Analyst October 1983 Vol. 108 pp. 1195-1208 1195 Discrimination Against Atomic-emission Spectral Interferences in Wavelength-modulated Continuum Source Excited Flame Atomic-fluorescence Spectrometry* John T. McCaffrey Man-Li Wang Wu and R. G. Michelt Defiartment of Chemistry University of Connecticut Storm CT 06268 USA The extent of spectral interferences at commonly used analytical atomic-fluorescence lines was demonstrated to be severe enough to necessitate the exploration of possible instrumental approaches to discriminate against the more serious type of spectral interference namely the atomic emission of concomitant metals in samples. Two novel methods of double modulation were incorporated into the instrumentation in order to allow such discrimina-tion. The two methods were evaluated with respect to their effect on detection limits and their effect on the particularly severe interference of potassium emission at 404.4 nm on lead fluorescence at 405.8 nm.Keywords A tomic-jluorescence spectrometry ; double modulation ; spectral interferences ; wavelength modulation ; continuum source Introduction Continuum source excited atomic-fluorescence spectrometry (AFC) has been proposed to offer many advantages in multi-element analysis.192 Amongst them are good detection limits in a system capable of determining quantitatively a large number of elements in samples using only one light source. Wavelength-modulated AFC permits the measurement of a combined emission and fluorescence signal hence permitting the determination of elements such as sodium and potassium which are more sensitively determined by atomic emission than by atomic fluorescence.The ability to measure both emission and fluorescence as combined signals has been found to be useful in the analysis of real samples. However atomic-emission measurements of any kind usually require the use of a narrow spectral band pass for the mono-chromator in order to discriminate against spectral interferences. In contrast fluorescence measurements particularly with line-source excitation exhibit few spectral interferences. Thus it is possible to use a large spectral band pass i.e. wide slit widths and a monochromator with a small f number in order to improve sensitivity by greater light throughput. The combination of emission and fluorescence in wavelength-modulated AFC must therefore, result in a number of spectral interferences unless the spectral band pass is made narrow.This would inevitably reduce sensitivities. Limited data presented in the literature2 and extensive data presented here demonstrate conclusively that the extent of spectral interferences that affect wavelength-modulated AFC is severe enough to necessitate the exploration of possible instrumental approaches to discriminating against such interferences. Instrumentation for AFC has been described el~ewhere.l-~ In such instrumentation the use of source intensity modulation alone allows for the discrimination against spectral inter-ferences caused by atomic emission of concomitant elements in the matrix but cannot dis-criminate against signals caused by the scatter of incident radiation from incompletely vaporised species in the flame.Conversely wavelength modulation can discriminate against ~ c a t t e r ~ - ~ but cannot discriminate against spectral interferences caused by atomic emission of concomitant elements. Neither approach will discriminate against spectral interferences caused by atomic fluorescence of concomitant elements. * Portions of this paper were presented at FACSS VIII Philadelphia PA on September 20 1981 and FACSS IX Philadelphia PA on September 20 1982. t To whom correspondence should be addressed 1196 MCCAFFREY et al. DISCRIMINATION AGAINST AES Analyst VoZ. 108 The combination of both source intensity modulation and wavelength modulation is called double modulation and discriminates against both scatter and spectral interferences caused by atomic emission of concomitant elements.Elser and Winefordners have described a double-modulation system that was used for both continuum source excited atomic absorption6 and atomic fl~orescence.~ The atomic-fluorescence instrument suffered from poor sensitivity. In this paper we describe two instrumental systems that achieve double modulation for AFC and give detection limits that closely approach those obtainable with the best source intensity or wavelength-modulated AFC instrumentation. The two novel methods of double modulation described here are called fixed-phase double modulation (FPDM) and dual lock-in double modulation (DLDM) and these were evaluated with respect to their effect on the particularly severe interference of potassium emission at 404.4 nm on lead fluorescence at 405.8 nm.Experimental Spectral Interference Studies The instrumentation used for the spectral interference studies of wavelength-modulated AFC was similar to that described pre~iously.~ The performances of the two instruments were identical in all respects. The instrumental conditions used for the interference studies are given in Table I. Wavelength modulation was accomplished with a rotating four-quadrant quartz chopper,3 also called a sectored wheel. The wheel was placed behind the middle slit of a double mono-chromator as described previ~usly.~ TABLE I GENERAL EXPERIMENTAL CONDITIONS Monochromator*-Entrancelexit slits . . . . 0.25 mm Middle slit . . .. 2.0mm Spectral band pass . . . . 0.5 nm Source*-Current . . 