Equations 2 to 6 were used to obtain values for the concentrations of the five components. These values were substituted in Equation 1. If the values did not fit, the areas on the right side of Equations 2 and 3 were readjusted, keeping the sum the same. New values were then calculated for [ a ] to [ e ] and substituted in Equation 1. E y repeating this procedure, five values for the component concentrations were obtained, which were consistent with the six equations. With experience in the use of the method, only one or two trials are necessary. The precision obtained in determining the composition of a typical 1 to 1 butadiene-isoprene copolymer is shown in Table 11. The error is estimated to be within 2 t o 3y0. No reference polymer was available for a check on the reliability of the method. Since pure natural rubber consists essentially of cis-l,4polyisoprene, known mixtures of natural rubber and 1 to 1 butadiene-isoprene copolymer were prepared. The composition of the copolymer was taken to be that shown in Table 11. The range of mixtures was rather narrow, because of obvious solubility limitations. The calculated and experimentally determined compositions of three independent mixtures are shown in Table 111. The
estimated error for each component is 2 to 3%. For quantitative analysis, proper operation of the spectrometer is critical. This subject has been discussed in detail (2-5). If the area of an absorp tion peak is to be proportional to the concentration of protons contributing to it, and independent of the relaxation times, rf power saturation must be avoided. However, the signal has to be strong enough to ensure accurate peak area measurements. The rf level chosen was a practical compromise between these two factors. Signal strength can also be increased by increasing the solution concentration or the inside diameter of the sample tube. For butadiene-isoprene copolymers, solution concentration is limited by solubility and viscosity. A 5y0 solution of copolymer in CC1, is barely pourable. The maximum diameter of the sample tube is determined by the design of the probe. The use of nonspinning tubes of up to 15-mm. outside diameter is feasible with the insert manufactured by Varian Associates for the V-4331A probe. Recently, the Wilmad Glass Co., Buena, N. J., has marketed a plastic spinner for use with the Varian probe which will take 8-mm. and 12-mm. tubes. The accuracy of the method depends
on the accuracy with which overlap ping peaks can be separated. In most cases, visual judgment was good enough for determining the areas to be cut out. Too great an error in separating the overlapping peaks produced inconsistencies in the six equations or negative values for some of the componenb. The separation of peaks 4 and 5 was the most critical and this is reflected in the method of calculation described previously. ACKNOWLEDGMENT
The author thanks Harry Greenberg for supplying the samples of copolymer and natural rubber for this work. LITERATURE CITED
(1) Bovey, F. A., Tiers, G. V. D., Filipcvich, G., J . Polymer Sci. 38, 73 (1959). ( 2 ) Reilly, C. A., ANAL. CHEM.32, 211R f1960). \ - - - - , .
(3) Varian Associates, Instrument Div., Instruction Manual, V-3521 NMR Integrator, Publ. 87-100-029. ( 4 ) VGian Assqyiates, “NMR and EPR Spectroscopy, Chap. 8, Pergamon Press. New York. 1960. (5) Williams, R. B., Ann. N . Y . Acad S a . 70, 890 (1958). RECEIVEDfor review April 16, 1962. .4ccepted June 11, 1962. Pittaburgh Conference on Analytical Chemistry and ,4pplied Spectroscopy, Pittaburgh, Pa., March 1962.
Dete rmina t io n of Unsatu ration a nd Ave rage Molecular Weight of Natural Fats by Nuclear Magnetic Resonance LEROY F. JOHNSON and JAMES N. SHOOLERY lnsfrurnent Division, Varian Associates, Palo Alto, Calif.
b
The proton NMR spectra of triglycerides dissolved in CCI, are characterized by four sets of signals including, respectively, the olefinic protons, the four glyceride methylene protons, methylene groups attached to two doubly-bonded carbon atoms, and the remaining protons on saturated carbon atoms. The area of the signals produced by the C1 and C3 glyceride protons is measured using an electronic integrator and a D.C. digital voltmeter. With this measurement as an internal standard, one can measure accurately the number of olefinic protons and the total number of hydrogen atoms in like manner. From these determinations the average molecular weight can be calculated. Finally, an iodine number is calculated using the number of
1136
ANALYTICAL CHEMISTRY
olefinic protons and the average molecular weight. The agreement between the NMR and Wijs iodine numbers is remarkably good in all fats studied except tung oil where conjugated bonds cause inaccuracy in the Wijs method.
