Anal. Chem. 1982, 5 4 , 1059-1064
L.
< - - _ _
I
300
I
400
Flgure 1. The siiver anomalous absorption peak: curve, (- - -) the c:urve fitted to the peak height
1
nm
(-)the experimental and position, (. the
curve fitted to the peak half-width.
-
e)
1059
sputtered atoms could then be involved in clusters because as the clusters are growing and slowing, they are more easily hit by other atoms and clusters. Of course, not all collisiions will result in silver atoms sticking to others and enlarging the cluster. There will also be some processes which will break up clusters. A good estimate of the proportion of sputtered atoms actually in clusters can be gained by comparing the numbeir of atoms per cubic centimeter calculated from the copper atomic absorption measurements with the number per cubic centimeter calculated from the agglomeration data. For the same sample in the same discharge, the ratio was found to be 2.2!5:1, or 30% in clusters. Thus, there is evidence to suggest tlhat when measurements are made, sizable fractions of the sputtered material will be present in a form rather than as single atoms. In absorption measurements made with most metals the absorption due to clusters is not as pronounced as it is for the silver anomalous absorption peak used in this study. For most metals ths effect of clusters in the light path is more alkin to “blocking” and a typical signal attenuation is 0.5% maximum. The analytical significance of clusters may be mlost pronounced in atomic fluorescence spectrometry.
ACKNOWLEDGMENT Previously, e.g., ref 6, experiments do not appear to have been The author thanks J. B. Willis for his careful reading of the performed in such an uncomplicated medium. manuscript and for his useful comments. A simple callculation produces an interesting result for silver-silver collisions. For (say) 30 X 1013silver atoms ~ m - ~ , LITERATURE CITED which is close to the total number of atoms sputtered from (1) Gough, D. S. Anal. Chem. 1976,48 1926-1931. (2) McDonald, D. C.Anal. Chem. 1977,4 9 , 1336-1339. the cathode (according to the copper atomic absorption de(3) Mle, G. Ann. f h y s . (Leiprlg) 1906,25, 377-445. termination), thle silver-silver mean free path would be about (4) Doyle, W. T.; Agarwal, A. J . Opt. SOC. Am. 1965, 55, 305-309. 1cm but the atoms may have undergone many free-paths and (5) Kawabata, A.; Kubo, R. J . f h y s . SOC.Jpn. 1966,21, 1765-1772. (6) Kreiblg, U.; v. Fragsteln, C. 2.f h y s . 1969,224, 307-323. collisions before drifting into the light path of the sputtering (7) Hannaford, P.; McDonald, D. C. J . f h y s . B 1978, 1 1 , 1177-1191. cell. Assuming tltlat there is a Maxwellian distribution of atom (8) Larkins, P. L. Anal. Chim. Acta 1981, 132, 119-126. (9) Smithard, M. A. SolidSfate Commun. 1973, 13, 153-156. velocities and therefore a distribution of free-path lengths, then only about, 20% of the silver atoms would not have collided with other silver atoms before being carried out of RECEIVED for review October 9, 1981. Accepted February 110, the light path b,y the flowing argon. More than 80% of the 1982.
Application of the Shpol’skii Effect to Quantitative Analysis of Monomsthylphenanthrene Isomers J. Rima, M. Lamotte,” and J. Joussot-Dubien Laboratoire de Chimie Physique A, ERA No. 167, Universit6 de Bordeaux
I, F 33405 Talence Cedex, France
The capablllty of highly resolved fluorescence emlsslon In frozen solutlons of alkanes (Shpol’skil matrlces) at 4.2 K for achlevlng quanitltatlve determinatlon of monomethylphenanthrene Isomers has been Investlgated. Inhomogeneity of the solutions, Interference effect between overlapplng fluorescence spectra, and Interactions with Inadequate Internal standard affected the fluorescence lntensltles. Nevertheless, the standard addition method, with a suitable lnternal standard inlnimlred substantlally the effect of these phenomena. Tests conducted on an artlflclal mixture of all flve monomethylphenanthrene Isomers In n-hexane lead to an accuracy better than 25 % for concentration determlnatlons. The method has been epplled to obtaln the dlstrlbutlon of these Isomers In a petroleum tractlon.
