In measuring dilute solutions of rhodamine B in water, it was observed that readings on the more dilute samples sometimes drifted for hours after the silica cuvette was inserted in the instrument. The direction of drift was always toward the concentration of the previous sample, suggesting adsorption on and desorption from the walls of the silica cuvette. The rate of drift was slow enough to permit useful measurements to be made by rinsing the cuvette and reading intensities immediately after filling. They formed an analytical curve similar to that shown in Figure 7. When a Pyrex cuvette was substituted for the silica one, drift was reduced to the point where it was possible to make repeat scans of dilute samples. A series of synchronous excitation scans for the most dilute solutions with 0, 1,10, and 100 parts per trillion of added rhodamine B appear in Figure 6. The zeros are offset to avoid confusion in drawing baselines for deflection measurements. All slit widths were 6 nm, time constants 2 s, and scanning rates 1nm/s. These and additional measurements on similar recordings at higher concentrations were used to construct the analytical curve shown in Figure 7 . A t the highest concentrations the curve was matched to the 1-ppm points and interpolated from there downward according to the exponential absorption law. At the lower concentrations the readings were corrected by subtracting the reading for pure solvent. The curve was drawn to fit the point at 1 ppt by assuming that all the nonlinearity resulted from the presence of 0.6 ppt more rhodamine in the water than was indicated by the baseline on the water spectrum. This implies that a spectrum for really pure water would be slightly concave upward a t 580 nm instead of being straight like the baseline or concave downward as recorded in Figure 6. Under these conditions, the curve is linear over four decades.
The minimum detectable signal is estimated from the signal to noise ratios of the lowest two curves in Figure 6 as less than 1 ppt, which is consistent with the lowest point’s deviation from the analytical curve. Excluding unused space near the walls and additional material needed for flushing, the actual used sample volume in the middle of the cuvette was 3 X 7 X 7 mm. Minimum detectable weight of sample was figured from the amount in this volume as about gram or lo8 molecules. T o utilize these sensitivities effectively will require additional work on pure solvents and sample handling techniques.
ACKNOWLEDGMENT The author thanks Ralph Eno of Hamamatsu Corp. for the 777 photomultiplier and Joe Vergato of Perkin-Elmer for building the electronics. He also acknowledges the advice of Hamilton Marshall of Perkin-Elmer in measuring the analytical curve and the help of Nancy Read in many phases of the work. LITERATURE CITED (1)J. B. F. Lloyd, J. ForensicSoc., 2, 83 (1971). (2)J. B. F. Lloyd, J. ForensicSoc., 2, 153 (1971). (3)J. B. F. Lloyd, J. forensic SOC.,2, 253 (1971). (4) R. J. Lukasiewicz and J. M. Fitzgerald, Anal. Chem., 45, 511 (1973). (5) N. J. Harrick and G. I. Loeb, Anal. Chem., 45, 687 (1973). (6)M. F. Bryant, K. O’Keefe, and H. V. Malmstadt. Anal. Chem., 47, 2324
(1975). (7)Unpublished work.
RECEIVEDfor review June 29, 1976. Accepted August 30, 1976.
Selective Fluorescence Quenching and Determination of Phenolic Antioxidants R. J. Hurtubise Department of Chemistry, University of Wyoming, Laramie, Wyo. 8207 1
Selective fluorescence quenching with 80% ethanoV20 % chloroform was used to analyze phenollc antloxldant mlxtures. With this solvent system the fluorescence of BHA (mlxture of isomers, 2-feft-butyl-4-methoxyphenol and 3-fert-butyl-4methoxyphenol), and TBHO (mono-terf-butylhydroqulnone) is quenched, but PG (propyl gallate) remains fluorescent. PG was determlned in the presence of BHA and BHT (2,6-diteff-butyl-4-methylphenol) using 80 % ethanoV20 YOchloroform as a solvent. BHA was determined in the presence of BHT and PG using ethanol as a solvent. The fluorescence quenching of BHA and TBHO with 80% ethanol/2O% chloroform lndicates that a slmple quenching mechanlsm is not followed. The fluorescence quenching of BHT appears to be due to steric crowding of the two teff-butyl groups.
Quenching can enhance greatly the selectivity of luminescence analysis. Quenching is the reduction or elimination of luminescence intensity of a molecule by either intermolecular or intramolecular processes. Parker ( I ) has suggested the use of quenching to increase selectivity, but there has been very little use of quenching phenomena in organic trace analysis.
