(5) A. Fontijn, A. J. Sabadell, and R. J. Ronco, Anal. Chem., 42, 575 (1970). (6) D. H. Stedman, E. E. Daby, F. Stuhl, and H. Nikl, J. Air Pollut. Control Assoc., 22, 260 (1972). (7) R . K. Stevens and J. A. Hodgeson, Anal. Chem., 45, 443A (1973). (8) H. V. Drushel, Anal. Len., 3 (7), 353 (1970).
LITERATURE CITED (1) R. B. Bradstreet, “The Kjeldahl Method for Organic Nitrogen”, Academic Press, New York and London, 1965. (2) R. L. Marlin, Anal. Chem., 38, 1209 (1966). (3) H. V. Drushel.. Preor.. . Div. Pet. Chem.. Am. Chem. Soc.. 21 (1). 146 (1976). (4) k. E. ‘Parks, “Analysls for Chemically Bound Nitrogen Using W o Chemiluminescence,” paper presented at the 27th Pmsburgh Conference on Analytical Chemistry and Applied Spectroscopy, Cleveland, Ohio, March 2, 1976.
RECEIVED for review November 23, 1976. Accepted March 149
1977.
Comparison of Multichannel SIT Image Vidicon and Photomultiplier Sequential Linear Scanning Systems for the Measurement of Steady-State and Transient Fluorescence of Molecules in Solution R. P. Cooney, 1.Vo-Dinh, G. Walden, and J. D. Winefordner” Department
of Chemistry,
University of Florida, Gainesville, Florida 326 1 1
The feasiblllty of uslng the SIT image vidicon as a potential detector for llquld chromatography Is demonstrated. Limlts of detection range from - 2 X lo-* ng for anthracene to - 6 X lo-‘ ng for riboflavln. A dlrect experlmental comparison Is made of a sillcon lntenslfled target vldlcon multichannel (SIT-MC) and a photomultipller sequential linear scanning (PM-SLS) system for the detectlon of steady-state fluorescence from molecules In solution. A dlscussion is given of the fundamental differences between the ways the two systems acqulre, process, and read out data. I t Is concluded that the PM system is slightly better than the SIT for measuring steady-state fluorescence; on the other hand, the multichannel advantage of the SIT would appear to make It the better system for measuring translent fluorescence slgnals.
(S/N) calculations (18))to the authors’ knowledge, there has been no critical, direct quantitative comparison of any image device with a photomultiplier tube for the detection of steady-state and transient molecular fluorescence signals. Nor has there been, to the authors’ knowledge, a thorough discussion given in any one place of the fundamental differences in the nature of the data and data collecting processes of the two types of detectors. Therefore, it is also the purpose of this paper to present the results of a comparison of a silicon intensified target vidicon multichannel system (SIT-MC) and a photomultiplier sequential linear scan (PM-SLS) for the measurement of steady-state and transient fluorescence of molecules in solution and to present the detailed discussion mentioned above. EXPERIMENTAL
Image devices are finding steadily increasing application as detectors in atomic and molecular spectroscopy (1, 2). Numerous papers describing the use of image device detectors have been published in the areas of molecular absorption (3-12)) molecular fluorescence (13-16), and Raman ( I 7) spectroscopy. Earlier publications by the present authors have reported analytical figures of merit (including limits of detection and linear dynamic range) for an instrument consisting of a commercial spectrofluorometer equipped with an SIT (Silicon Intensified Target) vidicon camera detector. This instrument has been used to measure steady-state fluorescence emission from molecules in solution (14)) and transient fluorescent emission from molecules in the vapor phase eluting from a gas chromatograph (15). A logical extension of these earlier investigations is to measure transient fluorescence emission from molecules in solution, such as would be the case for detection of fluorescent compounds being eluted from a liquid chromatograph. One of the purposes of this paper is to evaluate transient fluorescence signals giving limits of detection for several polynuclear aromatic and heterocyclic polyatomic molecules of environmental and biological interest Le., perylene, anthracene, chrysene, benzo[a]pyrene, riboflavine, acridine, and 4,5-diphenylimidazole. In addition, although image devices and photomultiplier tubes have been compared theoretically by signal-to-noiseratio
