Structure elucidation from the hydroxyl stretching region of vapor

Jan 27, 1981 - The hydroxyl stretching regionof over 400 vapor-phase in- frared spectra Is ... Fourier transform spectrometry (FTIR) has pushed detect...
1 downloads 0 Views 355KB Size
1460

Anal. Chem. 1981, 53, 1460-1462

(6) Peter, L. M.; SefaJa,J. Surf. Sci., in press. (7) Fleischmann, M.; Lller, M. Trans. Faraday Soc. 1958, 54, 1370. (8) Bewick, A.; Robinson, J. Surf. Sci. 1976, 55, 349, Bewick, A,; Gale, R. J., paper in preparation. (9) Born, M.; Wolf, E. ”Principles of Optics”, 5th ed.; Pergamon Press: New York, 1975; Chapter 13.

(10) Heavens, 0.S. “Optical Properties of Thin Solid Films”; Academlc Press: New York, 1955; Chapter 5. (11) Schopper, H. Z . Phys. 1952, 131, 215.

RECEIVED for review January 27,1981. Accepted May 5,1981.

Structure Elucidation from the Hydroxyl Stretching Region of Vapor-Phase Infrared Spectra Mlchael F. Delaney” and F. Vincent Warren, Jr. Department of Chem;stty, Boston University, Boston, Massachusetts 022 15

The hydroxyl stretching region of over 400 vapor-phase infrared spectra Is examined. Correlations between peak wavelength and functionality are observed for primary, secondary, and tertiary alcohols, phenols, carboxylic acids, and oximes. Peak shifts due to steric crowding and Intramolecular hydrogen bonding are seen. The use of the tabulated results for ldentlflcatlon of compounds separated by gas chromatography Is discussed.

The use of vapor-phase infrared spectrometry (VPIR) directly interfaced to gas chromatography (GC) continues to develop as a viable alternative or complement to mass spectrometry (MS) for identifying the separated components. VPIR can facilitate isomeric assignments which are difficult or impossible by MS, and the instrumentation based on Fourier transform spectrometry (FTIR) has pushed detection limits down to nanogram levels ( I ) . Applications of GC-FTIR have recently been reviewed (2). The interpretation of IR spectra can be approached in several ways, depending upon the needs and time constraints of the laboratory being served, the available instrumentation, the skill and experience of the chemist, and the availability of a large body of reliable and computer-accessible spectra. The capabilities of computerized spectral interpretation methods are rapidly progressing. Automated approaches range from systems employing artificial intelligence techniques, designed to mimic a chemist’s strategy (3), to the more abstract and mathematically based pattern recognition (4) and library searching (5) methods. Manual and artificial intelligence based spectra interpretation both employ some form of a “correlation table” which relates the wavelength of spectral peaks to specific structural fragments in the simple molecule. The interpretation consists of identifying the stuctural fragments present and combining these together to yield the correct structure. Recently a theory has been proposed (6-8) which seeks to firmly define the function and structure of a correlation table using information theoretic concepts to optimize the division of the wavelength axis into functional group categories. This theory is expected to formalize and improve the interpretation systems, both under development (3) and commercially available (9),which identify possible structure units using a correlation table. While structure elucidation by IR is routine in many laboratories and extensive correlations between structure and IR adsorption wavelength have been published (IO), there are significant differences in the vapor phase, primarily due to 0003-2700/81/0353-1460$01.25/0

the absence of intermolecular interactions. As an example, Figure 1 compares a portion of the VPIR spectra of 2-isopropylphenol with the same region of the liquid-phase IR spectrum. The hydroxyl stretching band is seen to be quite broad in the condensed phase due to extensive intermolecular hydrogen bonding, while the gas-phase 0-H band is considerably sharper. Published correlations for vapor-phase spectra (11) are based on only few examples due to the previous lack of suitable instrumentation. The focus of our research (12-15) has been to facilitate the use of VPIR to identify components separated by GC by providing a reliable computerized interpretation system. Reported herein are our observations of structural correlations for various hydroxyl group containing compounds drawn from a large commercial library of VPIR spectra measured on a FTIR instrument. Using a commercially available, high-quality spectral library with extensive “biographical” information provided for each spectrum and with computerized data handling, we were able to rapidly compile functional group categories and to study these spectral correlations extensively for a large number of spectra.

