Wavelength-modulated continuum source atomic ... - ACS Publications

Managing Editor: Josephine M. Petruzzl. Associate Editor: Andrew A. Husovsky ... goshes, Harry B, Mark, Jr., J. W. Mitchell,. Harry L. Pardue, Garry A...
0 downloads 0 Views 861KB Size
analytical chemistry Editor: H e r b e r t A. L a i t i n e n EDITORIAL HEADQUARTERS 1155 Sixteenth St., N.W. Washington, D.C. 20036 Phone: 202-872-4570 Teletype: 710-8220151 Managing Editor: Josephine M. Petrurzi Associate Editor: Andrew A. Husovsky Associate Editor. Easton: Elizabeth R. Rufe Assistant Editors: Barbara Cassatt, Nancy J Oddenino, Deborah C. Stewart Production Manager: Leroy L. Corcoran Art Director: John V Sinnett Designer: Alan Kahan Artist: Diane Reich Advisory Board: Donald H. Anderson, Peter Carr, Velmer Fassel, David Firestone, Kurt F. J. Heinrich, Philip F. Kane. Barry L.Karger, J. Jack Kirkland, Lynn L. Lewis. Marvin Margoshes, Harry B. Mark, Jr., J. W. Mitchell, Harry L. Pardue, Garry A. Rechnitz. W. D. Shuits lnsirumentaiion Advisory Panel: Gary D. Christian, Catherine Fenselau, Nathan Gochman, Gary M. Hieftje. Gary Horlick. Peter J. Kissinger, James N. Little, C. David Miller, Sidney L. Phillips. Contributing Editor. Claude A Lucchesi Department of Chemistry, Northwestern University, Evanston, lli. 60201 Published by the AMERICAN CHEMICAL SOCIETY 1155 16th Street, N W. Washington, D C. 20036

Books and Journals Division Director: D. H. Michael Bowen Editorial Charles R Bertsch Magazine and Production: Bacii Guiley Research and Development: Seldon W. Terrant Circulation Develooment: Marion Gurfein

The Toy Theory of Modern Analysis In an article entitled “The Toy Theory of Western History”,

[CHEMTECH., 7,595 (1977)], M. E. D. Koenig described the evolution of military weapons in terms of the desire for man to play with toys of an ever-increasing power and complexity. Even though the article has nothing to do with analytical chemistry, the concept of gadgetry seeming to feed upon itself to reach truly awesome proportions of complexity must on occasion haunt laboratory managers and research directors. This thought evidently was the main justification for publishing the article, for in the table of contents checklist was the query “Will that double overhead, quadriphonic, Fourier transform, gazzeloping hustang really solve The problem or do you just want to play with it?” Looking a t the past several decades of analytical instrumentation, one gets the impression that the trend toward complexity is not a linear but an exponential function of time. It is not difficult to recall examples of compounded intricacy that seem to support the idea of the toy theory. By and large, however, there is a close coupling between the complexity of equipment and the degree of sophistication of the information obtained. There seems to be no end to the demand for more and more detail of analytical information when it becomes possible to attain it. This is true both for complex systems in which components are being detected and measured a t higher and higher sensitivity and resolution, and also for relatively simple systems in which nontraditional analytical information is being sought higher and higher levels of detail and complexity. Up to a point, this is all to the good, for it stimulates analytical research which in turn pays dividends in useful output. The question a t some point arises as to whether this more detailed information is worth the investment in instrumentation and personnel. There is no absolute answer to this question. Rather, each specific situation needs to be examined in terms of the problems that need solution. If a problem can be adequately solved by a test tube observation, there is no justification for elaborate instrumental measurements; but if the problem involves intricate information about complex systems, there is no alternative to a correspondingly sophisticated measurement. We all need occasionally to sidestep the temptation to play with our toys unnecessarily, and to get on with the task of solving the problem.

Manuscript requirements are published in the January 1978 issue, page 189. Manuscripts for publication (4 copies) should be submitted to ANALYTICAL CHEMiSTRY at the ACS Washington address. The American Chemicai Society and its editors assume no responsibility for the statements and opinions advanced by contributors Views expressed in the editoriais are those of the editors and do not necessarily represent the official position of the American Chemical Society.