17A Power . . 300W Wavelength modulation frequency . . . . 30Hz Angle of incidence of light in monochromator Flame . . . . Nitrogen-separated air - acetylene operated as on wavelength modulation wheel . . 45" a stoicheiometric mixture * See Table I11 for further details. Table I1 lists the commonly used analytical atomic-fluorescence lines that were tested for spectral interferences together with those metals which have experimentally observed atomic-absorption -emission or -fluorescence lines (from reference 7) within 2 nm of the analytical fluorescence lines and which were tested as possible interferences. Lines within 2 nm were chosen because this allowed for interferences that were likely to occur on our instrument.This had a nominal spectral band pass of 0.5 nm that when using wavelength modulation, gave an effective spectral band pass of about 1.5 nm. Nominal band passes of about 0.5 nm are typical of some AFC instruments1s2 although not all such instrument^.^ Clearly those instruments with smaller band passes will not be as subject to interferences although in all probability they will not have as much light gathering power. Single element solutions of the analyte were used to obtain calibration graphs at the analyte wavelength. At the same wavelength calibration graphs for the interferents were obtained. An interference was deemed negligible if a 1000 pg ml-l interferent solution resulted in no detectable signal at the analyte wavelength.Double Modulation Studies The components common to both methods of double modulation are listed in Table 111 October 1983 INTERFERENCES IN CONTINUUM SOURCE EXCITED FLAME AFS TABLE I1 TESTED ANALYTICAL FLUORESCENCE LINES AND INTERFERENCES 1197 Fixed-phase Double Modulation Analyte Cd Ca Cr c o c u In In Fe Pb Mg . . Mn . . Analytical linelnm 228.80 422.67 357.87 369.35 240.73 242.49 252.14 324.75 327.40 303.94 410.18 451.13 248.33 283.31 405.78 285.21 279.48 279.83 280.11 Possible interferents Bi As Co Ni Sr Co Fe Mn Ni Cr Co Fe Mn Ni Pd Cr Ca Co Co Pt Sr Sn Co Fe Cd In Pd Cd In Ag Co Fe MG Ni Pt Sn Bi Co Cu. Pd. Fe Mg Pti T1 Sn K Sr ~ b Sn Bi Pb Mg Analyte Mn .. Ni Pd Pt Ag Na . . Sr . . T1 Sn ,. Zn Analytical linelnm 403.08 403.31 403.45 232.00 352.45 343.46 360.96 363.47 265.95 328.07 338.29 589.59 460.73 377.57 535.05 286.33 303.41 317.51 213.86 Possible inferents K Co Ni Bi Co Tl Pd Ni Co Fe Ni Pd Cr Co Mg Ni Pb Mg Ni Pt Cd Cu Bi Ni Mg Co In Fe Mn Ni Pt Pt Fixed-phase double modulation (FPDM) is a special case of the general method used by Elser and WinefordnerJ6 who used two modulation frequencies that had random frequency and phase relationships and required tuned filters to derive a reference signal in order to extract the fluorescence signal corrected for background scatter and structured emission.With FPDM two square-wave modulation frequencies which have the phase and frequency relationships shown in Fig. 1 allow the background-corrected fluorescence signal to be extracted with a conventional lock-in amplifier or a synchronous photon counter. FPDM depends on modulat-ing the source intensity at exactly double the frequency and 90" out of phase with the wave-length modulation. The required atomic-fluorescence signal is then detected at the wavelength modulation frequency but 90" out of phase with the wavelength modulation. For clarity the signal processing shown in Fig. 1 is in terms of synchronous detection of the type normally found in photon counters i.e. the signal plus background ( A ) and background (B) are routed TABLE I11 INSTRUMENTAL COMPONENTS COMMON TO BOTH DOUBLE MODULATION METHODS Component Type and/or description Source .. . . 300 W xenon arc (Model VIX300 Xenon arc power supply . . . . Model PS300-1 UV) Monochromator . Model DH-20A UV - visible 0.2 m double monochromator fl4.2 aperture Lenses . . . . Silica lenses 2 in diameter bi-convex focal length 50 mm Pre-mix burner chamber . . Burner head and nitrogen separator Wavelength modulation chopper . . Sectored wheel.3 30 Hz modula-tion frequency (rotating quartz chopper) Photomultiplier tube . . . . 9893QB/350 Supplier Varian Eimac Division, San Carlos CA Varian Eimac Division, San Carlos CA Instruments SA Inc., Metuchen N J Esco Products Oak Ridge, Perkin-Elmer Corp., Laboratory constructed NJ Norwalk CT Laboratory constructed EMI-Gencom Inc., Plainview NY Betram Assoc.Syosset, NY Photomultiplier power supply . . Model 21 1198 MCCAFFREY et al. DISCRIMINATION AGAINST AES Analyst Vol. 108 I I I S time I Source i intensity I modulation Fig. 1. Fixed-phase double modulation waveforms. FB = flame background emission; S = scatter of source radiation; AE = atomic emission from analyte; AE' = matrix emission a t a different wavelength to analyte A E ; AE" = matrix emission at same or similar wavelength to analyte A E ; and AF = wanted atomic fluorescence. The frequency relationships together with the 90" phase shifts place half of the various possible interfering signals (FB S AE AE' AE") into each of the background (B) and signal ( A ) channels.The subtraction (A - B) gives the wanted AF signal as follows ( A - B) = AF = (FB + FB + AE + AE'+ AE" + S + AF) - (FB + FB + AE + AE'+ AE" + S ) . into two separate counters. B is then subtracted from A to give the required signal. The discrimination against scatter and atomic emission is illustrated by the equation in the caption to Fig. 1. FPDM can be treated mathematically in a similar fashion to the treatment of