2040
Anal. Chem. 1981, 53, 2040-2044
more complicated in the near-IR area. Each quality control application envisaged for PAS must therefore be examined independently to assess the most suitable spectral region for analysis. Cogrinding of an absorbing material with a nonabsorbing material can considerably affect the photoacoustic response, and care must therefore be taken in sample preparation when quantitative analysis is required. ACKNOWLEDGMENT The authors thank R. R. Ford, J. H. W. Cramp, and C. M. Keary (IC1 Corporate Laboratory), P. Hampson (IC1 Pharmaceuticals Division), S. L. Castleden (Imperial College, London), and D. M. Spillane and G. F. Kirkbright (UMIST) for helpful discussions. LITERATURE CITED PA. signal
0
40
80
Yo propranolol F/u)
Figure 6. Corrected near-infrared photoacoustlc signal as a functton of % propranolol In unground mixtures: (a) at 2.2 pm; (b) at 1.72 pm.
region. The linear dependence found in our measurements in the near-infrared spectrum suggests that quality control measurements might most successfully be applied in this region. Spectral interferences could, however, make analysis
Rosencwaig, A. Anal. Chem. 1075, 47, 592 A. Rosencwaig, A. Opt. Commun. 1073, 7 , 305. Somoano, R. 0. Aigew. Chem., Int. Ed. Engl. 1078, 17, 238. Rosencwalg, A. Optoacoustic Spectroscopy and Detection"; Academlc Press: New York, 1977; Chapter 8. Lee, L. H. Org. Coat. flast. Chem. 1070, 40, 116. \hlong, K. Y. J . Appl. fhys. 1078, 49, 3033. Adams, M. J.; King. A. A; Kirkbrlght, G. F. Analyst (London) 1078, 101, 73. Adams, M. J.; Kirkbright, G. F. Spectrosc. Lett. 1076, 9 , 255. "The Merck Index", 9th ed.; Merck & Co. Inc.: Rahway, NJ, 1978; p 7628. Lln, J. W-p.; Dudek, L. P. Anal. Chem. 1970, 51, 1627. Fuchsman, W. H.; Silversmith, A. J. Anal. Chem. 1070, 57, 589. Rosencwaig, A.; Qersho. A. J. Appl. fhys. 1078, 47, 64. Freeman, J. J.; Friedman, R. M.; Reichard, H. S. J. fbys. Chem. 1080, 84, 315.
RECEIVED for review May 26,1981. Accepted July 30, 1981.
Precision and Accuracy of Absorption-Corrected Molecular Fluorescence Measurements by the Cell Shift Method D. R. Chrlstmann,' S. R. Crouch," and Andrew Tlmnick Department of Chemistry, Mlchlgan State Unlverslty, East Lansing, Michigan 48824
Interferences due to absorption of the exciting or fluorescence radiatlon by a fluorescence sample are corrected by the cell shift method with an accuracy of better than 2% for sample absorbances as high as 2.7 If scattered llght and reemission of the absorbed fluorescence are absent. For samples that exhiblt strong self-absorption and reemisslon, the accuracy of the absorption correctlon can be poor above a fluorophore concentration of about lo-' M. Poor accuracy of the corrected fluorescence also results if the fluorescence sample Is turbld or If the spectral bandwidths of excitatlon or emlsslon are large compared to the wldth of the interferlng sample absorption band. Theoretical conslderatlons and experiments show that the precision of the corrected fluorescence is aiways poorer than the precision of the raw fluorescence, but only by a factor of 2 or less.
In a recent article, Novak (I) introduced the cell shift method for correcting inner filter effects in right-angle fluo'Present address: Boeing Aerospace 42-29, Seattle, WA 98124.