T
of fatty acids and their triglycerides (commonly called natural fats) using high resolution nuclear magnetic resonance (NMR) has been reported by Hopkins and Bernstein (3). Their spectra of natural fats, taken at 40 Mc., showed signals assignable to olefinic protons in the fatty acid chains, protons in the glyceryl radical, and other groups of uniquely distinct proton environments. Area HE INVESTIGATION
measurements on the first two of these allowed a calculation .of degree of unsaturation with an accuracy of 5%. The present work was undertaken to determine the accuracy of unsaturation measurements using the latest instrumental equipment , and techniques. Measurement of the total number of hydrogen atoms present in the average fat molecule as well as the number of olefinic protons allows a determination of average molecular weight; this is used to calculate an iodine number which is compared to that obtained on the same sample by the Wijs method. EXPERIMENTAL
Quantitative information was obtained using a Varian HR-60 spectrometer equipped with a V-3521 NMR
integrator. Integrator output voltage was measured with a Hewlett-Packard 405AR D.C. digital voltmeter. Recorded spectra were obtained with a Varian A-60 spectrometer using 0.5-ml. samples containing a trace of tetramethylsilane as an internal reference. Signal positions were obtained directly from the precalibrated charts with an accuracy of f 1C.P.S. Matheson Coleman & Bell spectrograde CC14 used for all solutions was checked for proton background and had no detectable signals under conditions used for obtaining quantitative data. Fats used for this study are commercially available products. One of the fats, safflower seed oil, was dried over anhydrous CaS04 for several days and a subsequent solution produced an identical determination of unsaturation compared to that of the oil direct from the bottle. PROCEDURE
The spectrum of a typical fat, safflower seed oil, 50 vol. yo in CCh, is shown in Figure 1. The sharp signal a t the extreme right side of the spectrum (high applied field) is from a trace of tetramethylsilane added to the solution to serve as an internal reference standard. The spectrum is divided into seven groups of signals labelled A through G. These groups of signals are assigned as follows: A , hydrogen directly attached to doubly-bonded carbon (olefinic protons) and the methine proton in the glyceryl moiety; B, the two methylene groups in the glyceryl moiety [in each group the individual protons are magnetically nonequivalent because of hindered rotation of the C-C bond; hence these signals are characteristic of the AB portion of an -4BX group (5)] ; C, CH2 groups attached to two doubly-bonded carbon atoms; D, the three CH2groups alpha to carboxyl; E , CH2 groups attached to saturated carbon and doubly bonded carbon; F , CH2 groups bonded to two saturated carbon atoms; G, the three terminal CH3groups. The weak shoulder 65 C.P.S. from SiMel is the low field line of a triplet signal from CH3CH2CH=CH- as in linolenic acid. The other peaks of this triplet are 58 and 51 c.p.s. from SiMec. Quantitative measure of the amount of this grouping is not possible because of overlapping signals; however, semiquantitative information is available from the height of the peak a t 65 C.P.S. This signal is more apparent in the spectrum of linseed oil, Figure 2. Integration of the spectra was accomplished using a sweep rate of about 25 c.p.s./sec. and a radio frequency power level of approximately 100pgauss. Under these conditions no saturation effects are expected even for nuclei with long relaxation times ( 8 ) . The charge on the integrating capacitor is most conveniently read with a D.C. digital voltmeter. Readings were taken midway between areas A and B, between B and C, and after accumulating the total integral which occurred about 60 C.P.S. beyond area G. The integrator input level control was adjusted to provide for a total integral of approxi-
2co
300
A00
I 00
SAFFLOWER SEED OIL CYOCOR CHbCOR CH, o R
I
oc
CHOCOR
Figure 1 .
60-Mc. spectrum of safflower seed oil
mately 50 volts. The output level control was adjusted to attenuate the 50-volt charge to approximately 14 volts. The voltmeter can be fixed to read 0.00-9.99 volts and still perform measurements up to about 15 volts. Fixing the decimal point eliminates interruptions due to voltmeter scaling and, in this case, gives a better indication of the readings to be taken between A and B and between B and C. The first voltmeter reading is a measure of the number of olefinic protons and the methine proton in the glyceryl moiety. Because the integral voltage is cumulative, the second reading includes the signals contributing to area A as well as those from the four glycerine methylene protons producing The integral voltage per area B. proton is obtained from one fourth of area B and this number subtracted from area A leaves a number which is proportional to the number of olefinic protons. The total number of protons is simply obtained by dividing the total integral by the number corresponding to one fourth of area B. Given three digital voltmeter readB, 2 = total ings, X = A , Y = A (Figure l), taken as described, the following relationships apply:
+
Area per proton = ( Y - X)/4 (1) Number of olefinic protons = x - ( Y - X)/4 ( Y - X)/4 (2)
LINSEED
I
J 80
70
Figure 2.
