For the data a t hand, k = 1 and k = 3 both result in the classifications shown by the uncircled numbers of Figure 2 , except that A is again classified as belonging to class 3. It is, perhaps, worth pointing out that none of these methods are particularly sensitive to accuracy. Thus, adjacent points in region 3 o n the plots differ on the average by about 20 in each of ten element concentrations. CONCLUSION
Four independent computer methods have been applied t o the data processing task discussed herein. The display method represents a purely unsupervised learning approach because the source material and artifacts were not distinguished as such when input to the computer program. Cluster analysis was applied in both the supervised and unsupervised modes, demonstrating its flexibility. Finally, two classification methods were used, representing a completely supervised learning approach. The four methods were selected because of their popularity and usefulness.
It should not be surprising that the classifications achieved by the various pattern recognition algorithms d o not differ appreciably from that achieved by a knowledgeable scientist, for both use basically the same criteria. Rather, one should view pattern recognition as a method of automatic classification, automatic in the sense that classification proceeds once the rules are given. It may, thus, be said t o have two advantages over the classification ordinarily done by a scientist. First, by forcing the rules to be laid down explicitly (in the program), it removes subjective judgment, or at least lays it open to full view. Second, it can handle economically many more data than can be assimilated by a human classifier. In the present case, the problem is of such complexity as to be not far fromi the limit of human capability, and yet it seems almost trivial for the computer.
RECEIVED for review May 8,1972. Accepted July 13,1972,
Graphic Representation of Nuclear Magnetic Resonance Chemical Shifts for the Olefinic Protons Osamu Yamamoto, Masaru Yanagisawa, Hiroshi Tomita, Tomoko Ogawa, Yoshie Senga, Kikuko Hayamizu, and Fumiko Taka National Chemical Luboratory for Industry (Tokj.0 Kagyo Shikensho), 1-1-5, Honmachi, Shibuya-ku, Tokyo, Japan
A graphic representation of PMR chemical shifts for the olefinic protons is presented. About 400 7-values are compiled from about 200 spectra of olefinic compounds. The correlation charts presented here provide the ranges of the olefinic proton shifts arranged according to the neighboring substituents, and make it easier to deduce chemical structures from PMR spectra.
PROTON MAGNETIC RESONANCE (PMR) has been widely used for the structural determination and identification of organic compounds, and many attempts have been made to correlate the proton chemical shifts with the molecular structure having selected functional groups (1). Usually the proton chemical shifts can be easily obtained from the PMR spectra and the proposed structure of the molecule can be deduced from the chemical shifts of the signals by referring to suitable correlation charts or tables (2-12), as well as taking into considera(1) J. W. Emsley, J. Feeney, and L. H. Sutcliffe, “High Resolution
Nuclear Magnetic Resonance Spectroscopy,” Vol. 2, Pergamon Press, New York, N. Y., 1966, p 710. (2) G. V. D. Tiers, “Characteristic Nuclear Magnetic Resonance Shielding Values for Hydrogen in Organic Structures,” Part 1; Tables of 7-Values for a Variety of Organic Compounds, Minnesota Mining and Manufacturing Co., St. Paul, Minn., 1958. (3) L. H. Meyer, A. Saika, and H. S. Gutowsky, J . Arner. Chem. SOC.,15, 4567 (1953). (4) N. F. chamberlain, ANAL.CHEM.,31, 56 (1959). (5) K. Nukada, 0. Yamamoto, T. Suzuki, M. Takeuchi, and M. Ohnishi, ibid., 35, 1892 (1963). (6) bf.W. Dietrich and R. E. Keller, ibid., 36,259 (1964). (7) C. Heathcock, Can. J . Cliern., 40, 1865 (1962). (8) K. W. Bartz and N. F. Chamberlain, ANAL.CHEM.,36, 2151 (1 964). (9) F. C . Stehling and K. W. Bartz, ibid., 38, 1467 (1966). (10) 0. Yamamoto, T. Suzuki, M. Yanagisawa, K. Hayamizu, and M. Ohnishi, ibid., 40, 568 (1968). (11) N. F. Chamberlain, ibid., p 1317. (12) D. J. Frost and J. Barzilay, ibid., 43, 1316 (1971). 2180
~~
~