Apparatus. A block diagram of the instrumental system is shown in Figure 1. Most of the apparatus used in this study is the same as that employed in earlier works (14, 15), and is described in detail there. Thus the source, spectrometer, flow cell, SIT-OMA, oscilloscope, chart recorder, and syringes are as used and described previously (14, 15). In the comparison studies, the flow cell was replaced by a standard 10 X 10 mm i.d. spectrasil quartz cuvette, and the detector used was either the SIT or a Hamamatsu 1P21 photomultiplier (PM) tube. Signals from the PM were processed by an O’Haver (19) type nanoammeter. For the study of transient signals, a system simulating a liquid chromatograph was built. The system was designed to be simple and practical, its sole purpose being to provide a flow of solvent with which to carry injected samples to and through the detector. The pump was a peristaltic type (Ismatac, Saia AG, Murten, Switzerland, type AMY8-D-25SR). This type of pump produced a pulsating flow; however, this did not affect the performance of the SIT because the observation time was considerably greater than the changes produced by this pulsation. The injection port was constructed from a ‘/8-in. Swagelok Union Tee fitting and inserted between the pump and the flow cell. Solvent was fed into the branch portion of the union tee (see Figure l),while samples were injected through a rubber septum placed inside one of the hex nuts on the run portion of the union tee. This arrangement, as opposed to injecting through the branch portion of the union tee, allowed the sample to enter the solvent stream past the right angle bend, thus avoiding possible sample hang-up and excessive band broadening. Tygon (‘f8-in. 0.d.) tubing was ANALYTICAL CHEMISTRY, VOL. 49, NO. 7, JUNE 1977
939
decided to set T equal to 0.5 s because this is a value typically used in dc systems ( 8 s on the order of half a minute are not ordinarily available with commercial instruments), and because a time constant of 0.5 s was compatible with the -2 nm s-l scan rate, which was chosen so that the PM-SLS system’s scanning time and the SIT-MC system’s observation time of the 62-nm window would be approximately the same. The spectra obtained with both systems were plotted using the same chart recorder (T = 0.1 s, Af = 2.5 Hz). Signal-to-noise ratios (S/N) and limits of detection were calculated wth the data obtained from the printed spectra.
9
ILII-1 7
8
Fgure 1. Block diagram of the flow-cell and SIT/OMA detection system. (1) Light source, (2) Optics, (3) Excitation Monochromator, (4) Emission monochromator, (5) SIT image vidicon, (6) Optical mukhannel analyzer, (7) Strip-chart recorder, (8) Oscilloscope, (9) Peristaltic pump, (10) Injection port, (1 1) Flow cell, and (12) Carrier solvent reservoir
used between the solvent reservoir and the pump, and between the pump and the injection port, while ‘/8-in. 0.d. Teflon tubing was used between the injection port and the flow cell. Reagents. Double-deionized water, spectrograde methanol (Matheson, Coleman and Bell, Norwood, Ohio) and technical grade hexane were used as solvents. All compounds investigated were used without further purification: perylene, anthracene (Chem Service, West Chester, Pa.), chrysene and benzo[a]pyrene (Research Organic/Inorganic Chemical Corp., Sun Valley, Calif.), riboflavine (National Biochemicals Co., Cleveland, Ohio), acridine (Eastman Organic Chemicals, Rochester, N.Y.), and 4,5-diphenylimidazole (Aldrich Co., Milwaukee, Wis.). Procedure. Measurement of Transient Signals. A detailed discussion of the operation of the SIT-OMA system has been given earlier (14,15), and will not be presented here. In the study of transient signals, samples were injected after a uniform rate of sample flow had been established. The first few injections of each compound were made at sufficiently large concentrations (1100 ppm) so that real time signals could be observed on the oscilloscope output of the OMA. This allowed direct measurement of both the elution time and the sample peak width. Once the real time signal appeared on the scope, the solvent flow could be stopped (by temporarily turning off the pump), allowing the excitation and emission monochromators to be adjusted for maximum signal intensity and optimum placement of the emission spectrum in the -62-nm spectral window of the system. After these preliminary adjustment procedures, sample concentrations were reduced by a factor of 100 or more in order to determine limits of detection. Generally, data were accumulated over the entire sample peak width. Background correction and baseline measurements were made as reported previously (14, 15). Comparison of SIT-MC and PM-SLS. The spectral bandwidths of both systems were made the same; therefore, the arrangement and sizes of the slit widths