ner, Specfrochim. Acta, Part A, 30, 1459 (1974). (6)J. P. Perchard and M. L. Josien, J. Chim. Phys., 65, 1856 (1968)
LITERATURE CITED ( 1 ) J. I . Steinfeld, CRC Crit. Rev. Anal. Chem. (in preparation) (2)J. R. Allkins. Anal. Chem., 47. 752A (1975). (3) R. M. Measures and w. R. Houston, Opt. Eng., 13, 494 (1974). (4) R. E. Brown, K. D. Legg, M. W. Wolf, and L. A . Singer, Anal. Chem., 46, 1690 - i19741r (5) B W Smith, F W Plankey. N Ornenetto, L P Hart, and J D Wineford-
__
I
-
RECEIVEDfor review September 11, 1975. Accepted November 6, 1975. This work was supported by NIH GM 11373-12.
Horizontal Tungsten Coil Atomizer and Integrated Absorbance Readout for Atomic Absorption Spectrometry R. D. Reid and E. H. Piepmeier" Department of Chemistry, Oregon State University, Corvallis, Ore. 9733 1
An improved version of a tungsten filament atomizer for atomic absorption spectrometry has been built and studied. A signal processing circuit is described that eliminates the background blackbody emission signal, and integrates the transient absorbance peak. The integrated absorbance peak improves detection limits and improves the linearity of the analytical working curve compared to the use of the maximum value of the absorbance peak. Detection limits are for Ag, 9 pg; Ca, 1 pg; Cu, 2 pg; and Mg, 5 pg.
Electrically heated atomizers for atomic absorption spectrometry (AAS) and atomic fluorescence spectrometry (AFS) have received renewed attention since the initial work by L'vov (1 1 because of their ability to handle microgram samples and help to determine picogram quantities of trace elements. Robinson and Slavin (2) and Amos ( 3 )have briefly reviewed graphite furnaces and carbon rod atomizers used by several investigators (2, 4 - 7 ) . Donega and Burgess (8) have studied the use of tungsten and tantalum boats as atomizers. Platinum and tungsten loops have been studied for use in AFS by Bratzel, Dagnall, and Winefordner ( 9 ) and an automated platinum loop system of this has been investigated by Goode, Montaser, and Crouch ( I O ) . T h e work that preceded this study was reported by Williams and Piepmeier ( 1 2 ) and involved the use of an inexpensive 24-watt tungsten helix in AAS t o determine trace amounts of elements in relatively pure solutions. A rigid spiral wound tungsten filament from a commercially available light bulb was used to atomize 3-pl samples in an argon atmosphere. The central axis of the helix was vertical, and the atomic vapor was carried vertically upward from the helix into the hollow cathode beam where atomic absorption was observed. In the investigation reported here, several modifications were made to this instrument to improve absorption sensitivity, improve detection limits, reduce spectral interference due to tungsten blackbody emission, improve the linearity of the analytical curves, and increase the lifetime of the filament to 500-1000 runs. Time-integrating the entire absorbance peak is shown to significantly improve both the detection limits and linearity of the analytical curve up t o absorbance values of 0.8 compared to using the maximum value of the absorbance peak. The absorbance integration circuit described below 338
ANALYTICAL CHEMISTRY, VOL. 48,
NO. 2,
could be connected to the transmittance output signal of other transient atomizers (carbon furnaces, etc.) with proper choice of RC time constants.
