496
Anal. Chem. 1980, 52, 496-499
distilled water and filtered water stored in glass and Teflon are shown in Table 11. Most of the observed mercury loss occurred within the first 24 h of storage for both glass and Teflon. In overall mercury retention, glass was found to be far superior to Teflon for storage of both distilled and river water samples. Acidification with 170(v/v) HC1 was used to recover the mercury which was lost. Complexation with 2mercaptoethanol was also used to recover lost mercury. Only 75 and 78% of the mercury loss was recovered from Teflon and glass, respectively, for river water samples. Mechanisms responsible for mercury loss include absorption onto the container walls (9),volatilization, and complexation of the mercury in solution.
Table 11. Effect of Storage Time on the Amount of Mercury Remaining in Solutions of 80 ng L-' Hg(I1) in Distilled and Filtered River Water
Iiq uid distilled water
bottle
merc u r y storage remain time. h ing, 5;
glass
0 1
4 7 24 48 73
Teflon
0 1
-4
7 24 48 73 filtered river water
glass
0 1
3
6 23 47 96
Teflon
0 1 3 6 23 47 96
100 89.9 78.1 81.8 76.9 71.4 66.5 100 82.0 71.9 76.8 48.6 37.1
ACKNOWLEDGMENT The author thanks T. C. Wang for the opportunity to complete this study and R. S. Braman, University of South Florida, for assisting in the preparation of this manuscript. T h e author also thanks B. Herman for her assistance in the preparation of this manuscript.
23.0 100 93.5 84.5 82.1 62.9 60.1 58.2
LITERATURE CITED (1) Nishi, Sueo; Horimoto, Yoshiyuki. "Determination of Total Mercury Levels in Natural Water"; Water Quality Parameters, ASTM STP 573, American Socieity for Testing and Materials: Philadelphia, Pa. 1975; pp 25-29. (2) Ure, A. M. Anal. Chim. Acta 1975, 7 6 , 1. (3) Polnektov, M. S.; Vitkun. R. A. Zh. Anal. Khim. 1963, 18, 37. (4) Kunert, I.; Komarek. J.: Sommer, L. Anal. Chim. Acta 1979, 106, 285. (5) Braman, R. S. Anal. Chem. 1971, 43. 1462. (6) Braman, R. S.; Johnson, D. L. Environ. Sci. Technol. 1974, 8 , 996. (7) Johnson, D. L.; Braman. R. S. Environ. Sci. Techno/. 1974, 8. 1003. (8) Mahan, K. I.: Mahan, S. E. Anal. Chem. 1077, 4 9 , 662. (9) Lo, J.; Wai, C. Anal. Chem. 1975 4 7 , 1869. (10) Dequchi, S. T.; Urata, K.:Tomooka, J.; Nagain, H. Anal. Chim. Acta 1976, 8 7 , 479. (11) Feldman, C. Anal. Chem. 1974, 46. 99. (12) Heiden, R. W.; Aikens, D. A. Anal. Chem. 1979, 57, 151. (13) Matsumaga, K.: Konishi, S.; Nishimura, M. Environ. Sci. Technol. 1979, 13, 63. (14) Bothner, M. H.; Robertson, D. E. Anal. Chem. 1975, 47, 592.
100 97.5 90.7 66.9 58.8 54.6 52.4 -
Baker, Ultrex) and wrapped in aluminum foil and kept in a dark place. A sample a t time zero from each bottle was analyzed to confirm the initial mercury content. Twentymilliliter samples were analyzed a t various time intervals, ranging from 0 to 96 h. The results of mercury lost from
RECEIVED for review September 17, 1979. Accepted December 14, 1979. Harbor Branch Contribution No. 161.