were the same in both the PM-SLS and SIT-MC cmes, with one exception; in the SIT case, the exit slit of the emission monochromator was removed and the camera placed at the focal plane, so that a 1:l image of the limiting entrance slit was formed on the detector face. The spectral bandwidth used was 11nm, because narrower bandwidths would have decreased sensitivity, without improving resolution due to the broad bandwidths of the anthracene spectrum. Because measurements were made by both systems on the same solution, the spectral irradiancies to both detectors were the same. Observations of sample and blank with the SIT were made for 1000 accumulation cycles each, or 65.6 s, and covered a spectral window of 62 nm. The OMA plotted the contents of its 500channel memory (62-nm window) in 131 s, or at a rate of one channel per 0.26 s (or 0.47 nm s8). The PM system scanned and plotted a 62-nm window for sample (or blank) in 32 s, or at a rate of 1.9 nm s-’; a total time of 64 s. In order for the noise bandwidth, A f , of both systems to be the = 1/(2)(65.6) s = 7.6 X Hz, where t , same [AfsIT.Mc= ’/&, is the observation time, s; AfpM.SLS = ‘ / 4 ~ = 1/(4)(0.5) s = 0.5 Hz, where T is the electronic time constant of the system], T for the PM system would have to have been 32.8 s. However, it was 940
ANALYTICAL CHEMISTRY, VOL. 49, NO. 7, JUNE 1977
RESULTS AND DISCUSSION Theoretical Considerations. The comparison of the two systems becomes more meaningful if it is kept in mind that there is both a difference in the nature of the data obtained from each system, and a difference in the way the information is processed. The output signal from the P M is a current, or a charge flow. The current magnitude, or rate of charge flow is directly proportional to the rate at which photons strike the photocathode. Thus the basic information contained in the PM signal is rate information. Therefore, when the P M is not used as an integrating detector, its signal will be independent of the observation time, to,in the sense that for a constant photon flux, the P M current will have constant magnitude, for however long the measurement is made. Due to the electrical bandwidth, A f , of the system, the PM-SLS signal does possess some time dependency. For example, (20), it takes a finite time ( = A f ’ ) to reach a steady state value, and it must also be remembered that the magnitude of the output signal at any time, t, actually depends upon the average input signal from time, t - t,, t o time, t (where t , is the averaging time of the system, = l/zAj-). The SIT, on the other hand, is an integrating detector; therefore the output signal is in terms of total accumulated charge, or via an A to D conversion counts. The magnitude of the output signal depends upon both the rate at which photons strike the detector and the total observation time, to,regardless of whether or not the rate of photon arrival a t the detector is constant. In the spectrum generated from the PM, the wavelength axis is also a time axis because, as the spectrometer scans wavelength, the information for any given spectral interval is obtained a t a time different from that in which information for all other spectral intervals is obtained. In the spectrum generated from the SIT-MC, the wavelength axis is not also a time axis, as in the case of the PM-SLS, because the information accumulated in one spectral interval is accumulated simultaneously with the information in all other spectral intervals. Thus, spectra from the two systems, although having the same general appearance, will be different in that the PM-SLS spectrum plots current (rate) vs. wavelength (time) while the spectrum from the SIT-MC plots counts (total accumulated charge following an A to D conversion) vs. wavelength. The signal-to-noise ratio (S/N) expressions for image devices have been derived elsewhere (18) and therefore will be presented here without derivation. As seen from the above discussion, it will be most appropriate to give the S/N expression for the SIT in terms of counts, while the S/N expression for the P M in terms of currents. Thus,
+ to2gs2(RLi+ 2RIi + ZRs,)’ + 4NA2]1’2
(1)
where t , = observation time, s. R x , = count rate in channel i arising from process X (X= L for analyte luminescence, X = Z for source induced luminescence interference, X = S for source induced scatter, and X = D for detector half channel dark signal), (counts) s-’. Rx, = Ex,ALqi,where E , = photon
