Determination of Phenols in Water Using Raman ... - OSA Publishing

Determination of Phenols in Water Using Raman Spectroscopy N A N C Y A. M A R L E Y , C H A R L E S K. M A N N , a n d T H O M A S J . V I C K E R S Department of Chemistry, Florida State University, Tallahassee, Florida 32306

The potential of Raman spectroscopy for quantitative analysis of phenols in water solution has been investigated. A group of six compounds-phenol, o-chlorophenol, 2,4-dichlorophenol, 2,4,6-trichlorophenol, 2-chloro-5-methylphenol and 2-chloro-4-nitrophenoi--was studied with the use of the 514.5, 488.0, and 457.9 nanometer lines of an argon ion laser. Attention was given to the effects of source intensity, optical alignment, and background fluctuations on quantitative results. It was found that the use of an internal standard with each measurement made a significant improvement in the accuracy and precision of results. Two methods of quantitation, peak area measurement and cross-correlation, were used. Results were somewhat better for cross-correlation, presumably because of more effective exclusion of background interference. Limits of detection were calculated based upon the slope and the standard deviation of the intercept of the standard curve. These varied from the range of 100 ppm to 0.3 ppm, depending on the compound. The most important factor controlling sensitivity is occurrence of resonance enhancement. Index Headings: Raman spectroscopy; Phenols; Analysis; Computer methods.

INTRODUCTION Phenols are a class of compounds of industrial importance which constitute enough of an environmental hazard to be included in the List of Priority Pollutants, established by the Environmental Protection Agency. Their quantitative analysis is usually carried out by eolorimetric determination of the condensation product formed by reaction with 4-aminoantipyrine. ASTM Method D-1783-70 uses this reaction as the basis for a direct analysis that is recommended for samples that contain more than 100 ppb of phenols. For lower concentrations, the recommended method involves a preliminary steam distillation and extraction which concentrates the sample, extending the range to the low ppb level. 1 These methods ideally respond to the phenol functional group, providing a value which is proportional to the sum of the phenol groups present. In fact, the sensitivity is generally not equal for all phenol-containing compounds. When discrimination between different phenols is desired, analyses can be performed by gas chromatography, with a useful range that extends to the low ppm level. 2 A method based upon gas chromatography has been reported, which is suited for concentration of phenols from natural waters and can be used down to the ppb range2 Raman spectroscopy is the basis for a method which extends to the ppb range. It involves measurement of resonance-enhanced spectra of phenols and phenol derivatives. 4,~ The importance of this group of compounds makes desirable the development of alternative methods of Received22 August 1983;revision received 14 October 1983.

540

Volume 38, Number 4, 1984

analysis, which may for certain purposes be superior to those generally used. An analytical method based upon use of vibrational spectra is inherently very selective and would therefore be applicable to analyses of individual compounds. Water is an excellent solvent for Raman spectroscopy; accordingly, it is feasible to apply it directly to analyses of substances dissolved in water. The possibility exists, therefore, of designing a practical monitoring system which could be used in rivers, lakes, and marine waters, without the necessity for prior separations. In the work reported here, we have attempted to assess the usefulness of Raman spectroscopy as the basis for an automated phenol analysis. Six compounds have been studied. Optimum spectral conditions were identified and the inherent sensitivity of the measurements were evaluated. Any attempt to use Raman spectroscopy for analytical purposes must deal with the problem of fluorescence, which will invariably constitute the major spectral interference; we have examined this problem. In addition, we have given considerable attention to the problems of achieving an adequate level of reproducibility in quantitative measurements. EXPERIMENTAL I n s t r u m e n t a t i o n . All Raman spectra were recorded with a Spex Industries Model 1403 Raman spectrometer under control of a Spex "Datamate" system. Excitation was provided by a Spectra-Physics Model 164 argon ion laser. Radiant power at the sample was approximately 300 mW for the 514.5-nm and 488.0-nm lines and 45 mW for the 457.9-nm line. Spectra were usually scanned with entrance, exit, and intermediate slits at l-ram widths (nominal half-intensity bandpass of 10 cm '), with data acquisition at 1 cm -1 intervals with 1-s integration times. Spectra used for calculations were ensemble averages of five scans. A cooled RCA C31034 photomultiplier and photon counting were used. Except where noted below, spectra were recorded for samples held in a 1-cm fluorescence-type cuvette mounted in the Spex illuminator sample chamber; both a double pass mirror and rear scatter collection mirror were used. For a few cases, spectra were recorded for samples held in a Spex Model 1476 split cell, spun by a Model 1475 difference/ratio generator. This system provided for nearly simultaneous data acquisition from two samples. Neither double pass mirror nor rear scatter collection mirror is used with this assembly. Spectra were transmitted from the Datamate, by means of an RS232 port, to a nearby Data General Mi-