ANALYTICAL CHEMISTRY, VOL. 50, NO. 3, MARCH 1978

385

Wavelength-Modulated Continuum Source Atomic Fluorescence Spectrometer F. Lipari and F. W. Plankey" Department of Chemistry, University of Pittsburgh, Pittsburgh, Pennsylvania 15260

is employed, the ideal flame should be one of low background emission within the spectral response range of the detector. As a consequence, interference filters and solar blind photomultiplier tubes have been used to obtain increased collection of source radiation and increased spectral throughput (Jacquinot's advantage) (8-10). Vickers (11) in a recent review has shown that there are some drawbacks to the use of a nondispersive system. Among these include the relatively low transmission of interference filters below 300 nm and the relatively low sensitivity of the R166 solar-blind photomultiplier tube compared to the more commonly used 1P28 photomultiplier tube. Also, since most nondispersive systems are flicker noise limited ( 1 1 ) due to flame background emission or fluorescence (12),the S / N ratio may decrease in comparison to dispersive systems for elements This paper describes a simple atomic fluorescence specwhich have resonance lines in a region of high flame backtrometer employing a continuum source, interference filters, ground (e.g., Fe). At extremely low wavelengths (near 200 nm) and wavelength modulation as the method of reducing scatter, where both flame background and monochromator efficiency spectral, and flame background interferences. The appliare low, the S/N ratio may increase for a nondispersive system. cability to a real analysis of Cu and Mg levels in blood serum However, photomultiplier tube dark current (detector shot noise) may be the limiting factor in these regions. Scatter of is also described. Wavelength modulation involves the rapid scanning of a incident radiation can pose a serious problem to a nondissmall wavelength interval a t some modulation frequency (Io) persive system but the effect can be minimized through the while the subsequent photodetector signal is then measured use of an efficient atomization system such as an air-acetylene flame and appropriate matrix-matched calibrating solutions at two times the modulation frequency (2f0) with a phase and frequency selective amplifier (i.e., a lock-in-amplifier). (13). This type of modulation ordinarily has been accomplished A wavelength modulated nondispersive atomic fluorescence in atomic spectrometry by vibrating a quartz refractor plate system with a continuum source can offer some distinct just before the exit slit of a monochromator (1-3). The advantages for practical analyses. The feasibility of using only one source instead of several line sources and of using inmodulation has proved effective in reducing low frequency terference filters instead of a monochromator can result in additive noises, spectral interference, scatter, and broadband considerable savings in analysis time and cost. Also, the molecular absorption which has resulted in improved detection limits for flame emission ( I ) and continuum source atomic relative freedom from physical interferences of wavelength absorption spectrometry (AAS) ( 3 , 4 ) . Double modulation (Le., modulation can eliminate the need for carefully matched source and wavelength modulation) has also been applied to matrices and calibrating or blank solutions. Wavelength Modulation Using Interference Filters. continuum source atomic fluorescence spectrometry with fair The basic operation of a wavelength modulated filter system success owing to the elimination of analyte emission by source depends on the fact that the central or peak wavelength of modulation and the elimination of scatter and thermal flame transmission of an interference filter shifts to shorter emission by wavelength modulation ( 5 ) . More recently, wavelengths as the angle of incidence of radiation changes wavelength modulation in continuum source AAS with an from the normal angle. Ordinarily, the central wavelength echelle monochromator has resulted in detection limits better of an interference filter is specified for collimated radiation than or comparable to the best detection limits by AAS with normal to the filter surface. If the incident angle changes, a line source ( 4 ) . the filter's central wavelength is shifted to shorter wavelengths, Wavelength modulation can also be accomplished with an and in effect a certain wavelength region is scanned by varying oscillating or vibrating interference filter. The generally low the angle of incidence. The variation in angle of incidence cost and greater spectral throughout (Jacquinot's advantage) can be achieved by oscillating or rotating the filter about a of interference filters compared to monochromators are two vertical axis a t a fixed frequency, fa. Interference filters can potential advantages for their use in atomic fluorescence be purchased with their central wavelength coincident with spectrometry (AFS). Hieftje (6) has recently designed and an atomic resonance line. Therefore, if the filters are made demonstrated the use of a wavelength modulated filter t o oscillate equally on both sides of normal, the analyte line photometer in correcting for interferences and applied it to will be coincident with the bandpass maximum or central a real analysis. wavelength twice for each cyclic scan of the filter. The T h e potential advantages of nondispersive atomic modulated photodetector signal due to the analyte can be fluorescence spectrometry have been discussed by Vickers (7). measured a t two times the modulation frequency (2f0) with These advantages include increased spectral throughput and a lock-in amplifier. The extent of modulation can be chosen simultaneous collection of multiple fluorescence lines with such that the analyte line will be shifted out of the bandpass simple and rugged instrumentation. Since no monochromator The wavelength-modulation capability of an oscillating narrow bandpass interference filter has been combined with the versatility of an atomic fluorescence flame spectrometer with continuum source. The operating characteristics of this device are described for solutions of Cu and Mg in the presence of interfering elements Na, Li, and Pb. Detection limits are within an order of magnitude of other flame spectroscopic techniques and the background correction feature of wavelength modulation is effective in reducing spectral interferences. The system was applied, with good results, to a pooled blood serum sample. Because of the large throughput of the filter, the major noise was due to flame background flicker.