irradiance for process X in channel i, photons cm-2 s-’; A , = area of channel i, cm2; and qr = OMA conversion efficiency (photons to counts) in channel i, (counts) photon-’, i.e., A to D counts per incident photon. 5s = flicker (fluctuation) factor for the source dependent processes XL,I,s, dimensionless. N A = preamplifier noise, dimensionless (counts) = N , n:” where N , = preamplifier equivalent noise per half channel per scan in rms counts and n, = number of scans. The coefficients of 2 for RI,and Rs, arise from the fact that there are a sample and a blank measurement (with blank identical to sample except for the presence of analyte in the sample) for each analytical determination. The coefficient of 4 for RD,and NA2arise because the dark signal from a half channel of the SIT and preamplifier noise are accumulated four times for each set of sample-blank measurements. It is assumed the detector flicker noises and non-source induced background are negligible in all cases. Even though it is possible to sum the signal and noises over several channels (which generally results in improved S/N), this was not done, both for the sake of simplicity and because using one channel corresponds most closely to the “graphical” method of signal-to-noise ratio computation actually employed in this study. The noise expression in the denominator of Equation 1 yields an rms noise value which would be obtained if repeated measurements were to be made of the signal in a single channel and an rms deviation from the mean calculated. The same rms value could be obtained if, instead of making several measurements with one channel, one measurement were made with several adjacent channels, assuming a signal level constant over the channels employed (which holds fairly well for the case of either broad band scatter from a continuum excitation source or molecular luminescence interference). Graphically, this corresponds to taking the peak-to-peak baseline noise and dividing by five (for a 99% confidence level, where the baseline is obtained by using the A-B display mode of the OMA with blank signal in both the A and B memories). An implicit assumption used here is that at or near the limit of detection, the limiting noise will not be related to analyte luminescence. The expression for the S/Nof the P M has been shown to be (18):
(S/N),, = iL/[Ks* + 2 G 2 + 2 a 2 + (A= -t 2Ai1, + 2 A x ) 2 + 2 A ~ 2 ] l ” 2 I_
where i L = photon current due to analyte luminescence, A, = ExAeGqt where Ex = photon irradiance, photons cm-’s-’; A = surface area of PM struck by light, cm2; e = electron charge (1.6 X 1O-l’ C); G = average gain of the PM, dimensionless; 71 = efficiency of photocathode, dimensionless; 6 = efficiency of dynode chain, dimensionless; Aixs = shot noise current due to process X (where X is defined as in Equation l),A; = ( 2 e 2 G ~ c A f E ~ A ) Af ’ I 2=; noise bandwidth of system ( 1 /4~ where 7 is the limiting time constant of the system), Hz; AixF = flicker noise current due to process X, A, = [six, where 5s = flicker (fluctuation) constant (20) for any source dependent process, dimensionless. = detector dark current shot noise, A. Coefficients of 2 in all cases arise from making both a sample and a blank measurement. As in Equation 1, it is assumed detector flicker noise and non-source induced background are negligible. In addition, it is assumed that amplifier noise for the PM case is negligible. The PM-SLS system generates its plotted spectrum at the same time it scans. The scan rate (Rs, nm s?) is limited to a range (20) of values such that:
where 6Xb is the half width (in nm) of the narrowest spectral feature to be recorded, and t , is the response time of the system, equal to 1.6Af1, Af being the limiting, or smallest, bandwidth of the entire system, including recorder (Af = ~ / 2 t , )As . seen below, this can greatly affect the total analysis time of the PM system. The SIT system generates a plotted spectrum by sequentially reading the digital contents of each memory (A or B) location, or sequentially subtracting the digital contents of corresponding locations in both memories (A-B mode), and passing the desired information through a D-A converter to a chart recorder. Thus the spectrum actually is composed of 500 discrete points and, if the chart speed were fast enough compared to the OMA readout rate, the spectrum would consist of a series of steps. Usually, however, the chart speed is slow enough relative to the OMA readout rate that the spectrum has a “continuous” appearance. Also, if the readout time per channel of the OMA is shorter than the recorder response time, this will