Elser and WinefordneP and corresponds to their second-derivative difference mode of detection. Those additional components necessary for FPDM are listed in Table IV and a block diagram of the instrumentation is shown in Fig. 2. TABLE IV ADDITIONAL COMPONENTS NECESSARY FOR FIXED-PHASE DOUBLE MODULATION Component Type and/or description Source modulation motor .. . . Model BSMX 2038 115 VAC, 50-60 Hz polarised synchronous motor 1800 rev min-l Photon counter* . . . . Model 11 12 photon counter/ processor (in synchronous counting mode) Amplifier/discriminator . . . . Model 1120 Power amplifier to drive synchronous motor . . . . . . Two Model W5M 10 W audio amplifiers operated in parallel Supplier Elinco Inc. Norwalk CT Princeton Applied Research Princeton N J Princeton Applied Research Princeton N J Heathkit Benton Harbor MI Frequency doubler . . . . 30-60 Hz frequency doubler Laboratory constructed Digital phase shifters . . . . 60 and 30 Hz TTL phase shifters Laboratory constructed Chopper for source modulation . . Two-blade chopper Laboratory constructed * Occasionally the PAR lock-in amplifier (Table V) was used instead of the photon counter.FPDM required that source modulation occurred at exactly double the frequency of wave-length modulation. The synchronous motor was driven with a 60-Hz frequency doubled signal derived from the 30-H~ wavelength modulation reference signal. It was necessary to operate the system at these frequencies because the synchronous motor was designed only for 50-60-Hz operation. Work is progressing on replacing the synchronous motor with a high power bipolar stepper motor and this should allow FPDM over a much larger range of modula-tion frequencies October 1983 INTERFERENCES IN CONTINUUM SOURCE EXCITED FLAME AFS 1199 It can be seen from the phase relationships in Fig. 1 that FPDM requires accurate control of the phase of both wavelength modulation and source intensity modulation.This was achieved by separate control over the phase of the wavelength modulation reference signal and the phase of the source intensity modulation with respect to the wavelength modulation. The synchronous motor was of a polarised design to ensure that the source intensity chopper always synchronised in a particular position relative to the driving 60-HZ waveform. This was in an attempt to ensure that the phase relationships of the instrument would stay constant whenever the instrument was switched off and on again. ynchronous motor Double monochromator Fig. 2. Instrumental arrangement for fixed-phase double modulation. Experimental procedzcre for FPDM Firstly the correct wavelength was obtained by observing the output signal under wavelength modulation condi-tions.The reference signal was then phase shifted to place the detection phase 90" out of phase with the analytical signal by aspiration of an analyte solution and zeroing of the signal. Then the source modulation was switched on. The signal from the photomultiplier tube was then observed on the oscilloscope while phase shifting the source modulation signal with respect to the wavelength modulation signal until the waveform of Fig. 1 was obtained. Fine adjust-ment of the source-intensity modulation phase was possible by observing the output signal of the photon counter or lock-in amplifier because the proper phase corresponded to the maximum net signal when aspirating an analyte solution.The adjustment of the source modulation phase does not affect the wavelength modulation phase. An oscilloscope was necessary to obtain the desired phase relationships. Dual Lock-in Double Modulation Dual lock-in double modulation (DLDM) is similar to the method of Koizumi et aZ.* who used polarisation modulation of the light source (Zeeman effect background correction) combined with source intensity modulation in order to achieve background correction for graphite furnace atomic-absorption spectrometry. They used two widely separated modulation fre-quencies and discriminated between the two frequencies by the use of a filter tuned to the source intensity modulation frequency. The output of the tuned filter was then fed to a lock-in amplifier tuned to the Zeeman modulation frequency.The band pass filter discriminated against all signals not source intensity modulated and therefore did not respond to structured emission signals. Synchronous detection at the Zeeman modulation frequency served to discriminate against all remaining signals not Zeeman modulated i.e. molecular absorption and scattered light. The combination of the two modulation techniques allowed discrimination against both continuum-background and atomic-emission signals. In our work a sixth-orde 1200 MCCAFFREY et aE. DISCRIMINATION AGAINST AES Analyst Vol. 108 active filter was used with a lock-in amplifier but it was found impossible to resolve completely between source-intensity modulation at 350 Hz (the maximum frequency possible with our laboratory-built mechanical chopper) and wavelength modulation at 30 Hz.This was due to the closeness of our two modulation frequencies. Instead a second lock-in amplifier was used to resolve the two frequencies and directly replaced the tuned filter. The approximate wave-forms for DLDM are shown in Fig. 3. Source intensity modulation cycle + I + I I ! FB ! FB ! FB I I fT L. Wavelength modulation4 Time or wavelength cycle - 5 Fig. 3. Dual lock-in double modulation waveforms (for acronyms see Fig. 1). The first lock-in