~3-7o28/84/~8o4-054o~zoo/o

© 1984 Society for Applied Spectroscopy

APPLIED SPECTROSCOPY

TABLE I.

Internal standard fluctuations.

Compound measured 2-Chloro-4-nitrophenol Phenol o-Chlorophenol 2,4-Dichlorophenol 2-Chloro-5-methylphenol 2,4,6-Trichlorophenol

TABLE II.

Principal Raman lines of phenol and related compounds.

Water peak area 197,360 143,906 69,352 64,800 77,082 61,279

± 41% ± 16 ± 14 ± 8 ± 5 ± 7

cronova computer. Spectra were stored on 51A inch magnetic disc for subsequent processing on a Data General Nova 2 computer system. Solutions. Stock solutions containing 500 mg/L of phenol (Baker 2858) and o-chlorophenol (Aldrich 18,577-9) were prepared in distilled water. Stock solutions containing 2000 mg/L of 2,4-dichlorophenol (Aldrich 10,595-3), 2,4,6-trichlorophenol (Aldrich T5,5301), and 2-chloro-5-methylphenol (Aldrich 15,955-7), and 1000 mg/L of 2-chloro-4-nitrophenol (Aldrich C6,120-8) were prepared in 0.5% NaOH. Lower concentrations were prepared by appropriate dilutions with distilled water.

Compound Phenol

o-Chlorophenol

2,4-Dichlorophenol

2,4,6-Trichlorophenol

2-Chloro-5-methylphenol

2-Chloro-4-nitrophenol

RESULTS

Spectroscopic Reproducibility. In carrying out the measurements, we took care to adjust the power of the incident laser beam and the alignment of the optical system. In spite of this, entirely unacceptable fluctuat i o n s - f r o m day-to-day and from sample-to-sample-were observed; to obtain useful results, a spectroscopic standard, either internal or external, is necessary. Several approaches were examined. For the compounds of this study, the water peak at 1640 cm 1 serves as a suitable internal standard. It was necessary to measure a standard for each sample. This situation is illustrated in the data in Table I. Each of the entries represents the average obtained by the measuring of the area of the water peak in the solution of the indicated phenol. These data were obtained over a period of about eight months, in the order listed in the table. Each entry represents observations taken with the 514-nm laser line, usually on the same day; therefore, comparison between sets gives indication of long-term variation. The large differences between the water peaks shown for 2-chloro-4-nitrophenol and phenol and those for the others were caused by deliberate changes in operating procedure. The u n c e r t a i n t y ranges q u o t e d represent the standard deviations for an individual measurement in each set, and represent the size of sample-to-sample variations. It was apparent, the indicated corrections to the quantitative results having been applied, that observed water peak intensity variations did track variations in the intensities of the sample spectra and did provide an efficient basis for compensation. The design of the equipment used makes provision for two-channel operation. A divided sample cell can be rotated in the incident beam to generate spectra from both compartments; data acquisition is automatically synchronized with the movement of the cell. This permits the efficient use of an external standard and permits the user to compensate for short-term fluctuations in laser intensity. In this work, attempts to use this ar-

