2077
Anal. Chern. 1885, 57, 2877-2880
P.;Winefordner, J. D. Anal. Lett. 1982, 15, 373. (3) Bateh, R. P.;Winefordner, J. D. J . Pharm. Sci. 1983, 7 2 , 579. (4) Bateh, R. P.; Winefordner, J. D., J . Pharm. Blamed. Anal. 1983, 1 , 113. (5) De Silva, J. A. F.; Strojny, N. Anal. Chem. 1975, 4 7 , 714. (6) Long, W. J.; Su, S. Y. Anal. Lett. 1985, 18 (as), 543. (7) Su, S. Y.; Asafu-Adjaye, E. B.; Ocak, s. Analyst (London) 1984, f o g , 1019. (2) Bateh, R.
(8) Su, S. Y.; Winefordner, J. D. Can. J . Specfrosc. 1983, 2 8 , 21.
RECEIVED for review May 28,1985. Accepted August 5,1985. This work is supported by VCU Bio-Medical Grants-In-Aid. Parts of the material have been presented at the Pittsburgh Conference, paper 1117, March 1985, New Orleans, LA.
Phase-Resolved Fluorometric Determinations of Four-Component Systems Using Two Modulation Frequencies Frank V. Bright' and Linda B. McGown* Department of Chemistry, Oklahoma State University, Stillwater, Oklahoma 74078
The use of phase-resolved fluorometric measurements at two modulation frequencies and three emission Wavelengths to simultaneously determine 1-chioroanthracene (lCA), 9,lOdiphenylanthracene (9, lODPA), 9-phenylanthracene(QPA), and 9-vlnylanthracene (9VA) in mixtures of these four components is descrlbed. Relative errors below 2 % were obtained for each of the four components In ail the mixtures studied. Relative standard devlatlons of 1.15 %, 3.95 %, 1.27%, and 9.23% were obtained for lCA, QPA, 9,1ODPA, and OVA, respectively, for a mlxture in which the combined fluorescence intensity contrlbutlon from 9PA and 9VA (which have a fluorescence lifetime difference of 1.42 ns and very similar emission spectra) was less than 10% of the total fluorescence emission intenslty.
between components rather than on the actual values. The theory and instrumentation of phase-resolved fluorescence spectrometry (PRFS) have been described in detail elsewhere (2-4) and are based on the phase-modulation method for fluorescence lifetime determinations ( 5 , 6 ) . The phase-resolved fluorescence intensity (PRFI) is a time-independent signal proportional to the cosine difference between the detedor phase angle (i.e,, the relative position of the a-gate interval, aD,variable from 0 to 360') and the phase angle of the fluorescing species (a). For a sample containing j uncorrelated (noninteracting) fluorophores, the PRFIi as a function of emission wavelength (Aem) and excitation beam modulation frequency ( w ) takes the form PRFI(@D,hern,~) = I
C [Ae,(Aern)ime,(w)m,(w) COS (@D - @,)I i=l Fluorescence lifetimes have long been determined by using sinusoidally modulated excitation to produce fluorescence emission that is subsequently demodulated and phase-delayed to an extent depending on the fluorescence lifetime of the emitting species. Investigators have described the use of measurements at more than one excitation beam modulation frequency to allow the analysis of fluorescence lifetime heterogeneity, from which fluorescence lifetimes and fractional contributions of each fluorescent species to the total emission intensity are obtained. However, since both the phase-delay and the demodulation of fluorescence emission are in general independent of fluorophore concentration, component concentrations cannot be obtained from these analyses without additional information. Such determinations have not, to our knowledge been described. We have found that phase-resolved fluorometric measurements, which provide fluorescence intensities that are dependent both upon the concentration and the fluorescence lifetimes of the fluorescence emitters, provide a convenient means for exploiting fluorescence lifetime differences in combination with other selectivity parameters such as emission and excitation wavelength for multicomponent fluorometric determinations. Unlike conventional heterogeneity or software-based phase-resolved analysis (1)the fluorescence lifetimes of the components do not need to be determined, nor do values need to be assumed, since the determinations are based on the relative differences in fluorescence lifetimes Present address: Department of Chemistry, University of Indiana, Bloomington, IN 47405.