0003-2700/78/0350-0386$01.00/0

D

1978 American Chemical Society

ANALYTICAL CHEMISTRY, VOL. 50, NO. 3, MARCH 1978

387

Table I. Experimental Apparatus and Manufacturers Description

Component Xenon Illuminator Illuminator Power Supply

Manufacturer

Model VIX-150 150 W, collimated, high pressure short-arc, xenon illuminator Model P-150s-72

Photomultiplier

Perkin-Elmer Model 303 premix burnerlnebulizer equipped with (a laboratory constructed) 1.1-cm circular capillary burner head E.M.I. Model 9738B

Photomultiplier Power Supply

Keithley 240 Regulated High Voltage Supply

Current Amplifier

Keithley Model 427 Current Amplifier

Lock-In Amplifier

Keithley Model 8 4 0 Autoloc Amplifier

Filter Modulator System Interference filters Torque motor

Laboratory constructed Cu and Mg atomic absorption line interference filters Model S4-075B

Oscilloscope

Tektronix Model T935 35 MHz Oscilloscope

Strip Chart Recorder

Omniscribe Model A5123-51

Function generator

Heath-Schumberger Model EU-81A

Burner

Li I G ~ HnT p

CCL L l M l T l N G

-ENS

Varian, Eimac Div., San Carlos, Calif. 94070 Varian, Eimac Div., San Carlos, Calif. 94070 Perkin-Elmer Corp., Norwalk, Conn. 06856 E.M. I.-Gencon Inc., Plainview, N.Y. 11803 Keithley Instruments, Inc., Cleveland, Ohio 44139 Keithley Instruments, Inc., Cleveland, Ohio 44139 Keithley Instruments Inc., Cleveland, Ohio 44139 Corion Corp., Holliston, Mass. 01746 MFE Corp., Salem, N.H. 03079 Tektronix, Inc., P.O. Box 500, Beaverton, Ore. 97077 Houston Instrument, Austin, Texas 78753 Heath Company Benton Harbor, Mich. 49022

Table 11. Experimental Conditions Xenon lamp power supply 1 2 A Air 9.7 L/min Flame conditions C,H, 1 . 5 L/min Argon 15.5 L/min 4.0 mL/min Aspiration rate 1 6 mm above burner top Viewing height in flame Photomultiplier voltage -725 V 1O 6 gain (volts per ampere) Current amplifier 0.01-ms risetime sensitivity as required Lock-in amplifier 3-s time constant Modulation frequency ( f , ) 40 Hz

-

Figure 1. Block diagram of experimental apparatus

of the filter at the extreme end of the filter oscillation cycle. T h e effect of rotating an interference filter causes its effective peak transmittance to be reduced and a t high modulation amplitudes considerable bandpass distortion results. Therefore, continuum spectral features (i.e., flame background) are also modulated at two times the modulation frequency, 2f0.

EXPERIMENTAL A block diagram of the experimental arrangement which was used is given in Figure 1, along with a summary of the major instrumental components and manufacturers (Table I). The heart of the experimental system is the filter modulator system which consists of a 23/4-inchcubic block aluminum enclosure. The interference filters 3re mounted onto custom made machined aluminum holders which in turn are mounted onto the shaft of a limited rotation torque motor whose shaft protrudes through the base of the system. Incident radiation is limited to a '/,-inch circular entrance port on the front of the system. The inside of the modulator system contains appropriate baffles and light traps to minimize stray radiation. The photomultiplier tube (PMT) is mounted on the rear of the system with a '/4-inch circular aperture concentric with the entrance port. The top of the modulator system contains a demountable plate or shutter which can be lowered in front of the PMT when removing or replacing interference filters to avoid shutdown of the PMT power supply. The operation of the modulator system was as follows: A Perkin-Elmer model 303 burner/nebulizer was fitted with a laboratory constructed circular capillary-type burner and

