(17) R. G. Steinhardt, J. Hudis, and M. L. Perlman, Phys. Rev. B , 5, 1016 (1972). (18) D. R. Penn, J . Electron Spectrosc., 9, 29 (1976).
RECEIVED for review December 27,1976. Accepted March 28,
1977. We are very grateful to N.R.C. (Canada) for generous financial support, the award of a postgraduate fellowship to J. R. Brown, and the E. W. Steacie Fellowship to G. M. Bancroft.
Comparison of Image Devices vs. Photomultiplier Detectors in Atomic and Molecular Luminescence Spectrometry via Signal-to-Noise Ratio Calculations R. P. Cooney,' G. D. Boutlller, and J. D. Wlnefordner" Department of Chemistry, University of Florida, Gainesville, Florida 326 1 1
Several photomultiplier tubes and image devlces (Image dissector, silicon vidicon, silicon lntenslfied target vldlcon, intensifled silicon Intensified target vidicon, and secondary electron conductlon vldicon) are compared theoretlcally wlth respect to signal-to-noise ratio, S/N,In atomlc and molecular luminescence spectrometry. Equations describing the S/N for the image devices are developed and compared wlth the analogous equations for photomultlpller tubes. The equations are then used to calculate S/N for each device under conditlons likely to be encountered In analytlcally Important sltuatlons. Concluslons based on the calculations Include: (a) the sllicon vidicon Is not analytlcally useful for atomic fluorescence or molecular luminescence spectrometry; (b) the Image dlssector and secondary electron conduction vidicon are potentlally useful in atomlc fluorescence spectroscopy If used wlth echelle spectrometers; (c) In atomlc spectroscopy, whether the optimal system is an SEC echelle spectrometer, an I D echelle spectrometer, a multlchannel direct reader, or slewed scan slngle channel spectrometer wlll depend upon how many spectral lines are to be analyzed and whether or not It has been predetermined whlch lines wlll be examlned; and (d) in molecular lumlnescence spectrometry In the vldble (>350 nm), the Intenslfled hage devices possess considerable analytical potential, combining a multichannel tlme advantage with S/N approachlng those of photomultiplier tubes.
Talmi (1,2) has recently reviewed the principles of TV-type multichannel detectors (image devices) and their uses in analytical spectroscopy. It is quite obvious that multichannel detectors are becoming a greater and greater interest in analytical spectroscopy based upon the applications of such devices in atomic emission (3-20), atomic absorption (21-25), and atomic fluorescence (18, 19) spectroscopy, molecular uv-visible absorption spectroscopy (8, 26-38), molecular fluorescence spectroscopy (39-44), and Raman spectroscopy (45,46). These devices are now available as complete detector systems [silicon vidicon (V), silicon intensified target vidicon (SIT), and intensified SIT (ISIT)] from several commercial companies (Princeton Applied Research, Nuclear Data, EMR, RKB, and Quantex); the detector heads and electronic components are also available on a modular basis from several companies (Reticon, GE, Hamamatsu, EMI, Amperex, ITT, 'Present address, Exxon Chemical Co., Chemical Plant Lab, P.O. Box 241, Baton Rouge, La. 70821. 1048
ANALYTICAL CHEMISTRY, VOL. 49, NO. 7, JUNE 1977
RCA, Fairchild, Teltron, TI, and Westinghouse). In addition to the reviews by Talmi, five other review articles (47-51) have been either partially (51) or totally concerned (47-50) with image device detectors in atomic spectroscopy. In several of the research papers, experimental signal-to-noiseratios (S/N), relative standard deviations, and limits of detection for analyte atoms and molecules have been reported. In only a handful of the research papers, and few of the review papers, have the signal-to-noise ratios been given in terms of the parameters determining them. Indeed, in most of the experimental cases, it is not clear how the S I N has been obtained, i.e., whether the contents of one channel or of many channels are used, and whether there is direct use of information previously stored in digital form or whether the information is converted from digital to analog (graphic) form before use. As a result, it is difficult to compare analytical figures of merit on a systematic basis. In most of the manuscripts considering theoretical expressions for the signal-to-noise ratio, the expressions for signal and noise have been given in analog terms; Le., currents (rates), rather than in digital terms, despite the fact that digital readout is utilized in all of the image devices. The influence of preamplifier (electronic measurement) noise has been stressed as the most significant noise contribution near the limit of detection while the influence of background, luminescence interference, scatter, and dark count noise has usually been neglected. Finally, the present authors feel that there has been no fair comparison, based on calculated signal-to-noise ratios, of the image devices with photomultipliers or of the image devices with each other; also few fair comparisons have been made based on experimental figures of merit between various image devices and between image devices and photomultiplier tube detectors. Therefore, it is the purpose of this paper to compare the signal-to-noiseratios of several image devices (Si-vidicon, SIT, ISIT, SEC-vidicon,and Image Dissector) with each other and with several sensitive photomultiplier tubes commonly used in optical spectroscopy. The comparison will be made for several hypothetical, but realistic, experimental situations in atomic and molecular luminescence spectrometry. The S/N of the image devices will be estimated for these realistic experimental conditions either by utilizing the contents of a single channel or by pooling (summing) the contents of several channels (which generally leads to an increase in S/N). Finally, some discussion will be concerned with the calculation of S/N in terms of currents (see Appendix). Luminescence Signal-to-NoiseRatio Expressions. In luminescence spectrometry, the analyte luminescence signal SL is given by
SL =
(L,B,S,I,D) - S(B,S,I,D)
(1)
where S(L,B,SJ,D)is the signal resulting from the sample and contains contributions resulting from analyte luminescence (L),nonsource induced background (B,e.g. flame background emission in atomic fluorescence spectrometry), scatter of source radiation (S),source induced interfering luminescence background (I),and the background resulting from the detector and measurement system ( D ) ;and S(B,S,I,D) is the signal resulting from the blank, which ideally is the same in all respects as the signal resulting from the sample except for the absence of analyte luminescence (i.e., no analyte impurity is assumed present in the blank). Assuming negligible statistical dependencies between the noise sources (except for several of the flicker noises, discussed below), the total noise due to contributions from all sources is given (52, 53) by
where the terms are defined as: N x is~ the shot noise due to process X (X= L for analyte luminescence, X = D for dark current, X = B for nonsource induced background, X = I for source induced background, and X = S for source induced scatter); the corresponding N X Fterms represent the flicker type noises associated with the same processes giving rise to the N X Sterms; and NA represents the amplifier-electronic measurement system noises combined into one term. It is assumed that the analyte luminescence is present only in