amplifier tuned to the source intensity modula-tion frequency discriminates against all signals not source intensity modulated (AE AE' AE" FB). The second lock-in amplifier tuned to the wavelength modulation frequency discriminates against all remaining signals not wavelength modulated ( S ) .For clarity the relative frequency relationships (30 Hz wavelength modulation and 350 Hz source modula-tion) have not been represented exactly in this diagram. The components that were necessary for DLDM in addition to those in Table I11 are given in Table V. Simultaneous source modulation at about 350 Hz and asynchronous wavelength modulation at about 30 Hz were used. The commercial lock-in amplifier set a t minimum output low-pass filtering was used for synchronous detection at the source modulation fre-quency. The output of this lock-in was then fed to a second laboratory-constructed lock-in amplifier tuned to the wavelength modulation frequency. This instrumental arrangement is shown in Fig. 4. The first lock-in amplifier discriminated against all signals that were not source modulated (e.g.structured background and atomic emission). The second lock-in amplifier discriminated against those signals that were not wavelength modulated (e.g. scatter of incident radiation) as shown in Fig. 3. TABLE V ADDITIONAL COMPONENTS NECESSARY FOR DUAL LOCK-IN DOUBLE MODULATION Component Type and/or description Supplier Lock-in amplifier . . . . PAR Model 5204 lock-in Princeton Applied Lock-in amplifier . . . . X Y multiplier based lock-in Laboratory constructed Source modulation chopper . . . . Five-blade chopper Laboratory constructed Source modulation motor . . . . Variable speed 0-20 V d.c. motor TRW-Globe Dayton OH Variable d.c. power supply . . . . 0-20V d.c. Laboratory constructed analyser Research Princeton N J amplifie October 1983 INTERFERENCES IN CONTINUUM SOURCE EXCITED FLAME AFS Lock-in Recorder 1201 \ J ' I 7 : Lock-in In amplifier (2) Fig.4. Instrumental arrangement for dual lock-in double modulation. \ 7 t +j Lens Experimental procedzlre for DLDM The first lock-in amplifier was tuned using source modulation alone. The second lock-in amplifier was then tuned when using simultaneous source intensity modulation and wavelength modulation. The wavelength of interest was obtained using wavelength modulation alone. Reference Square Modulation Waveforms For both methods of double modulation square modulation waveforms were used to ensure maximum efficiency of detection of the fluorescence signals.It has been showns that square-wave modulation has a signal to noise ratio (SNR) advantage over sinusoidal modulation when using synchronous detection. The wavelength modulation system3 that was used in this work inherently gives square-waveform wavelength modulation and it is simple to achieve square-waveform source intensity modulation by the use of large mechanical chopper blades. These square waveforms are particularly important for double modulation because of the expected loss of signal inherent in using modulation techniques i e . a factor of two for any sort of 50% duty cycle modulation over direct current detection and a further factor of two after combining two types of modulation. Hence by using efficient square-wave modulation systems the loss in signal caused by double-modulation over single-modulation techniques can be restricted to the theoretical factor of two, The signal processing system of Koizumi et aZ.8 was inherently sinusoidal because the shape of the analytical signal was transformed by the tuned filter.The DLDM modification described allowed for square-wave signal processing. \ I Chopper General Experimental Conditions All measurements were made under the conditions given in Table I except where indicated in the text. Single element aqueous metal solutions were prepared in 0.04 M hydrochloric acid (to ensure maximum stability) from 1000 pg ml-l stock solutions made with analytical-reagent grade salts. / / 4 i NBS sample preparation Dried National Bureau of Standards (NBS) tomato leaves (SRM 1573) were weighed (0.1-0.4 g).Known amounts of aqueous standards were added to each sample to allow the con-struction of a calibration graph for analysis by the method of standard additions and then th 1202 MCCAFFREY et aZ. DISCRIMINATION AGAINST AES Analyst VoZ. 108 samples were digested in 10 ml of nitric acid for 4-8 h dpending upon the mass of the sample. The solutions were heated until dry 10 ml of aqua regia were added to each sample to continue the digestion and then the digests were brought to near dryness (approximately 1 ml). The residues were diluted with distilled water to 100 ml in calibrated flasks. A blank solution consisting of only the acids was prepared by the same digestion procedure. The samples were then analysed using the techniques of wavelength modulation and FPDM as described.Four samples were analysed by each method to obtain a determination of the precision of the analysis. Digested acid blanks were subtracted in all instances. All the standard additions graphs had a slope of approximately 1 indicating minimum chemical interference. Results and Discussion Spectral Interference Studies The interferences that were observed when using wavelength-modulated AFC are given in Table VI. The slope is a real slope in units of counts s-l per pg ml-l. This slope therefore could be viewed as the number of counts s-l obtained on the photon counter for a 1 pg ml-1 