The food antioxidant PG (propyl gallate) can be determined fluorometrically at the part-per-million level in food products in the presence of the other food antioxidants BHA (mixture of isomers, 2-tert-butyl-4-methoxyphenol and 3-tert -butyl4-methoxyphenol) and B H T (2,6-di-tert-butyl-ri-methylphenol) ( 2 ) .The fluorescence of BHA and B H T is quenched in chloroform, but PG remains fluorescent in chloroform and can be determined readily. Recently it was shown that P G could be determined in the presence of BHA, BHT, and TBHQ (mono-tert-butylhydroquinone)with selective quenching using chloroform as a solvent (3).TBHQ is a new food antioxidant approved by the federal government for use in food products (4).Sawicki (5) used the phrase “quenchofluorescence analysis” to describe the analysis of fluoranthenic hydrocarbons. He used nitromethane to selectively quench the fluorescence of over 25 non-fluoranthenic hydrocarbons; only hydrocarbons with a fluoranthenic ring as part of their structure were fluorescent. Lloyd (6) employed synchronously excited fluorescence and plots of the SternVolmer relationship resulting from the variation of quencher concentration for the characterization of engine oils and other complex mixtures for benzo[a]pyrene and other polycyclic aromatic hydrocarbons. Zander (7) has investigated the flu-
2092 * ANALYTICAL CHEMISTRY, VOL. 48, NO. 14, DECEMBER 197’6
orescence quenching of polycyclic aromatic hydrocarbons in a benzene/methyl iodine mixture and suggested that the unquenched fluorescent compounds in mixtures can be determined selectively by fluorometric analysis. This paper reports the quenching properties of BHA, BHT, PG, and TBHQ in 80% ethanol/20% chloroform and the analysis of food antioxidant mixtures for BHA and P G using selective fluorescence quenching and fluorometric analysis. Also, possible mechanisms for the fluorescence quenching are discussed.
60
1
EXPERIMENTAL Reagents. Ethanol and analytical reagent chloroform were distilled once, using a distilling column similar to the one described by Winefordner and Tin (8). BHA, BHT, PG, and TBHQ were food grade antioxidants and used without further purification in the analysis experiments. For some of the quenching experiments, samples of BHA and TBHQ were recrystallized from distilled ethanol. The antioxidants were obtained from Eastman Chemical Products, Inc., Kingsport, Tenn. Apparatus. All results were obtained with a Perkin-Elmer MPF-2A in the ratio mode with the excitation and emission slits set and then three a t 10 nm. All glassware was first rinsed in 50% "03 times in distilled water and finally dried in an oven. Procedures. Determination of B H A with No PG Interference. For these experiments two sets of solutions were prepared in ethanol, each set having three solutions. Each solution in the first set contained 0.10 pg/ml each of BHA, BHT, and PG, but two of the solutions were prepared to contain added concentrations of 0.10 pg/ml and 0.20 pg/ml BHA, respectively. Each solution in the second set contained 0.60 pg/ml BHA and 0.20 fig/ml PG, but two of the solutions also were prepared to contain added concentrations of 0.20 pg/ml and 0.40 pg/ml BHA, respectively. The relative fluorescence intensities of the three solutions in each set were measured a t the maximum excitation and emission wavelengths for BHA in ethanol; they are 297 nm and 323 nm, respectively. A standard addition curve was prepared for each set by plotting relative fluorescence intensity vs. pg/ml of BHA added. The line intersecting the relative intensity axis was extrapolated to zero relative intensity to obtain the BHA concentrations in the prepared solutions. Determination of PG. Three sets of solutions were prepared in 80% ethanol/20% chloroform on a volume-volume basis containing the same antioxidants and concentration levels described above and below for BHA determinations. Each set of solutions was treated in an identical way as described for BHA determinations above except PG was added in known amounts, and the maximum excitation and emission wavelengths for PG were used; they are 293 nm and 364 nm, respectively. A standard addition curve was prepared and the PG concentration obtained. Determination of B H A u t h PC Interference. A set of five solutions was prepared in ethanol each containing 0.060 pg/ml BHA and 0.60 pg/ml PG except two solutions contained added concentrations of 0.20 pg/ml and 0.40 kg/ml BHA, respectively, and two other solutions contained added concentrations of 0.20 pg/ml PG and 0.40 pg/ml PG, respectively. All solutions were excited a t 297 nm and the fluorescence was measured a t 323 nm. A standard addition curve for PG was prepared by plotting the relative fluorescence intensities obtained from the unknown solution and the two standard addition solutions with PG vs. pg/ml of PG added. The best straight line through the data points provides a calibration curve with a siope representative of fluorescence intensity of PG alone. An intensity value corresponding to the previously determined PG concentration in 80%ethanol/20% chloroform was obtained from this curve and subtraction of the unknown fluorescence intensity from this value gave the absolute intensity due to PG. The latter value was then subtracted from the intensities obtained from the unknown and the two standard addition BHA solutions. These values were plotted vs. pg/ml of BHA added to the unknown. Extrapolation of the best straight line through these points to zero relative intensity gave the concentration of BHA in the unknown solution (9). Quenching Studies. Three series of solutions were prepared. One series contained 0.20 pg/ml BHA, and another series contained 0.20 pg/ml TBHQ. The BHA and TBHQ series were prepared in duplicate and average relative fluorescence intensity values were used. Also, a series of blank solutions was prepared. In all series there were six solutions. One solution in a series was diluted to 50 ml with ethanol and the remaining five were diluted to 50 ml with ethanol after adding a known volume of chloroform. The amounts of chloroform added
4
6
12
16
20
C H C L 3 , VOL. %
Figure 1. Effect of chloroform concentration on the fluorescence intensity of BHA and TBHQ: (x) BHA, (*) TBHQ
sequentially to five of the solutions follows: 0.20,0.50,2.0,4.0, and 10.0 ml. The BHA solutions were measured at a 297-nm excitation wavelength and at a 323-nm emission wavelength. The TBHQ solutions were measured a t a 300-nm excitation wavelength and a t a 329-nm emission wavelength. The relative fluorescence intensities for the BHA and TBHQ solutions were corrected using the relative fluorescence intensities obtained from the blank solutions. Plots of Fo/F 1vs. quencher concentration were made. Fo is the relative fluorescence intensity without the quencher present and F is the relative fluorescence intensity with quencher present.
