Excitation resolved synchronous fluorescence analysis of aromatic

Sofian M. Kanan, Mohammad A. Omary, and Howard H. Patterson , Masaya Matsuoka and ... C. Ragupathi , J. Judith Vijaya , P. Surendhar , L. John Kennedy...
0 downloads 0 Views 959KB Size
Anal. Chem. 1987,59,2180-2187

2180

C

aJ c C

r

,e

serious qualitative problem is the relatively large number of emission lines from impurity elements. Attempts at reducing this impurity radiation by flame heating, acid washing, and ultrasonic cleaning were unsuccessful. The use of a higher purity graphite material should reduce the problem. Detection limits for Cr, Pb, and Mn in the Na-doped plasma are 6.0,12, and 1.4 ng, respectively. These values probably can be reduced by using a dual channel photoelectric detection system where a simultaneous background correction is obtained for every shot. Finally, it should be noted that while the use of the plasma current for magnetic field generation is very convenient, the oscillatory nature of the underdamped RCL discharge circuit results in periods of very low magnetic field strength. This reduces the effectiveness of the magnetic field confinement. The use of a unidirectional discharge circuit to improve magnetic field confinement will be described in a subsequent report.

Registry No. Na, 7440-23-5;graphite, 7782-42-5. LITERATURE CITED

Cr M a s s , ug

Figure 8. Analytical curves for the Cr 283.5-nm ion line from aqueous solution residue samples of Na,CrO, with the magnetic field (solid lines) and without the field (broken lines): (a) without Na doping; (b) with Na doping. For all curves, radiation was blocked from the region extending from the plasma axis to 9 mm on either side of the axis.

without and with Na doping, respectively. Analytical curves also were obtained for the elements P b and Mn with comparable results. While the data presented in this report are quite preliminary, they are very encouraging and do suggest that further study of this novel plasma device is warranted. The most

(1) Boyd, T. J. M.; Sanderson, J. J. Plasma Dynamics; Barnes and Noble: New York. 1969. (2) Albers. D.; Johnson, E.; Tisack. M.; Sacks, R Appl. Spectrosc. 1986, 40 . - , 60-70 -- . -. (3) Chen, F. F. Intrciduction to Plasma Physics; Plenum: New York, 1974. (4) Suh, S. H.; Sacks, R. D. Spectrochim. Acta, Part 8 1981, 3 6 8 , 1081-1096. (5) Walters, J. P. Appl. Spectrosc. 1060, 23, 317-331. (6) Johnson, E. T.; Sacks, R . D. Anal. Chem., preceding paper in this issue. (7) Suh, S. Y.; Collins, R. J.; Sacks, R. D. Appi. Spectrosc. 1981, 3 5 , 42-52. (8) J. Swan; Sacks, R. Appi. Spectrosc. 1085, 3 9 , 704-710. (9) Salmon, S.C.; Holcombe, J. A. Anal. Chem. 1078, 5 0 , 1714-1716. (10) Clark, E. M.; Sacks, R. D. Spectrochim. Acta, Part 8 1980, 3 5 8 , 471-488.

RECEIVED for review February 25,1987. Accepted May 4,1987. This work was supported by the National Science Foundation through Grant No. CHE 8411290.

Excitation Resolved Synchronous Fluorescence Analysis of Aromatic Compounds and Fuel Oil Todd A. Taylor and Howard H. Patterson* Department of Chemistry, University of Maine, Orono, Maine 04469

Excitation resolved synchronous fluorescence (ERSF) is described and used to study fluorescent aromatic compounds havlng a resolved 0-0 absorption transition in cyclohexane. ERSF shows hlgh resolution and sensltlvlty (wldth at half helght Is 3-8 nm, detection limit is parts per mllllon to parts per b i b ) . A compound’s absorptlon spectrum alone can be used to predict ERSF peak pasmOns and to estknate response factors. ERSF peaks from no. 2 and no. 6 oils and aqueous extracts of fuel oil are assigned to compound classes, and the degradation of oil on soil is Investigated.