Co.,P.O. Box
3999,
M/S
0003-2700/81/0353-2040$01.25/0
rometry. The method is so named because it involves shifting the sample cell in a spectrofluorometerto change the effective pathlength through which the exciting and fluorescence radiation must travel. Fluorescence intensities are measured at three cell positions. From these measurements corrections for absorption of the exciting and fluorescence beams are computed. The cell shift method offers several advantages over other absorption correction procedures in the literature (2-5). Direct measurement of the sample absorbance at the excitation and emission wavelengths is not required. A spectrofluorometer is, therefore, the only measurement device that is needed, and errors arising from the use of a spectrophotometer are avoided (I, 6). Commercial fluorescence instruments are suitable for the method after only minor modifications (6). The method has also been automated (7), resulting in a simpler calibration procedure, reduced measurement time, and reduced errors due to cell positioning. The cell shift method has been used to correct for selfabsorption and to improve the accuracy of solvent fluorescence subtraction in fluorescence emission spectra (I). Matrix absorption interferences in a fluorometric assay for aluminum have also been corrected with the technique (7). Although these applications have been successful, several factors that 0 1981 American Chemical Society
ANALYTICAL CHEMISTRY, VOL. 53, NO. 13, NOVEMBER 1981
2041
8200 ri:
5+
14
CK
0
3
e
0 0.0
ABSORBANCE AT 365 NM
q
L
00 Q
2
0.0
0.5
1.0 1.5 2.0 ABSORBANCE
, 2.5
: 2 0
3.0
Flgure 1. Correction for absorption of exciting radiation by the fluorophore. (A) Fluorescence of quinine sulfate vs. absorbance at the exciting wavelength: (m) corrected fluorescence, (A)measured fluorescence for cell positions 2 and 3, ( 0 )measured fluorescence for cell position 1. Straight line represents theoretical behavior for absorption-free fluorescence. (B) Relative standard deviatlon (RSD) of corrected fluorescence vs. absorbance. may affect the accuracy of the corrected fluorescence have not been thoroughly studied. These factors include the reemission of absorbed fluorescence by the sample, light scattering by the sample, and the spectral bandwidths of excitation and emission (2). In this paper, the effects of these factors on the accuracy of the absorption-corrected fluorescence by the cell shift method are investigated. Knowledge of these effects is valuable so that errors in the cell shift correction procedure can be minimized. Also of importance is the effect of the correction procedure on analytical precision. The factors affecting the precision of the absorption-corrected fluorescence intensity and the relationship between the measurement precision and the precision of the corrected result are considered in detail. EXPERIMENTAL SECTION Instrumentation. All fluorescence measurements were performed with a microcomputer-controlled spectrofluorometer that automatically positions the sample cell for the cell shift measurements, The instrumental parameters and the measurement procedure were the same as described previously (7). Spectrophotometric absorption measurements were made with this same instrument by changing the position of the reference detector to allow the intensity of the excitation beam transmitted by the sample to be measured. Appropriate cutoff filters were placed between the cell and reference detector to eliminate interferences from fluorescence by the sample. These measurements should be distinguished from the apparent absorbance values calculated from the fluorescence measurements (7). Reagents. All reagents were of analytical reagent grade and were used without further purification. Solutions were prepared in volumetric glassware with house distilled water. Quinine sulfate solutions were stored in hard polyethylene bottles to minimize adsorption losses. All solutions were stored in the dark to prevent photodecomposition. RESULTS AND DISCUSSION Accuracy When Reemission Is Negligible. In the theory of the cell shift method ( I , @ and in other absorption correction procedures (2, 3), the least discussed and, perhaps,
0
0.5 1.0 1.5 2.0 ABSORBANCE AT 313 NM
b
x
a
a
a
z
-
2
2.5
5
ABSORBANCE
Flgure 2. Correction for absorption of exciting radiation by sample matrix. (A) Fluorescence of a constant amount of quinine sulfate in the presence of increasing amounts of gentisic acid: (W) corrected fluorescence, (A)measured fluorescence for cell positions 2 and 3, (0)measured fluorescence for cell position 1. Straight llne represents theoretical behavior for absorption-free fluorescence. (B) RSD of corrected fluorescence vs. absorbance. z
24 0’35T A
M
0.30
tQ:
0.25
ABSORBANCE AT 436 N M
“I 8
i O O k C K
0!5
l!O 115 2!0 ABSORBANCE
215
Figure 3. Correction far absorption of exciting and fluorescence radiatlon. (A) Fluorescence of a constant amount of quinine sulfate in the presence of increasing amounts of fluorescein: (W) corrected fluorescence, ( 0 )measured fluorescence for cell position 1, (A) measured fluorescence for cell positlon 2, (+) measured fluorescence for cell position 3. Straight line represents theoretical behavior for absorption-free fluorescence. (B) RSD of corrected fluorescence vs. absorbance. most important assumption is that reemission of absorbed fluorescence by the sample is negligible. Several tests of absorption Correction accuracy in which this condition is satisfied have been described previously (2,Q).These tests were repeated in this study to illustrate the potential accuracy and precision of the cell shift method under ideal conditions.