50
40
(3)
These relationships are close approximations of the actual values which can only be obtained by correcting for satellite signals. Carbon-13 is naturally abundant to the extent of 1.108% and has a nuclear spin of The spin coupling constant between C13 and bonded protons depends upon the hybridization of the carbon orbitals (6). Olefinic proton signals have Cl8 satellite signals approximately 80 C.P.S. each side of the C l L H signals, while protons on saturated carbon atoms have satellites 4 -60 C.P.S. from the main signals. Consequently, the high field C13 satellites of signals in area A would be expected to fall in area B and the low field satellites of B in A . Likewise, the high field satellites of C are included with B. In all cases the low field and the high field satellites each amount to 0.005570 of the corresponding main band, Corrections for the satellite signals modify Equations 1, 2, and 3 to the following: Area per proton j1.0112 ( Y
=
- X) -
0.0055X -O.O055( C)] /4 = 0.2528Y - 0.2542X - O.O014(C) (4)
100
OIL
dU
60
- X)/4
(Y
200
300
400
5Y
Total number of protons =
30
,
,
20
/
,
10
I
0 PPM (1)
60-Mc. spectrum of linseed oil VOL. 34, NO. 9 , AUGUST 1962
1137
Number of olefinic protons
=
1.00556X - 0.0055( Y - X ) [1.0112(Y - X ) 0.0055X O.O055(C)]/4 [l.Oll(Y - X ) 0.0055X - O.O055(C)]/4
-
5W
TUNG
-
-
- 0.2583Y + 0.0014(C) - 0.2542X - O.O014(C)
1.2653X 0.2528Y
Total number of protons
-
(5)
I
The quantity C, which depends on the number of =C-CH2-CH2-C= groups, can be estimated from the spectrum in terms of area B-e.g., in safflower seed oil C is approximately equal to 1.O B and hence 1.O( Y - X ) . A general formula for natural fats may be written in the following manner:
J\
/,
I
4-
+ b + c) + 2 6 . 0 3 % ~+ 21 + z) (7)
The first term, 173.1, is the formula weight of the CeH60eglyceryl triester radical, while the second term, 45.1, is the formula weight of three methyl groups. The total number of protons, T, in the above formula is
+ 9 + 2(u + b + + C)
2(2
+ tl + z)
(8)
Iodine No. Values of Various Fats
Table 1.
NMR No.
Oil Cncnnut - -. . - -.
Olive Peanut Sovbean Suinflower seed Safflower seed Whale Linseed Tung
WIJS No.
10.5 i 1 . 3 8.0-8.7 80.8 =k 0 . 9 83.0-85.3 9 4 . 5 f 0 . 6 95.0-97.2 127.1 i 1 . 6 125.0-126.1 135.0 i 0 . 9 136.0-137.7 141.2 f 1 . 0 150.2 f 1.0 176.2 f 1 . 2 225.2 f 1 . 2
140.0-143.5 149.0-151.6 179.0-181.0 146.0-163.5
Table II. Average Molecular Weights of Fats from Saponification Values and
NMR
Oil Olive Peanut Safflower seed 1 138
0
Sap. Value
Mol.
Wt.
