Anal. Chem. 1985,57,436-439
436
Two-Photon Excitation of Fluorescence Spectrometry of Methylnaphthalene Derivatives Prepared in a Low-Temperature Durene Crystal Steven M. Thornberg and Jon R. Maple* Department of Chemistry, University of New Mexico, Albuquerque, New Mexico 87131
The analytical utility of Combining low-temperature sample preparation procedures with two-photon excitation (TPE) of fluorescence spectrometry has been explored. A 10 K durene crystal was used as the sampling medium for methyl derlvatives of naphthalene. I n order io minimize problems caused by the low TPE efflclency, a photomultlpller tube was placed directly In front of the sample, using a (UV transmitting) filter to protect the tube from laser scatter. Because of the high resolution of the TPE peaks (0.5-1.7 cm-’ fwhm bandwidth), each component of a nine-component methylnaphthalene mlxture could be Identifled from a single TPE scan. Likewise, the major methylnaphthalene components In a raw shale oil sample prepared In a durene crystal (without prior separations) could also be Identifled from a single TPE scan. It has been demonstrated that femtogram quantities of some methylnaphthalene derivatlves can be detected and that the TPE detection limits are 1 to 2 orders of magnltude higher than those obtained with UV excitation of fluorescence.
Relatively few analytical applications of two-photon excitation (TPE) of fluorescence spectrometry appear in the literature in spite of the fact that it is recognized that the polarization properties of TPE and the differing one- and twophoton selection rules can be exploited for qualitative analysis ( I , 2 ) . At present TPE of fluorescence spectrometry apparently has not been successfully utilized for the characterization of multicomponent, isomeric mixtures of organic molecules. The reason for this is that previous analytical applications of TPE utilized samples prepared in room-temperature solutions. The broad absorption and fluorescence bands which are characteristic of organic compounds prepared in solution greatly complicate multicomponent analysis, regardless of the method in which fluorescence is induced. In contrast, laser-induced fluorescence (LIF) spectrometry has proven to be highly useful for the characterization of multicomponent mixtures of polycyclic aromatic hydrocarbon (PAH) isomers when low-temperature sampling techniques are utilized (3-9). The success of LIF methodologies can be attributed to two primary factors (3, 5 , 7, 8 ) . First, highresolution absorption and fluorescence bands result when a PAH is prepared in a suitable low-temperature sampling medium, such as a Shpol’skii matrix. Second, a tunable monochromatic excitation source (Le., a dye laser) can be tuned to the narrow absorption line of the analyte, which will therefore be selectively excited. The presence of the analyte can be confirmed by the highly resolved and well-defined structure of the low-temperature fluorescence spectrum. Thus, each component of a complex mixture can, in principle, be identified by tuning the dye laser to a suitable absorption line and obtaining the resulting fluorescence spectrum. The primary purpose of this paper is to assess the analytical advantages of combining low-temperature sample preparation procedures with TPE of fluorescence spectrometry. In an0003-2700/85/0357-0436$01S O / O
ticipation of the weak signal caused by the low T P E efficiency, the emission monochromator has been removed from our LIF spectrometer, and the photomultiplier tube has been placed directly in front of the sample, using a (UV transmitting) filter to block laser scatter and to protect the tube. As a result all wavelengths of fluorescence are simultaneously detected and the signal is maximized. However, as will be shown, another primary advantage of this experimental configuration is that all of the components of a multicomponent mixture can be identified from a single T P E scan, in contrast to the selective excitation technique, which generally requires N scans for the identification of N components. T o accomplish this same task with a conventional UV (i.e., one photon) excitation scan, it would be necessary to scan over high energy vibronic bands, rather than the low-energy 0-0 bands, in order to filter intense laser scatter from the relatively weak fluorescence. Since there are many vibronic excitation bands for each component of a complex sample and since the bandwidths of vibronic bands are larger than those for 0-0 bands, vibronic bands from different components (particularly substitutional isomers) are much more likely to overlap than are 0-0 bands. Therefore, excitation scans over high energy vibronic states generally result in complex excitation spectra of multicomponent samples and are of limited utility for multicomponent analysis. One final advantage of utilizing TPE spectrometry for multicomponent analysis is the possibility of resolving overlapping bands from two components by altering the two-photon absorptivity. This can be achieved by changing the polarization (i.e,, linear vs. circular) of the excitation source ( I ) . Since the TPE scans that are presented in this paper have more than adequate resolution, this latter advantage of TPE has not been exploited. The prototype system used to demonstrate the applicability of T P E of fluorescence spectrometry to multicomponent analysis consists of methylnaphthalene (MN) samples prepared in a low-temperature durene crystal. This matrix choice was made since we have previously demonstrated that highresolution fluorescence and absorption bands result when MNs are prepared in a 10 K durene crystal (9).