measurements made on the sample while all other contributions are present both i n measurements made on the sample and (to the same extent) i n measurements made on the blank. Thus the coefficients of 1for NLs2and NLF:!arise from the assumption that analyte luminescence contributes only to the sample signal (and noise). The coefficients of 2 for the other shot noise terms arise from the fact that the shot noise arising from process X measured in the sample, and the shot noise arising from the same process X measured in the blank are independent, and therefore add quadratically. The coefficients of 4 or 2’ for the corresponding flicker noise terms arise from assuming that the flicker noise arising from process X measured in the sample and the flicker noise arising from the same process X measured in the blank are dependent, and therefore add linearly. In actuality, it is likely that the flicker noises arising from a given process X are to some extent both dependent and independent between sample and blank measurements. However, such considerations are beyond the scope of this paper and, even if treated here, would result in no significant changes in the conclusions reached. The flicker noises from the analyte luminescence, the interference luminescence, and the scatter are assumed to be dependent (Le., they all result from flicker in the source) and therefore will add linearly (summed before squaring) to each other. The signal-to-noise ratio (S/N) expressions below will follow those of the approaches taken by Winefordner et al. (54) in comparing multichannel vs. single channel vs. multiplex methods and by Boutilier et al. (55) in comparing pulsed source/gated detector systems (with and without time resolution) vs. the cw source/cw detector system. All calculations in the manuscript will be carried out in terms of total counts (digital) rather than in terms of currents (see Appendix) or in terms of count rates. For nonintegrating devices, such as photomultiplier tubes and image dissectors, the integral number of counts in an observation time tocorresponds to the product of the counting rate due to some process times to; while the same is true for integrating devices, such as conventional vidicons (Sb& or PbO photoconductors), Si target vidicons, SEC vidicons, and intensified target vidicons (SIT,
ISIT or EBS), the observation time is divided into integrating periods (when photon induced charge is stored by the image device) and scanning periods (when the stored charge is read by an electron beam). Thus,the total observation time is given by
to
=
tsn, + tdn,nd
(3)
where t , = time per scan, s (32.8 ms for vidicon with PARom);t d = time per integration, S (integrations called delays on PAR-OMA); n, = number of scans, dimensionless; nd = number of integrations (delays) before each scan (from zero to thousands with SEC or cooled vidicons) assumed to be the same for each scan, dimensionless. If td = t,, then Equation 3 reduces to
to = t,n,(l
+ nd)
(4)
Generally, the channels of an image device are scanned sequentially, and the number of integrations before each scan is constant. With computer control, it is feasible to have random access channel readout with variable numbers of integrations before a scan. The signal-to-noise ratio for image devices, (S/N) image, is given by
(S/N)mage = (toxE~,Atqz)/ {to [CA,qt(ELz
2&,
2E4 f 2Est) -k 4xRDj] + t o * [ ~ A t q t t s , ( E+~2Erl , +2~s~)l f
t o ’ ( x 2 A t q ~ t ~ f~ E4t0’E ~ ~ )(~& I ~ R D ~ ) ~
+ 4ZNA2}”’
(51
where Ex,= the photon irradiance in channel i arising from process X (X= L for luminescence irradiance of analyte, X = B for source independent background, X = I for source induced luminescence interference, and X = S for source induced scatter), photons s-’ cm-’; [x,= flicker (fluctuation) factor for process X in channel i, dimensionless; A, = area of channel i, cm2; qL= efficiency of conversion of photons to ~ counts in channel i, dimensionless, (counts photon-‘); R D = detector dark count rate in half (either upper or lower, as discussed below) of channel i, (count&-’; NA, = preamplifier-electronics noise per half channel i per observation time = (as defined in Equation 6), dimensionless (counts); and summation over the appropriate number of channels. The expressions for the dark current shot noise and for the dark current flicker noise arise because of the manner in which many of the image devices are normally operated. In some devices each channel may be divided in half, with radiation striking the “upper” half and no radiation striking the “lower” half. When a channel is read, the contents of the “lower” half (dark current) are subtracted from the contents of the “upper” half (dark current plus photon generated current) so that the counts stored in memory are those resulting from the photon generated current. This subtraction occurs for both the sample and blank measurements. Even though the dark currents in the two channel halves cancel out (ideally),the associated shot noises still add quadratically; the flicker noises are assumed to add quadratically between sample and blank and quadratically between channel halves and adjacent channels. Likewise, the preamplifier, which is assumed to be the principal source of electronic measurement noise, operates during the readout of each half channel, and thus will make four noise contributions per channel per pair of sample-blank measurements. Since the preamplifier electronic noises are assumed to be random (rms) noises, the total preamplifier noise is dNA2+NA2+NA2+NA2 or Finally, it should be noted that the analyte luminescence, interference luminescence, and scatter flicker noises are all given the same
m.
ANALYTICAL CHEMISTRY, VOL. 49, NO. 7, JUNE 1977
1049
Table I. Comparison of Detectors Parametera Internal gain Efficiency, Q, visible
uv
SiVidiconbic -1
-
103
x
5
x
2
2X
Dark count rate (20 “c)(s-l)e Area per channel, cmz f
S I T ~ ~ C IS IT^,^
5 2
x
- 2 x 103
SEC-Vidic0nb.C
Image Dissect0rb.c
PM(6256EMI)~
PM(9558EMI)~
PM(F4013ITT)~
-102
-lo6
-107
-107
-107
2 x 10-l
2 x 10-1
2
1 x 10-1 2 x
loF4
1X 5
10-3
1 x10-3
lo-’
2
x 10-1
x
4X < I x 10-2
2 X lo-’ < I x 10-1
1.7 X 10‘’ 1x 102
2.5 X 10” 1 x 104
5 x
1 x 10-3 1 x 10-3
< I x 10-4
...
...
...
2X 5
lo-’
10-1
1.9 X lo-’
loo
a Parameters: gain = number of charged pairs produced at target (V, SIT, ISIT, SEC) or electrons produced. q is in counts/photon. Count rate = photon flux X channel area X efficiency (9). Dark count rate = dark counts per second per Data taken channel for image devices, Data taken from OMA catalog, Princeton Applied Research, Princeton, N.J. from “The Optical Data Di itizer-An Applications Guide”, EMR Schlumberger, Princeton, N.J., and Y. T. Tdmi, Anal. Chem., 47, 697A (1975). Data taken from EM1 Catalog, EM1 Electronics, Ltd., Middlesex, 6B3, lHJ, England, and ITT Catalog. e Dark count rates are defined for an entire channel. On some commercial devices half the tube face is blocked off and used for automatic dark current subtraction; thus for half a channel which is illuminated, the dark count rate will be one-half of those in this table. Area per channel for SiVidicon, SIT, and ISIT taken from PAR data; area per channel for SEC-Vidicon taken from Talmi; area for image dissector taken as area of aperture; area for PM is same as slit area as long as slit area > means greater than a factor of 10 (usually greater than 100); > means a factor of 2 to 10; and 2 means within a factor of 2.