EXPERIMENTAL T h e glass bulb of an instrument grade light bulb (General Electric No. 1763, 6V-4A) was removed t o expose the precision wound tungsten filament. T h e filament was turned so t h a t the central axis of the helix was in line with the hollow cathode-monochromator axis and was mounted on an X-Y optical bench so t h a t the helix could be easily and precisely positioned. T h e argon flow and optical system were essentially the same as those of Williams and Piepmeier ( I I ) , except the newer system was single beam and did not incorporate the second monochromator. An aperture limited the viewing region to a volume inside the helix so t h a t direct viewing of the incandescent filament was eliminated. T h e viewed volume was limited to the lower half of the volume inside the helix because the best analytical precision was obtained in this region. T h e spectral bandpass for the monochromator for the Mg, Ag, and Cu lines was 0.7 nm, and 0.35 n m for the Ca line. A high-current operational amplifier (MP-1026, McKee-Pedersen, Inc., Danville, Calif.) was used in a standard voltage inverter circuit to drive the tungsten filament a t any one of three constant voltage settings. T h e constant voltage settings were determined by selecting with a 4-position switch. one of three precision resistors for the input resistance of the voltage inverter circuit. The input resistor was driven by a constant voltage source (MP-1008).T h e four switch positions were labelled OFF. DRY, ASH. and FIRE as indicated in the discussion below. T o discriminate against tungsten blackbody emission that impinges on the entrance slit of the monochromator, a modulated hollow cathode lamp coupled with an ac amplifier was used. T h e lamp was modulated with a 50% duty cycle by using a high voltage transistor as a parallel switch in series with a current limiting resistor as shown in Figure 1. A modulation frequency of 200 Hz was used because it was the highest frequency a t which this modulation circuit could drive a hollow cathode lamp without the lamp remaining on between pulses. Lamp currents of 5 mA rms were used t o keep self-absorption line broadening to a minimum. T h e signal processing circuit is shown in Figure 2. Figure 3 shows voltage signals a t points F and H in the circuit, for various conditions. T h e modulated photocurrent signal from the photomultiplier tube is converted to a modulated voltage by the currentto-voltage converter circuit of operational amplifier AI. A high pass RC filter with a corner frequency of 16 Hz essentially eliminates the tungsten blackbody emission which is low frequency in nature b u t passes the high frequency hollow cathode signal on to the precision voltage rectifier circuit of operational amplifier A2. T h e rectified hollow cathode emission signal is sent t o two low pass RC filter circuits of operational amplifiers A:i and A+ The time constant of the feedback loop of Ad is long enough to smooth t h e
FEBRUARY 1976
A
6
{
-2oov I5 KA 5w
6
Figure 1. Hollow cathode pulsing circuit using a parallel switch V
[A) EUW-15 Heath Universal Power Supply: (B) 0-100 mA current meter: (C) 0-20 mA current meter: (D) High voltage transistor, Delco 7223: (E) Hollow cathode lamp: (F) Negative-polarity square wave generator
(b)
c3
+IR6 R14
P M ANODE
R15
TIME (ma)
R13
Figure 2. Circuit used to process the anode current of the photomultiplier tube, to provide, finally, a time-integrated absorbance value, compensated for any drift in hollow cathode intensity or photomultiplier sensitivity -.- ......-....~. . ......- -. .- - ..-
(A) Amplifiers: (1) MP-103 1 chopper stabilized operational amplifier (McKeePedersen. Inc. Danville, Calif.): (2, 6, 7) MP-1006A operational amplifiers; (3, 4 ) MP-1039 dual FET operational amplifier. R1, 1 MR. R2-R8. R12. 100 k f l , 0.1%. R9. R10. 1 k R . 0 . 1 % . R l l , 100 Q , 0.1%. R13-Rl5, 10 K Q 0.1%. C1, 100 pf. C2. 0.1 pf. C3, 20 pf. C4, 0.25 p f . C5, 10 pf, MP-1012 integrator
rectified square wave signal h u t short enough t o allow t h e smoothed signal at point I to accurately follow the changes in t h e transmittance of t h e atomic vapor. T h e time constant in the feedhack loop of amplifier A:j is large enough so t h a t t h e output does not significantly respond to the atomization peak and thus the output of A,{ provides a lo@%T reference signal t h a t is used by t h e log-ratio amplifier A; t o automatically compensate for drift in t h e hollow cathode intensity. T h e log-ratio amplifier A3 provides a voltage proportional to the apparent absorbance of the atomic vapor. McWilliam a n d Bolton ( 1 2 )have shown t h a t for t h e 0.025-s time constant used in the circuit of amplifier Aq, Gaussian peaks having half-height widths greater than 0.20 s have peak absorbance values no less than 9 8 O h of the true value. Similar calculations indicated t h a t the circuit of amplifier A.3 with a time constant of 2.0 s gave a reduction in the 100?10transmittance signal a t its output of less than 2.590 a t the peak of the absorption profile. Since this reference signal changes a small amount during t h e peak profile, a slight skewing toward increased time values and a slight peak hroadening occurs. However. absorbance peak profile distortions are minimal, reproducible. and within t h e experimental noise level. Finally. the circuits of amplifiers As and A; are used to amplify t h e ahsorhance signal and provide a voltage proportional t o the integral of the ahsorbancr peak that can be monitored at point E by a strip chart recorder. 4 s with most nonflame atomizers, t h e transmittance peak observed at point I cannot he accurately monitored with t h e recorder because the halfwidth is only about 0.2 s. Therefore the voltage a t point I for this work was monitored with an oscilloscope and digitized with a 2 0 - p s , IO-bit. 0-1.25 V analog-todigital converter controlled by a PDP 11/20 computer. T h e digitized peaks were displayed on a Tektronix graphics terminal where permanent copies could be made with a hard copy unit. Procedure. A microliter syringe is used to place 2 - p l samples on the filament. Surface tension holds t h e sample droplet inside t h e helix. Sample sizes of 2 p i were used because droplets tended t o disperse outside t h e helix and adhere t o the filament supports. Atomization was less efficient outside the helix, and residues collected which were not readily vaporized. This resulted in “ghosting” effects and significantly larger standard deviations when larger droplets ivere used.