aromatic hydrocarbons (PAH). Among various fluorimetric methods, the detection of PAH in frozen crystalline solutions of appropriate n-alkanes, known as Shpol’skii matrices, has attracted much attention by analysts (1-3). The exceptionally well resolved fluorescence spectra (of aromatics in these matrices allow identification of individual PAH in complex media such as environmental samples. The method has been applied to the analysis of chromatographic fractions from air ( 4 , 5 ) ,automobile exhausts (6, 7), water (8, 9),crude oil (10,11,13,14),sediments (9,12), liquid fuels ( I s ) , etc. and was shown to be able to differentiate alkylated isomers (16, 17). The potential of the method in qualitative analysis of PAH has been discussed and tested in a few cases and is now unquestionable. However, its applications to quantitative analysis is still controversial (18) and has until now been limited to only a few examples (8, 15). The purpose of this paper is to investigate the capability of this technique for achieving quantitative determination of
Because of their high sensitivity, fluorimetric procedures have been widely investigated for the analysis of polynuclear
0003-2700/82/0354-1059$01.25/00 1982 American Chemlcai Society
1060
ANALYTICAL CHEMISTRY, VOL. 54, NO. 7, JUNE 1982
monomethylphenanthrene isomers. Additionally we report on some of the practical problems and difficulties encountered in applying the Shpol'skii effect to this analysis. The literature devoted to the Shpol'skii effect during the last 20 years shows that Shpol'skii matrices are complex solid solutions. The solute-solvent compatibility, the solute concentration, the freezing rate, the presence of other aromatics, and the formation of aggregates or microcrystallites are found to alter greatly the fluorescence intensity of the PAH of interest (19,20). At concentrations where crystallites are formed, excimer fluorescence or sensitization phenomena may be also observed (18, 21, 22). The drawbacks encountered with Shpol'skii matrices readily account for the suspicion of analysts about their potential for quantitative analysis by fluorescence measurements. In view of the difficulties inherent in these matrices, it is widely accepted that best experimental conditions require reproducible fast cooling, low concentration of working solutions to limit microcrystallite formation and subsequent phenomena as fluorescence quenching or sensitization, and the use of an internal standard to correct for irreproducibility of fluorescence intensity from one sample to another (15). Having provided such optimization of the experimental conditions, quantitative measurements have been attempted either by reference to a calibration curve (23)or by the method of standard addition ( 8 , 2 3 ) . In the cases where linearity of the fluorescence intensity is observed for a limited range of concentration, the latter method has been found to be the most convenient procedure (24). It has been employed successfully only for a limited number of quantitative analyses dealing mainly with the determination of benzo[a]pyrene (8), dimethylbenzo[a]anthraceneor 7,12-dimethylbenzanthracene (25). The fluorescence spectra of these PAH are easily distinguishable even at 77 K and negligible mutuai perturbation was found to affect the quantitative results even in a complex mixture of PAH (24,26). A different situation arises when the analysis concerns closely related compounds whose spectra are strongly overlapping. In that case highly resolved spectra that appear only at temperatures below 15 K are necessary in order to distinguish the spectra in the sample mixture. Recently we showed that methylphenanthrene isomers that have very similar mass spectra give rise to distinguishable emission spectra when included in n-hexane cooled to 15 K (27) or to 4.2 K. Since these compounds are being considered as geochemical markers (16),it becomes of interest to extend the qualitative analysis of these isomers to quantitative determination. Application of a simple and fast method allowing a quantitative determination of these compounds and constituting an alternative to the GS/MS method would be attractive. Phenanthrene and its derivatives present a particular challenge in fluorimetric analysis using Shpol'skii effect. Indeed previous work has shown that phenanthrene belongs to a class of aromatic compounds whose molecules are said not to fit properly in the lattice of any normal alkane. This conclusion comes from spectroscopic data (28-31) and from the fact that contrary to large molecules such as coronene, pyrene, perylene, etc. (32),no quasi-linear spectra are observed when an n-alkane solution of phenanthrene is cooled slowly to yield well-developed single crystals. Fast frozen solid solutions of phenanthrene in n-alkanes are supersaturated systems far from equilibrium which will manifest complex behavior characterized by inhomogeneity of the solutions and mutual interactions of the solute molecules due either to local changes in the crystal structure of the matrices (33) or to energy transfer in aggregates even at low concentration (22, 31, 33-36).
Solutions of pure individual compounds and synthetic samples were first investigated in order to estimate the precision attainable by fluorimetric analysis a t 4.2 K. A quantitative analysis of the five methylphenanthrene isomers in a petroleum fraction extracted by high-performance liquid chromatography (HPLC) is then presented and discussed.