provisions for inert gas sheathing. Source radiation from a 150-W Eimac lamp was focused by a 2-inch diameter 2l/,-inch focal length lens to about a '/4-inch spot onto the top of the burner. Fluorescent radiation was then collimated by a 1-inch diameter, 2-inch focal length lens onto the entrance port of the modulator system. The filter was made to oscillate an equal distance on both sides of the radiation normal by applying a 40-Hz sine wave from the function generator to the torque motor amplifier control. The amplitude of rotation was controlled by varying the amplitude of the sine wave voltage or by adjusting the gain on the torque motor amplifier control. The rest position of the filter was determined by turning the position control on the torque motor amplifier control until an output of zero volts was observed on the oscilloscope. The amplitude of oscillation was chosen so that the analyte line which was almost coincident with the central wavelength of the filter would be shifted out of the bandpass of the filter. The subsequent photosignal was fed into a current amplifier and then to a lock-in amplifier tuned to twice the frequency of oscillation. The photosignal from the current amplifier was also monitored on the oscilloscope t o ensure that a maximum 2f0 component of the signal was observed a t the modulation amplitude chosen. A maximum 2f0 component of the signal was observed by shining light from Cu and Mg hollow cathode lamps directly onto the respective filters and varying the amplitude of oscillation so as to obtain a maximum 2f0component. Once the proper modulation amplitude was chosen, the air, acetylene, argon, and aspiration flow rates were adjusted along with the viewing height in the flame so as to give a maximum S / N at the modulation frequency. For experimental conditions, see Table 11. The filter transmission characteristics (Table 111) were determined in another series of experiments in which light from an

388

ANALYTICAL CHEMISTRY, VOL. 50, NO. 3, MARCH 1978

Table 111. Interference Filter Transmission Characteristics

Element cu Mg

Transmission Central at normal wavelength, incidence, nm %a 324.5 285.2

13 10

Bandpass, nm 2.15 2.34

a Measured with Shimadzu Double-beam Spectrophotometer UV-200(Shimadzu Seisakusho Ltd., Japan).

Eimac lamp was directed onto the surface of the interference filters placed in front of the entrance slit of a GCA/McPherson EU-700 scanning monochromator. The effects of bandpass, transmission, and central wavelength changes were determined as a function of incidence angle of the light. A calibration plot of modulation amplitude with output voltage of the sine wave generator was made and therefore the exact position of the analyte line with respect to possible interfering lines and with respect to its position in the filter's bandpass was known as a function of modulation amplitude. The dc mode of operation of the system was also investigated. The dc method involved only locating the normal position of the filter, turning off the sine wave generator, and disconnecting the current amplifier to lock-in amplifier input. Reagents and Solution Preparation. Certified (lo00pgjmL) copper, magnesium, and lead atomic absorption standards were purchased from Fisher Scientific Co., Pittsburgh, Pa. 15219. Pooled blood serum samples were generously supplied by Presbyterian-University Hospitai of Pittsburgh, Pa. 15261. Standard Li, 1000 pg/mL and standard NaC1, 2500 Fg/mL, solutions were prepared by dissolving the appropriate amounts of dried Li2C03 (J. T. Baker Chemical Co.) and NaCl (Fisher Scientific) salts in HC1 and water respectively and diluting to 1 L with deionized water. Standard solutions for the analytical curves were prepared by serial dilutions of the copper and magnesium stock solutions. Blood serum solutions were prepared by a 1 to 5 dilution of serum with deionized water in a %dram vial for the copper analyses and a 1 to 100 dilution of serum with deionized water for the magnesium analyses. For the copper and magnesium analyses of blood serum by the standard additions method, known quantities of Cu and Mg standards were added to the serum and then diluted accordingly with deionized water. Waveform Figure Preparation Flame background, analyte, and interferent waveform figures were prepared with the aid of a PDP 11/10 computer (Digital Equipment Corp.) interfaced to the filter modulator system. The subsequent current amplifier signals were signal averaged by the computer through the use of SPARTA (Signal Processing and Real Time Analysis) Software supplied by Digital Equipment Corp. and the averaged signal displayed on a storage oscilloscope (Tektronix Model 603, Tektronix Inc.) Fourier transform spectra were also prepared with the SPARTA software.

RESULTS AND DISCUSSION Correction f o r Scatter a n d F l a m e B a c k g r o u n d Interferences. In order to investigate the utility of this system, solutions of varying concentrations of Cu and Mg were made up in 1000 pg/mL Na. T h e solutions were run in the wavelength modulation (WM) mode with the subsequent results compared to those obtained by their analyses via atomic absorption spectrometry (AAS). The modulation amplitude was chosen as described earlier. Modulation amplitudes of approximately 9.2' and 13' were chosen for Cu and Mg, respectively. T h e phase control of the lock-in amplifier was set with a standard solution of Cu or Mg, respectively. The effect of 1000 pg/mL Na does not change the slope of the analytical curve but only the intercept of the curve (additive interference) and can usually be corrected for by a suitable blank correction. The signal due to the 1000 pg/mL N a solution is probably stray light for it shows up as a positive