EXPERIMENTAL SECTION Reagents. Unless stated otherwise, all naphthalene derivatives were used as received and were obtained from Aldrich. 1,5-Dimethylnaphthalene (l,s-DMN) was purchased from Wiley Organics. Although several of these “pure” materials contained substantial amounts of naphthalene, 1-methylnaphthalene (1MN), 2-MN, and 2,3-DMN impurities, impurity peaks in the “pure”compound TPE spectra could be unambigously identified. The oil shale is an NBS Standard Reference Material (lot no. 1580) and was diluted (loo/ 1) in acetone prior to usage. Durene and 2,3-DMN were vacuum sublimed and then zone refined with an apparatus constructed in our laboratory. This apparatus is similar to the “NV 1” assembly described by Karl and utilized the “intermittent technique” (10). Two tubes containing ca. 100 g of durene (or 2,3-DMN) were subjected to 50 passes in two separate zone refiners to remove the bulk of the impurities. Then, the top half of each tube was combined in a new tube and refined for about 380 passes for durene and 180 0 1985
American Chemical Society
ANALYTICAL CHEMISTRY, VOL. 57, NO. 2, FEBRUARY 1985
passes for 2,3-DMN. Further details concerning the zone refining apparatus and procedures can be found elsewhere (11). Sample Preparation. All samples were prepared by adding appropriate amounts of MN stock solution into the stem of a disposable glass pipet which had been sealed at the bottom. About 50 mg of durene was added and the pipet was sealed with a torch about one-half inch above the durene. The crystals were prepared by the normal freezing technique ( I O ) , wherein the sample tubes were lowered by a motor through Bridgman furnaces heated to 90 "C at a rate of about 1 cm/h. Typically, six crystals were prepared at one time in a period ranging from 1 to 2 h, though many more crystals could have been prepared in roughly the same amount of time if this had been desirable. The resulting crystals were transparent and were 1-2 cm long and 2 mm in diameter. After removal from the glass stem, the crystals were glued to the copper cold finger of a closed-cycle helium refrigerator (Air Products, Model CSA-202). The refrigerator shroud was attached and the temperature was lowered to 250 K before a vacuum was applied. This minimized problems caused by durene and methylnaphthalene volatility at room temperature. The temperature was then lowered to 10 K prior to spectroscopic observation of the crystalline samples. Instrumentation. The excitation source for all experiments was an excimer laser (Lumonics TE-861T) pumped dye laser (Quanta Ray PDL-1E). The dye laser output was 1-2 mJ/pulse in normal operation with a temporal bandwidth of ca. 10 ns fwhm. According to Quanta Ray the dye laser spectral bandwidth is ca. 0.25 cm-'. Dye laser output in the wavelength range 610-660 nm was obtained with DCM dye dissolved in methanol. The dye laser scanning rate was typically 0.2 nm/s and was controlled by an MCI-1 motor controller/interface (Quanta Ray). Fluorescence was induced by focusing the dye laser beam to a spot (about 1 mm in diameter) on top of the crystal, where the maximum signal was obtained. The top of the crystal was the last portion to crystallize during the normal freezing process and always contained the highest concentration of MNs. Excitation of the top portion of the crystal was useful for quantitative analysis, since there is a well-defined concentration gradient in crystals prepared by normal freezing (10). In order to obtain precise quantitative data, it is necessary to excite the same portion of each crystal, and it is convenient to let this portion be the top of the crystal. An RCA 8850 photomultiplier tube (PMT) was used to detect the fluorescence and was placed as close as possible to the sample (Le., about 2 in. away). Three UG-11 filters (3 mm thick, Schott Optical Glass) were used to prevent scattered dye laser radiation from reaching the PMT and a WG320 filter (Schott Optical Glass) was used to attenuate W radiation below ca 320 nm. Undesirable sources of UV radiation below 320 nm were 