Fluorescence spectroscopy is very sensitive for the analysis of aromatic compounds (1, 2 ) and oils ( 3 ) . Synchronous fluorescence (SF) methods are more selective than ordinary 0003-2700/87/0359-2 180$01.50/0

fluorescence methods for the analysis of mixtures. SF has been used to characterize crude and refined oils (4-7), to identify and quantify polycyclic aromatic compounds (PACs) (7-12) and coal liquefaction products (13, 14), and to study oil weathering processes in seawater and sediments (15, 16). SF analysis involves scanning the excitation and emission wavelength drives of a fluorescence spectrometer simultaneously with a constant wavelength difference ( A h ) or energy difference (Av)between them. Small values of Ah produce the narrowest SF peaks and many compounds give a single SF peak at, for example, Ah = 3 nm (10, 17). A fluorescent mixture can be characterized more completely by using several S F scans at different values of Ah or AV (5, 7, 11, 12). It has been shown that SF peak positions, peak widths, and response factors of compounds each depend on the value of AX or Av. the excitation and emission instrumental band@ 1987 American Chemical Society

ANALYTICAL CHEMISTRY, VOL. 59, NO. 17, SEPTEMBER 1, 1987

widths, and the solvent employed (18,19). Thus, a spectral library is usually drawn up under a given set of conditions before unknown samples are analyzed (10-13,20). Experimental conditions should be chosen to minimize compound peak widths and to maximize compound sensitivity, when analyzing complex mixtures. In this paper, the method of excitation resolved synchronous fluorescence (ERSF) is described. ERSF spectra are scanned in the same manner that SF spectra are, except a narrow excitation bandwidth is used to resolve the excitation structure of a compound and a broad emission bandwidth is used for high compound sensitivity. The resulting spectra are plotted as a function of excitation wavelength. The ERSF properties of aromatic compounds are described and used in conjunction with the known composition of oils to assign ERSF peaks from no. 2 and no. 6 oils and the aqueous extracts of these oils. As an application of this method of analysis, oil decomposition in a field plot is monitored by ERSF to determine the rate of decline of different aromatic classes in oil.

EXPERIMENTAL SECTION Apparatus. Fluorescence measurements were made with a Perkin-Elmer MPF-44A spectrofluorometer using a Hamamatsu R928 photomultiplier tube detector. Absorption results were obtained with a Cary 17D absorption spectrometer. Cuvettes for absorption and fluorescencewere 1.00 cm in path length and made of Supracil quartz. Fluorescence spectra were recorded on a Bascom-Turner 4120 digital X-Y recorder and corrected for any light scatteringor solvent fluorescence by subtracting the spectrum of pure solvent. Fluorescence monochromators were calibrated by using the sharp lines of the 150-W xenon excitation lamp (21). Fluorescence spectra were not corrected for instrumental characteristics. Specified values in the tables were corrected for excitation instrumental characteristics by using a correction spectrum derived from the excitation and absorption spectra of anthracene and 1,1,4,4-tetraphenylbutadiene(21). ERSF AX = 3 nm spectra were acquired by using excitation\emission instrumental bandwidths of 1\2.5 nm, and AA = 7,11,15, and 20 nm spectra were acquired by using 1\8 nm bandwidths. Fluorescence samples were maintained at 20 f 0.2 “C with a thermostated sample holder. Deoxygenation was accomplished by bubbling cyclohexane-saturated nitrogen gas through the analyte solution in a cuvette. A fitted Teflon stopper was lightly coated with glycerol to seal the cuvette, and the ERSF intensity of deoxygenated solutions was stable for at least 1h. Response factor calculations and wavelength-energy axis conversion data files were made on an Apple IIe computer interfaced with the x-y recorder. Reagents. All solutions were prepared in spectrophotometric grade cyclohexane (Aldrich,Milwaukee, WI). Inner filter effects were avoided by using low concentrations (A < 0.05) in all studies. The no. 2 home heating and no. 6 boiler heating oil were obtained locally. Fluorescent model compounds were obtained in the highest purity available from Aldrich or Sigma (St. Louis, MO), and 3,6-dimethylphenanthrenewas obtained from Analabs (Foxboro, MA). A homogeneous NBS standard oil (SRM 1634a) was obtained from the US.Department of Commerce (Washington, DC). Aqueous Extractions. No. 2 oil was added to water at a 1:20 ratio, 25.0 mL (21.53 g) of oil to 500 mL of water. The no. 6 oil was weighed (21.5g) and added to 500 mL of water. Cyclohexane (10 mL) was added to the no. 6 oil to reduce its viscosity and allow the extraction to proceed reproducibly. The water for these extractions was distilled, deionized, and buffered with 10 mM cacodylic acid (Na salt, Sigma, 98%) adjusted to pH 6.0 with HC1. The oil/water mixtures were stirred slowly with Teflon-coated magnetic stirrers for 10 days in sealed bottles in the dark (22). Field Plot Studies. A field plot site was selected on a moderately drained pristine soil in Orono, ME. The selected site was tractor tilled to a depth of 15 cm and lime, phosphorus, and potassium were applied according to soil test recommendations for grass-clover mix seeding. Treatments at a level of approximately 3% and 6% (by surface soil weight) of no. 2 home heating oil and 3% of no. 6 fuel oil were uniformly applied over the 1.5