Figure 2. Signal averaged waveforms of current amplifier photocurrent during wavelength modulation with the Mg interference filter. (A) Modulated signal due to 1 pg/mL Mg being aspirated. (B) Sine wave driving waveform from the function generator to t h e torque motor amplifier control. (C) Modulated signal due to flame background with deionized water being aspirated

Figure 3. Signal averaged waveforms utilizing the Mg interference filter for (A) Flame background signal due to 1000 yg/mL Na being aspirated. (B) Sine wave driving waveform. (C) 1 pg/mL Mg solution being aspirated. Sensitivity 200-fold less than A

signal on the recorder and not as a negative signal as would be observed for an interfering line. As stated earlier, rotation of the filter causes the overall integrated transmittance of the filter to decrease, thereby modulating any continuous spectral features a t two times the modulating frequency as shown in Figure 2. A 1000 kg/mL N a solution (Figure 3) produces a small 2fo component in phase with the 2f0 component of Mg and therefore a 1000 pg/mL Na signal shows up as a positive deflection on the lock-in. However, the magnitude of the 1000 pg/mL Na stray light signal corresponds to an equivalent signal produced by 0.05 pg/mL and 0.1 pg/mL Mg and Cu solutions, respectively. The apparent sensitivity of the analytical curve is not affected by Na and its effect can be minimized by a suitable blank correction. However, even with a blank correction, a small but noticeable intercept occurs and thus may affect the accuracy of determinations near the detection limit. Background does not appear to be a major problem a t concentration levels removed from the detection limit. As pointed out by Vickers (111, flame background emission or fluorescence can be a serious limitation in nondispersive AFS. A sloping or changing background in the modulation interval will produce both fo and 2f0 signal components. This effect can be attributed to the derivative nature of the modulation which senses changes in intensity with wavelength such as that near an absorption line or curving background ( 2 4 ) . Normally, the modulation interval with wavelength modulation by oscillating refractor plates was chosen so that the modulation interval was approximately equal t o or two times the spectral slit width of the monochromator ( 4 , 15). Consequently, the modulation interval, Ah, was on the order of a few angstroms. The background over such a small region was essentially constant and did not contribute to the signal. A wavelength modulation interval with interference filters corresponds to a much larger modulation interval due to the much larger bandpass of a n interference filter compared t o the spectral slit width of a monochromator. Therefore, if the background changes over that interval, then a small 2fo

ANALYTICAL CHEMISTRY, VOL. 50, NO. 3, MARCH 1978

cc

,, 270

280

290

3 00

Wavelength (nrn)

389

Figure 5. Schematic diagram illustrating the effect of Pb and Mg on the resulting signal waveform at low modulation amplitudes (see text for explanation)

Selected portion of flame background molecular fluorescence spectrum obtained in the dc mode with a 150-W Eimac lamp, a GCA/McPearson EU-700 scanning monochromator, and stoichiometric air/C,H, flame Figure 4.

component will result. An examination of the flame background of an air-acetylene flame (Figure 4), obtained in the dc mode with a scanning monochromator reveals the presence of a small slope or change in intensity with wavelength. Therefore, a t reasonably high modulation amplitudes a small 2f0 component due to flame background can be seen (Figure 2). Also, as described earlier, the effective change in transmittance and bandpass shape of the filter induces a small 2f0component. Similarly, unequal tilting of the filter can lead to non-uniform background modulation and can result in a small fundamental (fo) signal as in Figure 2. It appears that because of the increased spectral throughput of interference filters as compared to monochromators, flame flicker is the major source of noise in the present system. This fact along with the various other factors just described ultimately affects the S / N ratio and detection limit for most elements. Correction for Spectral Interferences. Correction for spectral interferences can normally be accomplished by the use of the zero-crossing technique introduced by Epstein and O'Haver (15). The technique is based upon the fact that the second harmonic response function of a wavelength modulated system resembles a second derivative curve with two zero crossings (Le., points where the intensity goes through zero) on either side of the central maximum (16). The assumption that the wavelength positions of the inflection points (zero crossings) are independent of interfering species concentration is valid up to a point where a t high concentrations self-absorption broadening of the analyte line approaches the instrumental slit function. However, with the present system, the bandpass of the filter is so large compared to the absorption line width t h a t self-absorption broadening is not a limiting factor. The unique aspect of wavelength modulation is that the positions of the zero crossings are adjustable parameters and not a property of the analyte or interferent. The adjustment is usually made by aspirating a concentrated solution of the interferent and adjusting the modulation interval so that the point of zero crossing (or zero intensity) of the interfering line lies under the analyte maximum or 90" out of phase with the analyte, thereby resulting in a zero output from the lock-in amplifier. In a similar manner, it has been shown by Hieftje and Sydor (6) that wavelength modulation by interference filters results in a modulation pattern for the interfering line that differs from the analyte on the basis of phase and frequency. In order to test the effectiveness of our system from spectral interferences, the effect of the Li line a t 323.3 nm on the Cu 324.7 nm resonance line and the effect of the P b line a t 283.3 nm on the Mg 285.2 nm resonance line were investigated. Both of these interfering lines fell within the bandpass of their respective filters. Solutions of Cu and Mg were made up with