have the effect of “smoothing” out the spectrum, that is both making the spectrum look more “continuous” and also damping out the noise fluctations from channel to channel. Since the spectrum is plotted subsequent to the actual observation, the time it takes to print the spectrum must be taken into account if the total analysis times of the two systems are to be compared. For the OMA used in this study, the printout time ranges from 16.4 s to 3.00 min in 16.4-9 increments. Using other types of readout systems, such as a microprocessor or storage oscilloscope with fast printer, the printout time would be made insignificant compared to the observation time. In the A-B mode, the SIT-MC takes the sample spectrum (A memory) and subtracts the blank spectrum (B memory) to yield the “corrected” spectrum for a “complete” analytical determination (sample minus blank). This means that the A-B spectrum contains noise contributions from both the sample and the blank, as indicated in Equation 1. In contrast, the PM-SLS system prints either the sample or the blank spectrum, but it does not print a single spectrum containing sample minus blank information. Therefore (contrary to Equation 2, which contains noise contributions from both sample and blank), a spectrum generated from the PM-SLS system contains a noise contribution from either sample or blank, but not both. Thus a rigorous experimental comparison of the two systems requires that the noise measured on the PM-SLS spectrum be multiplied by a factor of between 2 and 2’12 depending upon whether the dominant noise is flicker or shot noise, respectively. It is not always convenient to determine the nature of the limiting noise, however, and failure to do so introduces at most an error of only 2’1’ in the comparison. Measurement of Transient Signals. Several authors (15, 21-29) have reported applications and discussed the relative merits of fluorescence detection in chromatography (i-e.,good sensitivity, selectivity, combined temporal and spectral resolution), and their discussions will not be repeated here. There have been several examples in the literature of fast scanning spectrometers and image devices used as absorption or fluorescence detectors for gas and liquid chromatography (12,27, 30-32). There has also been some discussion comparing both types of systems for the detection of transient signals, like those coming from chromatographic eluents (15). In Table I, limits of detection are given for several polynuclear aromatic and heterocyclic compounds of biological interest using the simulated liquid chromatograph with SIT detector discussed above. Here, the limit of detection is defined as the absolute amount (ng) of material injected which gives a S/N,, equal to 3. Heights of the most intense spectral bands were used as the measure of luminescence intensity, ANALYTICAL CHEMISTRY, VOL. 49, NO. 7, JUNE 1977
941
Table I. Limits of Detection of Several Polyaromatic and Heterocyclic Molecules with the Simulated Liquid Chromatograph and the SIT-MC System Compound Perylene Anthracene Chrysene Benzo[a]pyrene Riboflavin Acridine 4,5-Diphenylimidazole
Carrier solution Methanol Methanol Methanol Methanol Water
Solvent Hexane Hexane Hexane Hexane Watertmethanol (51100; VlV) Methanol Methanol
Water Water
hexa
kerna
Ata
toa
(nm) 410 250
(s)
(SI
350 430
(nm) 468 403 380 440 523
25 25 25 25 25
32.8 32.8 32.8 32.0 37.8
L. 0.Dea (ng) 3.0 X lo-' 2.0 X lo-' 8.0 X lo-' 1.5 X lo-' 6.5 X lo-'
346 285
419 390
25 25
37.8 32.8
2.3 4.5
280
X
lo-'
a h e x , hem = excitation and emission wavelengths giving maximum fluorescence intensity. A t = delay time between injecThe L.O.D. for anthracene via the same experimental system but using tion and measurement. t , = measurement time. the PM-SLS system is 1 X lo-*ng. Limits of detection for the other compounds should also be reduced by =2X with the PM-SLS svstem.
1 420
140
,/ I
100
lnllgrorlon
420
-
400 390
400
380
380
370
360
Wavelength (nml
Figure 3. Spectra of anthracene ppm) in methanol obtained with the SIT (observation time = 32.8 s) and the photomultiplier (1P21, system response time = 3.1 s, scan rate = 1.9 nm s-'). In both cases spectral bandpass = 11 nm
~
1000
cycle0
Plots of signal (Sl, S2, S,) and noise (Nl, N, N3,) vs. number of accumulation cycles at different count rates, Le., S1 and N1 made at 280 counts cycle-', S2 and N, made at 38 counts cycle-', S3and N3 made at 575 counts cycle-'. Curves S1, S2, and S3have slopes of 1, curves N, and N3 have slopes of 0.5, and curve N, has a slope of -0.6 Flgure 2.