NMR Mol. Wt,
189.3 887.1 873.7 i 5 . 3 188.8 891.5 882.3 f 7 . 4 191.5 879.0 874.9 f 9 . 3
ANALYTICAL CHEMISTRY
=
2(s
+y t
(9)
2)
+ c ) in terms of T and V. + b + c ) = (2' - V - 14)/2 (U
The molecular weight using this formula is :
= 5
//
Equation 8 may be solved for ( a
b
&Hnoco(cH2),(m = c H ),CHI
173.1 -I- 45.1
d"L.J*u?L
The number of olefinic protons, V , is simply:
v
CH--OCO( CHZ)b( CH=CH)&Ha
1 1
h '
CH,OCO(CHn)a(CH=CH)zCHs
T
OIL
-
-
=
,
i
[l.Oll(Y - X ) 0.0055X O.O055(C)]/4 2/[0.2528Y - 0.2542X - O.o014(C)] (6)
14.027(u
100
=
6
Mol. wt.
io:
300
400
-
+
(10)
Likewise Equation 7 may be expressed in terms of T and V using Equations 9 and 10. Mol. wt. = 218.2 7.013(T Mol. wt. = 120.0
+ + 7.013T +
- V - 14) + 13.019V 6.006V (11)
Thus, the quantitative data obtained with the digital voltmeter may be converted to average molecular weight through Equations 5, 6, and 1 1 . Finally, the degree of unsaturation may be expressed in a calculated iodine number. 126.91 IodineNo' = Equiv. wt. (fat) no. of olefinic H's X 100
Iodine No.
=
12691V
Mol. wt.
RESULTS A N D DISCUSSION
Table I shows the results of our study of nine natural fats. The errors listed for each NMR-iodine number are standard deviations derived from a t least 10 measurements on each fat. Iodine numbera determined by the Wijs method are given as a total range which is taken from duplicate runs done by two different analytical laboratories. The agreement between the two methods of determining iodine number is remarkably good for all fats except tung oil. The high concentration of eleostearic ester, containing conjugated double bonds, has been shown to interfere with the true measurement of unsaturation by the Wijs method (4, 7, 9). The NMR spectrum of this fat, Figure 3, differs from that of the other oils since the olefinic protons associated with
I
Jl
\J' LI \u
conjugated double bonds produce signals which are shifted toward lower field by 0.4-1.0 p.p.m. The total unsaturation is still readily measured by including these signals with those associated with the unconjugated double bonds. Thus, the NMR method provides an indication of total unsaturation, expressed as a calculated iodine number. I t is interesting to note that the theoretical iodine number for tung oil, calculated from gas chromatographic data ( I ) , is 229. While not from the same sample of tung oil, this value does support the correctness of the NMR value of 225. On a Dercentaze basis there is a fairly large discrepancy between the NMR and Wijs iodine numbers for coconut oil. Here again, calculations from gas chromatographic data reveal a theoretical value, 12.1, which is in better agreement with the NMR iodine number. The reason for the discrepancy in this case is not fully understood; however, it does appear to be real since the range in iodine number for coconut oil is reported to be within 7.5-10.5 ( 1 , 2 ) . Molecular weights calculated from NMR data using Equation 11 can be related to those obtained from chemical data through the saponification value. Table II shows the resl-'/q of such measurements on three of the oils. The agreement between the two methods is as good as in the corresponding iodine values. The total time spent in obtaining the NMR data is about 20 minutes per sample. Calculation time can undoubtedly be made quite shwt by using suitable nomographs prepared with the aid of Equations 5, 6, 11, and 12. -
0
ACKNOWLEDGMENT
The authors are indebted to DeSoto Chemical Coatings, Inc., Berkeley, Calif., for supplying samples of linseed oil and tung oil. Iodine numbers by the Wijs method were determined by Curtis and Tompkins, Ltd., and Pacific Chemical Laboratories, both of San Francisco.
LITERATURE CITED
(1) Archer-Daniels-Midland Go., Minneapolis, Minn., “Composition and Consia& of Natural F a 6 and Oils.” (2) Fieser, L. F., ~ieser,M,, “Organic Chemistry~”pa 4 0 8 ~ Reinhold, New York, 1956. (3) Hopkin% c. y., Bernskin, H. J., Can.J . Chem. 37, 775 (1959).
(4) Mikasch, J. D. von, Frazier, C., IND. ENG. CHEM., ANAL. ED. 13, 782 (1941). (5) Pack, F. C., Planck, R. W., Dollear, F. G., J . Am. Oil Chemists’ SOC.29, 227 (1952). (6) PoPle, J. A.7 Schneider, w. G.7 Bernstein, H. J., “High Resolution Nuclear Magnetic Resonance,” p. 98, McGrawHill, New York, 1959. (7) Shoolery, J. N., J . Chem. Phys. 31, 1427 (1959).