General Conclusions (common to both atomic and molecular luminescence spectrometry). (i) For integration times less than about 0.4 s [12 integrations (delays)] at 32.8 ms per integration (delay), preamplifier noise for image devices dominates over dark current shot noise at room temperature. The preamplifier noise can be reduced by the factor Jl‘i-+n, where n d is the number of delays (integrations; if nd = 0, this implies for the PAR-OMA that the time between consecutive readouts of a given channel is 32.8 ms). (ii) If detector shot noise, preamplifier noise, photon shot noise, or any combination of these shot noises is limiting, then the summation of signals and noises over n, channels of an integrating image device (V, SIT, ISIT, or SEC) will improve the S/N by a factor of assuming the signal and noise are constant over n, channels. (iii) Assuming that the PM is photon shot noise limited (i.e., a 1.05-s observation time), and that signal and noise are measured in one channel of the image device, then the S/N decrease in the following order: PM > ISIT k SIT, SEC R ID >> V. (iv) Assuming photon (shot or flicker) noise limitation, with an observation time of either 1.05 s or 32.8 s, and the signal and noise for the image devices being summed over n, channels, the S/N decrease in the following order: PM k ISIT k SEC, SIT 5 ID >> V. (v) If source related flicker noises (interferences, scatter, or luminescence of analyte) are dominant, then the S/N of image devices vs. PMs will not improve with integration, summation over channels, or longer observation time. (vi) The silicon vidicon, V, device is not analytically useful for atomic fluorescence or molecular luminescence spectrometry. Cooling would reduce the dark current and enable longer target integration times. This should reduce the preamplifier noise contribution which would result, in principle, in a much improved S/N. This improvement of course comes at the expense of longer analysis time and the accompanying problems with long term noise (drift). A cooled vidicon has great potential for measuring weak sources as in astronomy, but is predicted to be of little or no analytical use. Experimentally, the vidicon has been shown (18)to have little analytical use as a detector in atomic fluorescence flame spectrometry. Cooling the SIT will result in significant improvement in S/N in the visible region but at greater expense and difficulties. No S/N ratios were calculated for the cooled SIT due to a lack of reliable characteristics. (vii) The image dissector, ID, has a small format and small aperture requiring a spectrometer of small format, “high” resolution, and high luminosity (throughput) such as an echelle. For conventional unidirectional (not crossed) dispersion spectrometers, the ID is much less attractive (smaller S/N) than the P M even though the ID can slew scan at a
6
ANALYTICAL CHEMISTRY, VOL. 49, NO. 7, JUNE 1977
1051
Table IIA. Signal-to-Noise Ratios for Atomic Fluorescence Spectrometryi Single channel devicesC Conditions
Photomultiplier tubesd with synchronous photon counter EM1 EM1 ITT 9558QB 6256B F4013 1.05 6.56 32.8 1.05 6.56 32.8 1.05 6.56
Device Count time, sa log log log logb EL
EB
5 5 5 5 5 5 5 5 7 7 7 7 7 7 7 7
3 3 3 3
7 7 7 7 3 3 3 3 7 7 7 7
EI
3 3 7 7 3 3 7
7 3 3 7
3 7 3 7 3 7 3 7 3 7 3
I
7
3 3
3 7 3
I
0.90 2.30 0.57 1.10 0.57 1.10 0.41 0.68 0.48 1.20 0.40 0.88 0.40 0.88 0.34 0.62 69.0 157.0 48.0 85.0 48.0 85.0 37.0 57.0 44.0 105.0 36.0 73.0 36.0 73.0 31.0 53.0
I
7
Conditions Device Count time, sa log log log logb EL
EB
5 5 5 5 5 5 5 5 I 7 7 7 7 7 I 7
3 3 3 3 7 7 1 7 3 3 3 3 7 7 7
I
7
7
3 7 3
7
1
3 3 7 1 3 3 7 7 3 3
3 7 3 1 3 7 3 7 3 7 3 7
Device Count time, sa log log log logb EL
EB
5 5 5 5 5 5 5 5 7 7 7 7 7
3 3 3 3 7 7 7 7 3 3 3 3 7
5.0 1.40 1.40 0.78 2.70 1.30 1.30 0.76 260.0 104.0 104.0 64.0 200.0 99.0
5.10 13.0 28.0 9.20 23.0 51.0 0.91 2.30 0.62 1.20 1.20 1.50 0.66 1.50 0.07 0.18 0.62 1.20 1.50 0.66 1.20 1.50 0.07 0.18 0.41 0.68 0.80 0.44 0.70 0.80 0.05 0.13 0.47 2.60 0.51 1.20 2.80 0.05 0.13 1.20 0.37 0.83 1.40 0.41 0.87 1.40 0.04 0.10 0.37 0.83 1.40 0.41 0.87 1.40 0.04 0.10 0.31 0.17 0.09 0.59 0.77 0.33 0.60 0.04 90.0 191.0 280.0 96.0 200.0 25.0 286.0 10.0 49.0 86.0 104.0 51.0 88.0 105.0 15.0 5.9 49.0 86.0 104.0 51.0 88.0 105.0 5.9 15.0 36.0 56.0 64.0 57.0 4.6 12.0 37.0 64.0 42.0 100.0 192.0 44.0 106.0 200.0 4.6 12.0 34.0 70.0 97.0 72.0 98.0 9.7 36.0 3.9 99.0 34.0 70.0 97.0 36.0 72.0 98.0 3.9 9.7 63.0 28.0 51.0 62.0 30.0 52.0 63.0 3.4 8.3 Multichannel image devicesc,f 4 channels illuminated with 1 channel measured 0.0984-s integration SECg V SIT ISIT 1.05 32.8 1.05 32.8 1.05 32.8 1.05 32.8
5.10 0.39 0.39 0.21 0.29 0.23 0.23 0.20 56.0 31.0 31.0 24.0 26.0 21.0 21.0 18.0
EI
3 3 7
Conditions
1052
32.8
Synchronous photon counter Imagee dissector 1.05 6.56 32.8
0.59