Peak location cm 1

Relative intensity

820 1005 1285 1600 840 1040 1135,1165 1310 1580 860 1105 1150 1290 1580 870 1140 1310 1570 740 1060 t305 1390 830 900 1040 1130 1310 1520

4 6 1 2 3 7 1 1 3 3 2 2 1 2 3 2 1 1 7 4 2 1 4 11 1 2 12 1

rangement produced results that are marginally poorer than those obtained by the procedure described above. As noted, it is necessary to record a standard with each spectrum. However, significant fluctuations during a single run (usually about ten minutes) are rare. Accordingly, we conclude that short-term variations are caused by optical alignment problems, which are effectively compensated by an internal standard. Moreover, there was a noticeable increase in noise and decrease in signal intensity when the rotating cell was used. After the operating conditions of the spectrometer were optimized, noise was measured over time periods comparable to those of sample measurements, but without scanning. It was found that the RMS noise varied approximately with the square root of signal intensity, as would be expected for Poisson statistics in photoncounting experiments. We conclude that the major noise components remaining, after internal standard corrections are applied, are those originating in the counting statistics. Spectroscopic Behavior of Phenols. The ultraviolet absorption spectra of all the compounds in this study ex-

TABLE Ill.

Analytical peaks for phenols.

Compound

Peak position

Peak intensity area/concentration

Phenol o-Chlorophenol 2-Chloro-5-methylphenol 2,4-Dichlorophenol 2,4,6-Trichlorophenol 2-Chloro-4-nitrophenol

989-1018 c m a 1017-1065 720-764 850-878 860-890 1225-1329

146 114 122 60 78 7096


541

TABLE IV. Summary of quantitative results.

Compound Phenol

o-Chlorophenol 2-Chloro-5-methylphenol 2,4-Dichlorophenol 2,4,6-Trichlorophenol 2-Chloro-4-nitrophenol

Correlation coefficient

Laser line nm

Concentration range mg/L

Area

Correln

514.5 488.0 457.9 514.5 514.5 514.5 514.5 514.5 488.0 457.9

50-500 50-500 100-500 100-500 400-2000 400-2000 400-2000 1-100 1-10 1-10

0.9962 0.9866 0.9780 0.9670 0.9876 0.9881 0.9889 0.9993 0.9960 0.9881

0.9961 0.9880 0.9899 0.9895 0.9970 0.9892 0.9944 0.9991 0.9994 0.9997

Detection limit Area

Correln

Spacial frequenciesa for correln

25 49 61 55 102 151 153 0.92 0.42 0.30

1-19 2-14 3-19 2-19 2-59 1-5 6-15 5-19 3-19 2-39

mg/L 24 52 103 99 208 158 215 2.55 1.06 1.84

Transforms in 128-point window, except 256 for 2-chloro-4-nitrophenol.

hibit the typical phenolic absorption bands with maxima near 200 and 275 nmY For the 2-chloro-4-nitrophenol, these bands are shifted to larger wavelengths, and another band, due to the nitro chromophore, appears with a maximum near 400 nm. This band extends far enough into the visible that some resonance enchancement of the Raman spectrum is evident with the argon ion laser lines. The Raman spectra of phenol and related compounds have been reported elsewhereY 1~ Survey spectra for all of the compounds involved in this study were measured for the fundamental region; a summary of information about the principal lines has been collected in Table II. The intensities quoted have been normalized to unity for the weakest significant peak of each compound. The peaks selected for use are described in greater detail in Table III. Measurements were made at 514 nm. Peak positions and sensitivities are given. The sensitivities were normalized for light intensity fluctuations, with the water peak used as an internal standard. Vibrational modes responsible for the peaks in Table III have been assigned 7-15 as follows: phenol and o-chlorophenol, ring breathing; di- and trichlorophenol, CH bending; 2-chloro-4-nitrophenol, nitro stretch. The peak for 2-chloro-5methylphenol is most likely due to a CH bending mode, but could not be definitely assigned. Data Evaluation. Quantitative results are presented in Table IV for two data-handling procedures, peak area measurement and cross-correlation. Both were applied to data after smoothing by the application of 5- or 9-point Savitzky-Golay filters, and after correction for internal standard fluctuations. Areas were measured by selection of the peak ranges, given in Table III, by examination of high signal-to-noise spectra. In each case a linear background was assumed, and the area above this calculated. Cross-correlation was carried out after the same treatment, including subtraction of the linear background. In addition, an optimum spacial frequency range was selected for each set. This was carried out by repeated performance of the cross-correlation operation on a complete set of spectra for a given compound, with spacial frequency components deleted from both the high and low ends of the set. This amounts to a convolution of the data set with a rectangular function, the dimen542