0
200
100 TIME (ms)
Figure 3. Voltage signals at point F, Figure 2, for ( a ) tungsten filament off (expanded time scale), ( b )tungsten filament on but no sample, and (c) tungsten filament on with sample present. (d)Shows the upper half of the signal at point H [vertical scale charged compared to (c)]. The influence of scattered light from the emission of the tungsten filament is apparent in ( b ) and (c), but has been eliminated by the RC high pass filter in Figure 2 to give (d)
R‘ith argon flowing up from beneath the helix a t a rate of 2 I./ min. the filament is heated (drying step) by an applied voltage of 1.00 Y.T h i s voltage produced the highest temperature for rapid evaporation. yet did not cause losses due t o spitting or violent boiling. \Vhen the sample droplet is suspended inside the helix, virtually 100°/n of the hollow cathode beam is diverted by the droplet so that the intensity signal indicates O06 transmittance. T h e drying step is essentially complete when the signal level abruptly returns to 1009.0transmittance. T h e total drying time is 15 to 20 s. T h e filament is then heated at an applied voltage of 2.00 V in an ashing process. This intermediate step eliminated impurities which cause light scattering at the time of atomization, resulting in double peaks. T h e recording devices and integrator are activated and the sample is atomized by applying a voltage of 6.00 V when the signal at point K (viewed on an oscilloscope) returns t o 0 volts indicating 100%transmittance.
RESULTS AND DISCUSSION A
n u m b e r of t h e e x p e r i m e n t a l v a r i a b l e s s i g n i f i c a n t l y influenced t h e a n a l y t i c a l results. T h e a r g o n f l o w r a t e was t h e m o s t s e n s i t i v e v a r i a b l e t o a d j u s t . F o r 0 . 4 - n g A g samples, t h e p e a k a b s o r b a n c e a n d r e l ative s t a n d a r d deviations o f f o u r replicate r u n s were determ i n e d a t various argon f l o w rates a n d p l o t t e d t o d e t e r m i n e t h e b e s t values. T h i s g r a p h i n d i c a t e d t h a t t h e b e s t s e n s i t i v ANALYTICAL CHEMISTRY, VOL. 48, NO. 2, FEBRUARY 1976
339
/2/
u-
la
Ib
2
0.23s Half width
0% T TIME
Figure 6. Absorbance signal of Ag seen at point I in Figure 2. Peak 1 indicates absorption profile without prefire ashing step; peak 2, with
ashing step O
b
'
;
'
2'
'
3'
'
4'
Ag (ng)
Flgure 4. Silver calibration curves at argon flow rates of 1.5 I./min (0)and 7 . 5 Urnin ( 0 ) :wavelength = 328.0 nm. Error bars indicate dispersion of f 1 std dev based on 4 to 6 replicate runs
-20-
-i w
0
a
N (nd
Figure 5. Analytical curves using the integrated absorbance peak for Ag at argon flow rates of 1.5 I./min (0)and 7.5I./min ( 0 )
ity and reproducibility occurred over the range of 1.5 to 3.0 l./min, and that the reproducibility was four times poorer a t 1 and 7 l./min. The vertically positioned filament required 10 l./min (11). An analytical curve of peak adsorbance for Ag is shown in Figure 4 for two flow rates, 1.5 and 7 . 5 l./min. The curve for the high flow rate bends towards the concentration axis while the curve for the low flow rate is linear out to a peak absorbance value of 0.7. The faster flow rate decreased the maximum temperature of the filament from 2000 to 1900 "C, as indicated by a pyrometer. When the area of the absorbance peak rather than its maximum value is used, the analytical curves for both argon flow rates are linear, Figure 5. Also, the reproducibility of the measurements at low concentrations is improved, as is indicated by comparing the error bars in Figures 4 and 5 . A voltage proportional to the area of the absorbance peak is observed at point E, the output of the integrator in Figure 2. A similar improvement in linearity results for copper, 324.7 nm, when the output of the integrator instead of the peak absorbance is used to construct the analytical curve. One reason for the improved linearity obtained using the integrator is indicated in Table I where the width a t halfheight of the absorbance peaks for Ag is given for several amounts of Ag. The half-width of the peak increases by a factor of 5 for a decade increase in the amount of Ag in 2 ~1 of solution. This increase in width is most likely associated 340