EXPERIMENTAL SECTION Fluorescence spectra at 4.2 K have been obtained with a home-made fluorimeter built from commercial components. Excitation was provided by a xenon lamp (Osram 450 W) whose light was dispersed through a Jobin Yvon HRS monochromator equipped with a 1200 lines/mm grating. For most measurements, bandwidths for excitation were set to about 3 nm. The fluorescence emissions were observed at 90° through the same type of monochromator as for excitation. The slit width for analysis of the fluorescence intensity was set to give a resolution in the order of 0.1 to 0.06 nm which corresponds approximately to the natural bandwidth of the fluorescence bands. An EM1 9789 QB photomultiplier was used as detector and the spectra were recorded with a Servotrace recorder at a rate of 2.5 nm/min. The solution to be studied was contained in a 4 mm 0.d. 30 mm long fused silica tube attached at its top end to a stainless steel rodlike holder which fits inside a liquid helium cryostat (Meric) thus allowing rapid change of samples. Best reproducibility of the cooling rate was obtained by first immersing this assembly (sampletube plus holder) into a Dewar containing liquid nitrogen. The sample holder was introduced quickly inside the helium cryostat after a complete freezing of the solution to 77 K. The n-hexane and n-heptane used as solvent (Fluka spectroscopic grade) were dried and kept on molecular sieves (5-10 A). They were verified to be free of any undesirable fluorescence. 2-Methylphenanthrene (2-MeP)was of commercial origin (K and K) while the 9-methylphenanthrene (9-MeP) (37) and 1methylphenanthrene (1-MeP), 3-methylphenanthrene (3-MeP), and the 4-methylphenanthrene (4-MeP) (38)were synthesized according to the references cited. Ultimate purity of the authentic samples was tested by HPLC and gas chromatographic methods. RESULTS AND DISCUSSION (A) Choice of the Solvent. The best quasi-linear spectra for phenanthrene and its monomethyl isomers a t 4.2 K are obtained in n-hexane (nC6)and in n-heptane (nC,). However, because in n-hexane the multiplet structure is simpler and the phonon wings are weaker than in n-heptane, n-hexane appears to be the best solvent. Spectra of the five methylphenanthrene isomers in nC6 in the vicinity of the 0-0 band region are shown in Figure 1. For phenanthrene as well as for most monomethylphenanthrenes, the fluorescence intensity is localized in the (0-0) transition, which in each case exhibits a characteristic multiplet structure allowing an easy identification of all compounds even when they are mixed (Figure 2). The only exception is 2-methylphenanthrene (2-MeP) whose fluorescence intensity is distributed among many vibronic bands, some of them having an intensity larger than the (0-0)band. This feature makes the detection of this compound difficult at a concentration lower than 3 X M. The detection limit for the other isomers is less than 5 X lo-' M (-0.01 pg/g in nC6). (B) Reproducibility and Linearity Range of Fluorescence Intensity Measurements. Experiments on the reproducibility of intensity measurements were carried out with 9-methylphenanthrene in nC6 and in nC7 (C = 10%M) using anthracene as internal standard (C, = lo-' M). In all cases, the intensities were given by the peak heights. Anthracene was chosen first because of its spectroscopic properties: firstly, its fluorescence emission can be excited in a range (340 nm) where phenanthrene has a very small extinction coefficient (e N 300) and, secondly, its spectrum is well separated from the phenanthrene fluorescence spectrum. Moreover its rather good fluorescence quantum effi-
ANALYTICAL CHEMISTRY, VOL. 54, NO. 7, JUNE 1982 A.SYN'HET
3 55
3 SO
?bC
I
36s
'
170
1061
C MlXTJRE
7
2MeP nC6 4 2 r( AEX=297.0 nrn
B-OIL EXTRACT k E x 2 9 7 nm
I 2 MeP 3M.P nC6 4 . 2 K + x ' 2 9 9 5 nm
Figure 2. Part of the fluorescence spectra (0,Orange transition) obtained by selective excitations of monomethylphenanthrene isomers in n-hexane at 4.2 K: (A) from a synthetic mixture of the five isomers, C = 5 X IO-' M each; (B) from a petroleum fraction (see text). +os*
Figure 3. Dispersion of relative fluorescence intensity measurements done on three multiplet components of 9-MeP in nC, and in nC, (a, Ib, c lines in Figure 1). 1,' is th_e normalized intensity obtained from the experiment number N , and I,' = ~ , N I , * / is N the value of the average. C = IO-' M.
