calibration curves, having slopes of approximately one, were obtained with the continuum source when the echelle system was used, but when medium resolution instrumentation was employed, in conjunction with the continuum source, much lower slopes and sensitivities were observed.
Since the introduction of atomic absorption spectrometry as an analytical technique by Walsh ( I ) in 1955, it has been generally assumed that a sharp line spectral source, usually a hollow cathode lamp (HCL), was necessary for the practical utilization of atomic absorption spectrometry. The objections raised by Walsh ( I ) to the use of a continuum source were, first, that the very narrow width of the absorption profiles would require a monochromator with a resolution of a t least 500,000 (which would be much larger and more inconvenient to use than conventional spectrometers), and, second, that it was questionable whether the continuum source used would provide
enough radiation over so small a wavelength range (approximately 0.01 A) to give a favorable signal-to-noise ratio. With the introduction of high resolution, high luminosity monochromators using the echelle grating ( 2 - 7 ) . however. it is possible that the objections mentioned by Walsh ( 1 ) have been overcome. It is the objective of this paper to discover to what extent this may be true. The use of a continuum source in atomic absorption spectrometry enjoys certain potential advantages over spectral line sources, certainly the most obvious of which is the advantage of using only one source for a variety of elements, thus saving considerable expense. Also, a continuum source would be highly useful in qualitative analysis and would make background correction simpler. In order to exploit some of these advantages, a number of workers have employed continuum sources with medium resolution instrumentation (8-12). Although the spectral bandpass employed was always much greater than the (2) (3) (4) (5)
(6) (7)
(8) (9) (10) (11) (12)
( 1 ) A Walsh, Specirochm Acta 7, 108 (1955)
682
A N A L Y T I C A L C H E M I S T R Y , V O L . 46, N O . 6, M A Y 1974
G. R . Harrison, J . Opt. SOC.Amer.. 39, 522 (1949). D. Richardson, Specfrochim. Acfa, 6, 61 (1953) W. G . Elliott, Amer. Lab., 2 (3). 67 (1970) M. S. Cresser, P. N. Keliher, and C . C. Wohlers. Spectrosc. Lett.. 3, 179 (1970). F. L. Corcoran, Jr., P. N. Keliher, and C. C. Wohlers, Amer. Lab., 4 ( 3 ) ,51 (1972). M S Cresser, P. N . Keliher, and C C. Wohlers, Anal. Chem . 45, 111 (1973). J . H . Gibson, W. E. L Grossman, and W. D. Cooke, Appi. Spectrosc., 16, 47 (1962). N. P. lvanovand N. A . Kozireva, Zh. Anal. Khim., 19, 1266 (1964). V . A Fassel, V. G Mossotti, W E. L. Grossman, and R . N Kniseley, Spectrochim. Acta. 22, 347 (1966) W. W . McGee and J . D . Winefordner, Anal. Chim. Acfa. 37, 429 (1967). C W. Frank, W G Schrenk, and C. E. Meloan, Anal. Chem. 39, 534 (1967).