Figure 6. Signal averaged waveforms utilizing the Mg interference filter for (A) 1 pg/mL Mg solution being aspirated. (El) Sine wave driving waveform. (C) 200 pg/mL Pb solution being aspirated. Sensitivity is the same as in C as in A

varying concentrations of Li and Pb, respectively, and investigated in both the dc fluorescence and wavelength modulation modes. Ideally, the modulation interval should be chosen so that the interferent line would be shifted out of the filter bandpass and at the same time the analyte line will be a maximum (zero-crossing techniques). However, this was impossible in our case, since the central wavelength maximum of the filter was approximately coincident with the analyte line and both interfering lines lie at shorter wavelengths than the analyte line. Upon rotation, the shift to shorter wavelengths of the filter wavelength maximum would cause the analyte line to be out of the bandpass and the interferent line to be at a maximum. Therefore, it was decided to adjust the modulation interval such that both analyte and interferent lines were shifted out of the bandpass and discrimination could be achieved on the basis of the differing phase and frequency composition of the interferent peaks. The interference of P b on Mg was investigated by comparing the lock-in amplifier signals from 0.5 pg/mL Mg and 0.5 pg/mL Mg with 50 pg/mL P b added as a function of modulation amplitude. A t low modulation amplitudes, a 50 pg/mL P b solution gave a negative lock-in signal, indicative of a 180' phase shift with respect to the Mg signal. The signal due to 0.5 pg/mL Mg with 50 pg/mL P b was much smaller than the 0.5 pg/mL Mg signal. However, the sums of the 0.5 pg/mL Mg plus 50 pg P b and the 50 pg/mL P b signals were approximately equal to the amplitude of the 0.5 pg/mL Mg signal. At higher modulation amplitudes, the P b signal became progressively smaller and the amplitude of the 0.5 pg/mL Mg plus 50 pg/mL P b signal approached that of the 0.5 pg/mL Mg signal. At a modulation amplitude of about 21.4', 50 pg/mL P b produces about a 2.5% interference (or signal reduction) on 0.5 pg/mL Mg. These results are shown in Figure 5 and Figure 6. At higher modulation amplitude (-15') such that the Mg lines shifts out of the bandpass and the P b line remains in the bandpass but with greatly reduced intensity (Figure 7), the Mg line again appears at 2 f 0 but the P b line now begins to show noticeable 4f0 components superimposed on its main 21, waveform. This is because at that modulation amplitude

390

ANALYTICAL CHEMISTRY, VOL. 50, NO. 3, MARCH 1978

Relative Signal

Figure 7. Schematic diagram illustrating the effect of Pb and Mg on the resulting signal waveform at high modulation amplitudes ( - 15')

(see text for explanation)

Figure 8. Signal averaged waveforms utilizing the Mg interference filter at high modulation amplitudes ( - 15') for (A) 1 pglmL Mg solution being aspirated. (B) Sine wave driving waveform. (C) 200 pg/mL Pb solution being aspirated. Sensitivity is the same in C as in A

R e l a t i v e Signal

-

-

:

0I

Pb

Figure 9. Schematic diagram illustrating the effect of Pb and Mg on the resulting signal waveform at very high modulation amplitudes (-21.4') (see text for explanation)

the Mg line shifts out of the filter bandpass and appears at essentially zero intensity (Figure 8) while the P b line remains in the bandpass and appears at a finite but somewhat reduced intensity than when P b is at maximum intensity. At normal incidence, the Mg line is at maximum intensity while the P b line is a t minimum intensity (but not zero). Upon rotation, the P b line approaches a maximum and Mg approaches a minimum, further rotation causes the P b line intensity to be reduced slightly and the Mg line intensity to be essentially zero and therefore, a 4f0 component appears for Pb. At much higher modulation amplitudes (-21.4') both the Mg and P b lines are shifted out of the bandpass (Figure 9). As a consequence P b appears essentially at 4f0 and Mg a t 2f0 (Figure 10). From Figure 10, it is evident that when Mg is at a maximum, P b is a t a minimum and when Mg is shifted out of the bandpass (zero intensity), P b also appears with essentially zero intensity. The broadened base of the Mg 2f0 components is indicative of the fact that the Mg line stays out of the bandpass for a much longer time period than the P b line. Since P b appears at 4f0 and Mg at 2f0, discrimination by the lock-in can be achieved on the basis of frequency of the interferent peaks. The fact that the P h interference is not reduced to zero at this modulation amplitude is probably due to the extensive modulation of the flame background at such large amplitudes that distortion of the bandpass shape may result inducing 2f0 signal components (180' out of phase