308 nm stray light from the excimer laser and fluorescence from impurities in the durene and from the quartz windows used t o isolate the cold sample from the atmosphere. The PMT signal was integrated by a gated integrator (Evans Associates Model 4130) before digitation (by a Data Translation DT2781 Analog 1/0System) and storage on a floppy disk. Spectra were plotted by an X-Y recorder (via a D/A port on the DT2781) and were not corrected for a decrease in laser power near the ends of the dye tuning range. Further details concerning the data collection are described elsewhere (9, 11). When it was desirable to observe the fluorescence spectrum, fluorescence was dispersed by a 0.64-m grating monochromator (Instruments SA HR-640, f/5.2, equipped with a 1200 groove/mm holographic grating). Conventional UV excitation could be achieved by frequency doubling the dye laser pulses with an angle-tuned KDP crystal.
RESULTS AND DISCUSSION Qualitative Features of TPE Spectra. I t is well-known that the S1 electronic state of naphthalene has ungerade symmetry and that the So SI transition is one photon allowed and two photon forbidden ( I , 2,12,13). These same parity selection rules would be expected to be valid for centrosymmetric naphthalene derivatives (such as 1,5-DMN) prepared in a durene lattice, which is characterized by Cisite symmetry (13). As evidence of this we note that 1,5-DMN
-
437
a
1,B-DMN
I
1
2,3-DMN
b
I
!
-
~
645
-
- --
__
~~
625
Wavelength (nm)
Flgure 1. Two-photon excitation spectra of (a) 2 wg of 1,5dimethylnaphthalene and (b) 100 fg of 2,3dimethylnaphthaleneprepared in a durene crystal.
has a strong one photon 0-0 absorption band near 322.5 nm, while the corresponding two-photon transition a t ca. 645 nm is extremely weak in the T P E spectrum illustrated in Figure la. The numerous two-photon transitions shown in Figure l a undoubtedly represent absorptions by vibrational levels (in Sl) with ungerade symmetry. Unlike 1,5-DMN, however, most MNs are not centrosymmetric. Consequently, the parity selection rules are not operative and the 0-0 two-photon transition should be allowed. This expectation is verified by the T P E spectrum of 2,3-DMN in Figure l b , wherein the 0-0 transition a t 642.5 nm is much stronger than the higher energy vibronic transitions and the 0-0 line is therefore the most useful for the identification of noncentrosymmetric MNs from T P E spectra. Most of the T P E spectra of MNs (in durene) are characterized by two intense 0-0 bands, which presumably result from MN molecules occupying both of the two possible positions in the durene crystal unit cell (9, 13). Because of site splitting, fluorescence spectra of MNs tend to be rather complicated and consist of a series of doublets. The components of each doublet can be selectively excited by tuning a dye laser to the appropriate 0-0 band, and the fluorescence spectra obtained by dye laser excitation of the 0-0 absorption bands are identical except for a shift in the peak position which corresponds to the energy difference between the 0-0 bands (9). The two-photon excitation 0-0 wavelengths for each of the MNs which have been examined are given in Table I. From this table it can be seen that all of the MNs except for naphthalene, 1,4-DMN, l,&DMN, and 2,3-DMN have two 0-0 bands. Apparently, these four compounds reside in only one of the two possible lattice sites. Although neither site splitting nor the two 0-0 bands for 1,5-DMN can be observed in the TPE spectrum, they can be distinguished in fluorescence spectra (11). All of the MN 0-0 T P E bands are highly resolved and have bandwidths in the range 0.5-1.7 cm-' fwhm.Apparently, these bandwidths are not limited by the excitation source. In contrast, the resolution of MN fluorescence spectra obtained with our LIF spectrometer is monochromator limited at about 10 cm-l. Consequently, there is an inherent resolution advantage associated with TPE spectra (as well as with conventional excitation spectra). It can be seen that the difference
438
ANALYTICAL CHEMISTRY, VOL. 57, NO. 2, FEBRUARY 1985