while the rms noise was 0.2 times the peak-to-peak baseline noise, as discussed in the section on theoretical considerations. All compounds but one were detected in the subnanogram range. Thus, considering its good sensitivity and multichannel advantage (discussed previously ( 1 4 , 1 5 ) and below), there is obviously great potential use for the SIT and similar devices as detectors for liquid chromatography, even more so than for gas chromatography, where the spectra are less detailed than they are for molecules in the condensed phase. In Figure 2, log-log plots are given of signal (curves SI, S2, 942
410
ANALYTICAL CHEMISTRY, VOL. 49, NO. 7, JUNE 1977
and S,) and noise (Nl, Nz, and N3) vs. time for steady-state fluorescence emission of intensities of 38,280, and 575 counts per accumulation cycle, respectively. Curves S1, Sz, and S3.allhave slopes of unity, demonstrating that the response of the SIT is linear with signal intensity over the range studied, which is what was expected, because the stated range of linearity of the OMA is close to 780 counts per accumulation cycle. For the two highest count rates, log-log plots of noise vs. observation time (Nz,N3) have slopes of 0.5, showing that at relatively high signal levels, shot noise (either photon or detector) predominates. Because the highest count rate observed is very close to the saturation rate (780 counts per accumulation cycle), it may be that for this system, photon flicker noise will never become dominant. The noise vs. observation time plot for the lowest count (NJ rate has a slope of 0.6, indicating that some type of noise, in addition to shot noise, is present. Because it is present a t low signal levels, it cannot be source related flicker noise, and detector electronics flicker noise is likely to be negligible. It seems likely
that the additional noise is a coherent pattern noise observed previously in the authors’ laboratory (15). However, this noise has not been fully characterized, so it is not possible to say what the exact time dependency of this type of noise is, other than as the observation time increases, photon or detector shot noise eventually becomes dominant. Comparison of SIT-MC and PM-SLS. Figure 3 shows the spectra obtained from the 4 X ppm anthracene solution using the SIT-MC and PM-SLS systems, respectively. Both SIT-MC spectra in Figure 3 were obtained using the A-B display mode of the OMA. In the upper spectrum, sample emission was contained in memory A and blank emission was in memory B, while in the baseline (lower spectrum) blank emission was in both memories. In the PM-SLS spectra, the solid line is the sample spectrum while the lower plot shows the blank spectrum. The limits of detection (for S/N,,, = 3) calculated from the spectra are 1.2 X ppm for the SIT and 4.2 X lo4 ppm for the PM. Several points should be kept in mind concerning this comparison, however. First, the bandwidth of the SIT-MC (AfsIT.MC = 7.6 X Hz) is about 65 times smaller than that of the PM-SLS ( A f p M . s u = 0.5 Hz). Because shot noise is directly proportional to Af1l2, the S/N of the SLS would have improved by a factor of 65112 or =8 if AfpM-SLS had been made equal to AfSIT.MC, and therefore the limit of detection with the PM-SLS would have been lower by a factor of -8 (assuming shot noise predominance). However, if the PM-SLS had had the same bandwidth as the SIT-MC, as discussed above, its maximum usable scanning rate would have been
Rs =
~
11nm 206 s
or 0.052 nm s-l, and the scanning time for a 62-nm spectral window would become over 19 min, which is inordinately long. Secondly, the readout time of each memory channel of the OMA was ~ 2 6 ms, 0 while the response time of the recorder, t,, was ~ 4 6 0ms, so the recorder will have smoothed out somewhat the noise inherent in the SIT spectrum. In the PM-SLS case, the scan rate, R,, was 1.9 nm s-l, well below maximum usable rate of
-
-
t,
11 nm 4.6 X 0.5 s
=
4.8 nm s-’
It is also interesting to compare the analysis time for the two systems studied in our laboratory (disregarding sample preparation and handling, etc.). The analysis time for our SIT-MC system was about 3.5 min (about 0.5 min each for data accumulation into memory A and memory B, and a little over 2 min for spectrum printout which could be made negligible by a faster readout). The analysis time for the PM-SLS system was a little over 1 rnin (scanning -62 nm twice at 1.9 nm s-’) and could have been even less if it had scanned at the maximum useable rate (4.3 nm s?). Thus, the PM-SLS system has a slight advantage over the SIT-MC system (in the measurement of steady state fluorescence)
having both a lower limit of detection and a faster analysis time (the analysis times of the two systems, however, would be essentially identical if the readout time for the OMA could be reduced to < l o s via a faster method; however, this may degrade the S/N). Thus it can be seen that there is a trade-off between analysis time and limit o f detection for both t h e PM-SLS and SIT-MC systems. For the SIT-MC, increasing the signal-to-noise ratio by a factor of n requires increasing the accumulation time by a factor of n2, assuming shot noise dominance. For the PM-SLS, increasing the S / N by a factor of n requires decreasing the noise bandwidth-decreasing the scan rate-by a factor of n2. This trade-off between analysis time and limit of detection may or may not be important with the PM-SLS system for the analysis of steady-state signals, but it certainly must be considered when making measurements of transient signals such as those discussed earlier in this paper.