2181

by 3.0 m plots in mid June, 1983. Nitrogen as urea was applied to the plots receiving oil to achieve a 1OO:l carbon to nitrogen ratio in the oil amendment. Because these oils were found to clump and aggregate upon weathering, 40 70-cm3cores were taken from each plot and mixed and sieved thoroughly to produce a sample for further analysis. Total oil was determined by extracting 30 g of soil with reagent grade methylene chloride in a Soxhlet extractor for 4-5 h (23). The methylene chloride was rotary evaporated from the oil and the oil was determined as percent of dry weight soil. Solutions of oil for fluorescence analysis were prepared from 10.0 g of dry weight soil by drying the soil with CaSO, and extracting with cyclohexane. The extract was diluted so that oil from the soil had a concentration of approximately 2.0 mg/L at the start of the experiment. All soil samples from the same plots were extracted and diluted identically so that the fluorescence results could be compared.

RESULTS AND DISCUSSION ERSF Peak Positions and Response Factors. The following is a simplified derivation of ERSF peak properties; more detailed mathematical treatments of synchronous peak properties are available (18,19). The SF intensity, I,, can be calculated from the equation

I, = k Ex@) Em(X’)

(1)

where Ex(X) and Em(X’) are the excitation and emission intensities as functions of the excitation and emission wavelength, respectively, and k is a constant (8). SF peak wavelengths and intensities can be calculated when the excitation and emission spectra are represented as Gaussian curves (18, 19). By use of the approximation that excitation and emission spectra are Gaussian as a function of wavelength, eq 1becomes

I, = k exp[-c(a

- A)’] exp[-d(b - A’)’]

(2)

where a and b are the maxima of excitation and emission and c and d are (2ue:)-l and (2uem2)-*, respectively, with u being the standard deviation of excitation or emission. By substituting A’ = X AX and setting dI,/dX or dI,/dX’ equal to 0, the SF peak maxima are as follows:

+

Xmax

=

A‘max

=

uc

+ bd - dAX c+d

ac

+ bd + cAA c + d

(3) (4)

In the ERSF technique we assume PACs fall into two classes: those compounds with prominent vibronic structure in their lowest electronic absorption transition (group A) and those without vibronic structure (group B). Group A includes a large variety of aromatic compounds and examples are given in Figure 1and Tables I and 11. Group B includes biphenyl, terphenyl (Figure 2), fluoranthene, and many phenyl-substituted compounds (24). Emission vibrational structure is not usually observed in ERSF studies because of the wide emission band-pass employed; thus emission peaks tend to be broad and featureless. Group A compounds have narrow vibronic excitation peaks and a broad emission peak (composed of the entire emission manifold) under ERSF conditions. Thus, c >> d and eq 3 and 4 reduce to A, = a and,,’A = a + AA, respectively (analogous to “case 2” in ref 18). In this case the SF peak maximum should coincide with the excitation spectrum’s maximum and be independent of AA when the x axis is plotted as excitation wavelength. In the more general case where n = uex/uem,eq 3 becomes Am,

a

- Ax) + n2(b n2+1

=n’+1

(5)

2182

ANALYTICAL CHEMISTRY, VOL. 59, NO. 17, SEPTEMBER 1, 1987

Table I. Comparison of SF 0-0P e a k Positions a n d Intensities at Different Slit Settings

compound

SF 0-0 peak propertyo

phenol

mean range mean range mean range mean range mean range mean range

indole fluorene naphthalene carbazole anthracene

wavelength,* nm 1/8 nm 3/3 nm 278.2 0.3 287.5 0.2 300.8 0.3 311.6 0.4 331.5 0.5 376.0 1.4

intensity ratios E ~ ( l ) / E x ( 3 ) ~ 110(1/8j/110(3/3je

1(1/8)/1(3/3)'