width of the absorbing line, the results obtained were at least moderately successful. However, detection limits were generally 10-100 times worse than those presently reported using hollow cathode lamps, which was generally attributed to the poor signal-to-noise ratio of the continuum systems. This has recently been partially overcome by the use of wavelength scanning (13-15). Veillon and Merchant (16) have recently used a piezoelectric FabryPerot interferometer in conjunction with a conventional monochromator for atomic absorption measurements with a continuum source. Results reported for copper and silver showed sensitivities and calibration curves comparable to results obtained with line sources. Expected analytical results for sensitivity and shapes of calibration curves have been described theoretically for both atomic absorption using a narrow line source (AAL) and for atomic absorption using a continuum source (AAC) (17-19). The model for AAL is based on a sharp line spectral source (such as a hollow cathode lamp) with line widths much narrower than those of the absorption line profiles of the element in the flame. Shapes of theoretical calibration curves using this model are identical to those experimentally obtained in conventional atomic absorption---i.e., a slope of one for a log-log plot of concentration us. absorption, with a sloping off at high concentrations. The theoretical model for AAC presumes a medium resolution monochromator such that the spectral bandwidth of the source, as defined by the spectral slit width, is much greater than the absorption line profile of the element in the flame. The calibration curve expected for a log-log plot of concentration LE. absorption has a slope of one a t low concentrations and one-half a t high concentrations, which is identical to that expected in emission work. This model should fit the experimental conditions used in previous work with continuum sources (8-12). If, however, the dispersion of the monochromator is high enough such that the spectral slit width is significantly smaller than the absorption line profile in the flame (-i.e., similar to the width of lines from a hollow cathode lamp), the theoretical model for AAL would apply, no matter what type of source was used, and results for AAC using such a monochromator would be similar to atomic absorption results using a sharp line source. In other words, if the spectral slit width of a high resolution, high dispersion monochromator was similar to the widths of lines emitted from a hollow cathode lamp, the sensitivities and shapes of calibration curves would also be similar for AAC using such a monochromator and for atomic absorption results obtained with typical commercial instrumentation. Unfortunately, prediction of results for AAC using high resolution monochromators is made difficult by scanty experimental reports of accurate measurements of absorption line profiles in flames or of hollow cathode lamp spectral line widths for comparison with AAL data. One element for which such measurements have been made, however, is calcium. Bruce and Hannaford (20) have measured the half-intensity line width of the calcium 422.7(13) W . Snellernan. Spectrochim. Acta, Part B , 23,403 (1968). (14) R. C. Elser and J . D. Winefordner, Anal. Chem., 44, 698 (1972) (15) G . J. Nitis, V. Svoboda, and J. D Winefordner, Spectrochirn. Acta, Part 6, 27, 345 (1972). (16) C. Veillon and P. Merchant, Jr., Appl. Spectrosc., 27 361 (1973) (17) L. deGalan, W . W . McGee, and J . D. Winefordner. Anal. Chirn. Acta, 37,436 (1967). (18) C . Th. J. Alkernade. Appl. Opt., 7 , 1261 (1968) (19) J. D. Winefordner, V . Svoboda. and L. J. Cline, CRC Crit. Rev. Anal. Chem., 1, 233 (1970) (20) C. F. Bruce and P. Hannaford, Spectrochim. Acta, Part E , 26, 207 (1971)
nm line emitted by an HCL and obtained a value of 0.0092 A a t a current of 5 mA. Kirkbright and Troccoli (21) have studied the width of this absorption line, in an air-acetylene flame, using a Fabry-Perot interferometer, and reported a value of 0.034 A . Both of these results have been confirmed by Wagenaar and DeGalan (22) who obtained values of 0.0112 A for the HCL and 0.037 A for the absorption profile. Thus, the spectral slit width needed to obtain results for AAC similar to those for AAL would seem to be approximately 0.01 A or at least significantly less than 0.034 A for the calcium 422.7-nm line. For the echelle spectrometer used in this work, the spectral slit width for a 25-pm slit a t this wavelength is 0.023 A. As will be shown, this is adequate for sensitive results in AAC comparable to results obtained by AAL on the same instrumentation.