Analysis of Multicomponent Fluorescent Mixtures through Temporal Resolution L. J. Cline Love* and Linda M. Upton Department of Chemistry, Seton Hall University, South Orange, New Jersey 07079
tentimes makes it the method of choice for the analysis of fluorescent compounds such as drugs and metabolites present in low concentrations. When two or more of these compounds are the components of a mixture, time consuming chemical steps, such as extraction or chromatography, are often taken to separate the compounds prior to fluorescence analysis. If the excitation or emission bands are sufficiently separated, the components may be selectively analyzed through spectral resolution without chemical separation. As the excitation and emission bands of the components begin to overlap, this technique is no longer possible. Several techniques have recently been developed which avoid the chemical separation of mixtures with overlapping spectral bands. The selective modulation technique utilizes a slight difference in the excitation spectra of the mixture components. Through a wavelength modulation, the emission intensity of each component modulates at a different fre-
A quantitative fluorescence analysis method through temporal resolution using time correlated single photon counting and reiterative convolution (RC) was developed and evaluated. A proof of the mathematical relationship between the amplitude constant " a " from the RC technique, where E = and concentration is derived. Several two-component mixtures of various antimalarial drugs were analyzed, with most devlations from actual concentration ratios falling between 10 and 50 %. The components' relative contribution to the decay curve, rather than to the fluorescence intensity, was found to determine the success of the analysis. The method works best for mixtures of compounds having superimposable spectra, thus complementing spectral resolution analysis methods.
T h e high sensitivity of the fluorescence measurement of0003-2700/80/0352-0496$0 1 OO/O
C
1980 American Chemical Society
ANALYTICAL CHEMISTRY, VOL. 52, NO.
quency ( 1 ) . T h e use of synchronous excitation allows qualitative and quantitative analysis of mixtures through a synchronous scan of the excitation and emission with a constant difference in wavelength (2, 3 ) . Again, this technique fails if the spectra of the components are superimposable. The video fluorometer provides quantitation of fluorescent mixtures through a principal component analysis of a n excitation-emission matrix (4-6). All of these methods make use of spectral resolution. Another solution to the problem of multicomponent fluorescent analysis arises through temporal resolution. The lifetime of the excited state can be a distinguishing feature. Winefordner and co-workers have utilized phosphorescent lifetime differences in analyzing the components of phosphorescent mixtures (7-9). In the same manner, fluorescent lifetime differences have been utilized for the identification of components of fluorescent mixtures (10-12). Hiraki and co-workers have reported a study in quantitating fluorescent mixtures through time resolution (13). They used a graphical slope method with sequential subtraction of components. This method is limited to compounds whose lifetimes are appreciably longer than the excitation pulse and has limited accuracy for short-lived species. This paper presents the development of a method to quantitate the components of a mixture without prior separation using time resolved fluorescence through time correlated single photon counting (TCSP) (14) and reiterative convolution (RC) (10). T h e precision, accuracy, and limitations of the technique will be discussed.
PRINCIPLES Several mathematical curve fitting methods have been developed which correct for the error in fluorescence lifetimes due to the pulse width of the excitation source (15-18). Some of these methods also allow the mathematical separation of species in multicomponent mixtures when the components have lifetimes which are too close to be resolved by the simple graphical slope method. None of these methods has been used for quantitative analysis to date, however. This laboratory deconvolutes decay curves through reiterative convolution (10, 19) in order to qualitatively identify species in two-component mixtures as well as to correct for the distortion in short lifetimes caused by the finite pulse width of the excitation light source (12). This method has been deemed superior to the others in its ability to deconvolute simulated multi-exponential decays (17). I t was felt that its mathematical formulation would be especially amenable for quantitation of two-component mixtures of fluorophores. When using reiterative convolution, the deconvolution results in a lifetime, T , and an amplitude constant, "a". The amplitude constant, "a" is a n expression of the intensity of the fluorescence emission as it relates to the excitation source intensity, Io, shown below
E =
(11
where E is the total emission intensity and t is the time elapsed since excitation. T h e fluorescence lifetime, T , is defined as the time it takes for the signal to decay by l / e . Rigler and Ehrenberg have suggested that the amplitude constant for a compound in dilute solution, exhibiting little absorption, is proportional to its extinction coefficient, e , its quantum yield, @, its concentration, c, and the inverse of its lifetime (201, as expressed below.
a
3
ct+/r
(2)
This relation has never been proven or used for quantitation. A proof of Equation 2 follows. T h e area, A , under a decay curve is proportional to the photon count rate, R , and the time of counting, t as for ex-
3, MARCH 1980
497
ample in the T C S P method (14).