Figure 10. Signal averaged waveform utilizing the Mg interference filter at very high modulation amplitudes (- 21.4') for (A) 1 pg/mL Mg solution being aspirated. (E?) Sine wave driving waveform. (C) 200 pglmL Pb being aspirated. Sensitivity is the same in C as in A

with respect to Mg) due to flame background (Figure 2). As a consequence, zero suppression is needed for the Mg analysis. The Fourier transform power spectrum of intensity vs. frequency for 200 pg/mL P b a t a modulation amplitude of 21.4' shows a rather large 4f0 component due to P b and a much smaller 2f, component. A similar investigation for the Li interference on the Cu signal was also conducted and it was found that at a modulation amplitude of approximately 14.5', 100 pg/mL I,i produces about a 5% interference on 0.1 pg/mL Cu. The somewhat higher interference due to Li is due to the fact that 100 pg/mL Li produces a fairly intense emission signal which is modulated by the filter resulting in a negative deflection (180' quadrature) of the lock-in amplifier. In the dc mode of operation, both the Li and P b effect on Cu and Mg, respectively, were more pronounced along with a great deal more interference from scatter due to Na and flame background. It is apparent that wavelength modulation can correct for spectral interferences. However, the magnitude of the correction is based on the differing frequency composition of the interferent peaks with respect to the analyte peaks and thus dependent on the modulation amplitude chosen. Therefore, very narrow bandpass interference filt,ers should he used to limit the bandpass distortion that results at high modulation amplitudes and help eliminate foreign lines that might fall in the filter bandpass. It should be cautioned, however, that high modulation amplitudes increase the likelihood of foreign lines falling in the filter bandpass and bandpass distortion which induce 2f0 signal components due to flame background modulation; both interferences limit the magnitude of the correction. Linearity and Detection Limits. The extent of linearity for Cu and Mg by WMAF'C is about the same as that obtained by AAS. Cu appears to be linear up to 7.5 gg/mL and Mg up to 1.0 pg/mL. The value for Mg appears a little low; however, this may be due to appreciable self-absorption of fluorescence and the range may be extended by focusing the source radiation onto the portion of the burner nearest the detector. The detection limits are within an order of magnitude of the best reported detection limits by AAL, AFL, AFC (Table IV) and comparable to those obtained by high resolution wavelength AAC with an echelle monochromator. The average S / N for a particular Cu concentration was found to be relatively independent of modulation amplitude, probably owing to the essentially constant flame and source background in that region (12). The average S / N for Mg, however, was found to increase with increasing modulation amplitude to a maximum near an amplitude of 15' and remain essentially constant at higher amplitude. This effect can he explained by the fact that the Mg line was shifted completely out of bandpass at that modulation amplitude. However, noticeable 2f0 components of the flame background are present at high modulation amplitude (Figure 2).

A N A L Y T I C A L CHEMISTRY, VOL. 50, NO. 3, MARCH 1978

*

391

Table IV. Comparative Detection Limitsapb for Cu and Mg Element

WM AFCC

DC AFCd

cu

0.02 0.0025

0.08 0.01

Mg

AALe

AAC~ 0.018' 0.002'

0.004' 0.0003'

AFLg 0.001' 0.0003'

AFC~ 0.0015' 0.0003'

Detection limits are defined as that concentration which gives a signal 2 x rms noise level. &I Detection limits are in WMAFC denotes wavelength modulation atomic fluorescence flame spectrometry with a continuum source (prepg/mL. sent work). DCAFC denotes dc atomic fluorescence flame spectrometry with the radiation normal to the filter surface with the dc mode of detection and no modulation (present work). e AAL = atomic absorption flame spectrometry with a line source. f AAC = atomic absorption flame spectrometry with a continuum source and a wavelength modulated echelle monochromator. g AFL = atomic fluorescence flame spectrometry with a line source. AFC = atomic fluorescence flame spectrometry with a continuum source. Reference 17. Reference 18. I Reference 4. a

'

Table V. Comparitive Determinations of Cu and Mg Levels in Pooled Blood Serum in pg/mL . WMAFC Ele- standard ment addtions Cu Mg

1.28 23.5

WMAFC anaAALa lytical standard curve additions 1.34 25.1

1.30 22.3

AFC (19) 1.31 i 0.35 24.6 i 6.1

a = Perkin-Elmer Model 305 Atomic Absorption Spectrophotometer equipped with Cu and Mg Hollow Cathode Lamps.