sions of which are selected after the whole concentration range has been examined. The criterion for retention of components was that of minimum limit of detection. For evaluation of results, the data were plotted as area or correlation number vs. concentration. The Pearson correlation coefficient, given in Table IV, provides a measure of linearity. 16An estimate of detection limit was obtained by calculation of the 95% confidence limit for the difference in means for two sets which have nearly the same concentrations. The standard deviation of the intercept of the curve of area or correlation number vs. concentration is taken as an approximation of the uncertainty experienced near the limit of detection, and it is assumed that this value will not change for small variations in concentration. Accordingly, the detection limit shown in Table IV represents an estimate of the smallest difference in concentration that can be distinguished with 95 percent confidence, assuming that one measurement is to be made. Spectral and Chemical Interference. All of the spectra obtained in this study were heavily contaminated by fluorescence originating both from the phenols and from small concentrations of contaminants. The intensity of fluorescence varied considerably from sample to sample, generally amounting to five to ten times the observed peak heights. The data processing procedures largely eliminated effects of fluorescence from either the area or cross-correlation results. Examination of the results in Table IV shows that the cross-correlation procedure generally gives somewhat better results. The convolution with a rectangular function provides more efficient exclusion of high frequency noise than does convolution with just the filter function, because it is fitted to the specific set. More importantly, exclusion of low frequencies greatly reduces the effects of background curvature introduced by fluorescence and the water background. We believe that this data treatment should be effective in dealing with samples of natural origin because the major contributions to interference are likely to be fluorescence by organic contaminants that will show reliably wider spectral bandwidths than those of the Raman peaks. Accordingly, it should always be possible to discriminate against them by the low frequency selection procedure described above. For example, an examination of the fluorescence spectra of terrestrial and

riverine humic acids, which presumably typify those which must be expected to be present in natural samples, shows relatively intense but featureless spectra very similar to the background of a phenol solution. The major deleterious effect is likely to be increased noise caused by the general increase in signal level. No attempt has been made in this work to evaluate interferences that will be caused by the Raman spectra of sample concomitants, other than the data presented in Table II. However, one of the advantages of using a method that is directly based upon a fundamental physical observation is that the effects of potential interferences can be evaluated by examinination of their spectra. Just as importantly, the high information-content of vibrational spectra provides an inherently favorable situation since the average width of peaks is quite small compared with the spectral frequency range in which they may occur. It will be noticed that for the six phenols examined in this work, spectral interference would be encountered only for solutions containing both dichlorophenol and trichlorophenol. Effect of Laser Wavelength. If a substance exhibits absorbance in the range of the incident radiation, there may be increased scattering caused by resonance enhancement. Considering that the shortest principal wavelength that is available with an argon ion laser is 458 nm, the only compound examined in this work which could be so affected is 2-chloro-4-nitrophenol, which has an absorbance maximum at 400 nm. In the absence of resonance enhancement, it would be expected that Raman signals will vary as the fourth power of incident frequency. To examine this, we have calculated the response from the principal peak of phenol, which would not show enhancement, to compare it with that from 2-chloro-4-nitrophenol, for the available wavelengths. After normalization for variations in incident radiant

power and concentration, the values obtained for phenol are 146, 123, and 100 area units for 514, 488, and 458 nm, respectively. The corrsponding values for 2-chloro4-nitrophenol are 7096, 14723, and 57172. This indicates that the increased analytical sensitivity for the chloronitrophenol in the results outlined in Table IV is caused by resonance enhancement. ACKNOWLEDGMENT Partial support of this work by The National Oceanic and Atmospheric Administration through Contract NA82AA-D-00012 is gratefully acknowledged.