LITERATURE CITED Y. Talmi, Anal. Chem., 47, 658A (1975); 47, 697A (1975). J. D. Winefordner, J. J. Fitzperaid, and N. Omenetto, Appl. . . Spectrosc., . 29, 269 (1975). M. J. Milano and H. L. Pardin, Anal. Chem., 47, 25 (1975). G. Horiick and E. G. Codding, Anal. Chem., 46, 133 (1974). T. E. Cook, M. J. Milano, and H. L. Pardue, Clin. Chem. ( Winston-Salem, N.C.1. 20. 1421 (19741. M. J. ‘Milano, H. C: Pardue, T. E. Cook, R. E. Santino, D. W. Margerum, and M. J. T. Raycheba, Anal. Chem., 46, 374 (1974). M. J. Milano, H. L. Pardue, and R. E. Santino, Anal. Chem., 48, 452 (1976). 21, M. J. Milano and H. L. Pardue, Clin. Chem. ( Winston-Salem, N.C.). 211 (1975). T. A. Nieman and C. G. Enke, Anal. Chem., 48, 619 (1976). T. A. Nieman, F. J. Holler, and C. G. Enke, Anal. Chem., 48, 899 (1976). D. A. Yates and T. Kwana, Anal. Chem., 48, 510 (1976). R. E. Deny, W. G. Nann, G. A. Titus, and W. R. Reynokls, J. Chromatogr. Sci.. 14. 195 (19761. I. M.’Warner J. ’e. Calk, E. R. Davidson, M. Goulerman, and C. D. Christian, Anal. Lett., 8, 665 (1975). T. Vo-Dinh, D. J. Johnson, and J. D. Winefordner, Spectrochim. Acta, Part A , in press. R. P. Cooney, T. Vo-Dinh, and J. D. Winefordner, Anal. Chlm. Acta, in press. J. M. Warner, J. B. Callis, E. R. Davidson, and G. D. Christian, Clin. Chem. ( Winston-Salem, N . C . ) ,22, 1483 (1976). W. H. Woodruff and G. H. Atkinson, Anal. Chem., 48, 186 (1976). R. P. Cooney, G. D. Boutilier, and J. D. Winefordner, to be published. T. C. O’Haver and J. D. Winefordner, J . Chem. Educ., 46, 241 (1969). J. D. Winefordner, Ed., “Trace Analysis, Spectroscopic Methods for Elements”, John Wiley, New York. 1976. R. C. Toiiey and C. D. Scott, Clln. Chem. ( Winston-Salem, N . C . ) ,16, 687 (1970). P. R. Brown, J. Chromatogr., 52, 257 (1970); P. R. Brown, Science, 118, 158 (1972). P. R. Brown, “High Pressure Liquid Chromatography”, Academic Press, New York, 1973. M. C. Bowman and M. Beroza, Anal. Chem., 40, 535 (1968). H. P. Burchfield. R. J. Wheeler, and J. B. Bernos, Anal. Chem., 43, 1976 (1971). D. J. Freed and L. R. Faulkner, Anal. Chem., 44, 1194 (1972). J. W. Robinson and J. P. Goodbread, Anal. Chim. Acta, 66, 239 (1973). H. P. Burchfield, E. E. Green, R. J. Wheeler, and S. M. Biiledeau, J . Chromatogr., 99, 697 (1974). E. D. Pellizzari and C. M. Sparacino, Anal. Chem., 45, 378 (1973). Mark S. Denton, T. P. De Angeiis, A. M. Yacynyck, W. R. Heinman, and T. W. Gullbert, Anal. Chem., 48, 20 (1976). A. McDowell and H. L. Pardue, Anal. Chem., 48, 1815 (1976).
RECEIVED for review December 22,1976. Accepted March 17, 1977. Work supported solely by NIH GM-11373-14.
ANALYTICAL CHEMISTRY, VOL. 49, NO. 7, JUNE 1977
943