with the finite time required to vaporize a larger quantity of material from the tungsten filament particularly a t lower temperatures. Integrating the absorbance signal helps to compensate for the change in peak width with concentration, and thereby helps to produce a more linear analytical curve. A prefire ashing step was added to the procedure to eliminate an interfering peak which was observed on the absorption vs. time profiles of Ag, and to a lesser extent, Ca, Cu, and Mg. Figure 6 illustrates a typical Ag profile a t 328.1-nm run a t a 6.0-V firing voltage and a 2 l./min argon flow rate. Two absorption maxima are found. Peak l a is due to interfering absorption or scattering by some unknown species, and peak l b is due to silver absorption. A prefire ashing step where 2.00 V is applied to the filament for 5 s followed by a 20-s cooling off period before atomization resulted in absorption peak 2. The shape and size of the analyte absorption peak was apparently not significantly affected by the ashing step as shown when the curves are superimposed. Similar results were obtained for the 338.3-nm line of Ag except that the second peak was about half as big as for the 328.1-nm line, while the first peak was the same size. Since the analytical absorption sensitivity of the 338.3-nm line in flame work is only half that of the 328.1-nm line ( I 3 ) , this suggests that the second peak is due to Ag absorption while the first is not. The first peak was also observed a t the nonabsorbed 324.8-nm Cu line for an Ag sample, suggesting t h a t the first peak is due to the presence of light scattering impurities such as water held by the sample residue after solvent evaporation occurs. The value of the filament voltage used for the prefire ashing step was the highest voltage that could be used without causing a significant atomization of the analyte atoms. This was determined by comparing the absorbance of the analyte peak at 0.5-V intervals of the prefire ashing voltage. For one element, a t 422.7 nm, the prefire ashing step significantly improved the linearity of the analytical curve obTable I. Effect of the Quantity of Silver on Absorbance Peak Half-Width at a High Argon Flow Ratea Standard deviation,
Ag, ng
Absorbance peak halfwidth, ms
0.08
18
0.24 0.4
17
81
3 1 2 4 4 4
98
5
Quanrity of
0.8
1.2 2.0 4.0
22 59 69
ins
a Analyzed a t 3 2 8 . 1 n m with an argon flow r a t e of 7.5 l . / m i n a n d a firing voltage of 6.0 V. __
ANALYTICAL CHEMISTRY, VOL. 48, NO. 2, FEBRUARY 1976
~
Detection limits (in pg) and absorption sensitivities (in pg/l% absorption) are shown in Table I1 for Ag, Ca, Cu, and Mg. These values compare favorably with the best values obtained using other resistance heated atomizers ( I , 3-7, 9) for Ca and Cu, but are one to two decades poorer than the best values for Ag and Mg. They are 2 to 20 times better than those obtained with the vertical filament technique ( I 1).
Table 11. Peak Absorption Sensitivitiesa and Detection Limits Sensitivit p
Detection l i m i t
Ag 328.1 nm 13.0 pg 1 6 Pg (9.0 Pg)b 0.8 1 Ca 422.1 nm Cu 324.8 nm 4 3 (2.0) 30 5 Mg 285.2 nm a Sensitivity in p g / l % absorption. b Values in parentheses were obtained using time integrated absorbance values.