Table I. Gas Chromatograph Columns Used for Purification of Materials Column Purified material Hydrocarbons, haloalkenes, Al, lGft, Carbowax 4OOO; 30x/Chromosorb W (AW) vinyl compounds, (60-80 mesh) oxygenated compounds with bp below 200 “C.. aromatics, and most of allylic compounds. Allylic, aromatic and Al, 20-ft, FFAP; 2073 Chromosorb W (60-80 oxygenated compounds mesh) with higher bp, and unsaturated fatty acids. AI, 10-ft, DC-550;3 0 z / Compounds with higher bp, Chromosorb W (AW) halogenated styrenes, and (60-80 mesh) unsaturated fatty acids. Al, 10- and 20-ft, SE-30; Hydrocarbons, halogenated styrenes, and unsaturated 30~/ChromosorbW (AW) fatty acids and their (60-80 mesh) derivatives. Amines, nitro compounds, Glass, 10-ft, Carbowax 20 M; 20% NaOH; 5 7 3 vinyl pyridines, and nitriles. Chromosorb W (AW) (40-60 mesh) Maleic acid derivatives. Al, 20-ft, Carbowax 20 M; 30x/Chromosorb W (60-80 mesh)
+
tion splitting of the signals by using the simple spin multiplicity rules. Olefinic compounds are an important class of organic compounds, and the correlations of their chemical shifts with the molecular structure in a convenient chart form have been made by some workers. Thus, Stehling and Bartz (9) made a correlation chart of the chemical shifts for olefinic hydrocarbons, and Chamberlain (11) compiled the chemical shifts of
ANALYTICAL CHEMISTRY, VOL. 44, NO. 13, NOVEMBER 1972
2-Value
I
I
I
I
I
I
I
1
I_
4
I3 13 1
1
I
17
I
63
63
1 1
2 2 10 e
10
I
1 h
16
I
16
9 9 I
Figure 1. The first subdivision chart for the olefinic proton shifts of the type H '\
/'
H
/
c=c
\
( a ) acylated, (b) -COOR, -COC6H5,and -COCI, (c) -COR, -CHO, and -COW*, (d) pyridines and 2,4,6-trimethglstyrene, ( e ) 2-vinylpyridine, (f)4-vinylpyridine, (8) 2,4,6-trimethylstyrene, (h) HzC=C(CH ,)CHO, (i) acrylyl chloride, and ( j )acrolein
oxygenated unsaturated aliphatics, also in a chart form. Frost and Barzilay (12) presented the PMR parameters for the identification of non-conjugated cis-unsaturated fatty acids and esters from 220 MHz spectra. Although these publications are very comprehensive, not all of the functional groups usually met in organic compounds are involved in their charts; therefore, it is desirable to obtain more general correlation charts for the olefinic proton chemical shifts. Furthermore, the usefulness and the convenience of such correlation charts depend, to some extent, on the classification system of the functional groups. Thus, in order to meet the requirement in a variety of fields, it seems useful to have several correlation charts of different classification systems. In our previous paper, we proposed a classification system for compiling proton chemical shifts for analytical purposes, comprising main classification groups in which some subdivisions are included (5). The main classification is based on differences in the bond type of the carbon atom to which the proton is bonded, and on differences in the number of
protons attached to that carbon atom. The first subdivision is made by both species and bond type of the atoms which are substituted at the a position. Similarly, the second subdivision is made by species and bond type of the substituent at the 0 position, and so on. In terms of this classification system, we have already presented the graphic representations of PMR chemical shifts for methyl, methylene (9,and acyclic methine (IO) protons. In the present paper, we have obtained about 200 PMR spectra of olefinic compounds and have summarized about 400 7 values of the olefinic protons in the form of chemical shift charts according to the classification system described above. The manner in which the correlation charts are applied to the actual structural determination will be evident. EXPERIMENTAL Apparatus and Procedure. PMR spectra were obtained by a Varian HA-IOOD spectrometer at 100 MHz. The conin CCll centration of the samples was from 3.5 to 7.5 mol
ANALYTICAL CHEMISTRY, VOL. 44, NO. 13, NOVEMBER 1972
2181
-
n -
I
1
8
2
4 27
+-
17 63
5
3
13
Figure 2. The first subdivision chart for the olefinic proton shifts of the type H
\
/
/
\
c=c
R
(b)(CH3)C=CH(CH&C(CH3)=CHCHrOH, (c) 4-methylpentene-2, (d) 1,4-dichlorobutene-2, and (e) l-chloro-
butene-2
Table 11. Comparison between the Chemical Shift Values Obtained from Direct Reading and the Exact Calculation (T values) Ha X
\
/
/
\
c=c
H.