0.006 0.03 0.20 1.1 0.69 3.8 0.006 0.01 0.41 0.11 0.59 0.03 0.07 0.006 0.03 0.01 0.41 0.11 0.59 0.05 0.05 0.29 0.006 0.03 0.08 0.40 0.07 0.006 0.03 0.07 0.32 0.11 0.39 0.05 0.006 0.03 0.05 0.26 0.08 0.33 0.05 0.006 0.03 0.05 0.26 0.08 0.33 0.04 0.006 0.03 0.05 0.22 0.07 0.28 10.0 3.1 10.0 55.0 0.55 16.0 86.0 5.9 0.52 9.3 47.0 2.9 6.3 33.0 0.52 6.3 33.0 9.3 47.0 5.9 2.9 4.6 0.49 2.8 7.2 34.0 5.0 26.0 0.52 6.2 27.0 9.1 36.0 5.9 2.9 4.6 0.49 2.8 1.1 30.0 4.9 24.0 0.49 4.9 24.0 7.1 30.0 4,6 2.8 3.8 0.46 2.7 4.2 20.0 6.0 26.0 Multichannel image devicesc,f 4 channels illuminated with 4 channels measuredh 0.0984-s integration SECg V SIT ISIT 1.05 32.8 1.05 32.8 1.05 32.8 1.05 32.8 0.01
5.4 0.39 0.39 0.27 0.31 0.25 0.25 0.21 56.0 32.0 32.0 24.0 27.0 22.0 22.0 19.0
EI
3 3 i 7 3 3 7 7 3 3 7 7 3
3 7 3 7 3 7 3 7 3 7 3 7 3
1.2 0.14 0.14 0.10
0.14 0.10 0.10 0.08 20.0 12.0 l2.0 9.1 11.0
11.0
0.72 0.72 0.47 0.43 0.37 0.37 0.32 108.0 57.0 57.0 41.0 40.0
0.011 0.011 0.011
0.010 0.011 0.010 0.010 0.010 1.1 1.0 1.0
0.98 1.0
ANALYTICAL CHEMISTRY, VOL. 49, NO. 7, JUNE 1977
0.07
0.06 0.06 0.06 0.06 0.06 0.06 0.06 6.2 5.7 5.7 5.3 5.7
0.40 0.15 0.15 0.11
0.14 0.11 0.11 0.09 20.0 12.0 12.0 9.8 12.0
2.2 0.75 0.15 0.49 0.43 0.38 0.38 0.33 106.0 59.0 59.0 43.0 40.0
1.4 0.22 0.22 0.16 0.21 0.15 0.15 0.13 32.0 18.0 18.0 14.0 17.0
7.7
1.o
1.0
0.61 0.46 0.42 0.42 0.37 156.0 76.0 76.0 51.0 45.0
Table IIA (Continued) Conditions Device
Count time, sa log log log logb *L
ES
Multichannel image devicesc,f 4 channels illuminated with 4 channels measuredh 0.0984-s integration SECg V SIT ISIT 1.05 32.8 1.05 32.8 1.05 32.8 1.05 32.8
EI
9.0 34.0 0.98 5.3 9.7 35.0 14.0 40.0 9.0 34.0 0.98 5.3 9.7 35.0 14.0 40.0 7 7 7 7 7.6 29.0 0.94 5.2 8.3 31.0 12.0 35.0 a 1.05 s corresponds to 32 machine cycles and 32.8 s to 1000 machine cycles at a cycle time of 32.8 ms. Photon irradiance units of photons s-’ ern-', Flicker (fluctuation) factors t s = 0.003; t B = 0.01 for multichannel image devices and 0.00 for PM and ID. See Table I for detector efficiency and dark count rate used in calculations. Slit height 1 cm, slit width 0.01 cm. e ID aperture of cm’. f Devices assumed to be 1 cm high, 1.25 cm wide, and to contain 500 channels with lower half of each channel used for dark current subtraction. See g for SEC. g Since the SEC has “no” dark current, the entire channel height is used for measurement of luminescence. Also, since the SEC is strictly an integrating device, it is read out only at the end of a measurement. 4 channels have the same width as the slitwidth used for the photomultiplier tubes (0.01 cm). Tables with no on-target integration are available o n request from one of the authors (JDW). 7
7
3
7
7
7
7
3
much higher rate than conventional spectrometers. For a complex spectrum requiring high resolution and many spectral measurements (many resolution elements), the ID with an echelle spectrometer becomes more attractive (9,10) compared to the P M or to the integrating type image devices (SIT, ISIT, and V); however, for such situations the SEC would appear to be the most generally useful detector (11,56, 57). (viii) The linear dynamic range (52,58) of the P M exceeds that of all the image devices because of its high sensitivity, which results in a lower limit of detection, and its immunity to target saturation which results in a higher upper limit compared to the SIT, ISIT, and SEC. The ID has a linear dynamic range less than that of the photomultiplier, but greater than that of the integrating image devices. (ix) Image devices, except for the ID, are plagued with background noise problems (Le., radiation not connected with the source), particularly in atomic fluorescence flame spectrometry. Synchronous operation reduces it, but is not highly efficient because of the limited repetition rate imposed by the scan time of vidicon devices. (x) The photon irradiance (photons cm-’ s-’) necessary for a detector to become photon noise limited (shot or flicker) is lower for more sensitive detectors, is greater for photon flicker than photon shot noise, and is greater in atomic than molecular luminescence spectrometry (because of the larger slit widths typically used in molecular luminescence spectrometry). (xi) It should be stressed that all of the S/N results for the P M cases given in this paper are concerned with the measurement of a single spectral interval. If m spectral intervals are measured via a sequential linear scan spectrometer (SLS) or r(m 2 r ) spectral components are measured via a sequential slew scan spectrometer (SSS) during a given measurement time to,the S/N for any given spectral interval (channel) will necessarily decrease (or at best not change). The effect of the optical measurement systems upon S/Nhas already been discussed by Winefordner et al. (54).
Specific Conclusions Regarding Atomic Fluorescence Spectrometry. (i) Assuming source related photon shot noise for the P M (1.05-s observation time), if one channel is measured for the integrating image devices, then the S / N decreases in the following order: P M > ISIT 2 SIT, SEC > ID >> V. If four channels of the integrating image devices are summed, or if an observation time of 32.8 s is used (in which case, the PM becomes primarily flicker noise limited), the PM and ISIT are within a factor of 2.