EXPERIMENTAL SECTION The spectral library available was the Sadtler Research Laboratories VPIR collection (Sadtler Research Laboratories, Inc., Philadelphia, PA). Five thousand spectra were obtained on magnetic tape. Each spectrum was measured from 4000 to 450 cm-I at a sampling rate of 2 cm-’ and a resolution of 4 cm-I using a Digilab FTS-14 spectrometer and a CIRA GC (Sadtler Research Laboratories, Inc., Philadelphia, PA). The resulting spectrum contains 1842 data points which are background corrected with intensities digitized in milliabsorbance units. For compounds numbered 2001-4000, the information record, which contains the compound number, molecular formula, Chemical Abstracts Service (CAS) name and registry number, etc., were text searched by computer to find hydroxyl group containing compounds. For each compound the Wiswesser line notation (WLN) (16)was examined by computer to place spectra into the categories: primary, secondary, and tertiary alcohol, phenolic, oxime, and carboxylic acid. The members of each category were examined by using the CAS name to verify correct category assignments. In this manner 416 hydroxyl-containing compounds were found. For each compound, the spectral zone from 3400 to 3800 cm-l, encompassingthe hydroxyl group stretching region, was examined. Each spectrum was studied by using several programs to prepare tabulations of spectral information. The location of peak maximum positions using maximum absorbance and a peak picking algorithm (15) was examined, as were average spectra and a histogram of peak positions vs. number of spectra, for each functional group class. A visual assessment of a large fraction 0 1981 American Chemlcal Society

ANALYTICAL CHEMISTRY, VOL. 53, NO. 9, AUGUST 1981

1461

Table I. Observed Positfon of Hydroxyl Stretching Bands for Several Functional Groups this work Welti(l1) intramol H no. of steric crowd. regular range, bond limit, no. of regular range, functional group spectra limit, cm-' cm- ' cmspectra cm-' 3429 5 -3670 91 a 3616-3666 primary alcohol 3639 a 3664-3649 benzyl alcohol 3529 9 -3660 84 3668 3658-3645 secondary alcohol 3545 2 -3640 24 a 3644-3640 tertiary alcohol 3402 16 3655-3650 101 3610 3651-3645 phenolic b 6 b 3641-3640 oxime carboxylic acid a 14 3515-3570 68 a 3580-3568 saturated 3512 11 3585-3580 42 a 3586-3581 apunsaturated ~

a

Minimal effect observed.

Insufficient data.

Includes aromatics.

IWICROMETERS

4000 !

-

~

O

-5

O

CM-

Flgure 1. Liquid- and gas-phase IR spectra for 2-isopropylphenol: (A) vapor phase, (B) neat liquid.

A HO-"Y

0, .N, H'

lou ALCOHOL

H

-7

2 O b--

3 O CI PHENOL

+-

OXIME

H

-8

CARBOXYLIC ACID

3ioo

3650

c . (

3600

WAVIENU MBERS

Flgure 2. Correlation of hydroxyl functional group type with peak maximum location.

of the spectra was facilitated by display on a Tektronix graphics terminal. A cursor was useld to cross check computer derived vs. manual peak location.