1. American Society for Testing and Materials. Method D-1783-70, Philadelphia, Pa., 1970. 2. American Society for Testing and Materials, Method D-2580-68, Philadelphia, Pa., 1968. 3. C. D. Chriswell, R. C. Chang, and J. S. Fritz, Anal. Chem. 47, 1325 (1975). 4. R. J. Thibeau, L. Van Haverbeke, and C. W. Brown, Appl. Spectrosc. 32, 98 (1978). 5. L. Van Haverbeke and M. A. Herman, Anal. Chem. 51,932 (1979). 6. R. M. Silverstein, G. C. Bassler, and T. C. Morrill, Spectrometric Identification of Organic Compounds (Wiley, New York, 1974), 3rd. ed., p. 249. 7. P. K. Mallick, Ind. J. Phys. 48, 1089 (1974). 8. J. A. Fanuan and H. F. Shurvell, J. Ram. Spec. 9, 73 (1980). 9. J. H. S. Green, D. J. Harrison, and W. Kynaston, Spectrochim. Acta 27A, 2199 (1971). 10. J. C. Evans, Spectrochim. Acta 16, 1382 (1960). 11. K. Kumai and P. R. Corey, J. Chem. Phys. 63, 3697 (1975). 12. J. H. S. Green, Spectrochim. Acta 26A, 1503 (1970). 13. A. Stojilijkovic and D. H. Whiffen, Spectrochim. Acta 12, 47 (1958). 14. J. H. S. Green, D. J. Harrison, and W. Kynaston. Spectrochim. Acta 28A, 33 (1972). 15. F. R. Dollish, W. G. Fateley, and F. F. Bentley. Characteristic Raman Frequencies of Organic Compounds (Wiley, New York, 1974). 16. B. W. Brown and M. Hollander, Statistics (Wiley, New York, 1977), pp. 286-287.

Coherent Anti-Stokes Raman Spectroscopy of Polycyclic Aromatic Hydrocarbons D. R. V A N HARE,* L. A. CARREIRA, t L. B. ROGERS, and L. A Z A R R A G A University of Georgia, Department of Chemistry, Athens, Georgia 30602 (D.R.V., L.A.C., L.B.R.); and U.S. Environmental Protection Agency, College Station Road, Athens, Georgia 30605 (L.A.)

Coherent anti-Stokes Raman spectroscopy (CARS) was used to obtain Raman spectra of thirteen polycyclic aromatic hydrocarbons (PAHs) composed of between three and seven fused rings. The compounds were pumped in the resonance and preresonance regions to obtain the sensitivity necessary for examination at concentrations down to 1.5 × 10 5 M. A simple mixture of three PAHs was optically separated through the wavelength selectivity of the resonant enhancement process. Com-

parisons between the spectra of the mixture and the spectrum of each pure component showed essentially no differences when pumped at the same wavelength. Finally, lineshape analysis was performed on the CARS spectra to facilitate comparisons between classes of compounds and for quantitative purposes. Index Headings: CARS; PAHs; Lineshape analysis.

Received 9 December 1983. * Present address: Savannah River Laboratory, E. Io DuPont de Nemours and Co., Inc., Aiken, SC 29801. t Author to whom correspondence should be addressed.

INTRODUCTION


Coherent anti-Stokes Raman spectroscopy (CARS) is one of many recently developed nonlinear spectroscopic

0003-7028/84/3804-054352.00/0

(c) 1984Societyfor AppliedSpectroscopy


54,3