(ii) Assuming detector-preamplifier noise limitation and an observation time of 32.8 s for the image devices, and an observation time of 6.56 s for the PM, and summing over four channels with three integrations between scans, the S/N decrease in the following order: F4013 2 6256B 2 SEC 5 ISIT 5 SIT, 9558QB, ID >> V. This would be the case for 5 spectral lines measured by a slew scan system with P M or ID and all five measured simultaneously with the integrating image devices. This is approximately the number of useful analytical lines in a 20-nm spectral window (19). (iii) Assuming background noise limitation (ELI10-2E~, e.g., EL = lo5 and EB = lo‘ photons s-l crn-’), for an observation time of 6.56 s for P M and ID and 32.8 s for integrating image devices, the S/N decreases in the following order: P M > ISIT, SEC, SIT > ID >> V. This is again the case for 5 spectral elements measured by slew scan compared to simultaneous measurement with integrating image devices. (iv) (a) The maximum number of resolution elements required for the measurement of complex atomic spectra as in atomic emission spectrometry of high temperature plasmas is approximately 6 X lo4 (800-200 nm divided by the spectral bandwidth or line width, 6X 0.01 nm). The maximum number of resolution spectra (AAS) or atomic fluorescence spectra (AFS) is approximately 6 X lo3 (i.e., the value of 6X can increase to 0.1 nm with AAS and AFS and still not have significant spectral overlap in most cases). (b) Because the number of effective channels (1,2,31,51) of V, SIT, and ISIT is = 200 (500 actual channels, but spreading of the images results in a loss here, one channel is taken to spread into “on the average’’ 2.5 channels), the number of spectral windows needed to cover 200-800 nm in AES is -300 and in U S , AFS, or flame AES is -30 assuming all windows contain sufficient spectral information to warrant measurement. Although the AES high temperature plasma case is hardly a tenable one, the AAS, AFS, and flame AES cases appear to be potentially useful. However, such an image device system must be weighed against a direct reading multichannel system as well as a slewed scan single channel P M case. Finally, the usefulness of the image device may increase in AAS, because the windows can be increased considerably with the concomitant loss in S/N assuming appropriate line sources with simple spectral outputs are available. (c) For the case where a large number (say 10-100) of spectral lines in a complex spectrum, e.g., emission spectrum from a high temperature (>4000 K) plasma, must be resolved and measured, an echelle spectrometer with the SEC is a more
-
ANALYTICAL CHEMISTRY, VOL. 49, NO. 7, JUNE 1977
1053
Table IIIA. Signal-to-Noise Ratios for Molecular Luminescence Spectrometry’ Single channel devicesC Photomultiplier tubesd with photon counting Conditions EM1 EM1 ITT 9558QB 6256B F4013 Device 1.05 32.8 1.05 32.8 1.05 32.8 Count time, sa log log logb EL Es EI 4 2 2 8.1 10.0 56.0 14.0 75.0 1.4 1.6 4 2 6 1.6 0.87 1.6 0.87 0.75 4 6 2 0.87 1.6 0.87 1.6 1.6 0.75 4 6 6 0.81 0.55 0.81 0.55 0.81 0.51 6 2 2 98.0 290.0 130.0 130.0 310.0 310.0 6 2 6 61.0 67.0 67.0 110.0 110.0 110.0 6 6 2 61.0 67.0 67.0 110.0 110.0 110.0 6 6 6 44.0 65.0 66.0 46.0 66.0 46.0 Conditions Device Count time, log EL 4 4 4 4 6 6 6 6
log Es 2 2 6 6 2 2 6 6
Conditions
SECg sa
1.05
32.8
0.33 0.05 0.05 0.04 7.1 4.2 4.2 3.2
3.7 0.28 0.28 0.20 40.0 23.0 23.0 17.0
Multichannel image devicescgf 40 channels illuminated with 1 channel measured 0.0984-s integration V SIT 1.05 32.8 1.05 32.8
Photon counting Imagee dissector 1.05 32.8
0.32 0.03 0.03 0.02 4.6 2.6 2.6 2.0
1.8
0.18 0.18 0.13 25.0 15.0 15.0 11.0
ISIT 1.05
32.8
0.36 0.08 0.08 0.06
2.0 0.43 0.43 0.29 60.0 35.0 35.0 26.0
logb
EI 2 6 2 6
2 6 2 6
0.0006 0.003 0.10 0.56 0.0006 0.003 0.05 0.27 0.0006 0.003 0.05 0.27 0.0006 0.003 0.04 0.20 0.06 6.3 35.0 0.31 0.06 0.31 4.2 23.0 0.06 0.31 4.2 23.0 0.06 0.31 3.4 18.0 Multichannel image devicesc*f 40 channels illuminated with 40 channels measuredh 0.0984-s integration SECg V SIT 1.05 32.8 1.05 32.8 1.05 32.8
11.0
6.5 6.5 5.1
Device ISIT Count time, sa 1.05 32.8 log log logb EL Es EI 4 2 2 0.004 0.02 2.1 24.0 0.63 3.5 2.3 13.0 0.004 0.02 0.31 1.2 0.31 1.4 1.2 0.48 4 2 6 0.004 0.02 0.31 1.2 0.31 1.4 4 6 2 1.2 0.48 0.004 0.02 6 0.22 0.70 4 6 0.23 0.76 0.70 0.33 6 2 2 45.0 200.0 40.0 190.0 68.0 250.0 0.35 2.0 6 2 6 26.0 89.0 100.0 26.0 89.0 39.0 0.35 2.0 26.0 89.0 6 6 2 26.0 89.0 39.0 100.0 0.35 2.0 21.0 0.35 20.0 58.0 58.0 29.0 62.0 6 6 6 2.0 a 1.05 s corresponds to 32 cycles and 32.8 s is 1000 cycles at the standard cycle time of 32.8 ms. Photon irradiance, units of photons s-’ cm-a, Flicker (fluctuation) factor ts = 0.003. See Table I for detector efficiency and dark count rate used in calculations. cm’. f Devices assumed to be 1.0 Slit height 1 cm, slit width, 0.1 cm. e ID aperture of cm high, 1.25 cm wide, and to contain 500 channels with the lower half of each channel used for dark current subtraction. See g for SEC. g Since the SEC has “no” dark current, the entire channel height is used for measurement of luminescence. Also, since the SEC is strictly an integrating device, it is read out only at the,end of a measurement. 40 channels have the same width as the slit width used for the photomultiplier tubes (0.01 cm). Tables with no integration available on request from JDW. analytically useful detector because the number of channels is greatly increased (>2000) (11,56,57). (d) For the case where a smaller number of spectral lines (=IO) in a complex spectrum must be resolved and measured, an echelle spectrometer with ID would be an analytically useful system. Because of the rapid slew rate of the ID, the additional use for observing rapid transients is also possible
(9,10). (e) For the case where a predetermined large number (10-50) of spectral lines in a complex spectrum must be resolved and measured, a multichannel direct reader is 1054
ANALYTICAL CHEMISTRY, VOL. 49, NO. 7, JUNE 1977
certainly optimal (54), whereas for the measurement of a smaller number (210)of predetermined spectral lines in a less complex spectrum, a slewed scan single channel spectrometer with PM is equivalent to the direct reader (54). (v) I t should be again noted that in the calculations for atomic fluorescence we have assumed use of transitions in the ultraviolet. For transitions occurring at wavelengths greater than 350 nm, Winefordner et al. (53) have shown that atomic emission should be superior to atomic fluorescence for analytical purposes with conventional light and atomization sources. Howell and Morrison (59) have extensively compared
experimentally the V, SIT, and PM (1P28) for flame atomic emission and concluded that the SIT provides excellent analytical performance for elements with analytical lines above 380 nm. Specific Conclusions Regarding Molecular Luminescence Spectrometry. (i) Assuming photon noise limitation (shot or flicker), if either a 1.05-s or a 32.8-s observation time is used, and 40 channels are summed for the integrating image devices, then the S / N decrease in the following order: P M 5 ISIT 5 SIT, SEC > ID >> V. (ii) Assuming photon noise limitation (shot or flicker), if only one channel of the image device is measured, then for a 1.05-s or 32.8-s observation time, the S/N of the integrating image devices become several times poorer than the S / N of the PM. (iii) Assuming preamplifier noise limitation, if a 32.8-9 observation time is used, and there is summation over 40 channels of the integrating image device, then the S / N decrease in the following order: P M > SEC > ISIT L SIT L ID >> V. (iv) Because the S/N of image devices (ISIT, SIT, and SEC) are close to the S / N of P M for the case of photon (shot and flicker) noise limitation (which is common in analytical luminescence spectrometry), these image devices possess a time advantage since they record many spectral components simultaneously (multichannel advantage). With regards to quantitative analysis, similar detection limits should be achieved with the SIT, ISIT, SEC, or PM. Vo-Dinh (40) et al. have obtained similar detection limits for several organic molecules by conventional spectrofluorimetry using either an SIT or PM detector (PM about 5-fold better). (v) The integrating image devices (ISIT, SIT, SEC, and V for high incident light photon fluxes) have considerable analytical potential. They perform adequately compared with a P M when measuring steady-state molecular luminescence or absorption. In addition, they function particularly well when measuring transient molecular luminescence or absorption, e.g., when acting as a detector for liquid or gas chromatography. Rapid scan spectrometers with PM or ID would be expected to result and generally have resulted in reduced S/N ratios for the large wavelength regions scanned in molecular luminescence (absorption) spectrometry. In comparison, the integrating image detectors are capable of giving more spectral information, with greater spectral fidelity and little loss in S / N ratio as compared to single channel measurements with a P M system.