RESULTS AND DISCUSSION Quenching Properties. BHA, PG, and TBHQ are highly fluorescent in ethanol and BHT is very weakly fluorescent (2, 3 ) . The fluorescence of BHA, BHT, and TBHQ is quenched in chloroform, but PG remains fluorescent. Earlier a method was reported for determining PG in chloroform with BHA, BHT, and TBHQ present by selective fluorescence quenching ( 3 ) .The sensitivity of the PG determination is increased and the fluorescence of BHA, BHT, and TBHQ is still quenched by using a solvent mixture of ethanol and chloroform. After trying several chloroform-ethanol mixtures to determine the relative fluorescence intensities of BHA, BHT, TBHQ, and PG it was decided to use 80% ethanol/20% chloroform for studying the fluorescence of the antioxidants. With this solvent combination the fluorescence of BHA and TBHQ is almost quenched completely and PG remains fluorescent. BHT is unique among the four antioxidants studied because it is very weakly fluorescent in ethanol, chloroform, and chloroform-ethanol mixtures. In fact, for this study it was considered nonfluorescent because no useful analytical data could be obtained from it. This has been previously reported ( 2 ) . Figure 1 shows the variation of relative fluorescence intensity of solutions of BHA and TBHQ vs. vol % of chloroform. The relative fluorescence intensity of a 0.20 pg/ml BHA decreased from 78.8 in ethanol to 0.5 in 80% ethanol/20% chloroform. TBHQ at the same concentration decreased from a relative fluorescent intensity of 16 in ethanol to 0.2 in 80%ethanol/20% chloroform. It can be seen that the relative fluorescence intensity of BHA and TBHQ decrease in a similar manner as the amount of chloroform increases, In other experiments with BHA, PG, and TBHQ at 1pglml each in separate 80% ethanol/20% chloroform solutions, it was found the relative fluorescence intensity of BHA decreased by 97.1%, PG by 27.0%, and TBHQ by 97.0% compared to ethanol solutions of the antioxidants. Even though the fluorescence of PG was quenched somewhat in 80%ethanol/20% chloroform, the fluorescence of BHA and TBHQ was in effect
ANALYTICAL CHEMISTRY, VOL. 48, NO. 14,DECEMBER 1976
2093
Table I. Analysis of Mixtures of Antioxidants" Found, ,ug/ml
Present, pg/ml BHA
BHT
PG
BHA
PG
BHA
PG
0.10
0.10
0.10
0.10 0.61
0.12 0.18
0.0
0.20
20 10
0.60
O.07Ob 0.59
0.60
0.060 b
Av error, %
., . .,,
1.7 17
30
1.7
a Results based on the average of duplicate determinations. Corrected for contribution due to PG.