solution. This aids in interpreting the magnitude of an interference. For example for cadmium at 228.8 nm (the first element in Table VI) 506 counts s-l could be obtained for a 1 pg ml-l solution assuming that the calibration graph is linear at that concentration.For the same concentration of nickel at 228.8 nm 1.1 counts s-l could be obtained. Hence at the same concentration the interference of nickel on cadmium could be considered insignificant but assuming the linearity of the calibration graphs a &fold excess of nickel would give an error of about 1%. TABLE VI OBSERVED* SPECTRAL INTERFERENCES IN WAVELENGTH-MODULATED AFC Cd Cr Cr c u c u In Fe Pb Pb Mg Analyte * . Analyte wavelength/ nm 228.8 357.9 359.3 324.8 327.4 410.2 248.3 283.3 405.8 285.2 Slopet /counts per pgml-l 506 1260 1.1 7.5 - 1.4 - 15 1060 - 15 2.8 5.3 -0.7 3.7 - 0.7 -0.85 1370 - 38 850 - 860 2 900 242 -2.8 -0.7 265 2.8 - 6.3 990 - 180 4.9 x 104 -0.6 - 5.2 Limit of detection / pg ml-l 0.05 12 0.05 5 10 5 0.1 5 35 15 0.1 100 1 .o 10 0.1 30 125 0.1 0.003 0.2 1.0 50 50 100 26 0.15 1.3 0.001 175 25 Interferents and wavelengthslnm Ni 229.0 [AF] Co 357.9 357.8 Fe 357.0 358.6 Ni 357.2 Fe 358.1 358.7 [AE] Co 359.5 [AE] Ni 359.8 Cd 326.1 [AE] [AF] In 325.6 [AE] Pd 324.3 325.2 In 325.9 [AE] Cd 326.1 [AE] [AF] Ag 328.1 (AE) (AF) Co 409.2 Pd 247.6 [AF] Pt 283.0 [AF] Sn 284.0 [AE] [AF] K 404.4 404.7 CAE] Pb 286.4 Sn 286.3 284.0 [AE] [AF October 1983 INTERFERENCES IN CONTINUUM SOURCE EXCITED FLAME AFS 1203 Mn Mn Ni Ni Pd Pd Pd Ag Sn Sn Analyte * .. . Analyte wavelength/ nm 279.5 279.8 280.1 403.1 403.3 403.4 232.0 352.5 343.5 361.0 363.5 328.1 286.3 303.4 TABLE VI-continued Slope? /counts per pg ml-l 7 200 -0.5 12 3 020 - 65 2.2 - 0.8 290 14 - 8.9 15.5 28 35 - 5.7 1 50 - 39 - 3.7 68.5 0.4 190 - 1.4 2 640 - 540 17 -2.65 x 104 16.9 - 22 - 23 55 3.1 - 1.2 Limit of detection$/ pg ml-1 0.008 160 0.5 0.1 8 0.5 10 0.05 2 20 1.3 5 1.3 0.5 1 .o 5 100 3 > 100 0.5 100 0.05 0.1 1.5 0.001 2 3 2 1 10 350 Interferentss and wavelength§/nm Pb 280.2 [AF] Mg 279.6 280.3 K 404.4 [AE] Co 230.9 Co 352.7 (AE) Pd 351.7 [AE], T1 352.9 Co 343.2 343.3 [AE] Ni 344.6 [AE] [AF] Ni 361.9 [AE] Co 360.2 Cr 360.5 MgOH 362.4 [AE] Pb 364.0 [AE] [AF] Cu 327.4 (AE) (AF) Mg 285.2 (AF) Co 304.4 Fe 302.1 [AF] In 303.9 [AF] Ni 303.2 Pt 304.3 [AF] * For experimental conditions see text and Table I.t Slope of linear region of log - log plot of signal us. concentration. $ Approximate “limit of detection” corresponding to the concentration a t which the calibration graph intercepts the background level. This is within a factor of 2 of the calculated detection limit based on SNR - 2. Count time 1 s; and lamp current 17 A. The symbols (AE) and (AF) indicate that the wavelength is an analytical line that is used routinely for atomic emission or atomic fluorescence respectively The symbols [AE] and [AF] indicate that AE or AF has been observed at the indicated wavelengths but that they are not usual analytical lines.Element and most likely wavelength(s) in nanometres causing an interference. The sign of the slope indicates the nature of the spectral interference. A positive slope means that the interfering spectral line was within the 0.5 nm nominal spectral band pass of the monochromator and added to the signal. A negative slope means that the interfering line occurred during the background measuring part of the wavelength modulation cycle. There-fore the interfering line occurred near the nominal analyte wavelength but outside the 0.5 nm nominal band pass of the data measuring part of the wavelength modulation cycle and thus was subtracted from the signal.This resulted in the interferent signal becoming more nega-tive with increasing concentration hence the assigned negative slope. In the real negative slope situation which was tested here for the interference of potassium on lead at 405.8 nm the net analyte signal decreased with increasing interferent concentration. The magnitude of the interference clearly depends on interferent concentration. However, at equal concentrations of analyte and interferent three quarters of the analyte metals listed were affected by at least one interferent with a signal change of more than 1%. The limits o 1204 MCCAFFREY et al. DISCRIMINATION AGAINST AES Analyst VoZ.108 detection give a further indication of the magnitude of interference. For example a small number indicates a worse interference than a large number and any concentration of inter-ferent below its limit of detection does not cause an observable interference. Table VI only serves as a guide to the interferences likely to be encountered in wavelength-modulated AFC. This is particularly so in view of the usual difficulty in reproducing flame and instrumental conditions. Flame conditions will of course significantly alter the relative magnitudes of the atomic signals from all metals. Nature of the Spectral Interference The interferences in Table VI could be due to either atomic emission (AE) of the interferent or atomic fluorescence (AF) of the interferent excited by the continuum light source.These were not differentiated in this work. However an interference due to atomic emission can be distinguished from atomic fluorescence by using source intensity modulation alone or by making a blank measurement after blocking the light source. An indication of the sources of the inter-ferences (AE or AF) is given in Table VI. The symbols (AE) and (AF) indicate that the wavelength