2042
ANALYTICAL CHEMISTRY, VOL. 53, NO. 13, NOVEMBER 1981
s
2400;I
"
$350-
t
2 300-t
i
:
I ' I'
I
!
'
I
~
I
AE / I
I
: :
I
I
I :
3.0 1
2.5
8
" 1 600
w
-2.0 b
0
1 21.5 .0 1.0
2
0.5
$
200
0.5 FLUORESCEIN CONCENTRAT/ON, M x IO4
Flgure 4. Effect of reemission on the corrected fluorescence (H)and apparent sample absorbance at the excitation wavelength (A)when fluorescence is monitored outside the overlap region.
The results are shown in Figures 1-3. Figure 1A illustrates the accuracy of the correction for absorption of exciting radiation by the fluorophore itself. For lo-' to 4 X lo4 M solutions of quinine sulfate in 0.1 N H2S04, absorption-corrected fluorescence intensities and uncorrected fluorescence intensities from the three cell positions (7) are plotted against the solution absorbance at the excitation wavelength. Curvature of the measured fluorescenceresponse due to absorption of the exciting light is evident. Figure 2A illustrates the accuracy of the correction for absorption of exciting radiation by sample components other than the fluorophore. Corrected and uncorrected fluorescence intenM quinine sulfate sities for a series of solutions of 2 X and 0 to 4 X M gentisic acid in 0.1 N H2S04are plotted against the solution absorbance at the exciting wavelength. In Figure 3A, raw and corrected fluorescence intensities for M solutions of 2 X 10" M quinine sulfate and 0 to 6.5 X fluorescein in 0.1 N H2S04are plotted as a function of the solution absorbance a t the emission wavelength. The uncorrected data in Figure 3A reflect the effects of absorption of exciting radiation by both the fluorophore and the sample matrix and the effect of absorption of the fluorescence emission by the sample matrix. The straight lines in part A of each figure represent a theoretical absorption-free fluorescence response which was determined by extrapolating from the corrected data at low absorbances (50.01). Part B of each figure shows the relative standard deviation (RSD) of the corrected fluorescence as a function of the sample absorbance. The RSD in each test ranged from 0.2 to 2.0%. The correction precision vs. absorbance is discussed in detail in a later section. In each of these tests, the cell shift corrected results do not differ significantly ( a = 0.05) from the theoretical lines up to an absorbance of 2.0. In Figure lA, the corrected results are in error by less than 2% up to an absorbance of 2.7. The decrease in accuracy at higher absorbances is most likely due to errors in the extrapolation of the theoretical fluorescence response and in the values of the instrument window parameters used in the correction calculations. No fundamental limitation of the accuracy of the cell shift method is indicated under the conditions of these experiments. Effect of Reemission. Solutions of to 5 X lo4 M fluorescein in 0.1 N NaOH were prepared and excited at 313 nm to investigate the accuracy of the cell shift method under conditions in which the reemission of absorbed fluorescence is not negligible. Fluorescein has a relatively small Stokes shift of 1100 cm-l (10) and, therefore, a large degree of overlap between its absorption and fluorescence emission bands. Fluorescence emission was monitored at 510 nm (in the overlap region) and also at 540 nm (outsidethe overlap region).
"-++I
I
I
: I
I
I
~
0.0
1.0 2.0 3.0 4.0 5.0 FLUORESCEIN CONCENTRATKN, M x 104
0.0
Figure 5. Effect of reemission on the corrected fluorescence (A)and apparent sample absorbance at the excitation (H)and emission (+) wavelengths when fluorescence is monitored in the overlap region.