quenched completely and allowed for direct accurate determination of P G in the presence of BHA and TBHQ (Table I). Previously P G was determined in chloroform ( 3 ) .However, PG determinations in 80%ethanol/20% chloroform are more sensitive than in chloroform and still allow for selective quenching of BHA and TBHQ. The relative fluorescence intensity per microgram per milliliter of PG in ethanol, 80% ethanol/20% chloroform, and chloroform are in the ratios 3.0:2.2: 1.0, respectively. The maximum excitation and emission wavelengths found for BHA, PG, TBHQ in ethanol were used throughout this work. Experiments showed that the variation in the maximum excitation and emission wavelengths for BHA, PG, and TBHQ was not significant when varying amounts of chloroform were added to ethanol solutions of the antioxidants. Actually the maximum excitation and emission wavelengths of PG are more important because BHA and TBHQ are essentially nonfluorescent a t 80% ethanol/20% chloroform. The maximum excitation and emission wavelengths for PG in ethanol are 293 nm and 364 nm, respectively, and in 80%ethanol/20% chloroform are 294 nm and 364 nm, respectively. Analysis of Mixtures. T o demonstrate the analytical usefulness of selective fluorescence quenching, mixtures of antioxidants were analyzed using 80% ethanol/20% chloroform. The standard addition method was used and the advantages of this approach were discussed previously ( 2 , 3 ) .A complete fluorometric analysis of a mixture containing BHA, BHT, TBHQ, and PG cannot be made with 80%ethanol/20% chloroform and ethanol solvents. B H T cannot be determined because of its low fluorescence quantum efficiency. If BHA and TBHQ are present in the same mixture, they cannot be determined individually because of their similar fluorescence properties in ethanol. Several combinations of antioxidants can be determined by using both 80% ethano1/20% chloroform and ethanol as solvents because of selective quenching and the fluorescence properties of the antioxidants in the solvents (2, 3 ) :a) BHA and P G in BHA, B H T and PG and simpler combinations. b) TBHQ and P G in TBHQ, B H T and P G and simpler combinations. c) P G in P G and BHT. If just chloroform is used as a solvent, P G can be determined in a mixture of BHA, BHT, TBHQ, and PG, and simpler combinations. Table I shows the analysis of some of these combinations using 80% ethanol/20% chloroform and ethanol solvents. Previously it was shown that BHA could be determined directly in food samples containing BHA and P G or BHA, BHT, and PG by first determining P G in chloroform and then determining BHA in 50% ethanol-ligroine (9). Because both BHA and PG are fluorescent in 50% ethanol-ligroine, a correction was made for the contribution of P G to the total fluorescence from the mixture. A similar approach was used in this study except ethanol was used in the determination of BHA. In the determination of BHA in 50% ethanol-ligroine, a 1P21 photomultiplier tube was employed which has a maximum response a t approximately 400 nm (10). In this work a 1P28 photomultiplier tube was used which has a maximum response at approximately 340 nm (10). With the 2094
0
0.20
0.40
0.60
0.80
1.0
CHCLa, M
Flgure 2. Stern-Volmer graphs for BHA and TBHQ: (x) BHA, TBHQ
(a)
1P21 photomultiplier tube, the maximum excitation and emission wavelengths for PG in ethanol were 286 nm and 372 nm, respectively ( 2 ) ;with the 1P28 photomultiplier tube and P G in ethanol, they are 293 nm and 364 nm, respectively. Because the 1P28 photomultiplier tube is more sensitive a t shorter wavelengths, its relative response to the fluorescence of P G will be less than the relative response of the 1P21 photomultiplier tube a t the same concentration of P G and using the corresponding maximum PG emission wavelengths for the two photomultiplier tubes. In the earlier work with the 1P21 photomultiplier tube, the correction for PG fluorescence was larger than for this work; thus there is less interference from P G in the determination of BHA when a 1P28 photomultiplier tube is used. Interestingly, for the first two concentration levels of BHA and PG in Table I, no correction was needed for the fluorescence due to P G when determining BHA. This can be explained by a combination of three facts: a ) the concentration level of P G was not great enough to yield significant fluorescence, b) the maximum emission wavelength of BHA is a t 323 nm compared to P G a t 364 nm, and c) the 1P28 photomultiplier tube is more sensitive at shorter wavelengths. The standard addition plots for P G for the first two samples in Table I gave lines that were parallel to the concentration axis indicating the fluorescence intensity did not change with concentration of PG. When the concentration of P G was ten times that of BHA, as indicated by the last concentration level in Table I, the standard addition plot gave a line whose slope was representative of P G concentration. Thus, the total fluorescence a t the maximum excitation and emission wavelengths for BHA had to be corrected for the fluorescence due to PG. Without the correction a value of 0.084 ,ug/ml BHA was obtained, or an error of 40%. With the correction the error was reduced to 17%. With BHA, B H T and PG or BHA and P G mixtures it can be decided easily whether PG will interfere in the BHA determination. If the slope of standard addition plot for P G is zero there will be no interference from PG. If the slope is greater than zero, a correction has to be made for the P G fluorescence. Table I also shows that P G can be determined accurately in antioxidant mixtures by using 80%ethano1/20% chloroform. The main advantages t o using 80% ethanol/20% chloroform are that P G can be determined with greater sensitivity compared to previous work and the fluorescence of BHA and TBHQ is quenched by this solvent combination. Fluorescence Quenching. In general, there are two types of fluorescence quenching: dynamic quenching and static quenching. In dynamic quenching a diffusion-controlled in-
* ANALYTICAL CHEMISTRY, VOL. 48, NO. 14, DECEMBER 1976