Table I. Two-Photon Excitation (TPE) 0-0 Wavelengths for the Methylnaphthalenes compound naphthalene 1-methylnaphthalene 2-methylnaphthalene 1,3-dimethylnaphthalene 1,4-dimethylnaphthalene 1,5-dimethylnaphthalene 1,6-dimethylnaphthalene 1,7-dimethylnaphthalene l,%dimethylnaphthalene 2,3-dimethylnaphthalene 2,3-5-trimethylnaphthalene
a
1,6
TPE 0-0 band," nm
9-COMPONENT 2,3 MIXTURE 1,4
625.96*
640.21, 639.17 641.13, 639.17 645.02, 642.35
642.02 622.40b 654.05, 650.83
6 55
625
643.09, 643.51 643.49 642.52 649.65, 648.80
,2,3
"Absolute accuracy = f0.15 nm. bThis is an analytically useful two-photon excitation wavelength when DCM laser dye is used. The 0-0 bands were too weak to observe. in 0-0 wavelengths between any two pairs of the MNs (listed in Table I) is sufficiently greater than the T P E bandwidths, and it should be possible to resolve each 0 4band from a T P E spectrum, even when all of these MNs are present in the sample. One minor exception to this general rule is the fact that the 639.17-nm 0-0 bands of 1-MN and 2-MN cannot be resolved. This creates no problem, since they both have a second 0-0 band which can be readily resolved. Another exception to the rule is that the 0-0 band of l,&DMN cannot be distinguished from the 643.51-nm 0-0 line of 1,7-DMN. All other possible combinations of 0-0 wavelengths can be adequately distinguished, including the 0-0 lines for 1,3-DMN and 2,3-DMN at 642.35 nm and 642.52 nm, respectively. In order to test the utility of TPE for qualitative analysis, a mixture consisting of naphthalene (12 pg), 1-MN (2 pg), 2-MN (4 pg), 1,3-DMN (6 pg), 1,4-DMN (1wg), 1,5-DMN (3 wg), 1,6-DMN (24 wg), 2,3-DMN (0.5 wg), and 2,3,5-trimethylnaphthalene (8 fig) was prepared in a durene crystal. As evidenced by the TPE spectrum in Figure 2a, a 0-0 band for each (noncentrosymmetric) component is clearly resolved and can be unambiguously identified from a single T P E scan. The two centrosymmetric compounds can be identified from higher energy vibronic transitions. In order to fully appreciate this result it should be noted that with the "conventional" selective excitation technique, it is generally necessary to selectively excite the component of interest and then obtain a fluorescence spectrum to confirm the presence of this component. Thus, in order to identify the nine components of this mixture, nine different fluorescence spectra must be obtained, With the T P E technique utilized here, only one TPE scan was needed. The success of this method in characterizing the ninecomponent mixture in a single T P E scan is due both to the high resolution of the peaks and to the fact that all wavelengths of fluorescence can be observed simultaneo>:sly. Consequently, the question arises concerning whether 0, not this T P E technique is applicable to complex samples with a large number of components (i.e., interferents) with broad absorption bands in the wavelength region of analytical interest. In order to partially answer this question, 30 pg of a raw shale oil sample was prepared in a durene crystal. No attempt was made to remove a light yellow coating which formed on top of the crystal. The T P E spectrum of this crystal is presented in Figure 2b. From this spectrum naphthalene, 1-MN, 2-MN, 1,7-DMN, 1,8-DMN, and 2,3DMN can be readily identified and it can be concluded that T P E is indeed applicable to the analysis of very complex samples. Although the 1,7-DMN and 1,8-DMN bands (at 643.51 nm an 643.49 nm) cannot be adequately resolved, we were able to unambiguously determine that the peak (at 643.5 nm) labeled as 1,8-DMN in Figure 2b was predominately due
S H A L E OIL
b
1,7 1 I
655
625 Wavelength (nm)
Flgure 2. Two-photon excitation spectra of (a) a nine-component methylnaphthalene mixture and (b) 30 pg of an unfractionated shale oil sample prepared In a durene crystal: N, naphthalene; 1, 1methylnaphthalene: 2, 2-methylnaphthalene; 1,3, 1,3-dimethylnaphthalene(1,3-DMN); 1,4, 1,4-DMN; 1,5, 1,B-DMN; 1,6, 1,B-DMN; 1.7, 1,7-DMN; 1,8, l$-DMN; 2,3, 2,3-DMN; 2,3,5, 2,3,5-trimethylnaphthalene.