In Figures 4 and 5 the corrected fluorescence intensities at each emission wavelength are plotted against the fluorescein concentration. Also plotted are values of the apparent solution absorbance calculated by the cell shift method (7) and used to compute the absorption correction factors. The straight lines in each figure are the absorbance analytical curves determined from spectrophotometric data and extrapolated fluorescence analytical curves. Just as Novak (1)has shown with rhodamine B, these data indicate that the accuracy of the cell shift method is good at concentrations of 1 X low5M and below. However, at higher concentrations there is a definite decrease in the accuracy of the absorption-corrected fluorescence. Two effects are apparent from Figures 4 and 5. First, the presence of a small reemission component in the measured fluorescence signals causes a negative error in the sample absorbance calculated by the cell shift method. This is similar to the effect of stray light on a spectrophotometric absorbance measurement. The absorbance error tends to cause a negative error in the corrected fluorescence. Opposing this effect, however, is the reemission component in the measured signals which tends to make the corrected fluorescence larger than the theoretical value. The nature of the net error due to reemission depends on the emission wavelength chosen. Away from the region of overlap (Figure 4), where corrections for the absorption of fluorescence radiation are minor, the absorption-corrected fluorescence shows a positive error. The error increases in magnitude as the fluorophore concentration is increased. However, if fluorescence is monitored in the region of overlap (Figure 5), where the correction for absorption of fluorescence radiation can be significant, the error in the corrected fluorescence is first positive but eventually decreases and becomes negative as the fluorophore concentration is increased. Apparently, the combined negative errors in the apparent safnple absorbances at the excitation and emission wavelengths become large enough to dominate the effect of the extra component of fluorescence due to reemission. Further evidence of this is given in Figure 6. Absorptionand corrected fluorescence emission spectra of 5 X lo4, 3 X 10" M solutions of rhodamine B in ethanol are shown normalized for concentration. Clearly, as the fluorophore concentration is increased, a decrease in fluorescenceintensity occurs on the low wavelength side of the emission band and an increase in intensity occurs on the high wavelength side. Regardless of the emission wavelength that is chosen, the accuracy of the cell shift method can be poor when the absorption and emission bands of the fluorophore overlap. Reemission of the absorbed fluorescence imposes a concentration limit above which the corrected fluorescence will be
ANALYTICAL CHEMISTRY, VOL. 53, NO. 13, NOVEMBER 1981 2043 400 t - 3 . 0
T
o.ooot m.fib;I I 530
~
~
550
I
I
I
I
570
I
I
I
~
590
I1 I
I
610 WAVELENGTH, NM
I : I
630
Normalized absor tion-corrected fluorescence emission M (0)rhodamine B spectra of 5, X 10" (0),10- ?(A),and 3 X in ethanol. Figure 0.
2
0
0.80
0 1 2 3 4 SCATTERING AGENT CONC., REL. UNlTS
Corrected fluorescence (A)and apparent absorbance (0), normalized to zero scattering agent, of a constant concentration of quinine sulfate as a function of increaslng amounts of soluble starch. Figure 7.
in error. I3ecause the example shown here for fluorescein approacheti the worst case, however, it is expected that the cell shift method will generally be accurate at fluorophore concentrations of M and below. Effect of Light Scattering. Scattering of the exciting and fluorescence radiations by the sample matrix is a potential problem common to all fluorometric methods. To determine the effect of matrix scattering on cell shift corrected fluorescence measurements, we prepared a series of solutions of 4 X M quinine sulfate in 0.1 N H2S04 containing increasing amounts of soluble starch. The solutions were excited at 365 nm and fluorescence was monitored at 450 nm. In Figure 7 the corrected fluorescence intensities and solution absorbances, normalized to the values in the absence of the scattering agent, are plotted as a function of the scattering agent concentration. Clearly, light scattering by the sample matrix produces a positive error in both the apparent sample absorbance and the corrected fluorescence. The magnitude of the error increases as the scattering agent concentration is increased. Effect af Spectral Bandwidth. Because the spectral bandwidths of excitation and emission have been shown to affect the accuracy of other absorption correction procedures (3), the effect of this measurement parameter on cell shift corrected fluorescence measurements was investigated. The excitation monochromator of our spectrofluorometer, which has a maximum bandwidth of 4 nm, was replaced with a Corning 7-60 band filter which has a bandwidth at half-height of approximately 40 nm. Quinine sulfate standards from lo-' to 4 X 10"' M were excited with a 200-W Xe-Hg arc lamp, and
QU~NINESULFATE CONCENTRATION, M x 104
Corrected fluorescence (A)and apparent absorbance at the exciting wavelength (0)vs. concentration for quinine sulfate excited with a Xe-Hg arc lamp and Corning 7-60 filter. Figure 8.