278.3 0.8 287.6 0.5 300.6 0.8 312.1 1.4 331.2 1.3 376.1 2.9

1.29

1.35

1.13

1.40

1.35

1.24

1.22

1.24

1.39

1.22

1.30

1.17

1.08

1.12

1.31

1.14

1.15

1.30

"Measured at AA values of 7, 11, 15 and 20 nm. bExcitation/emission slits were 1/8 or 3/3 nm. Excitation spectral correction shifts the peak positions of phenol and indole to the blue by 0.6 and 0.4 nm, respectively, and does not significantly affect other peak positions. CActual values were multiplied by (9/8)' to correct for the spectral band-pass difference. dRatio of the 0-0 excitation intensities a t slit resolutions of 1 and 3 nm, with values corrected for band-pass difference. 'Ratio of the I,' values at slits 1/8 and 3/3 nm; IIo is the ratio of the intensity of the 0-0 SF peak to the next most intense SF peak. Table 11. ERSF 0-0P e a k Properties for Aromatic Compounds Found in Oils response factorsa compound

hoo nm

1 benzene 2 toluene 3 o-xylene 4 m-xylene 5 p-xylene 6 phenol 7 1,2,4,5-tetramethylbenzene 8 p-cresol 9 anisole 10 p-methylanisole 11 indole 12 2-methylindole 13 fluorene 14 dibenzofuran 15 naphthalene 16 1-methylnaphthalene 17 2-methylnaphthalene 18 2,3-dimethylnaphthalene 19 acenapthene 20 2,6-dimethylnaphthalene 2 1 1-naphthol 22 2-naphthol 23 dibenzothiophene 24 carbazole 25 2,3-benzofluorene 26 4H-cyclopenta[deflphenanthrene 27 phenanthrene 28 3,6-dimethylphenanthrene 29 chrysene 30 benzo[e]pyrene 31 pyrene 32 anthracene 33 9-methylanthracene 34 benzo[a]pyrene 35 perylene

269 269 271 273 275 277 280 286 277 287 287 288 301 302 312 314 319 319 321 325 322 328 326 331 340 345 346 352 361 367 372 376 386 403 437

width a t half height, nm

calcd R(7)

4.5

37

59

3.5 4.0 5.9

420 320 310

510 410 410

3.5

890 2900 2900 34000 7100 24

1100 3300 3300 25000 5200 66

3.6

330

350

5.8

2500

2100

5.2 3.0

7700 23000 130

5100 14000 150

4.0

290

310

4.5 4.5 5.7 4.6

oxygen sensitivity rod

2.0 15OC 130 140 1oooc 490b 560 1600 1400 3800' 24000b 5700 91 113 360 270 1800 1300 1400 1500b 730b 3200b

5.7 4.9 0.65 4.2 1.6 1.8 1.2 1.5 1.4 11.7

8.4

2.0 2.6 3.5 6.5

33 6.4 82 4.6

6.6

14

5.0

8600

4400

3.9 7.3

1100

780

8.3 41 3200 3800 16000

1.0

5.6 0.9

"Measured experimentally at AA = 3 and 7 nm or calculated from spectra (24) by using eq 6 (AA = 7 nm). AA = 3 nm responses were multiplied by (8/2.5)2 to make response independent of spectral bandwidth. Values are corrected for instrumental excitation characteristics. b"Dilute" solution (24) response; spectra from "concentrated" solutions gave 10% lower response than "dilute" solution spectra. Values for the "dilute" and "concentrated" solutions were the same. dRatio of the oxygen sensitivity to that of anthracene (measured).

thus, SF peak position is related to the square of n. The close correspondence of ERSF peaks with the excitation peaks, especially the 0-0 peak, is very useful for qualitative analysis b y ERSF. Group B compounds have excitation and emission spectra of approximately the same width (composed of the entire

emission and lowest energy excitation manifolds). If c = d , then e q 3 and 4 combine-to give

(1/2)(Amax

+ A'm*x)