HI From
-X
-CH?CI -CHzNH?
-0 -O(CH?)zCI -COCI -C N -COCH? -COOCH s
2182
direct reading 4,076 4.833 4.692 4.055 4.972 4.862 3.401 4.603 4.116 3.616 6.007 5.858 3.684 3.886 3.419 4.362 3.992 3.837 3,762 4.220 3.908 3.944 4.255 3,667
From the exact calculation
4.059 4.837 4.703 4.042 4.976 4.869 3.391 4.609 4.127 3.610 6.008 5.860 3.691 3.903 3.397 4.384 3.983 3.822 3 712 4.226 3.949 3.935 4.250 3.699
Difference 0 017 -0 004 -0 011 0 013 -0 004 -0 007 0 010 -0 006 -0 011 0 006 -0 001 -0 002 -0 007 -0 017 -0 022 -0 022 0 009 0 015 0 050 -0 006 -0 041 0 009 0 005 -0 032
when permitted by solubility. When the solubility was not sufficiently high to form solutions in the above concentration range. a saturated solution was used, and when a good SIN ratio could not be attained with the saturated solution, deuterated chloroform was employed as a solvent. The chemical shift is referred to internal TMS and is expressed in r value. Reagents. All of the samples were obtained from commercial sources. Olefinic compounds often contain impurities, particularly their geometrical isomer counterparts, and the fairly complex PMR patterns of the olefinic proton region tend to be disturbed by the impurities. Then purification of the materials was inevitable in order to obtain reliable results. The purification was carried out, where possible, by preparative gas chromatography for liquid samples. A Varian Autoprep A 700 Preparative Gas-Chromatograph was used with helium as a carrier gas at a flow rate of 180-200 ml/min. The columns used for the purification are listed in Table I, with the types of compounds for which the columns were employed. Naturally, details of the purification conditions vary from compound to compound, and the summarized results will be published later. The purity of the purified samples was above 99%, as determined with a Shimazu 4ATF Gas Chromatograph. Some of the solid samples were purified by the zone-melting technique. RESULTS AND DISCUSSION
Data Compilation. It is well known that the concentration of the samples and the solvent used afTect the chemical shifts, and the accuracy of the obtained shift values is mainly determined by the sampling conditions rather than the manner of the measurement. Thus, it is reasonable to set the precision of the obtained shift values at i 0 . 0 2 ppm, based o n the concentration range used in this work. The signals of the olefinic protons vary from simple t o very complex patterns. In general the vinyl and allyl type patterns are fairly complex, but the chemical shifts can be extracted
ANALYTICAL CHEMISTRY, VOL. 44, NO. 13, NOVEMBER 1972
I
I
I
I
I
I
I
I
I
'
d
'I
3 13
2
4 1
3 9 6
5
'd
13
0
8
I
2 1 1 1
3 10 1
1 5 1 1
1
I
0
1
I
1 ~~
I
7-VaLue
I
I
I
I
I
I
I
I ,
-
Figure 3. The first subdivision chart for the olefinic proton shifts of the type H \ ./ C=C / \ X (a) acylated, (b) -COOR, -COCsHs, and -COCI, (c) -COR, and -CHO, (4 (CH&C= CHCOOCHI, (e) H*C=CHCON(CH&, cf) divinyl maleate, (g) CBHd-IC=CHCOCBH5, (h) (CeHs)CH=CHCH=CHCOCH=CHCH==CH(CeHs), (i) G - C H - C H C O O H ,
u) CH~COCHECHCOOCH~,and ( k ) CH&OCH=CHCOOCHa with ruler and compasses from a spectral chart by use of the simple spin multiplicity rule as far as the 100 MHz spectra are concerned. The values obtained in this manner agree with those obtained by the exact computer analysis to within about 1 0 . 0 2 ppm, except for the case of very strong coupling. This is illustrated in Table 11. Thus, in most cases we used, in our graphic representation, the chemical shifts obtained directly from the first-order spectra.