their fluorescence emission was measured at 450 nm. In Figure 8 the corrected fluorescence intensities and solution absorbances determined by the cell shift method are plotted against the quinine sulfate concentration. The extrapolated fluorescence response from low concentration and the absorbance analytical curve determined by spectrophotometry are also shown. Clearly, the increased excitation bandwidth produces negative deviations of both the sample absorbance and the corrected fluorescence from their theoretical values at concentrations above 1.5 X 10"' M. The effect is similar to the effect of polychromatic radiation in spectrophotometry. Although the effect on the corrected fluorescence in this example is not severe, it would be expected to be worse if a continuum radiation source had been used or if the emission monochromator had also been replaced with a band filter. Correction Precision. Although Novak (I)has discussed the effect of the instrument window parameters on the precision of the absorption correction factors, the results shown in Figures l B , 2B, and 3B are better understood through a simple extension of his discussion. Ignoring the corrections for reflections within the sample cell, the corrected fluorescence intensity from the cell shift method F, is given, to a good approximation, by Fc = FiXF2"F3'
(1)
where F,, F2,and F3 are the measured fluorescence intensities with the cell in positions 1, 2, and 3, respectively. The exponents, x , y, and z are given by x = -(wm wa2)/26w (2)
+
The window parameters wm, wa2, 6w, % ,, ga2, and 66 have been defined previously (7). Since Fl, F,, and F3 are not measured simultaneously,the random errors in these measurements will not be correlated, and the relative uncertainty of the corrected result will be given by
If the relationships FI = F2WA6"and F3 = F210-A'8eare used and it is noted that alp,,/F,,is proportional to F,,f where n = 1, 2 , or 3, eq 5 can be reduced to 'JFz -aFc- - ('2102AE6" + y2 + ~2102A'E68)1/2
(6) Fc Fz where A is the sample absorbance at the exciting wavelength
2044
ANALYTICAL CHEMISTRY, VOL. 53, NO. 13, NOVEMBER 1981
"O.'O
0:5
I:O 115 2:O ABSORBANCE
2:5
Figure 9, Normalized RSD of corrected fluorescence as a function * ) , -) theoretical result from of absorbance: (0)( u F ~ / F ~ ) / ( u F J F(-eq 6 for E = 0.5. and A'is the sample absorbance at the emission wavelength. The parameter E is the slope of the log-log plot of fluorescence signal-to-noise ratio vs. fluorescencesignal. Values for 4 are in the range of 0 to 1and depend on the noise characteristics of the spectrofluorometer. Typically 5 will be about 0.5 due to the dominance of signal shot noise. Equation 6 predids that the RSD of the cell shift corrected fluorescence depends directly on the RSD of the measured fluorescence signal from cell position 2 and will always be greater by a factor that is a function of the instrument window parameters, the sample absorbances at both the excitation and emission wavelengths, and the noise parameter [. The data in Figures lB, 2B, and 3B confirm that the relative precision of the measured fluorescence intensity is the major factor affecting the relative precision of the corrected result. In each experiment, the RSD of the corrected fluorescence increases as the RSD of the raw measurements increases, Le., as the measured signal magnitude decreases. The effect of all other parameters on the precision of the corrected fluorescence is illustrated in Figure 9. The experiment shown in Figure 1 was repeated five times and values of the RSD of the corrected fluorescence were divided by the RSD of the measured signal from cell position 2. The average values of the normalized RSD are plotted against the solution absorbance at the excitation wavelength along with a theoretical curve calculated from eq 6 for E = 0.5. Figure 9 shows clearly that the cell shift correction process reduces the experimental precision by a small amount (a factor of 2 or less in this example) and that the effect becomes greater as the sample absorbance increases. This is predicted by eq 6 (broken curve in Figure 9). The effect also becomes greater at low sample absorbances, contrary to eq 6. No explanation for this behavior is currently offered. The imprecision introduced by the absorption correction process can be minimized by making the window parameters for cell position 2 as small as possible as Novak (I)has stated. Maximizing the distance between cell positions, Le., the parameters 6w and 60, will also reduce the imprecision a t low sample absorbance although the precision at high sample absorbancewill suffer. This may often not be a disadvantage, however, because of limitations of the accuracy of the cell shift method a t higher absorbances due to the factors discussed previously.