EXPERIMENTAL Instrumentation. A prototype Spectraspan Model 101 Echelle Spectrometer (SMI, Inc., Andover, Mass.) was used fitted with a Hamamatsu R446 photomultiplier and a Polaroid film attachment for rapid qualitative analysis of spectra. Polaroid Type 57 packet film (A.S.A. 3000) was used with typical exposure times of less t h a n 1 minute. The Spectraspan utilizes a n echelle grating in its monochromator to obtain approximately an order of magnitude greater resolution and dispersion t h a n monochromators used in typical atomic absorption instrumentation. The focal length of the Spectraspan is 0.75 m . Characteristics of this instrumentation have been previously reported ( 5 - 7 ) . T h e Spectraspan is designed primarily for emission work and has a fixed vibrating reed chopper placed immediately after the exit slit. In atomic absorption spectrometry it is, of course, desirable t o modulate t h e source b u t not the flame and it was found, in this study, more convenient to bypass the Spectraspan electronics rather t h a n to remove t h e fixed chopper. Accordingly, the vibrating reed chopper was left stationary--i.e., moved out of position of t h e optical path-and a Model LPA-1 Photometric Amplifier (Spectrogram Corporation, North Haven. Conn.) was used in conjunction with a Spectrogram Model IDV-1 Intergrating Digital Voltmeter. Conventional sealed hollow cathode lamps manufactured by Perkin-Elmer (Atomax series), Atomic Spectral Lamps. Pty., or Westinghouse Corporation were powered by a Spectrogram Model LPS-1 Hollow Cathode L a m p Power Supply. The spectral sources were electronically modulated a t 400 Hz in phase with the photometric amplifier. Lamps were run a t a n average current of a p proximately 5 mA which amounts t o IO-mA peak current pulsed above a direct current of about 0.2 mA. T h e d.c. base current is added to improve stability. T h e continuum source used was a Hanovia 150-W xenon arc lamp powered internally by a power supply contained within the Spectraspan. T h e l a m p was run a t 6 A and 35 V d . c . a n d was mounted on a micrometer positioning device for easy adjustment. The lamp was water cooled. Since it was not convenient t o electronically modulate t h e xenon lamp, a mechanical chopper (Model 125, Princeton Applied Research, Princeton, N. ,J.) was placed between the continuum source and t h e flame. This chopper w2s modulated a t 400 Hz in phase with the photometric amplifier. A conventional pre-mixed burner system taken from a Varian Techtron AModelAA-4 was used for all measurements. An AB-50 (6-cm) burner head was used for nitrous oxide-acetylene. a n AB-51 (10-cm) for air-acetylene, a n d an AB-52 (10-cm) for airpropane. For comparison purposes, some AAC d a t a was collected on conventional atomic absorption instrumentation. A Varian Techtron Model AA-4 was used in conjunction with the 150-W xenon l a m p and a n S M I Model 11-B d.c. Plasma Power Supply. This power supply is similar to the power supply contained within the SpectraSpan used for the other measurements. A mechanical chopper (Varian Techtron Model MK-2) modulated a t 285 Hz was placed between the xenon l a m p and the flame. Procedure. Stock solutions were prepared from reagent grade chemicals and distilled water. From these stock solutions, working solutions with relative concentrations of 1.00, 2.00, and 5.00 (21) G. F . Kirkbright and 0 . G . Troccoli. Spectrochim. Acta, Part B, 28, 33 (1973). (22) H. C. Wagenaar and L. deGalan, Spect;ochirn. Acta, Part E , 28, 157 (1973). A N A L Y T I C A L C H E M I S T R Y , VOL. 46, NO. 6, M A Y 1974
683
XD
C.w.rr.t*..
Figure 1. Calibration curves for sodium 589.0 nm. ( A ) cathode lamp; ( B ) Continuum
Hollow
cc
Figure 2. Calibration curves for calcium cathode lamp; (8)Continuum
422.7
nm. ( A )
Hollow
for each decade range were prepared. All atomic absorption measurements were performed with the gas flow rates and burner position optimized for each element. Slit widths used on the SpectraSpan were 25 pm for AAC, unless otherwise stated, and 200 pm for AAL. Ten-pm or 25-pm slit widths (0.033-nm and 0.082nm spectral bandpasses) were used for AAC measurements on the Techtron AA-4. Direct measurements were made on the AA-4. On the SpectraSpan, measurements were made by integrating a blank (distilled water) and then a sample solution for 10 seconds each, and repeating this four times. Each set of blank and sample readings were then divided to obtain a per cent transmittance, and these results averaged and the logarithms taken to obtain the absorbance.