to 1,8-DMN, rather than 1,7-DMN. In the absence of 1,8DMN, the peak height ratio of the TPE 0-0 bands of 1,7DMN at 643.51 nm and 643.09 nm, respectively, is consistently about 0.1. In Figure 2a the peak height ratio is greater than 1. In situations where 1,8-DMN is present in small amounts compared to 1,7-DMN, it probably will be necessary to selectively excite the 1,S-DMNand determine its presence from a LIF spectrum. The reason for the small T P E background signal from the multitude of components in the shale oil sample is that methylnaphthalenes exhibit high-resolution absorption bands when prepared in a durene crystal, in contrast to the vast majority of other components. As we have previously noted, this is because MN molecules are approximately the same size as the durene molecule (9). Thus, the two-photon peak excitation efficiency for MNs in durene should be much greater than those for compounds with broad absorption bands. Because of this, TPE should be applicable to the characterization of the PAH content of very complex samples, provided that an appropriate low-temperature matrix is utilized. Shpol'skii matrices (3, 7, 8) would therefore appear to be well-suited for the characterization of complex mixtures by T P E of fluorescence spectrometry. Quantitative Results. The T P E detection limits for pure MNs in durene generally range from somewhat less than 100 fg for 2,3-DMN to 100 ng for 1,5-DMN and are significantly lower than the MN results obtained by alternative, highresolution LIF methods, such as rotationally cooled laser induced fluorescence (14)or fluorescence line narrowing spectrometry (15). However, the detection limits are generally lower with our LIF spectrometer when UV excitation is employed. For example, with UV excitation the detection limits for 2,3-DMN, 1-MN, and 1,B-DMNwere approximately 1 fg, 50 fg, and 1 ng, respectively, while the T P E detection limit for 1-MN was ca. 500 fg. This indicates that the very low detection limits that we have observed are due in part to the normal freezing sample preparation procedure and that 1 to 2 orders of magnitude of sensitivity must be sacrificied when
ANALYTICAL CHEMISTRY, VOL. 57, NO. 2, FEBRUARY 1985
utilizing T P E , rather than conventional UV excitation. Nevertheless, the TPE spectrum of 100 fg of 2,3-DMN in Figure l b provides a vivid example of the potential of T P E of fluorescence spectrometry for trace analysis. We have noticed several peculiar phenomena that occur when the samples are prepared by normal freezing, rather than by an alternative method previously described (9). For example, the detection limit for 2,3-DMN is 6 orders of magnitude lower than those for some of the other MNs, even though we previously (9) did not notice significant differences in the detectability of 2,3-DMN and the other MNs. This phenomenon cannot be due to an unusually high two-photon absorptivity since similar effects occur when UV excitation is employed. We have made new solutions and have employed several blanks (i.e., a durene crystal with no 2,3-DMN inside) to verify that the detection limits for 2,3-DMN are indeed extremely low with respect to some of the other MNs. Another peculiarity that has been observed is that several (but not all) of the MNs can also be detected at femtogram levels, provided that 2,3-DMN is present in at least a 20-fold excess (11). In addition a calibration curve obtained for 2,3-DMN (using 1-MN as an internal standard) was linear from 100 ng to 1 mg but was nonlinear for less than 100 ng of 2,3-DMN. At present, 2,3-DMN is the only compound for which a calibration curve has been obtained. We believe that