(1)
where me,(w) is the frequency-dependent excitation beam modulation, m,(w)is the frequency-dependent demodulation factor, and Aern(Aern), is the dc emission component. These three fractors can be combined into a single term \k(Aem,w), I
PRFI(@D,Xern,U) =
c*(Aern,a), i=l
cos
(@D
- at)
(2)
which is the amplitude of the phase-resolved fluorescence emission function PRFI(@D,Aem,w).The PRFI is therefore a function of both the concentrations and the fluorescence lifetimes of the contributing emitters. Recently, the simultaneous determinations of two fluorescent species with essentially identical emission and excitation spectra using measurements at two or more non-nulling detector phase angles have been described (7). Simultaneous three-component determinations using the non-nulling detector phase angle approach in combination with wavelength selectivity have also been demonstrated (8). In the work described here, results are shown for the analysis of a four-component system (1-chloroanthracene (lCA), 9-phenylanthracene @PA),9,lO-diphenylanthracene (9,1ODPA),and 94nylanthracene (9VA)) using phase-resolved fluorescence measurements at combinations of three emission wavelengths with eight detector phase angles at two modulation frequencies. We have previously described the determination of a four-component syste: containing anthracene, 1-chloroanthracene, 2-chloroanthracene, and 9-chloroanthracene for which a single modulation frequency provided adequate resolution of the components (9). The four-component system described here could not be adequately resolved 0 1985 American Chemical Society
2878
ANALYTICAL CHEMISTRY, VOL. 57, NO. 14, DECEMBER 1985
by use of a single modulation frequency due to the particular fluorescence lifetimes involved and the spectral overlap of the components, which was especially extensive in the case of 9PA and 9VA. The generation of simultaneous equations for multicomponent phase-resolved fluorometric determinations using measurements at more than one modulation frequency has not been previously described.
EXPERIMENTAL SECTION Materials. 1-Chloroanthracene, 9-phenylanthracene, 9,lOdiphenylanthracene, and 9-vinylanthracene were all purchased from Aldrich and each was recrystallized once from absolute ethanol, (U.S. Industrial Chemical Co.). Dimethylbis(5phenyl-2-oxazoly1)benzene(Me2POPOP, Aldrich, Scintillation Grade) was used as the reference fluorophore for all fluorescence lifetime determinations (IO). Standard solutions of the individual compounds were prepared by diluting the appropriate weight of the compound to 1.000 dL with absolute ethanol and sonicating for 30 min, followed by a 100-fold dilution with absolute ethanol. Mixtures were prepared by combining the appropriate volumes of the standard solutions of the components with no further dilution. All fluorescence measurements were made with disposable polyethylene cuvettes (Evergreen Scientific). Data Collection. All fluorescence measurements were made with an SLM 4800s spectrofluorometer (SLM/Aminco, Urbana, IL) with a 450-W Xe arc source and Hamamatsu R928p photomultiplier tube (PMT) detection. Two modulation frequencies (18 and 30 MHz) were used for the phase-resolved measurements, which were all taken in the delta phase mode (an instrumental mode that allows simultaneous comparison of the sample beam with a portion of the modulated excitation beam that is split and diverted to a reference quartz cuvette containing water and detected by a phase sensitive PMT). Sample chamber temperatures were maintained at 0.0 A 0.1 'C, with a Haake A81 temperature control unit, which increased the quantum yield for all components by a factor of -2-3% relative to room-temperature measurements. Lowering of the sample temperature also reduced the noise fluctuations within any measurement set. Dry nitrogen was directed through the sample chamber housing and around the capped sample cuvette to minimize the water condensation on the exterior of the cuvette surface. Monochromators were set at 360 nm for excitation and 385, 395, or 425 nm for emission. Slit settings included 16 nm and 0.5 nm for the excitation monochromator entrance and exit, respectively, 0.5 nm for the modulation tank exit, and 16 nm for both the