A
0:
Rt
(3)
The count rate, R , is proportional to the fluorescence intensity which is, in turn, proportional to C@C 1\21).Therefore, the area under the decay curve is related to these molecular constants and the concentration as shown in Equation 4.
A
s:
C+ct
(4)
The area, A , under the decay curve is given by the integral of the decay curve equation, shown below.
Thus, integration of a given decay curve yields Equation 6,
A =alo~
(6)
and, finally, combining Equations 4 and 6 gives the useful relationship in Equation 7.
alo7 = tgct
(7)
If two samples, 1 and 2, were counted for equal times using the same I,, then the deconvolution would result in a l and T~ for sample 1, and a2 and T~ for saimple 2. These would be related as shown in Equation 8.
If samples 1 and 2 were combined in a mixture, then t and lo would be identical and the concentration ratio of the mixture components, 1 and 2 , could be found from Equation 9. Cl/C:!
= alTlf242/QT2tl+l
(9)
This is true only if components 1 and 2 are truly independent, Le., they do not react or interact with one another in the ground or excited state. Equation 9 expresses the relationship upon which this quantitation work is based. T h e values al and u p obtained through the reiterative convolution of the mixture decay curve may be used, then, in principle, to quantitate the ratio of the components, 1 and 2: identified through rl and r2 if (t@:ll and ( e @ ) * are known or can be determined. A similar expression may be used to quantitate three-component mixtures. However, experimentally, three components have never been adequately resolved by reiterative convolution (10).
EXPERIMENTAL Instrumentation. The instrumentahion of the time-correlated single photon technique used in this laboratory has been described elsewhere (12, 14,21). Cline Love and Shaver have evaluated the reiterative convolution program utilized in this work (10,221. The extinction coefficients were measured on a Bec kman Acta 111. Reagents. The compounds quinine bisulfate (Aldrich) and the atabrine homologous series (G. V. Downing, Merck, Inc., Rahway, N.J.) were used as received. Their properties have been previously discussed ( 1 2 ) . Mixtures. The components were chosen so that there were lifetime differences of at least 3 ns between them ( I O ) . Known amounts were weighed and mixed together in different proportions keeping the total concentration less that 1 X 10-j M so that quenching and interactions would be negligible. Since there were no interactions, quenching or self-absorption in the pure compounds at this concentration level. there should be none in the mixture of the homologous series. Csually the component of lower @ was present in higher concentration since it has been stated that reiterative convolution cannot recover a minor component which contributes less that 10% to the total fluorescence intensity (10). The solvents used were 0.1 N H2S04and 0.1 N HC1 because these acidic materials were found to b'e optimum for analysis of
498
ANALYTICAL CHEMISTRY, VOL. 5 2 , NO. 3, MARCH 1980
fluorescence intensity is also presented. The calculated c1:c2 ratio is compared to the known cI:c2 to test the accuracy of the method. The percent error in c1:c2 is given in the last column of the table. It can be seen by comparing values of T in Table I to those in Table I1 that when the components are correctly identified through their lifetimes, the concentration ratios are usually within 20% of the real mixture composition. However, when the lifetimes found are incorrect, the quantitation also fails. In addition, it was found that the mixtures which resulted in inaccurate lifetimes did not fit as well to a two-component decay curve according to the x’ criteria, as did the accurate T mixture curves. Table I1 shows mixtures of the same two components with increasing concentration of one component. In each set of compounds, A:B, D:B, and C:E, as the contribution of one component to the total composition decreases, the qualitative and quantitative analyses are no longer operative. T h e calculated lifetimes and concentration ratios deviate increasingly from the actual values. As found by Cline Love and Shaver (IO), the accuracy of the lifetime values is affected by the relative amounts of the components in the mixture. As one component contributes increasingly to the total fluorescence decay, it begins