In the case of a,B-disubstituted olefins where the two substituents are both alkyl or substituted alkyl (for example, octene-2), very complex patterns are frequently obtained due to the small difference in the chemical shifts of the two olefinic protons and also due to the many couplings with the substituent protons. The double or triple resonance technique may be used to solve the difficulty, but we sometimes encountered cases in which these decoupling techniques cannot be
ANALYTICAL CHEMISTRY, VOL. 44, NO. 13, NOVEMBER 1972
2183
Table 111. Long Range Coupling Constants (Hz)and the Chemical Shifts (7)for Some Terminal Olefinic Compounds
Compounds
/.C=C, H2 Hi
‘c=c
HZ’
/
7
~
JCE~-H
IJCH~HI
&bsda
h o d b
1.55
4.99
5.16
0.81
4.72
4.77
1.68
4.35
4.31
1.14
4.25
4.25
1.60
4.54
4.59
0.97
3.98
4.09
1.12
5.43
5.86
0.52
5.40
5.63
CH2COOCHs
1.12
4.35
4.59
COOCH3
0.48
3.77
4.09
1.11
0.81
5.21
5 . 1tlC
0.48
1.54
5.17
5.25
1.49
0.0
5.11
5.16
0.77
1 .oo
4.97
5.11
1.17
4.42
4.86
0.56
4.62
4.89
3
‘CH2COOCH3
Compound from which &&led is obtained Styrene
Acrylonitrile
Ethyl acrylate
Vinyl acetate
Ethyl acrylate
Methyl allylacetate
Allyl chloride
Vinyl chloride
Assignments are made assuming the rule for the long range coupling constants. See text. Assignments are made assuming the rule by Reddy and Goldstein (14). See text. The assignment is not consistent with that obtained from the rule for the long range coupling constants. But the difference in chemical shifts between the two olefinic protons is very small, and the error in acalod of this order of magnitude would be well expected. We prefer the assignment from the rule for the long range coupling constants. a
successfully applied. Then we were obliged t o carry out computer analysis, but even in this case some trial calculations without iteration techniques may suffice to obtain the chemical shifts with an accuracy of about 10.02ppm. As a guiding principle, we use the values directly read out from the chart as much as possible, and, if required, the double or triple resonance technique is employed. When a proton gives a complex signal with a relatively narrow band width, the center of gravity of the signal is taken as the chemical shift value. Only in cases where all of the above methods fail is computer simulation without iteration techniques used to extract the chemical shifts with reasonable accuracy. The computer simulation of the spectra was carried out by use of the LAOCOON I1 program (13), modified by the authors. Assignment of the signals in the olefinic compounds is generally straightforward. However, for the compounds, H X
\
/
may be determined by handling many data for the related compounds. When either X or Y is an alkyl or substituted alkyl group, we may use the following method. Reddy and Goldstein (14) found that the effects of the methyl substitution HI H
\ C=C
for the a-proton in vinyl compounds,
,’, are a
/’ \ Hz X $0.35 ppm shift for the H1 proton and a $0.41 ppm shift for the H2proton. For example, since the HI and H? proton
HI
H
\ shifts in
/
C=C
/
are 4.84 and 4.37
7 ,respectively,
\
Hz CBH5 the predicted values of the Hi and Hi CHI
H2
proton shifts in
, it is not so easy to determine which proton is
\
/
H Y trans or cis to the particular substituent X. Frequently, this
/
\
H2
(13) S. Castellano and A. A. Bothner-By, J , Chem. Phys., 41, 2796 (1 964).
(14) G. S. Reddy and J. H. Goldstein, J. Amer. Chem. SOC.,83, 2045 (1961).
C=C
/
2184
\
c=c
ANALYTICAL CHEMISTRY, VOL. 44, NO. 13, NOVEMBER 1972
are 5.16 and 4.77
7,
respectively.
The
C6H5
C Q :,c; H'
H3 COOH
Qc;c:CH2cO~ H' COOR
i
-s-1-
- coo -
I:
Fc:c:
COOR
C HQ
Ii '
I
10
3
I
-Ne
6
-0-
-1
19
-Hal
I
3
7-Value
Figure 5. The second subdivision chart for the olefinic proton shifts of some unsaturated carboxylic compounds
-cc
21
-c=c-
I
- s8 --
II I
-coo-
j
1
2
L
\
/
C=C
~
-Hal
1
I
i-
-cc
1
I
+
- c;c -
methyl compounds. Thus, we may conclude that only the long range coupling constants can be used to determine the assignment of the signals in the type of compounds H R
3 6 19
3 -
(R = alkyl or substituted alkyl).