The detection limits for the dc mode of operation are about an order of magnitude worse than the WMAFC. Flame background, scatter, and spectral interferences are more pronounced in the dc mode. In order to ascertain the relative accuracy of our system in a real analysis, a sample of pooled blood serum was analyzed for Cu and Mg levels (Table V). Human blood serum affords a fairly complex background matrix to test the effectiveness of our system. The analysis was performed utilizing both the standard additions and analytical curve methods with the results compared to the results obtained by the standard additions method and atomic absorption spectrometry. The results obtained by the WMAFC standard addition and AAS standard additions differ by about 1.5% and 5.5% for Cu and Mg, respectively. The values obtained by the WMAFC analytical curve and AAS differ by about 3.1 % and 5.1 % for Cu and Mg, respectively. The values obtained by WMAFC and AAS standard addition methods are in complete agreement with the mean values obtained for Cu and Mg serum levels in our laboratory (19). The values obtained by the WMAFC analytical curve method and WMAFC standard additions method differ by about 4.7% and 6.7% for Cu and Mg, respectively, and are within the experimental error of the technique. This is probably due to the difference between the water standards and the serum matrix and can be eliminated by dilution of both serum and standards with a 10% glycerol solution (19). As can be seen by the accuracy of our results, the WMAFC approach does not necessitate the use of complex blank solutions or matched matrices, and accurate results can be obtained with the analytical curve method and simple water standards.

proved effective in reducing scatter, flame background, and spectral interference. It has resulted in reasonably good detection limits and accurate analyses for Cu and Mg levels in such complicated matrices as human blood serum without the need for complicated blank solutions. The fact that the wavelength modulation can be accomplished with interference filters is important for it allows the filter to act as both the spectral isolation device and the background correction device eliminating the need for a monochromator. Also, since a continuum source can be used instead of several line sources, it allows for a low cost, simple, dedicated system to be built for several elements. As a consequence of the high spectral throughput of interference filters, flame flicker appears to be the major noise source and an atomization system of low background, e.g., a graphite furnace cell, may prove most effective. Also, the use of very narrow bandpass interference filters is recommended to reduce any modulated 2f0 background signal components and eliminate the likelihood of foreign spectral lines falling in the filter bandpass. The system is currently being improved and its applicability to the analysis of several trace elements in diverse matrices with a graphite atomizer is being investigated.

LITERATURE CITED (1) W Sneileman, Spectrochim Acta, Part 8,23, 403 (1968) (2) W Snelleman, T C Rains, K W Yee, H D Cook, and 0 Menis, Anal Chem., 42, 394 (1970). (3) R. C. Elser and J. D. Winefordner, Anal. Chem., 44, 698 (1972). (4) A. T. Zander, T. C. O'Haver, and P. N. Keliher, Anal. Chem.. 48, 1166 (1976). ( 5 ) W.K.'Fowler, D. 0. Knapp, and J. D. Winefordner, Anal. Chem., 46, 601 (1974). (6) R. J. Sydor and G. M. Hieftje, Anal. Chem., 48, 535 (1976). (7) T. J. Vickers, P. J. Slevin, V. I. Muscat, and L. T. Farias, Anal. Chem., 44, 930 (1972). ( 8 ) D. G. Mitchell and A. Johanasson. Snectrochim. Acta. Part 6. 25. 175

(12) (13) (14) (15) (16) (17) (18) (19)

Spectrosc., 29, 5: , W. K. Fowler and J. D. Winefordner. Anal. Chem., 49, 944 (1977). P. L. Larkins and J. 6.Willis, Spectrochim. Acta, Part6, 29, 319 (1974). R. N. Hager, Anal. Chem., 45, 1131A (1973). M. S.Epstein and T. C. O'Haver, Spectrochim. Acta, Part 6, 30, 135 11975). A. T. Zander. T. C. O'Haver, and P. N. Keliher, Anal. Chem., 49, 838 (1977). F. S.Chuang and J. D. Winefordner, Appl. Spectrosc., 29, 412 (1975). D. J. Johnson, F. W. Plankey, and J. D. Winefordner, Anal. Chem., 47, 1739 (1975). R. D. Dresser and F. W. Piankey, unpublished work, this laboratory.

CONCLUSION A wavelength modulated continuum source atomic fluorescence spectrometer shows considerable promise as a dedicated instrument for the analysis of several trace elements in diverse matrices. The wavelength modulation feature has

for review August 1, 1977. Accepted December 14, 1977. Acknowledgement is made to the Donors of The Petroleum Research Fund, administered by the American Chemical Society, for the support of this research. RECEIVED