APPENDIX I. Signal-to-Noise Ratio: Currents. The data of interest to the user of commercial image devices is in the form of counts rather than in the form of currents (or counting rates). Nevertheless, much commercial literature and some research articles give the signal-to-noise ratio of image devices in terms of currents, i.e.,
and the noise current is increased from ZT to d z ~The . fallacy of this argument is twofold: (1)as seen above, increasing the number of scans increases the total charge or counts accumulated but has nothing to do with the rate a t which the charge (counts) is read out, i.e., current; ( 2 ) if the current noise increases by d& then the implicit assumption is that the noise is random noise and not proportional noise (such as flicker or whistle). As will be seen below, even though AiT does not depend upon the number of scans, it does depend upon the exposure time (the time between scans). Therefore AiT will depend upon td and nd as well. At any rate, most users of image devices are interested primarily in the signal-to-noise characteristics of the total number of counts (equivalent to the total charge accumulated) in memory rather than in the signal-to-noise characteristics of the rate (current) at which charge is transferred from the target and stored in memory (in the form of counts); and such (current) information, even when available is seldom of use to the analytical spectroscopist performing the measurements. Regardless, in order to complete the discussion undertaken in this paper, and hopefully to prevent some confusion by others, we will present a signal-to-noise expression in terms of currents. The expression for the signal-to-noise ratio is:
(S/N)image=
(iL)/[(ELsi2 + 2iiiBsi2 + 2ails,2
where the numerical coefficients of the different terms result from the same experimental measurement -considerations given in the discussion of Equation 5 . Aixs, represents the shot (rms) noise current due to process X in channel - i, A; represents the preamplifier noise current, A; AixFLrepresents the proportional (mainly flicker) noise current due to process X in channel i, A. The value of signal-to-noise obtained using this expression will be smaller than the value obtained using the count expression (Equations 5 and 59, if n, > 1. The current terms in Equation A2 can be expressed in terms of the operating parameters of an image device as follows:
where e = charge of electron, C; qi = photon to photoelectron conversion efficiency of the photocathode, dimensionless; G = gain of device, dimensionless; Ex,= photon irradiance, photons cm-’s-’; t, = scan time, s; t , = exposure time, s ( t , = t, tdnd with td and nd as defined earlier); t, = read out time of channel i, s; A, = area of channel i, cm’.
+
-
where i~ is the signal current due to the analyte luminescence and AiT is the total rms noise current due to all sources. These currents represent the rate at which charge is read from the target during any one individual scan. The magnitudes of iL and ZTdepend upon the amount of charge stored on the target from one readout to the next and on the time it takes to read out the accumulated charge during each scan, and not on how many scans are made. However, some readers may infer from various articles that if the number of scans is increased by a factor of n,, then the signal current iL is also increased by the same factor to nsiL,
where Af = noise equivalent bandwidth of measurement system ( A f = f,- f i or 1/24, where t , is the averaging time of the system), h; ix, = current given in Equation A3 from a photon induced process, or from dark current, A;
where &, = flicker (fluctuation) constant for process X in channel i; dimensionless;
where A, = the equivalent target electron uncertainty per ANALYTICAL CHEMISTRY. VOL. 49, NO. 7, JUNE 1977
1055
channel (for PAR-OMA, it is -2500 rms electrons per channel), dimensionless. The signal-to-noise expression in terms of current for a photomultiplier (nonintegrating) detector yields the same results as the signal-to-noise expression in terms of counts (or count rates) and so will not be given here. It should be noted that the analogs of Equations A2-A5 for the PM differ from the ones in the text in that the sub is are omitted. In addition, for the PM analog of Equation A3, t,/t, = 1;and there is no analog to Equation A6 because is or can “always” be made negligible for PM.