emission monochromator entrance and exit, respectively. At each modulation frequency, phase-resolved fluorescence intensities (PRFI) were measured for all solutions (four standards and six mixtures) at each A,, A,, pair (360, 385; 360, 395; 360, 425 nm) for detector phase angles (OD)equally spaced at 45' intervals between 0' and 315'. Phase-resolved fluorescence intensities were measured in the "100 average" mode in which an average value is obtained by integration of 100 samplings over approximately 30 s. All PRFI values were measured in triplicate and averaged. Blank contributions due to ethanol solvent were negligible so that no correction to the measured PRFI was necessary. Data Analysis. The collected data were entered by hand into an Apple iIe microcomputer (Apple, Inc., Cupertino, CA) interfaced to an Okidata b 9 2 ) printer (Okidata, Inc., Mount Laurel, NJ). In general, for L emission wavelengths (A,,) at m detector phase angles (OD)using n modulation frequencies ( w ) for our four-component system, the augmented matrix takes the form shown in eq 3, where I values are the-PRFI of the mixtures for each (OD,A,, w ) combination. The I values are the PRFI per micromole per liter for each pure component, determined by measuring a standard solution of each component at each set of conditions (OD,A,, w ) . The PRFI vs. QD curves were generated for each (A,, w ) pair by computer for the raw data collected at the eight equally spaced detector phase angles from 0' to 315'. The curves for the individual components were generated by use of a BASIC leastsquares cosine-fit routine described in detail elsewhere (9). The curves for each of the mixtures were generated by use of a
x
(3)
fifth-order polynomial fitting routine since this approach yielded slightly better results than the cosine-fit or lower-order polynomial fits for a previous four-component study (9). In this earlier study (9) we erroneously noted that the cosine-fit was not appropriate for mixtures of fluorophores when in fact the resulting curve for a mixture is also a true cosine function. The I and I values were found at the first four detector phase angles (O', 45', go', and 135') from the fitted curves and input manually into the Apple IIe. The resulting augmented matrices (eq 3, square or overdetermined) were "solved" for the analytical concentrations of lCA, 9PA, 9,1ODPA, and 9VA in each of the quaternary mixtures by a Gauss-Newton least-squares minimization (11) iterative program capable of solving all possible combinations of 4 X 4, 5 X 4, 6 X 4, ..., and 24 X 4 matrices. Values of I and f for the second portion of the fitted curves (180°, 225', 270°, and 315') were not used since the symmetry of the fitted curves dictates that these points are redundant with the first half of the curve, differing only in sign.
RESULTS AND DISCUSSION The steady-state emission spectra are shown in Figure 1 for lCA, 9,1ODPA, 9PA, and 9VA. The extensive spectral overlap is evident for the four components, most noteworthy between 9PA and 9VA. Fluorescence lifetimes of 1.48 (&0.02), 5.10 (hO,Ol),6.12 (f0.02), and 7.53 (h0.04) ns were determined for lCA, 9,1ODPA, 9PA, and 9VA, respectively, using MezPOPOP (T = 1.45 ns) as the reference fluorophore, and using 80 sample/reference phase shift measurement pairs. Values determined using demodulation measurements gave similar results, indicating a single exponential fluorescence decay for each component. The PRFI vs. OD curves are shown in Figure 2 (18 MHz) and Figure 3 (30 MHz) for each of the four individual components a t A, = 360 nm, &, = 395 nm, generated from measurements of PRFI a t eight detector phase angles (0'-315') as described above. The effects of the number and choice of emission wavelengths, modulation frequencies, and detector phase angles were studied. Six mixtures were used for this study and are described in Table I. Errors of determination (not shown)
ANALYTICAL CHEMISTRY, VOL. 57, NO. 14, DECEMBER 1985 I
2879
0.61-
I
I
240
320
I
80 I
-0.60 V
160
DETECTOR PHASE ANGLE
315
500
EMISSION WAVELENGTH Cnm)
Figure 1. Steady-state emission spectra of 1-chioroanthracene (-), 9-phenylanthracene 9,1O-diphenyianthracene (- -), and 9-vinylanthracene (- - -); excitation wavelength, 360 nm.