RESULTS .AND DISCUSSION The major correlations 'between peak location and hydroxyl functional group type from this work and from the study by Welti (11)are reported in 'Table I. A graphical representation is presented in Figure 2. The most obvious observation is that the carboxylic acids give peaks in a region well separated from the alcohols and phenols. All functionalities studied are noted to be shifted to frequencies 20-60 cm-l higher than corresponding absorptions in condensed phase spectra due to the elimination of intermolecular interactions in the gas phase. The hydroxyl region is also well separated from the vapor-phase amine and amide stretching frequencies which range from 3330 to 3500 cm-l and which are often of low intensity (11). As demonstrated below, the position of the 0-H stretching band can be shifted due to intramolecular interactions of two major types: 1. Steric Crowding. For many of the functional group categories, shifts to slightly higher frequencies (larger wavenumbers) are observed when the hydroxyl group is surrounded

Flgure 3.

2

Representative compounds dlscussed In the text.

by nonpolar side chains. This steric crowding results in restricted movement of the hydroxyl group, increasing the apparent force constant for the vibration, which produces a higher energy absorption. 2. Intramolecular Hydrogen Bonding. While intermolecular interactions are virtually eliminated in the vapor phase, intramolecular interactions are still routinely observed. For molecules which can intramolecularly hydrogen bond to form five- and six-membered rings, a new absorption is usually observed. The 0-H bond is slightly weakened by hydrogen bonding, so the new peak is seen at lower frequency since the apparent force constant is decreased. This effect will often produce a new peak in addition to the nonhydrogen bonded form, and the spectral shifts are usually much larger than for steric crowding effects. Alcohols. In the absence of steric hindrance and hydrogen bonding, the 0-H stretching regions for primary, secondary, and tertiary alcohols are seen to be resolvable, facilitating spectral identification. Only for secondary alcohols were steric crowding effects observed with shifts to higher frequency of about 10-15 cm-l. A correlation between the degree of steric crowding and the 0-H peak location is seen. For example, the unhindered 6-methyl-2-heptano1, 1 (Figure 3), absorbs at 3655 cm-l, while 2-methyl-3-heptanol,2, peaks at 3658 cm-l, 3,5-dimethyl-4-heptanol, 3, a t 3662 cm-', 2,2-dimethyl-3-0~tanol, 4, a t 3664 cm-l, and 2,2,4-trimethyl-3-pentanol, 5, a t 3668 cm-l. Displacement of peaks to lower energy by intramolecular hydrogen bonding can be up to 100 cm-' or more and are

1462

ANALYTICAL CHEMISTRY, VOL. 53, NO. 9, AUGUST 1981 HYDROGEN BONDED

FREE

3647

6 r

3653

3653

6 6

3595

3570

3574

CI

3651

@ r

3562

&

pr

354?GBr

Flgure 4. Comparison of the 0-H stretching frequency for free vs. hydrogen bonded phenols.

observed routinely when five- or six-membered rings involving electonegative atoms or double bonds can be formed. For example, 4-diisopropylaminobutanol, 6 (Figure 3), exhibits a free 0-H stretch at 3670 cm-l, while Z-isopropylaminoethanol, 7, has a peak a t 3566 cm-l due to the formation of a hydrogen bonded six-membered ring. In contrast to Welti's observation (11)that the bonded 0-H band is usually more intense, we found a fairly even distribution with the free 0-H peak being the more intense in about half of the observed spectra. However, these relative peak intensities are expected to be a function of both the gas temperature and the strength of the hydrogen bond. Higher temperatures will favor the free 0-H while a stronger hydrogen bond will favor the bound peak. Benzyl alcohols as a group exhibit a free 0-H peak at somewhat lower frequencies than normal alcohols due to the inductive effect of the aromatic ring (IO). The 0-H stretch for benzyl alcohols can also be shifted by hydrogen bonding. For example, 4-methylbenzyl alcohol, 8 (Figure 3), has a free 0-H peak at 3653 cm-l, while 2-hydroxybenzyl alcohol, 9, has a peak at 3639 cm-l due to the formation of a six-membered ring. Phenols. The hydroxyl stretching region for phenols closely parallels that for secondary alcohols, being from 3645 to 3657 cm-' for unshifted compounds. Steric hindrance moves the peak up to 3670 cm-' in 2,6-di-tert-butylphenol. Substantial shifts to smaller energies are seen for ortho-substituted phenols. As seen in Figure 4, this "ortho effect" is quite well-behaved, with the magnitude of shift for the bonded 0-H