in question is an analytical line used routinely for AE or AF respectively. The symbols [AE] and [AF] indicate that AE or AF have been observed at the indicated wave-lengths but that they are not the usual analytical lines. Two symbols together indicate that the interference is probably a combination of both AE and AF. One symbol indicates that the interference is probably dominated or entirely due to the indicated process.The relative magnitudes of the contributions of the two processes to the interferences will depend on the intensity and age of the light source as well as the character of the lines in question. The potassium interference on lead at 405.8 nm was verified on our system to be entirely due to potassium atomic emission at a 1000 pg ml-l potassium concentration. Double Modulation The merits of the two methods of double modulation were evaluated by studying the spectral interference of potassium emission at 404.4 nm on the measurement of lead fluorescence at 405.8 nm when using wavelength-modulated AFC. Wavelength modulation alone was used and signals were measured at 405.8 nm while varying in separate experiments the concentra-tion of lead or potassium.Lead produced the expected linear calibration graph with increas-ing positive signals (Fig. 5). Potassium also gave a linear calibration graph but the signals decreased with increasing concentration (Fig. 5). This was a result of the potassium signal appearing in the background part of the wavelength modulation cycle because the potassium 1 0s c 'a 105 2 100 1 03 U 3 C 0 m .-7 lo4 cn 3 U 3 z S 103 1 o2 1 02 0.1 1.0 10 100 1000 104 Fig. Concentration/pg ml- ' 0 10' 1 02 1 03 Lead concentration/pg ml-' 6. Fixed-phase double modula-tion calibration graph showing the elirni-Fig. 5. Spectral interference nation of the potassium interference at the lead 405.8 nm line.Both the (0) aqueous lead data and the (0) data obtained with 100 pg ml-l potassium added are shown. The solid line corresponds to the least-squares log - log regression line for the points with (0) 100 pg ml-I of potassium added. from potassium at the lead 405.8-nm line. Potassium (A) and lead (B) calibration graphs a t 405.8 nm. The broken line indicates an increasingly negative signal October 1983 INTERFERENCES IN CONTINUUM SOURCE EXCITED FLAME AFS 1205 wavelength was to one side of the lead line. A concentration of 1000 pg ml-l of potassium gave a signal count equivalent to -200 pg ml-l of lead. A calibration graph for lead at 405.8 nm was obtained with and without the addition of 100 pg ml-1 potassium using FPDM (Fig. 6).Similar data were also obtained for FPDM at the 1000 pg ml-1 level (Fig. 7). The results indicated that the system did not respond to potas-sium because the calibration graphs did not change after addition of potassium. Similar 0 I I 10’ 1 o2 103 Lead concentration/pg ml-’ Fig. 7. Fixed-phase double modula-tion calibration graph showing the elimi-nation of the potassium interference at the lead 405.8nm line. Both (0) the aqueous calibration data for lead and (0) the data obtained with 1000 pg ml-l of potassium added are shown. The solid line corresponds to the least-squares log -log regression line for the points with (0) 1000 pg ml-l of potassium added. behaviour was observed for DLDM. However the experiments with varying concentrations of potassium demonstrated as expected that the noise on the lead signal was increased in the presence of potassium.This is because the techniques of double modulation do discriminate completely against the interfering emission signal but do not discriminate against the accomp-anying noise. This is evident in Fig. 7 from the large scatter of the points about the calibra-tion graph when 1000 pg ml-l of potassium are present relative to the scatter observed when 100 pg ml-l of potassium are present (Fig. 6 ) . Table VII gives the log - log regression equations for the lead calibration graphs with and without potassium present. The standard deviation about the regression line for the lead calibration graph almost doubled when 100 pg ml-l of potassium were present relative to the calibration graph when no potassium was present.TABLE VII EFFECT OF ADDED POTASSIUM ON THE DETERMINATION OF LEAD AT 405.8 nm Potassium concentration*/ pg nil-’ s,.,.t y s Regression equations 0 0.06 0.996 y = 0.84(f0.07)~ + 1.9 100 0.10 0.992 y = 0.87(&0.11)x + 1.9 1000 0.38 0.894 y = 0.83(&0.42)~ + 2.0 * Concentration of potassium added to aqueous soh tions of lead. t Standard deviation about the regression line; y = fluorescence signal; x = $ Correlation coefficient of the regression equation. 5 Log - log regression equation with the 95% confidence limit for the slope given All equations are based on a total of eight lead concentrations, lead concentration in micrograms per millilitre. in parentheses. varying from 5 to 1000 pg ml-1. See discussion in text 1206 MCCAFFREY et al.DISCRIMINATION AGAINST AES Analyst VoZ. 108 Further when 1000 pg ml-l of potassium were present the standard deviation about the regression line increased 6-fold relative to the aqueous lead calibration graph without potas-sium added and lead detection limits were degraded to approximately 50 pg ml-1 when 1000 pg ml-l of potassium were present. The points below 50 pg ml-l in Fig. 7 were not significantly different from the blank but were included in Fig. 7 and in the regression analysis (Table VII) in order that a direct com-parison of the