-
(..e),
(DEGREES)
Figure 2. Phase-resolved fluorescence intensity (PRFI) vs. detector phase angle for 1-chloroanthracene (0),9-phenylanthracene (e), 9,lO-diphenylanthracene (O), and 9-vinylanthracene (m); excitation wavelength, 360 nm; emission wavelength, 395 nm; modulation frequency, 18 MHz.
~~
Table I. Concentration of the Four Anthracene Derivatives Used in the Study" mixture
1CA
9PA
9,lODPA
9VA
1 2 3 4 5 6
1.175 1.175 1.175 0.235 0.235 0.235
0.129 0.129 0.043 0.129 0.043 0.129
0.245 0.082 0.327 0.327 0.245 0.245
0.130 0.173 0.130 0.130 0.013 0.173
5
&
"pmol/L in cuvette. 80
0
for the six mixtures using the four phase angles with a single emission wavelength were very poor (>2000% in best case). Results for the least favorable case, solution 5, in which the two most similar components 9PA and 9VA together contribute less than 10% to the total emission, are shown in Table
160
DETECTOR PHASE 'ANGLE
240
320
(DECREES)
Figure 3. Phase-resolved fluorescence intensity (PRFI) vs. detector phase angle: same symbols and wavelengths as in Figure 2; modulation frequency, 30 MHz.
Table 11. Error of Determination for the Individual Components in Solution 5 Four-Component Mixture as a Function of Both Modulation Frequency and Emission Wavelength eq
1CA
8c
241 287 0.41 17.8 12.5 0.25
8d 12e 12f
168 24h
relative error, % 9PA 9,lODPA -16100 -95.0 -91.0 40.9 36.8 1.96
-1970 -14600 14.5 -11.9 -9.87 -0.19
9VA
%E"
I%Elb
366 38.0 416 247 59.3 -1.42
-4370 -3600 84.9 73.4 24.7 0.15
4670 3770 130 79.4 29.6 0.96
aAverage % error. *Average 1% errorl. c18 MHz, emission at 385 and 425 nm. d30 MHz, emission at 385 and 425 nm. '18 MHz, emission at 385,395, and 425 nm. 130 MHz, emission at 385,395, and 425 nm. 818 and 30 MHz, emission at 385 and 425 nm. h18 and 30 MHz, emission at 385, 395, and 425 nm.
Table 111. Best Case Errors of Determination for the Simultaneous Quantitation of Four Components Using the Information at Both 18 and 30 MHz with Emission Wavelengths at 385,395, and 425 nm Generating 24 Simultaneous Equations in Four Unknowns relative errog, % mixture
1CA
9PA
9,lODPA
9VA
%E"
I%Elb
1 2 3 4 5 6
-1.29 -0.66 1.97 -1.16 0.25 1.84
1.66 -1.42 1.55 -0.21 1,96 -1.84
0.91 0.04 -1.46 1.36 -0.19 1.07
-1.94 1.11 1.62 -0.97 -1.42 -0.94
-0.17 -0.23 0.92 -0.25 0.15 0.03
1.45 0.83 1.65 0.93 0.96 1.42
"Average % error. bAverage 1% errorl.
Anal. Chem. 1985, 57, zaao-zaa5
2880
11. These results are tabulated as a function of both modulation frequency and emission wavelength. As shown in Table I1 results obtained for solution 5 using three wavelengths (385,395, and 425 nm) are better than those using two wavelengths due to the additional discrimination provided by the 395-nm emission wavelength. Errors of determination using both 18- and 30-MHz excitation beam modulation frequencies at two emission wavelengths (385 and 425 nm) are better than those using either of the modulation frequencies alone, but accuracies are still very poor. The other five solutions were also analyzed under all the conditions described in Table I1 for solution 5, and a similar trend in accuracy improvement is observed. Errors of determination using all three emission wavelengths and both 18-and 30-MHz modulation frequencies are shown for solution 5 in Table 11, and for all of the solutions in Table 111, in which the percent errors are much lower in all cases (