Figure 1. Fluorescence spectra of monomethylphenanthrene isomers in n-hexane at 4.2 IK and C = 5 X IO-, M, a, y for 1-MeP and a, b, c for 9-MeP r e f w to the multiplet components used for testing intensity measurements.
ciency (@PFE 0.36 (39)) allows the use of very low concentrations (C, = M). For each multiplet component (Figure l),the ratio of the fluorescence intensity over the intensity of the standard (anthracene: A, = 340 nm, Aobd = 387 nm) has been calculated for different positions of the sample tubes and for several melting-freezing cycles. The distribution of the experimental results are schematically represented in Figure 3. With respect to the mean value, the range of deviations is about *25% in nC7 and less than f15% in nC6 confirming than nCe is the most suitable solvent. The reproducibility appears to be different frorn one
multiplet component to another. This effect may be related to the sensitivity of multiplet intensity to the cooling rate. Kt should be noted that the poor reproducibility is due mainly to variations in the physical structure of the frozen solution, rather than to instrumental instability. The snowlike polycrystalline samples are made of crystal grains of different sizes that constitute an inhomogeneous diffusing medium which can account for the fact that the Beer-Lambert law cannot apply strictly (40). Furthermore R significantly higher fluorescence intensity is observed inside the sample than a t the periphery. This anomaly was found by introducing a very thin Teflon sheet inside the sample tube. After freezing, the solid solution was broken and the fluorescence intensities emitted from the outer surface and from the inner surface in contact with the Teflon sheet were compared. Because of the low concentrations used in these experiments (C I lo4 M) the effect can be traced to self-absorption phenomena only to a very minor extent. It
1062
ANALYTICAL CHEMISTRY, VOL. 54, NO. 7, JUNE 1982
1
9MeP
-7
-6
l0gC
5
-4
1
-a-
,A
301nm
bX301nm
Change in the multiplet intensity distribution of 1-MeP fluorescence (0,O transition range) induced by the presence of 9-MeP: (left) spectra of pure solutlons; (right) spectrum obtained from mixture of the two compounds in the same conditions of concentration and excitation. Figure 5.
(i) on solutions containing only 9-MeP, 1-MeP, or phenanthrene, (ii) on solutions containing phenanthrene in the presence respectively of 9-MeP and 1-MeP, and (iii) on solutions of 9-MeP in the presence of l-MeP. Extrapolation was performed from the regression curve IF* = AC E representing formally the function
+
-bFigure 4. (a) Concentration dependence of the fluorescence intensity of 9-MeP in nC, A (, = 348.2 nm; a line in Flgure 1). (b) Regression
curve obtalned from the standard addition method for concentration determination of 9-MeP in the presence of 1-MeP. The fluorescence Intensities are corrected by using anthracene as the internal standard. Solvent was n -hexane. shows evidence of a rather nonuniform spatial distribution of the solute molecules and shows that during the freezing process the guest molecules migrate toward the center of the tube, causing a concentration gradient. This gradient is expected to depend sharply on parameters affecting the rate of freezing and particularly on the thickness of the tube walls which appears to be an important factor. The use of an internal standard can minimize this effect if it has about the same distribution inside the sample as the solute. However, the distribution is also expected to depend on the solute due to different solute-matrix interactions so that this phenomenon can also contribute to the dispersion of the results. For phenanthrene derivatives (Figure 4a), the linear dependence of the fluorescence intensity vs. the concentration occurs only for a reduced range, namely, between M or lower and 3 X lo4 M. The upper limit may be due to increasing perturbation related to the complexity of phenanthrene solutions in frozen alkanes as the concentration rises. To a degree difficult to estimate it is also due to the reabsorption effect in the (0-0)transition, which obviously also limits the linearity range of others aromatics (24). Observation either of a fluorescence vibronic line or of a phosphorescence line when possible must overcome the problem of the reabsorption. However for phenanthrene derivatives the vibronic structure as well as the phosphorescence emission have substantially lower intensity than the (0-0)fluorescence line so that these possibilities have not been considered. (C) Quantitative Analysis on Test Solutions. (1) Anthracene as Internal Standard. The method has been tested