= (a + b ) / 2

(6)

and t h e wavelength axis m u s t be plotted as the average of the excitation and emission wavelengths for SF peak positions to

ANALYTICAL CHEMISTRY, VOL. 59, NO. 17, SEPTEMBER 1, 1987

260

300

280

300

320

2183

340

Nnml Figure 1. Excitation, emission, and ERSF spectra of aromatic compounds: (A) phenol, (B) 2-methylnaphthalene,and (C) anthracene. AX = 3 nm intensity was multiplied by 2. ERSF spectra are plotted as a function of excitation wavelength. A

I’

280

h,,(nm)

340

h(nm) Flguro 2. ERSF spectra of p-terphenyi plotted as a function of excitation wavelength (A) or average of excitation and emission wavelengths (e).

be independent of AX. This is shown in Figure 2 for p-terphenyl. When ERSF spectra are plotted as a function of excitation wavelength, the ERSF peaks shifts to shorter wavelengths (Figure 2A). When spectra are plotted as in eq 6 (Figure 2B), the ERSF peak positions have less dependence on AX. Group B compounds often have a low quantum yield or a low 0-0 transition probability, which makes their SF intensity low, especially a t small values of AX.

Some aromatic compounds have greater vibrational structure in their emission spectra than in their excitation spectra (e.g., biphenyl, terphenyl(24)) and show peak alignment when the instrumental emission band-pass is narrow and the excitation band-pass is broad (“case 3“ in ref 18). The spectral properties of these compounds were not investigated since oil solutions under these experimental conditions show poorly resolved peaks. When the excitation and emission spectra in eq 1 are expressed in terms of energy, then Au is a function of X when Ah is held constant. However, the 0-0 excitation peak width is approximately 10 nm and Au changes less than 8% over this range (for wavelengths longer than 260 nm). ERSF 0-0 peak positions were calculated for the first five compounds listed in Table I by using eq 5 and experimental excitation and emission spectra. The values calculated from a, b, and n from spectra plotted as functions of either wavelength or energy agreed with the observed ERSF 0-0 peak positions to within f0.8 nm. This accuracy is within the accuracy of estimating peak positions and peak widths from excitation and emission spectra and supports the assumption used to derive eq 5, i.e., that peaks can be considered Gaussian on wavelength (or energy) axes to a good approximation. It should also be apparent that the choice between keeping AX or Au constant is not very important when a wide emission band-pass is used. These results suggest that the intensity of the lowest energy ERSF peak should be useful for quantitative analysis of compounds having a resolved 0-0 absorption peak. The lowest energy ERSF peak response factor, R(AX), can be calculated from

R(AX) = ZC(&)~(&

+ AX)@

(7)

where 1 is the path length, e(&,) is the molar absorptivity of the absorption 0-0 peak, I is the fluorescence intensity from a fluorescence spectrum of unit area, Am is the wavelength maximum of the 0-0 transition, and is the quantum yield. Equation 7 is derived from the general SF intensity equation (8, 17) for group A compounds under ERSF scanning conditions. SF and ERSF Spectra of Aromatic Compounds. The ERSF peaks of pure compounds occur at nearly the same wavelengths in the different AX spectra, as shown in Figure