to overwhelm the second component and the temporal resolution is affected. Inaccurate lifetime values are obtained when the decay curves are not sufficiently resolved for the RC method to handle them. Also, it is again apparent that inaccurate lifetime values imply inaccurate quantitative analysis. The RC method cannot produce accurate T values when the lifetimes of the components are less than 3 ns apart ( 1 0 ) . I t appears from the data in this study that there are other restrictions, also, concerning the resolution of the “a” values. A comparison of the percent error in the experimental cI:c2 to the fluorescence contribution ratios, ( t @ c ) , : ( t @ ~shows ) ~ , that the failure of the analyses does not depend on the relative contributions of the components to the fluorescence signal. It appears, however, that a good indication of inaccurate results is the amplitude values, a, given by t @ c / r . When a1:a2 approaches 1:3, the qualitative and quantitative analyses fail. This implies that each component of the mixture must contribute a t least of the total decay curve, aloe-t’r, where a and k are some combination of a l and a 2 , and r 1 and T ~ or , it will not be adequately resolved for analysis. Thus, the relative values of lifetimes and fluorescence intensities interact to affect the accuracy of the qualitative and quantitative measurement. For example, a longer lived species must be proportionately more intense to retain accuracy in its analysis. Simulation studies have shown that the “a” ratios must be at least 1:15 for 10% accuracy in the fluorescence lifetime calculations (IO). These simulated decay curves contained
Table I. Fluorescence Lifetimes and Quantum Yields of Single Components extinc-
c,
compounds
tion
0,
7,
lifetime, quantum ,sa yieldb
A . q u i n i n e bisulfate
B. atabrine, ri = 3, R = CH,‘ C. homologue, 2, R - H‘’ D. homologue, 11’ 1, R = H “ E. homologue, n = 6, R = H‘
coefficients
19.4 2.9
0.510 0.056
3000 2000
6.7
0.090
2000
12.7
0.234
2000
1.6
0.013
2000
a i 0 . 3 ns; all dissolved in 0.1 N HCI except A which is in 0 . 1 N H 2 S 0 , . The lifetimes of B, C, D , and E are identical in 0.1 N H,SO,? to above. +0.001. ‘ See reference 1 2 for structures.
atabrine mixtures through fluorescence ( 1 2 ) . Procedure. Prepared mixtures of known composition were irradiated in the nanosecond TCSP fluorometer using the 337-nm excitation of a nitrogen flash lamp. Emission filters were used (Corning 0-52) which transmitted the spectra of both components equally so that the count rate, R , is truly proportional to the fluorescence intensity, t@c ( 2 1 ) . The total counts taken were approximately 10OOO and blanks were subtracted when necessary. The lifetimes T~ and T ~ and , the amplitude constants a , and a2 were determined using reiterative convolution. The criteria of good fit used was a reduced chi square test (19). The components of the mixture were qualitatively identified through T , and r 2 . The comparative quantum yields (12) and extinction coefficients of the components were then measured. The t’s of the atabrine series at 337 nm were all equivalent. That of quinine bisulfate was found to be 1.5 times greater. Using Equation 9, the concentration ratio of component 1 to component 2 in each mixture was then calculated through the use of the amplitude constants a, and a2 obtained from the reiterative convolution procedure.
RESULTS AND DISCUSSION Table I presents the experimental lifetimes ( l a = h0.3 ns), quantum yields, and extinction coefficients of the pure components in 0.1 N acid solution (12). These values are assumed to be best estimates of the lifetimes obtainable under the experimental conditions employed, and are used as reference values. T h e results of the quantitative analysis of 1 2 prepared mixtures are given in Table 11. This table presents data on mixtures of varying known composition ratio, c1:c2. The lifetimes, and T ~ and , the amplitude constants, a l and a L , determined through deconvolution are shown. The actual fluorescence contribution of each component, t@c,to the total _____-__~_____ Table 11. Quantitative Analysis of Two-Component hlixtures ~
actual mixture
C: B“ A:B“ ( 0 . 1 N H , S O , ) D: B
c,:cl
(c
Vc):
1:l
Y:2
1:lO
3:2
1:2
8:l
1:50
3: 10
1:l 1:2 1: 4
4: 1
6.5 19.1 19.9 13.1 12.b 12.5 10.6 11.0 6.6 6.7
1:1.3 1 :2..i
2:l 1:l 4 :5 6:1 5:2
1:5
7 :