11. Regarding Definitions, Terms, and Units with Image Devices. A. Preamplifier noise for image device systems is often recorded in terms of rms counts (or electrons) per channel per frame (scan). It should be stressed that the designation of 1 count (-2500 electrons for PAR-OMA system) per channel per frame is an rms noise count rate and not a signal count rate. B. Dark currents for image devices are listed typically either as currents, readout count rates, or electron fluxes. Several calculations of dark current magnitudes (for the PAR-OMA system) will be given which illustrate, among other things, that care must be taken to distinguish between the rate at which dark current charge accumulates, and the rate at which it is read out from the V, SEC, SIT, ISIT target. (i) If the dark current read out for the entire image device is 8 X lo3 A, then, assuming a 500-channel device, the conversion to dark current in terms of the number of counts accumulating (on the target surface) per channel per second is given by:
lo-’- Cs
8X
X
1electron 1 X 1.6X lo-’’ C 500 channel
32.8 X s 1r e a d o u t X readout accumulation 1accumulation 1 count X X 32.8 X s 2500 electrons 40 counts channel s (ii) The conversion from a readout count rate to an accumulation count rate is performed similarly, i.e., assuming a readout count rate of 4.0 X lo4 counts channel-’ s-’, then the accumulation rate is given by:
counts
32.8 X s readout 1readout 1accumulation X X accumulation 32.8 X s - 40 counts -
x io4 channel s
4.0
X
channel s (iii) A readout electron flux of 8 X lo1’ electrons cm-’s-l can be converted to an accumulation rate (counts channel-’ s-l) as follows: electrons 12.8 X cmz X channel cmz s 32.8 X 1readout
8 X 10” X
lom6
readout accumulation 1 accumulation 1 count X 32.8 X s 2500 electrons 40 counts channel s 1056
ANALYTICAL CHEMISTRY, VOL. 49, NO. 7, JUNE 1977
The above calculations are meant only to illustrate the procedure used to interconvert the various literature values for dark count rate. The authors do not imply, or mean to imply, that the values chosen in the examples above are the actual values for any specific image devices. C. Fixed pattern noises do exist with image devices and have been given names such as coherent noise, hash noise, sinusoidal noise, etc. These noises seem to depend to a great extent on the specific instrument used (its adjustment, grounding, and so forth). As a result, such noises are difficult to characterize in terms of mathematical expressions. Therefore, even though these noises can become the limiting noises (under conditions of low light flux measured for periods of a few seconds or less) when the instrument has not been set up and adjusted optimally, they have not been considered in this discussion. D. The detector flicker factor was assumed to be zero in calculations and detector flicker noise was assumed to add quadratically for sample and blank measurements with integrating image devices. If the detector channels are not split in half with one half blocked for dark charge subtraction, then detector flicker noise will add linearly from sample to blank. Assuming detector flicker noise is the same within a channel, then operation in the “on-target’’ dark charge subtraction mode has the same effect in reducing detector flicker noise as optical modulation has in reducing flame background flicker in atomic fluorescence and absorption spectrometry. A. van der Ziel has extensively discussed noise in photoconductive detectors (60).
LITERATURE CITED Y. Talmi, Anal. Chem., 47, 685A (1975). Y. Talmi, Anal. Chem., 47, 697A (1975).
K. W. Busch, N. G. Howell, and G. H. Morrison, Anal. Chem., 48, 575 (1974). K. W. Busch, N. G. Howell, and G. H. Morrlson, Anal. Chem., 46, 1231 (1974). K. W. Busch, N. G. Howell, and G. H. Morrison, Anal. Chem., 48, 2074 (1974). N. G. Howell, J. D. Ganjei, and G. H. Morrison, Anal. Chem., 48, 319 (1976). J. D. Ganjei, N. G. Howell, J. R. Roth, and G. H. Morrison, Anal. Chem., 48, 505 (1976). M. J. Milano, H. L. Pardue, T. E. Cook, R. E. Santlni, D. H. Margerum, and J. M. T. Raycheba, Anal. Chem., 48, 374 (1974). A. Danielsson, P. Lindblom, and E. Sijderman, Chem. Scr., 8, 5 (1974). A. Danielsson and P. Lindblom, Phys. Scr., 5, 227 (1974). D. L. Wood, A. B. Dargls, and D. L. Nash, Appl. Spectrosc., 27, 310 (1975). G. Horlick and E. G. Coddlng, Anal. Chem., 45, 6490 (1973). E. G. Codding and G. Horlick, Appl. Spectrosc., 27, 386 (1973). E. G. Codding and G. Horlick, Spectrosc. Lett., 7, 33 (1974). G. Halick, E. G. codding, and S. T. Leung, Appl. Specbpsc.,29, 48 (1975). J. A . C. Broekaert, Bull. SOC. Chlm. Be@., 84, 1159 (1975). F. L. Fricke, 0. Rose, and J. A. Caruso, Anal. Chem., 47, 2018 (1975). D. 0. Knapp, N. Omenetto, F. W. Plankey, and J. D. Wlnefordner, Anal. Chim. Acta, 60, 455 (1974). T. L. Chester, H. Haraguchi, D. 0. Knapp, J. D. Messman, and J. D. Winefordner, Appl. Spectrosc., 30, 410 (1976). A. Danielsson and P. Lindblom, Appl. Spectrosc., 30, 151 (1976). K. W. Jackson, K. M. Aldous, and D. G. Mitchell, Spectrosc. Lett., 8, 315 (1973). K. W. Jackson. K. M. Aldous. and D. G. Mitchell.. ADD/. Soectrosc.. 28. , 569 (1974). K. M. Aldous, D. G. Mitchell, and K. W. Jackson, Anal. Cbem., 47, 1034 (1975). G. Horlick and E. G. Codding, Appl. Spectrosc.. 29, 167 (1975). G. Horlick and E. G. Codding, Anal. Chem., 48, 133 (1974). M. J. Milano and H. L. Pardue, Anal. Chem., 47, 25 (1975). G. Horlick and E. 0.Codding, Anal. Chem., 48, 133 (1974). T. E. Cook, M. J. Milano, and H. L. Pardue, C/in. Chem. ( Winston-Salem, N.C.), 20, 1422 (1974). M. J. Milano and H. L. Pardue, Clln. Chem. ( Wlnston-Salem, N.C.), 21, 211 (1975). T. E. Cook, H. L. Pardue, and R. E. Santini, Anal. Chem., 48, 452 (1978). T. A. Nieman and C. G. Enke, Anal. Chem., 48, 619 (1976). D. A. Yates and T. Kuwana, Anal. Chem., 48, 510 (1978). T. A. Nieman, F. J. Holler, and C. G. Enke, Anal. Chem., 48, 899 (1976). R. E. Dessv. W. G. Nunn. C. A. Tltus. and R. Revnolds. J. Chromatow. Sci., 14, i 9 5 (1976). (35) R. E. Dessy, W. D. Reynolds, W. G. Nunn. C. A. Titus, and G. F. Moler, Clin. Chem. ( Winston-Salem, N.C.), 22, 1472 (1976). (36) H. L. Pardue, A. E. McDowell, D. M. Fast, and M. J. Milano, Clln. Chem. (Wlnsfon-Salem, N.C.), 21, 1192 (1975).
..