LITERATURE CITED 8. V. L'vov, Spectrochim. Acta, 17, 761 (1961). J. W. Robinson and P. J. Slavin. Am. Lab., 4, 10 (1972). M. D. Amos, Am. Lab., 4, 57 (1972). T. S. West and Y . K . Williams, Anal. Chim. Acta, 45, 27 (1969). J. E. Cantle and T. S.West, Talanfa, 20, 459 (1973). H. Massmann, Spectrochim. Acta, Part 8, 23, 215 (1968). R. Woodriff and G. Rameiow, Spectrochim. Acta, Part 8, 23, 665
tained by plotting peak absorbance vs. concentration. Both curves were linear and superimposed up to an absorbance of 0.15 (30 pg Ca). Above this value, the curve obtained without the ashing step bent towards the concentration axis while the curve obtained with the ashing step remains linear to A = 0.85. The improved linearity is probably related to faster and more efficient atomization that would be expected when the tungsten filament, already hot due to the ashing step, reached its maximum temperature more rapidly during the atomization step. A brief study of the interference of phosphate on calcium was conducted, using 0.08 ppm calcium. It was found t h a t phosphate suppression increased up to a phosphorus-tocalcium mol ratio of about 3 and then leveled off a t about 80% of the original absorbance value where there was no phosphate present. This suggests that the influence is due to the formation of a calcium-phosphorus compound of low volatility as is the case in flames ( 1 4 ) .
(1968). H. M. Donega and T. E. Burgess, Anal. Chem.. 42, 1521 (1970). M. P. Bratzei, R. M. Dagnall, and J. D. Winefordner, Appl. Spectrosc., 24, 518 (1970). S.R. Goode, A. Montaser, and S.R. Crouch, Appl. Spectrosc., 27, 355 (1973). M. Williams and E. H. Piepmeier, Anal. Chem.,44, 1342 (1972). i. G. McWilliam and H. C. Bolton, Anal. Chem., 41, 1755 (1969). W. J. Price, "Analytical Atomic Absorption Spectroscopy", Heyden and Son, Ltd., London, 1974, p 185. L. Leyton, Analyst (London),79, 497 (1954).
RECEIVEDfor review August 6, 1975. Accepted November 13, 1975. Work done in partial fulfillment of requirements for a master's degree (RDR).
Atomic Fluorescence Spectrometry with a Premixed Argon/Oxygen/Acetylene Flame D. J. Johnson and J. D. Winefordner" Department of Chemistry, University of Florida, Gainesville, Fla. 326 7 1
An Ar/02/C2H2 flame is compared with an air/C2Hp flame (both flames with Ar sheath) for use in atomic fluorescence flame spectrometry. Flame temperatures, noise levels, and signal-to-noise ratios in the two flames are measured. The Ar/02/C2H2 flame is highly versatile in that the flame temperature and composition can be varied considerably by variation of the Arlo2 ratio. As a result, in the Ar/02/C2H2 flame, easily atomized elements produce greater fluorescence signals in a cooler flame primarily due to a reduction in quenchers and difficultly atomized elements produce greater fluorescence in hotter flames primarily due to an increased atomization fraction. Although the Ar/02/C2H2 flame is in all cases equivalent to or better than (in terms of detection limits and linearity of analytical growth curves), it is a more costly flame and slightly more hazardous than the air/C2H2 flame.
For a flame to be suitable as an atom reservoir in atomic fluorescence spectrometry (AFS), there are a number of requirements which should be met.
(i) As in other flame spectrometric techniques, the flame should both vaporize and atomize the analyte efficiently, which necessitates a high temperature and reducing atmosphere. This aspect is particularly important for atomic fluorescence with a continuum source of excitation, because any particulate material in the flame will give rise to scattered source radiation, which will be present across the whole monochromator bandpass (as opposed to only across the line width when a line source of excitation is used). (ii) The flame should have a low background emission because the limit of detection obtainable by AFS is highly dependent upon the flame background noise. (iii) There should be a low concentration of quenching species in the flame (assuming the use of a non-laser source) in order to optimize the fluorescence yield. The extensive use of hydrogen-based flames in AFS ( I , 2) has been largely based on items (ii) and (iii) above, but the cool temperatures (typically