where IF* is the normalized fluorescence intensity equal to the ratio of the solute fluorescence signal over the internal standard fluorescencesignal. C, and IFo* are respectively the solute concentration to be estimated and the normalized fluorescence intensity of the starting solution. C is the known concentration added. The concentration Co is determined by the negative of the intercept of the curve with the abscissa axis (8). The plotted values for IF* were the average from several measurements. For cases (i) and (ii) the errors estimated from several experiments were found to be about 25% of the real concentrations. This value is significantly larger than the errors we may expect considering the inhomogeneity of the sample and the good correlation factor, which was larger than 0.998. A similar error is obtained whatever multiplet component is chosen for the analysis. In case (iii) it is evident (Figure 4b) that an important effect alters the fluorescence intensity of 9-MeP in the presence of 1-MeP so that it is impossible to make any significant extrapolation of the curve from any multiplet component. The same difficulty arises with component p (see Figure 5 ) of 1-MeP. For the other components the error is more than 25%. These results suggest that beside the source of errors exposed above, at least two other phenomena introduce further errors. The first one is traced to weak interactions between anthracene and phenanthrene molecules, giving rise to a partial energy transfer from phenanthrene to anthracene. In practice, addition of phenanthrene to an anthracene solution (C M) was found to increase substantially the quasi-linear fluorescence intensity of anthracene even at concentrations of phenanthrene (C I M) for which crystallites are not observed. Under our experimental conditions, anthracene fluorescence is excited at a wavelength (340 nm) which in principle is very weakly absorbed by phenanthrene molecules responsible for the quasi-linear spectra. This fact added to the low concentration used makes radiative transfer very improbable.
-
ANALYTICAL CHEMISTRY, VOL. 54, NO. 7, JUNE 1982
11363
-
-.
Table I. Experimental Data and Results for Concentration Determination from Artificial Mixtures of All Five Monometh:yl Isomers of Phenanthrene (C, , Real, and C, , Measured, Concentrations, Respectively) 10-7co, 1 0 - 7 c m , a hexc,
1-Mep
2-MeP 3-MeP 4-MeP 9-MeP
nm
3 01 297 299.5 299.2 298
hbsd,
nm
34 8 358.5 350 349.5 348.5
M 3.0 3.0 3.0 2.0 2.25
M 3.53 3.50 3.57 2.30 2.80
(c, - coy c,, t 17.6
t16.6 t19 -1-15 + 24
Po
b
0.226 0.226 0.226 0.150 0.170
a Values obtained from the standard addition method with a theoretical precision of about 6%. tivelv. the real and the determined ratios of concentrations over the total absolute concentration.
However the existence of associations of phenanthrene molecules a t concentration as low as lo4 M (36)with diffuse absorption spectra may be responsible for exchange (and/or coulombic) energy transfer to some associated anthracene molecules. This5 phenomenon, which has also been observed with methylphenanthrene isomers, has been studied in more detail in ref 36 but has not yet been fully interpreted. The second cause of error is manifested in case (iii) where, in addition to the above phenomena, an interference effect between 9-MeP and 1-MeP must be considered. In that case, strong overlapping of certain of the multiplet components occurs (Figure 5). The most striking perturbation is clearly observed in the multiplet structure of the (0,O) fluorescence band of 1-MeP. Despite the fact that selective excitation of 1-MeP fluorescence can be achieved in the presence of 9-MeP, the component /3 (Figure 5) is substantially enhanced compared to its intensity for pure 1-MeP solution excited under the same condiitions. In the case of selective excitation of 9-MeP, no relative increase (or decrease) of one of its multiplet components is observed but in view of the poor results obtained for quantitative analysis of this compound in the presence of 1-MeP (Figure 4b), the overall intensity fluorescence of 9-MeP is obviously also affected. In view of these results, it is clear that anthracene appears as an unsuitable internal standard for fluorescence intensity calibration of phenanthrene derivatives. ( 2 ) Acenaphthene as Internal Standard. Some of the difficulties encountered in these experiments point to the importance in the choice for the internal standard and of possible interferences (or interactions) between solute molecules. Identical quantitative tests have been carried out, with acenaphthene as internal standard instead of anthracene. In a spectroscopic point of view, acenaphthene appears to be less convenient than anthracene because its absorption bands as well as part of its fluorescence spectra overlap with the absorption spectra of phenanthrene. However considering the low concentrations used (