(37) A. E. McDowell, R. S. Harner, and H. L. Pardue, Clin. Cbem. (Winston-Salem, N.C.), 22, 1862 (1976). A. McDowell, H. L. Pardue, Anal. Chem., 48, 1815 (1976). I. M. Wamer, J. B. Callis, E. R. aavidson,M. (;outerman, and G. D. CMstlan, Anal. Lett., 8, 665 (1975). T. Vo-Dlnh, D. J. Johnson, and J. D. Winefordner, Spectrochim. Acta, Part A, submitted. R. P. Cooney, T. Vo-Dinh, and J. D. Winefordner, Anal. Chim. Acta, in press. (42) I.M. Warner, J. B. Callis, E. R. Davidson, G. D. Christian, Clin. Cbem., ( Winston-Salem, N.C.), 22, 1483 (1976). (43) J. P. Fillard, M. de Murcla, J. Gasiot, and S. Chor, J. fbys. E., 8, 993 (1975). (44) E. Kohen, C. Kohen, J. M. Salmon, Mkrmbim. Acta, 1978 11, 195 (1976). (45) W. H. Woodruff and G. H. Atkinson, Anal. Cbem., 48, 186 (1976). 146) . . R. Wilbrandt. P. Paasbera. K. B. Hansen. and C. V. Weisbera. - Chem. Phys. Lett., 36, 76-(1973). (47) R. E. Santini, M. J. Milano, and H. L. Pardue, Anal. Chem., 45, 915A 119731 \ -I.
(48) D. G. Mitchell, K. N’.Jackson, and K.M. Aldous, Anal. Chem., 45, 1215A (1973). (49) K. W. Busch and 1s. H. Morrison, Anal. Cbem., 45, 712A (1973). (50) G. Horlick, Appl. Spectrosc., 30, 113 (1976). (51) J. D. Winefordner, J. J. Fltzgerald, and N. Omenetto, Appl. Spectrosc., 29, 369 (1975).
(52) E. J. Bair, “Introduction to Chemical Instrumentation”, McGraw-Hlll, New York, 1962. (53) J. D. Winefordner, V. Svoboda, and L. J. Cline, Crif. Rev. Anal. Chem., 1, 233 (1970). (54) J. D. Winefordner, R. Avni, T. L. Chester, J. J. Fitzgerald, L. P. Hart, D. J. Johnson, and F. W. Plankey, Specfrochim. Acta, Part B, 31, 1 (1976). (55) G. D. Boutilier. J. D. Bradshaw, S. J. Weeks, and J. D. Wlnefordner, Appl. Spectrosc., in press. (56) M. Margoshes, Spectrocblm. Acta, Part B, 25, 113 (1970). (57) M. Margoshes, Pittsburg Conference on Analytical Chemistry and Applied Spectroscopy, 1970. (58) D. 0. Knapp, Ph.D. Thesis, University of Florida. Gainesville, Fla., 1973. (59) N. G. Howell and G. H. Morrlson, Anal. Cbem., 49, 106 (1977). (60) A. van der Zkl, “Noise in Measurements”, Wiley-Interscience, New York, N.Y., 1976.
RECEIVED for review November 15, 1976. Accepted March 30, 1977. One of the authors, G. D. B., acknowledges the support of the American Chemical Society Analytical Division Summer Fellowship sponsored by the Society for Analytical Chemists of Pittsburgh. This work supported by NIHGM-11373-1 and AF-AFOSR-F44620-76-C-0005.
Instrumental Effects on Limits of Detection in Gas Phase Fluorescence Detection of Gas Chromatographic Effluents R. P. Cooney and d. D. Winefordner” Department of Chemistry, University of Florida, Gainesville, Florida 326 1 1
Several dlfferent opi lcal systems and excltatlon sources are compared for the datectlon of gas phase fluorescence from gas chromatographic effluents. I t Is determlned that, for one of the systems (system V) studled and llkely for the other systems utilizing conventional nonlaser sources, the limiting noise comes from stray llght at the fluorescence emission wavelength whlch olriginates at the excltatlon source and Is passed by the systeni optlcs and scattered by the sample cell. General concluslons; are drawn for Improving the system signal-to-noise ratio In the case of elther background shot or background flicker nolse Ilmltatlon. For the systems studied, It is concluded that Innproved performance can be attained by the use of sources wlith Intense output In the UV or by the use of conventlonal sources with exdtation monochromators which are most efficient in ithe UV and have improved (lower) stray light levels.
Molecular fluorescence, as well as being a powerful analytical tool by itself, hm become well-established as an ancillary technique to both liquid (1-10) and thin layer (11-15) chromatography. There have also been several examples of molecular fluorescence being used to identify and quantify gas chromatographic eluents (16-21). The limits of detection of gas phase fluorescence detectors have been excellent (20). For example, many polynuclear aromatic compoundo have been detected in subnanogram amounts (19). While previous studies have discussed to some extent how limits of detection might be improved, there has been little critical examination of the various systems in order to define precisely the origin and nature of the limiting noise, nor has there been much discussion on how this type of information could be used to improve the performance of the system. It is the purpose of this communication to report, the
results of such a study, including a comparison of two different spectrometer systems and several different excitation sources.
EXPERIMENTAL Apparatus. For the comparison studies reported in this work, several combinations of optical components and excitation sources were employed. In all the studies: the gas chromatograph used was a Varian Model 1400 (Varian Instrument Division, Palo Alto, Calif.), containing a 3-ft length of ‘/a in. 0.d. stainless steel tubing used to simulate a chromatographic column; the sample compartment was a rectangular quartz flow cell 50 mm long, having 3 mm and 5 mm i.d. and o.d., respectively (Model B 16-63019, Aminco, Silver Spring, Md.); the photomultiplier tube was a Hamamatsu 1P21 (Hamamatsu Corp., Middlesex, N.J.); in the dc systems, signals from the PM were fed to an O’Haver type (22) nanoammeter, and in the pulsed systems, signals were processed by a boxcar integrator (Model 160, Princeton Applied Research Corp., Princeton, N.J.); signals from the nanoammeter and boxcar integrator were recorded by a strip chart recorder (Model SRG, Sargent Welch Scientific, Skokie, Ill.). Figure 1A shows a block diagram of systems i, ii, and iii, consisting of the gas chromatograph connected by a heated transfer line to the quartz flow cell, held in a modified heated cell compartment of an Aminco spectrophotofluorimeter( -5.5 nm/mm). This arrangement has been described in detail (21) elsewhere. The only modifications made in this study were in the excitation sources, which were: in system (i), a 150-W Eimac xenon arc lamp (Varian Eimac Div., San Carlos, Calif.) was operated in a continuous (cw) mode at 11.5-A dc with a Varian power supply (Model 2505-2); in system (ii), the 150-W Eimac lamp was operated in a pulsed mode, using a pulsed power supply (23) (Velonix, Div. of Pulse Engineering, Inc., Santa Clara, Calif.); and in system (iii), a 200-W Hanovia mercury-xenon arc lamp was operated cw at 150 W with a standard Aminco power supply, and was placed in a standard Aminco ellipsoidal condensing system. Figure 1B shows a block diagram for systems (iv) and (v). In both of these cases, the quartz flow cell was placed in the oven of the gas chromatograph, and was connected to the simulated ANALYTICAL CHEMISTRY, VOL. 49, NO. 7, JUNE 1977
1057