CONCLUSIONS The results show that for fluorescence samples that are nonscattering and do not exhibit reemission phenomena, the cell shift method very accurately corrects for inner-filter effects. To ensure this accuracy it is necessary that the spectral bandwidth of excitation and emission be narrow with respect to the absorption bands of the sample. The best general accuracy could be expected by following the rules of molecular absorption spectrophotometry when selecting the spectral bandwidth. Because the accuracy is degraded most for highly absorbing solutions, a wider monochromator bandwidth or even interference filters in place of monochromators might be useful if the sample absorbance is low enough. These steps would also help alleviate any sensitivity problems created by modifying the spectrofluorometerfor cell shift measurements (7). Our results indicate, however, that the accuracy of the method using a standard filter fluorometer with glass cutoff and band filters would probably be very poor due to the very large excitation and emission bandwidths. For samples that are turbid it has been shown that the cell shift method gives erroneously high results and that the magnitude of the error increases with the turbidity. Although these errors may often be small enough to be acceptable, this problem is expected to preclude the use of the cell shift method with many real samples such as body fluids for which absorption corrections would be very useful. Because of reemission phenomena it is also expected that the cell shift method will be of limited use with samples in which the fluorophore is present at concentrations higher than 1X M. Below this limit the method should be accurate for any fluorophore solution that is nonscattering. Above this concentration limit the accuracy of the method will vary with the degree of overlap of the absorption and emission bands of the fluorophore. Although a fundamental loss of precision results from the use of the cell shift correction procedure, the magnitude of the loss can be kept small, a factor of about 2, and usually insignificant. The major factor that determines the precision of the absorption-correctedresult is the measurement precision of the spectrofluorimeter. LITERATURE CITED (1) Novak, A. Collect. Czech. Chem. Commun. 1978, 43, 2869-2878. (2) Chrlstmann, D. R.; Crouch, S. R.; Holland, J. F.; Timnlck, A. Anal. Chem. 1980, 52, 591-595. (3) Leese, R. A.; Wehry, E. L. Anal. Chem. 1978, 50, 1193-1197. (4) Ham, R. A.; Rosseinsky, D. R.; White, T. P. J . Chem. Soc., Faraay Trans. 2 1974, 70, 1522-1525. (5) Vlgney, P.; Duquesne, M. Phofochem. Photoblol. 1974, 20, 15-25. (6) Brltten, A.; Archer-Hall, J.; Lockwood, G. Anakst (London) 1978, 703, 928-936. (7) . . Christmann. D. R.: Crouch, S. R.: Tlmnick, A. Anal. Chem. lg81, 52. 278-280. (8) van Slageren, R.; den Boef, G.; van der Llnder, W. E. Talanfa 1973, 20, 201-212. (9) Holland, J. F.; Teets, R. E.; Kelly, P. M.; Tlmnick, A. Anal. Chem. 1977. 49. 706-710. (10) Beritkn, 'I. B. "Handbook of Fluorescence Spectra of Aromatlc Molecules"; Academic Press: New York, 1971; p 410.
RECEIVED for review April 6, 1981. Accepted July 7, 1981. This work was partially supported by NSF Grants CHE 7681203 and CHE 79-26490 and by an American Chemical Society Analytical Division Fellowship sponsored by PerkinElmer Corp.