2184

ANALYTICAL CHEMISTRY, VOL. 59, NO. 17, SEPTEMBER 1, 1987

1 for phenol, 2-methylnaphthalene, and anthracene. Compound excitation and fluorescence spectra are shown below the ERSF spectra for reference. When the ERSF wavelength axis is excitation wavelength as in Figure 1, the position of the lowest energy peak (the ERSF 0-0 peak) in these compounds occurs very near the wavelength of the lowest energy excitation peak. At Ah = 3 nm, only one major ERSF peak is observed for each compound. As AA increases, shorter wavelength peaks are observed, and these also align with the excitation peaks. ERSF peak maxima align in this manner for any value of Ah chosen and not just for the 4 A values shown. ERSF 0-0 peak width is the same as the excitation 0-0 peak width in these compounds. Table I compares the SF properties of several aromatic compounds under ERSF and ordinary SF conditions (Le., excitation\emission spectral bandwidths of l\8 and 3\3 nm, respectively) without changing any other instrumental settings. The peak intensities under these conditions are nearly the same. Absorption measurements showed that >90% of the maximum 0-0 absorption peak intensity was resolved in these compounds when a 1-nm spectral bandwidth was used. Table I shows that the S F 0-0 peak positions for the l \ 8 nm results occur within one-half of the wavelength range of the 3\3 peak positions. Anthracene has the widest range of peak positions because its emission has strong vibrational structure (Figure IC). The wavelengths of the SF 0-0 peaks (Table I) and the absorption 0-0 peaks (Table 11) coincide within the range of the S F 0-0 values. The SF 0-0 peak intensities are 8-40% larger for the l\S nm results than for the 3\3 nm results when corrections are made for spectral band-pass difference. This increase in SF intensity results largely from the greater excitation resolution of the 0-0 transition, as shown by comparing columns 5 and 6 of Table I. The 1\8 nm spectra also have 10-40% greater SF 0-0 intensity relative to the S F intensity of the next highest peak in the spectrum, as shown in column 7 . This gives the 1\8 SF spectra less interference from peaks other than the SF 0-0 peak in the spectra. ERSF Peak Widths and Response Factors. ERSF peak widths a t half height for AA = 3 nm and response factors for a series of aromatic compounds are shown in Table 11. The observed peak widths are similar to the SF peak widths observed a t low temperature (12). Relative R(AA) values were measured at AA = 3 and 7 nm and calculated for 4A = 7 nm from spectra (24) by using eq 7 . ERSF response factors span more than 4 orders of magnitude. Alkyl-substituted compounds have a greater response than their parent aromatic compounds, and this is due to their greater 0-0 absorptivity and greater fluorescence efficiency. Alkyl-substituted aromatics are of interest because of their relationship to mutagenicity (25). The agreement between the calculated and observed response factors is good in view of the experimental difficulties with quantum yield measurement (I), self-absorption (24),spectral correction (21),and oxygen removal ( 5 ) . The detection limits for fluorene and anthracene were and 3 X M, respectively (Ah = 7 , scan rate = 8X 15 nm/min, time constant = 1.5 s). Figure 3 shows that the 0-0 molar absorptivity is the most important factor in determining the ERSF response of compounds from Table 11. The quantum yields of these compounds (24)are similar (0.2-0.8) and the fluorescence intensity values vary from 0.40 to 0.95 in relative intensity. The compounds with a low quantum yield (0.05-0.20) are shown as x values and have lower response factors (dashed line). In addition to the compounds in Table 11, condensed aromatic compounds usually exhibit a resolved 0-0 absorption in cy), clohexane, and this absorption can be very strong (e.g., & = 23 800 M-' cm-I for 10-azabenzo[a]pyrene; and 23 800 M-' cm-I for benzo~]fluoranthene)(26).

51

l o g A(OO1

Flgure 3. Relationship between ERSF 0-0 response at AA = 7 nm calculated from spectra ( 2 4 ) and 0-0 molar absorptivity: (e)compounds with 9 = 0.8-0.2 and ( X ) compounds with 9 = 0.15-0.05. The numbers refer to the compounds in Table 11.

ERSF Oxygen Sensitivity. The fluorescence of aromatic compounds can be quenched by dissolved oxygen as described by the Stern-Volmer equation (24). The relative oxygen sensitivity of compound x to that of anthracene, rO, is derived by taking the ratio of the Stern-Volmer equations

where I and Io are the ERSF intensities in the presence and absence of oxygen, respectively, k is the bimolecular rate constant for oxygen quenching, and 7 is the fluorescence lifetime. The value of k is similar for aromatic compounds, and oxygen sensitivity has been correlated with fluorescence lifetime (24). In this study, anthracene is doped into solutions and acts as an internal standard for measuring oxygen sensitivity. ERSF is convenient for determining the rO values of several compounds in the same solution because of the narrow peak widths observed, and Table I1 shows the ERSF determined values for several aromatic compounds. Although rO should be independent of oxygen concentration by eq 8, experiments did show some dependence on initial oxygen concentration. All Io values were determined by bubbling nitrogen through the solution until the fluorescence intensity did not increase further (4min). Oil ERSF Spectra and Peak Assignments. The ERSF spectra of 2 mg/L solutions of no. 2, no. 6, and NBS standard no. 6 oil for AA values of 3, 7 , 11, and 15 nm are shown in Figure 4. The no. 2 oil has six peaks labeled in Figure 4 a t excitation wavelengths of 275 (a), 280 (b), 305 (d), 318 (e), 325 (0,353 (h), and 380 (i) nm. Region e at 318 nm is not a peak in the ERSF spectra but is the wavelength of maximum oxygen sensitivity of the oil solutions. The no. 6 oils also have 275,280,305, and 318 nm peaks and less well resolved peaks at 327-330,356, and 398 nm. Resolution between E M F peaks is greatest for small values of AA. The fluorescence intensity of the 305-nm peak was linear with concentration from 2.0 mg/L to the detection limit of 0.02 mg/L (AA = 7 nm). The wavelength regions that exhibit uniform sensitivity to oxygen are shown with their rO values in Table 111. Figure 4 suggests that oil ERSF peaks are derived predominantly from compounds with resolved 0-0 transitions. This follows from the coincidence of the different A i spectra peaks when the x axis is plotted as the excitation wavelength. Peak alignment is not observed when the x axis is plotted as