decreased precision resulting from high potassium concentrations could be made. The slope of the lead calibration graph (Table VII) is less than one because the poor sensitivity shown by lead results in a linear range of only about two orders of magnitude.For metals with better sensitivity than lead the linear range will be longer. We are currently carrying out detailed studies concerned with the effect of spectral interferences on the signal to noise ratio of double-modulated AFC during the analysis of various real samples (standard reference materials). However it is clear from Fig. 7 and from Table VII that the determination of lead at 405.8 nm in the presence of concentrations of potassium above about 100 pg ml-l would be affected by decreases in precision. The lead line at 283.3 nm is not affected by potassium atomic emission and would in fact be the wavelength of choice for lead determina-tion. The 405.8 nm lead line was chosen here only to demonstrate the effectiveness of double modulation in discriminating against spectral interferences of which the lead/potassium situa-tion is probably an example of the worst that such an interference could be in samples other than those in which the major constituent results in a spectral interference.Detection Limits for the Double Modulation Systems The detection limits given in Table VI were not obtained under optimum conditions and are consistent only in the relative magnitudes of each associated interferent and analyte detection limit and only indicate the severity of a spectral interference on a particular metal. Detection limits obtained under conditions close to the optimum are given in Table VIII for source intensity wavelength and double modulation as described and for the double modulation method of Elser and WinefordneF as applied to AFC.* It can be seen that the double modula-tion methods described here gave detection limits that were improved over the previous double modulation AFC figures.6 Double modulation halves the amount of time spent measuring the analyte fluorescence compared with single modulation.Therefore double modulation would be expected to produce detection limits about a factor of 2 worse than source modulation alone or wavelength modulation alone. Only FPDM achieved this goal being at most a factor of 3 worse than either method of single modulation. The calculated detection limits in Table VIII were based on measurements of aqueous solu-tions of pure metal salts with no interferents present and on the assumption that the noise was TABLE VIII LIMITS OF DETECTION USING SOURCE INTENSITY WAVELENGTR AND DOUBLE MODULATION FOR CONTINUUM SOURCE EXCITED ATOMIC-FLUORESCENCE MEASUREMENTS Wavelength/ Element nm Zn .. . . 213.9 co . . . . 240.7 Pb . . . . 283.3 Mg . . . . 285.2 Cr . . . . 357.9 Pb . . . . 405.8 Sr . . 460.7 7 SMt 9 20 200 0.4 4 200 2 Limit of detection*/pg 1-I A 1 WM FPDMS DLDMT DMAFFSII - 10 20 50 30 30 100 -200 500 700 -0.6 0.9 0.9 30 4 10 20 600 300 500 2 000 5 000 1 3 20 -* A 300-W xenon arc operated at 20 A and 10 s count time was used in this work for all experiments except SM DMAFFS when a 900-W xenon arc was used and the data were taken from reference 4. t SM source intensity modulation. WM wavelength modulation.FPDM fixed-phase double modulation. 7 DLDM dual lock-in double modulation. 11 Double modulation by the method of reference 4. DMAFFS is double modulation atomic-fluorescence flame spectrometry October 1983 INTERFERENCES IN CONTINUUM SOURCE EXCITED FLAME AFS 1207 white (shot noise) and could be calculated by taking the square root of the background. The detection limits were demonstrated to be shot noise limited by standard deviation measure-ments of example metals at concentrations close to the detection limit. The detection limits for DLDM were up to eight times lower than those for FPDM. How-ever they were still an improvement over previously published double modulation results. The reasons for the low detection limits for DLDM were not clear but it was thought that they were due to the use of the laboratory-constructed lock-in amplifier which may have been less efficient at filtering noise than a commercial instrument.Also a photon counter was used for FPDM which in our experience is usually more sensitive than a lock-in amplifier for flame spectroscopic measurements. In the low background regions of the flame this sensitivity improvement can be a factor of 3-5. Real Sample Analysis Table IX shows the results obtained for the analysis of NBS tomato leaves. The results are in good agreement with the NBS certified values in all instances except for the determination of iron by wavelength modulation at 248.3 nm. The low value obtained for iron by wave-length modulation indicates that a spectral interference occurred during the background measuring part of the wavelength modulation cycle.The interference was corrected by FPDM because the FPDM result was in agreement with the NBS certified value. The exact nature of the interference is unknown and does not correspond to any of the atomic line inter-ferences shown in Table VI. Based on the number and severity of the spectral interferences shown in Table VI it may appear that the determination of any metal in a real sample would be impossible by wavelength modulation. However as shown in Table IX it is often possible to determine a metal in a complex sample using wavelength modulation alone but the technique of double modulation may be necessary to ensure accuracy when determining a large number of metals in a complex sample.The ability to carry out both wavelength modulation and double modulation in a single instrument is advantageous for real sample analysis. Wavelength modulation can exhibit higher sensitivity by providing a combined emission and fluorescence signal which is useful for those elements which are more sensitively determined by atomic emission than by atomic fluorescence. However as shown here for the determination of iron in tomato leaves double modulation is required in certain situations in order to obtain accurate results. TABLE IX ANALYSIS OF NBS TOMATO LEAVES (SRM 1573) BY WAVELENGTH MODULATION AND FPDM Contentlpg g-l Wavelength/ - Element nm NBS value* WMt FPDMS Zn . . . . 