RESULTS AND CONCLUSIONS Calibration curves for various elements are presented in Figures 1-5. In all cases the sensitivity and linear range for AAC is comparable to, and sometimes better than, that for AAL. It should be noted, however, that detailed comparisons of sensitivities using these curves is not possible owing to inevitable small changes in burner position, gas flow rates, etc. These changes should not markedly affect the shapes of calibration curves, however, and approximate comparisons of sensitivities should still be possible. For all five elements, the calibration curve for AAC is similar in shape to that for the HCL source. This apparently confirms the application of the theoretical model for AAL for high resolution AAC. For sodium 589.0 nm (Figure l), the sensitivity is apparently greater for AAC than when using AAL, and the linear range is a t least as good. For calcium 422.7 nm and aluminum 396.1 n m (Figures 2 and 3), the shapes of curves for AAC and AAL are essentially identical, with the apparent sensitivities being equal for calcium and somewhat less for AAC in the case of aluminum. With chromium 357.9 nm (Figure 4), a noticeably poorer result is obtained for AAC; the calibration curve slopes off a t lower absorbance for AAC than for AAL, and the apparent sensitivity is less. This trend is continued for copper 324.7 nm (Figure 5), where the calibration curve for AAC slopes off a t a still lower absorbance value. Thus, the results for AAC on this experimental system seem to become progressively worse with decreasing wavelength until a point is reached, as for copper 324.7 nm, where it is not clear whether or not the theoretical model for a sharp line source (that used for AAL) still applies. One would assume that the spectral bandpass is no longer significantly smaller than the line width of the absorption line in the flame a t these lower wavelenghts. This conclusion is confirmed in Table I. The line width of the HCL, line width in the flame, and spectral bandpass all decrease with decreasing wavelength. However, the spectral bandpass decreases a t a much slower rate until, as in 684
A N A L Y T I C A L C H E M I S T R Y , VOL. 46, NO. 6, M A Y 1 9 7 4
crmrr.tkm.
m
Figure 3. Calibration curves for aluminum 396.1 nm. ( A ) H o l l o w cathode lamp; ( B ) Continuum. Nitrous oxide-acetylene flame.
Figure 4. Calibration curves for chromium 357.9 nm. ( A ) H o l l o w cathode lamp; ( B ) Continuum
the case of copper, it is essentially the same as the absorption line width in the flame. This results in poorer response, as is shown in Figure 5 , although the calibration curve obtained is still analytically useful. Although the theoretical model for AAL presumes a source bandwidth significantly smaller than the absorption line width, it is apparent from the comparison of the calibration curves in Figures 1-5 and Table I that the spectral slit width for high resolution AAC need not be much smaller than the absorption line width in the flame for acceptable analytical results. Although the sensitivity and linear range decrease as the spectral slit width becomes larger relative t o the absorption line width, the lower portion of the calibration curve is still linear with a slope of one over a significant concentration range. Data from Figures 1-5 and Table I
Table I.
Summary of Line Width Data
Element and line
Sodium 589.0 nm.
Hyperfine splittingo pattern, A
Relative intensity
0.0000
5 (24) 3
-0,0193
(f)
Calcium 422.7 nm Aluminum 3 9 6 . 1 nm
- 0 .0035
4 (26) 16 38 13 16 13
-0.0020 0.0000 f0.0028 +0.0039 +0.0053
Chromium 357.9 nm Copper 324.7 nm
(g) 0.0000 + O ,039 + O ,043
5 (27) 2.1 0.9
H C L line width, A
0.0197 (a) 0.0112 0.0092 0.0108 0.0135 0.0122
(22) ( d ) (20) ( d )
SpectraSpan spectral bandwidth: 25pm slit, A
Absorption line width in flame, X
Theoretical line width in flame, .4 ( 2 3 ) ( b )
0.032
0.070 (25) (c)
0.048-0.081
0.023
0,034 (21) ( d ) 0.037 (22) ( d )
0.025-0.039
0.0215
0.047 (22) (d,e)
0.028-0.040
0.019 0.018
... 0.019 (28) (c)
0.018-0.022 0.015-0 ,022
(a)
(22) ( d ) (a)
0.0079 (a) 0.0066 ( a )
*
'' Assuming pure Doppler broadening and Doppler temperature of 500 OK (22). Assuming pure Voigt profile with no hyperfine splitting; air-acetylene flame (including aluminum). e From curve of growth measurement. From direct interferometric measurement. e Nitrous oxide-acetylene flame. Isotopicabundance: 97% 'Oca, 2% 44Ca,