being: F C phenyl C OCH3 < C1 C Br < SCH3 < NOz. Compounds doubly ortho-substituted show only a slight tendency to move to lower frequencies relative to the corresponding mono-ortho-substituted analogues. Oximes. The small number of oximes included were observed to be at the same frequency as tertiary alcohols, from 3647 to 3640 cm-l. In addition to the utilization of other spectral features, oximes can be distinguished from tertiary alcohols since the hydroxyl stretching band is seen to be much more intense for oximes. Carboxylic Acids. The carboxylic acid 0-H stretching bands range from 3586 to 3568 cm-l, in a spectral region well separated from the hydroxyl stretching of alcohols. This can presumably be ascribed to inductive effects (IO). A careful examination of the acid 0-H bands shows that the region can be further subdivided, to distinguish aliphatic from aromatic carboxylic acids. Aromatic and a@-unsaturatedcompounds have 0-H peaks from 3586 to 3581 cm-' while saturated acids have peaks from 3580 to 3568 cm-'. For example, dodecanoic acid absorbs at 3576 cm-', while 2-dodecenoic acid has a peak at 3586 cm-l and 3-bromobenzoic acid peaks a t 3583 cm-l. Although the difference in peak location for these two subgroups is small, it is expected to be of utility in structure elucidation due to the sharpness of these bands in the vapor phase. Only small spectral shifts, if any, were observed for carboxylic acids. For example, benzoic acid peaks at 3585 cm-l, while hydrogen bonding shifts the band to 3576 cm-l for 2-chlorobenzoic acid and 3572 cm-' for 2-trifluoromethylbenzoic acid.

ACKNOWLEDGMENT Acknowledgment is made to the donors of the Petroleum Research Fund, administered by the American Chemical Society, for the support of this research. LITERATURE CITED (1) Coffey, P.; Mattson, D. R.; Wright, J. C. Am. Lab. (FalrfleM, Conn.) 1978, 10, 126. (2) Erickson, M. D. Appl. Spectfosc. Rev. 1979, 15, 261. (3) Woodruff, H. B.; Smith, G. M. Anal. Chem. 1980, 5 2 , 2321. (4) Kowaiski, B. R.; Jurs, P. C.; Isenhour, T. L.; Reiiley, C. N. Anal. Chem. 1989, 41, 1945. (5) Grotch, S. L. Anal. Chem. 1974, 48, 526. (6) Veszpremi, T. J.; Csonka, G. I. J. Mol. Stfuct. 1980, 60 249. (7) Veszpremi, T.; Csonka, G. J. Chem. Inf. Comput, Scl. 1980, 2 0 , 234. (8) Veszpremi, T.; Csonka, G. J . Chem. Inf. Comput. Scl. 1980, 2 0 , 239. (9) Anacreon, R. E.; Pattachini, S. C. Am. Lab. (Falrflekl, Conn.) 1980, 12, 97. (IO) Beiiamy, L. J. "The Infra-red Spectra of Complex Molecules", 2nd ed.; Wiiey: New York, 1958. (1 1) Weiti, D. "Infrared Vapour Spectra"; Heyden: New York, 1970. (12) Deianey, M. F.; Uden, P. C. Anal. Chem. 1979, 51, 1242. (13) Deianey, M. F.; Denzer, P. C.; Uden, P. C.; Barnes, R. M. Anal. Lett. 1979, 72A, 963. (14) Deianey, M. F.; Uden, P. C. J . Chromatogr. Sci. 1979, 17, 428. (15) Warren, F. V.; Deianey, M. F., submitted to Appl. Spectrosc. (16) Smith, E. G. "The Wiswesser Line-Formula Chemical Notation"; McGraw-Hili: New York, 1968.

RECEIVED for review January 12,1981. Accepted May 11,1981.