213.9 62 ( f 6 ) 60 (*2) 58 ( f 4 ) Mn . . . . 279.5 238 ( f 7 ) 230 ( f 9 ) 227 (f7) cu .. 324.8 11 (&I) 11 ( f 2 ) 11 (f1) Fe . . . . 248.3 690 (f25) 614 (f14) 685 (h50) * Certified value. t WM wavelength modulation f (standard deviation of four samples) FPDM fixed-phase double modulation f (standard deviation of four samples). Conclusion The extent of spectral interferences listed in Table VI indicates that wavelength-modulated AFC is unlikely to be a technique of choice for a wide variety of analyses. The removal of the potassium atomic emission interference on lead determinations at 405.8 nm by the application of either of the double modulation techniques indicates that it is possible instrumentally to discriminate against all spectral interferences caused by atomic emission of concomitant ele-ments in any matrix.This is because there is unlikely to be a worse example of atomic emission interference than the potassium in lead situation except where a major element cause 1208 MCCAFFREY WU AND MICHEL a spectral interference. Only in the latter instance is the decrease in precision likely to make the technique unusable for trace analysis. Double-modulated AFC is therefore more likely to be of general use than wavelength-modulated AFC. The fixed-phase version is more suit-able for routine analyses because of its sensitivity advantage over dual lock-in double modula-tion although the reason for this advantage is not clear. The general advantages of AFC lie primarily in the ability to carry out sequential multi-element analyses with the use of only one light source.ly2 Various tables of detection limits in the literat~rel-~ define the typical range of sensitivity to be expected of AFC.The example detection limits in Table VIII indicate that double-modulated AFC (FPDM) falls mostly within that range. One reason for the favourable sensitivity of the double modulation techniques described probably lies in the use of efficient square waveforms for the modulation devices. However a monochromator with a small f number a modern design of xenon arc and for FPDM photon counting probably also made a contribution to the high sensitivities that were achieved relative to previously published4 work on double modulation. There is no question that there are more sensitive techniques available to the analyst. However where sensitivity is adequate the sequential multi-element capability of AFC will have advantages for some analyses over single element techniques such as line source excited atomic absorption or fluorescence.Although double modulation does discriminate against atomic emission spectral interferences, there remains the problem of the noise associated with the spectral interferences. We are currently involved in studies to characterise more fully the seriousness of the decrease in pre-cision caused by this noise and by scatter noise present during the analysis of real samples and at varying concentrations of analyte and interferent in aqueous solutions. We are also in the process of evaluating the importance of spectral interferences caused by atomic fluorescence of concomitant elements in the matrix. However Table VI does give some indication that this is much less of a potential problem than atomic emission spectral interferences. Finally there is the possibility of improved sensitivity in AFC by the replacement of the flame with a furnace and by the use of more powerful xenon arcs. The authors express their appreciation to John E. Gammerino in the Chemistry Department Instrument Shop for construction of the wavelength modulation unit. Acknowledgement is made to the donors of the Petroleum Research Fund administered by the American Chemical Society for the support of this research. Acknowledgement is also made to Jenny Jones for assistance during the digestion of the Standard Reference Materials. 1. 2. 3. 4. 5. 6. 7. 8. 9. References Johnson D. J. Plankey F. W. and Winefordner J. D. Anal. Chem. 1975 47 1739. Ullman A. H. Pollard B. D. Boutilier G. D. Bateh R. P. Hanley P. and Winefordner J . D., Michel R. G. Sneddon J . Hunter J . K. Ottaway J . M. and Fell G. S. Analyst 1981 106 288. Fowler W. K. Knapp D. O. and Winefordner J . D. Anal. Chem. 1974 46 601. Lipari F. and Plankey F. W. Anal. Chem. 1978 50 386. Elser R. C. and Winefordner J. D. Anal. Chem. 1972 44 698. Parsons M. L. Smith B. W. and Bentley G. E. “Handbook of Flame Spectroscopy,” First Edition, Koizumi H. Yasuda K. and Katayama M. Anal. Chem. 1977 49 1106. O’Haver T. C. Epstein M. S. and Zander A. T. Anal. Chem. 1977 49 458. Anal. Chem. 1979 51 2382. Plenum New York 1975 Chapter 2. Received December 31st. 1982 Accepted May loth 198