ANALYTICAL CHEMISTRY, VOL. 59, NO. 17, SEPTEMBER 1, 1987

2185

Table 111. ERSF Oxygen Sensitivity in Oil and Oil Extracts compound type

region (peak)

ERSF wavelength range, nmo

alkylbenzene alkylbenzene 0,N-subst benzene alkylfluorene alkylnaphthalene mixture alkylcarbazole unidentified alkylphenanthrene

no. 2 oil

273-276 279-282 285-292 299-310 319-321 322-331 329-340 336-344 350-355

4.5 3.8

rO values no. 2 oil extract

no. 6 oil extract

4.9

4.0

3.2

S 1.2

1.0 1.3 7.3 Sb

1.4 6.8 Sb 2.7

1.2 5.0 Sb

2.9

2.7 5.4

"rO was constant within 6% over the reported wavelength range. *rO varied continuously from a maximum at region e to a minimum near 330 nm. S = slope to 0 sensitivity of a region. 0.N-ben

ben

W

ind f l u

H H

&

nap

&c

chr

H

ant

H

"'3

N0.2 OIL

molar absorptivity. The range of the 0-0 absorption peaks for a class of aromatic compounds in hydrocarbon solvents can be obtained from absorption data (24, 26-28). For instance, the alkyl derivatives of several aromatic compounds exhibit their 0absorption peaks within the ranges given at the top of Figure 4. Aromatic compounds with an electrondonating group substituted on the ring system (e.g., -OH, -OCH3, -phenyl) have spectra shifted 10-20 nm to longer wavelengths, shorter fluorescence lifetimes, and higher molar absorbances than the corresponding alkyl derivatives of the same aromatic compound (24). The compounds in Table I1 represent the types of aromatic compounds found in high concentrations in oils (29-31) and aqueous extracts of oils (22, 32-35).

N B S N0.6 OIL

3

340

420

h,(nrn)

Flgure 4. ERSF spectra of oil solutions. The ranges of 0-0 peak wavelengths for alkyl substituted aromatic compounds in nonpolar solvents are shown at the top: ben, benzene; i d , indole; flu, fluorene; nap, naphthalene;car, carbazole; phe, phenanthrene;chr, chtysene; ant, anthracene; 0,N-ben, oxygen- or nitrogen-substituted benzene compounds having a distinct 0-0 transition.

the emission wavelength or as the average of the excitation and the emission wavelengths. This peak alignment property is observed when the instrumental bandwidths are equal (3 nm) but is more apparent and peaks are better resolved under ERSF conditions. Substitutions on an aromatic compound by alkyl- or heteroatom-containing groups shift the position of the 0-0 absorption peak to longer wavelengths and usually increase the

Table I11 shows the assignments of the no. 2 oil ERSF peaks made on the basis of the model compound properties described above: peak positions, response factors, oxygen sensitivities, and spectra at different values of Ax. All four ERSF criteria must be consistent for a peak assignment. For instance, the peak assigned to fluorene shows a constant sensitivity to oxygen (rO = 1.3) over a range of 11 nm. The wavelength position of this peak is also consistent with an assignment of dibenzofuran, 9,10-dihydrophenanthrene,or p-terphenyl compounds. However, these compounds do not have the same ERSF profie as a function of Ax as the oil spectra, are present in lower concentrations, and have lower response factors (