\
X It is to be noted, however, that the comparison between the two long range coupling constants should be made in the same molecule. When the comparison is made between two molecules which are geometrical isomers of each other, J c H ~ - may H ~be~ larger ~ ~ ~than J C H ~ - Hwhich ' ~ ~ , is indeed the case for, for example, crotonic and isocrotonic acid esters
(16). 21
I
-s-
2
-1-
L
3 6 19 3
Figure 4. The second subdivision chart for the olefinic proton shifts of allylic compounds observed values for the two olefinic protons in the latter compound are 4.99 and 4.72 r , respectively, and then the signal a t 4.99 7 is assigned t o HI and the other to Hs. If this assignment is correct, the long range coupling constant between the methyl protons and H1proton (the long range cis coupling constant, 1.55 Hz) has a larger absolute value than the long range trans coupling constant (0.81 Hz). The examination for the other compounds shows that the parallelism between the above two facts always exists (see Table 111). This has also been confirmed by Kowalewski (15) for the __ (15) D. G. de Kowalewski, J. Mug. Resonance, 4,249 (1971).
/ H
Correlation Charts. Results obtained are shown in chart form according to our classification system. Figures 1 to 3 are the first subdivision charts of the olefinic proton chemical shifts. The a-substituent is a hydrogen atom in Figure 1, a n alkyl carbon atom (R) in Figure 2, and the other functional groups in Figure 3. T h q Figure 1 shows the terminal olefinic proton shifts, while in Figures 2 and 3, vinylic, a,& disubstituted, and trisubstituted olefins are included. If two classification groups in the cis and trans relationship for a particular substituent are available, they are shown in pairs in the charts. In the figures the range of the chemical shifts obtained for each type of compound is represented by a horizontal thick line, with a vertical short line expressing an average of the T values of the samples in a classification group. The figures in the right-hand column are the numbers of the samples in the groups. As in our previous paper (IO),the shifts of the compounds which show the acylation shift are separately shown in the 'same row of the group, denoted by the symbol u in the charts. 'The ranges shown separately from the main with notations (5, c, and so on are the exceptions in the classification group, which otherwise unduly extends the entire shift range of the group. Although the general features of olefinic proton shifts have already been studied extensively ( I ) , and the detailed discussions about exceptions are outside the scope of this paper, some remarks can be made here on the general be(16) R. R. Fraser and D. E. McCreer, CarzJ. Cliem., 39, SO5 (1961).
ANALYTICAL CHEMISTRY, VOL. 44, NO. 13, NOVEMBER 1972
2185
Figure 6. The first subdivision chart for the olefinic proton shifts of vinyl compounds (a)acylated, (b)-COOR, -COCsHs, and COCI, (c) -COR, -CHO, and -CONR2, (d)acrylyl chloride, (e) acrolein, (1)HzC= CHCON(CH&, (g) pyridines and 2,4,64rimethylstyrene, ( h ) 2-vinylpyridine, (i) 4-vinylpyridine, and ( j ) 2,4,6-trimethyl styrene
havior of the olefinic proton shifts. On the average, the shift of an olefinic proton trans to a particular functional group is at a higher field side than the proton cis to the substituent. This is easily seen from the two groups shown in pairs in Figures 1-3. Second, a comparison between Figures 1 and 2 shows that the change in the substituent from hydrogen atom to an alkyl carbon at the a-position tends to make the olefinic proton shift to a higher field side. This is a more general illustration of the rule found by Reddy and Goldstein for the vinyl compounds (14). Figures 4 and 5 are the second subdivision charts for allylic and some unsaturated carboxylic compounds, respectively. In allylic compounds the substituent X affects the olefinic proton shifts only to a small extent. It is interesting to note that the range of the a-proton ( p to the substituent X) shifts
2186
is comparable to those of the @-protons ( y to X) in allylic compounds. In Figure 5, it is seen that the olefinic protons in acid resonate at a lower field than those in corresponding esters do. Figure 6 is a chart for vinyl compounds. Although there is some overlapping with Figures 1-3, because of their importance, they are given here for the reader’s convenience. ACKNOWLEDGMENT
The authors express their hearty thanks to Tokyo Kasei Kogyo Co. Ltd., from which a majority of the samples were obtained.
RECEIVED for review April 21, 1972. Accepted July 17, 1972.
ANALYTICAL CHEMISTRY, VOL. 44, NO. 13, NOVEMBER 1972