these unusual phenomena are due to competition between analyte and impurity molecules for the occupation of lattice sites during the normal freezing process. We are currently attempting to significantly improve the purity of the durene matrix in order to test this assumption. Improvements in the durene purity should minimize these effects and should also lower the detection limits, which are currently limited by fluorescence from impurities present in the durene. It is also possible that some MNs alter the local structure of the host in a manner which could affect the solubility of other analytes in the solid host. The usage of a deuterated analyte as internal standard should help to substantiate this possibility. The results that we have obtained indicate that T P E of fluorescence spectrometry is a highly sensitive and selective analytical procedure which can be used for the rapid characterization of complex isomeric samples. The quantitative problems that have been encountered can probably be minimized by utilizing a purer grade of matrix material or may
439
even be eliminated by utilizing alternative sample preparation procedures, such as vapor deposition (Le., matrix isolation) techniques ( 3 ) ,provided that the resulting deposits are not amorphous. We are currently exploring this possibility and have found that crystalline durene deposits are formed as long as the temperature of the surface (upon which the durene is deposited) is higher than ca. 240 K. Finally, we conclude by noting that the T P E methodology developed herein is applicable to any organic (or inorganic) sample in which it is possible to find a suitable low-temperature matrix which results in sufficient reduction of the excitation bandwidth. Registry No. Naphthalene, 91-20-3; 1-methylnaphthalene, 90-12-0;2-methylnaphthalene, 91-57-6; 1,3-dimethylnaphthalene, 575-41-7; 1,4-dimethylnaphthalene,571-58-4; 1,5-dimethylnaphthalene, 571-61-9;1,6-dimethylnaphthalene,575-43-9; 1,7dimethylnaphthalene, 575-37-1; 1,8-dimethylnaphthalene,56941-5; 2,3-dimethylnaphthalene,581-40-8; 2,3,5-trimethylnaphthalene, 2245-38-7;durene, 95-93-2.
LITERATURE CITED (1) Wirth, M. J.; Lytle, F. E. In "New Applications of Lasers to Chemistry"; Hieftje, G. M., Ed.; American Chemical Society: Washington, DC. 1978; ACS Symposium Series 85; pp 24-49. (2) Wirth, M. J.; Koskelo, A. C.; Mohler, C. E.; Lentz, B. L. Ana/. Chem. 1981, 5 3 , 2045-2048. (3) Maple, J. R.; Wehry, E. L.; Mamantov, G. Anal. Chem. 1980, 52, 920-924. (4) Maple, J. R.; Wehry, E. L. Anal. Chem. 1981, 53, 266-271. (5) Conrad, V. B.; Wehry, E. L. Appl. Spectrosc. 1983, 37, 46-50. (6) Brown, J. C.: Duncanson, J. A,. Jr.; Small, G. J. Anal. Chem. 1980, 52. 1711-1715. (7) Yang, Y.; D'Silva, A. P.; Fassel, V. A. Anal. Chem. 1981, 53, 894-899. (8) D'Silva. A. P.; Fassel, V. A. Anal. Chem. 1984, 5 6 , 985A-1000A. (9) Thornberg, S. M.: Maple, J. R. Anal. Chem. 1984, 56. 1542-1544. (10) Karl, N. I n "Crystals: Growth, Properties, and Applications"; Freyhardt, H. C., Ed.; Springer-Verlag: New York, 1980; pp 1-90. (11) Thornberg, S. M. Ph.D. Dissertation, University of New Mexico, 1984. (12) Hochstrasser, R. M.; Sung, H. N. J . Chem. Phys. 1977, 66. 3276-3296. (13) McClure, D. S. J. Chem. Phys. 1954, 22, 1668-1675. (14) Hayes, J. M.; Small, G. J. Anal. Chem. 1982, 54, 1202-1204. (15) Conrad, V. B. Ph.D. Dissertation, University of Tennessee, 1983.
RECEIVED for review August 20, 1984. Accepted November 13, 1984. Acknowledgment is made to the Sandia UNM Research Program and to the donors of the Petroleum Research Fund, administered by the American Chemical Society, for the partial support of this research.