(8) Stiihli, H., Mitt. Gebiete Lebensm. Hyg. 46, 121 (1955). (9) Varian Associates, Palo Alto, Calif., Tech. Info. Bull. Vol. 3, No. 1 (1960).
RECEIVEDfor review March 5, 1962. Accepted May 21, 1962. Pittsburgh Conference on Analytical Chemistry and Applied Spectroscopy, Pittsburgh, Pa., March 5, 1962.
Purification of Acetonitrile as a Solvent for Exact Measurements J. F. COETZEE, G. P. CUNNINGHAM, D. K. McGUIRE, and G. R. PADMANABHAN Deportment of Chemistry, University o f Pittsburgh, Pittsburgh 7 3, Pa.
b Acetonitrile is attracting increasing attention as a solvent for electrochemical and other reactions. Since it is a comparatively inert solvent, several possible impurities are sufficiently reactive to modify its propertiessignificantly, even in very low concentrations. The conventional method for the purification of acetonitrile (repeated distillation from phosphorus pentoxide) is not entirely satisfactory. Two alternative methods that give superior results are described. The results of polarographic and gas chromatographic tests for logical impurities are given. The persistent polarographic wave at -2.2 volts vs. the saturated calomel electrode is not caused by acetic acid, as previously assumed, but by unsaturated nitriles.
D
the last several years there has been a rapidly increasing interest in the use of acetonitrile as a solvent for electrochemical and other reactions. Various theoretical studies of a quantitative nature have been carried out in acetonitrile. URING
TYPICAL EXAMPLES. Polarography of inorganic substances (11, 12, 16) and various classes of organic compounds (18), as well as solid electrode voltammetry of inorganic substances (11, 16). Measurement of electrode potentials of common inorganic couples (16). Conductometry of salts, particularly substituted ammonium salts (4, 17). Quantitative studies of dissociation equilibria of acids (6, 10) and aminetype bases (14), using conductometric and spectrophotometric methods, Measurement of the autoprotolysis constant of acetonitrile (7). Electron spin resonance studies of electrolytically generated transient free radicals (9). Aromatic chlorination rate studies (9)
-
In addition to these quantitative theoretical studies, numerous empirical and semiempirical investigations (particularly acid-base titrations) have provided results that are useful for practical and comparative purposes (14). Since acetonitrile is a comparatively inert solvent, the presence of many possible impurities, even in very low concentrations, would make it unsuitable for exact studies. Several authors have commented on difficulties encountered in the purification of acetonitrile for such purposes. EFFECT OF IMPURITIES
Numerous methods are available for the preparation of nitriles (3, 13). We were unable to obtain information from the manufacturers about the methods actually used in the commercial preparation of acetonitrile and other nitriles. Technical data reports are available from the Eastman Kodak Co. for n- and isobutyronitrile (8), but not for acetonitrile. Polarographic and other tests have been carried out for a number of substances that logically may be present as impurities in acetonitrile, particularly isonitrile, water, and the various hydrolysis products of acetonitrilenamely, acetamide, ammonium acetate, ammonia, and acetic acid. Acetonitrile is a relatively weak base, much weaker than water (11,12), and an extremely weak acid (7,14). Several possible impurities are acids or bases sufficiently strong to modify the properties of the solvent significantly, even when present in very low concentrations. The presence of acetic acid would be particularly objectionable in exact studies of bases in unbuffered solutions, and that of ammonia in similar studies of acids, even if the concentrations of these impurities are as low as 10-6M. Although the effect of water is less
marked, the presence of relatively low concentrations of this persistent impurity can cause large errors in certain measurements. EXAMPLES. A. The polarographic half-wave potential of the solvated proton (as 1 X 10-*M perchloric acid) in acetonitrile as solvent becomes 0.15 volt more negative on adding 1 X 10-2M water (11). This large shift is mainly the result of the pronounced increase in the solvation energy of the proton which occurs when the relatively strong base water converts the species CH&NH+ into H2OH+ (11,12). B. The dissociation constant of the protonated form of the Hammett indicator 4chloro-%nitro-N-methylaniline increases by a factor of 20 when 2 X 10-2M water is added (as .shown in unpublished results, this laboratory). The cause of this large effect is basically the same as for example A. C. The conductance of 0.1M nbutylamine is doubled by adding 8 X 10-2M water ( I C ) , because the acid properties of water (proton donor strength, as well as solvation of anions by hydrogen bonding) are much stronger than those of acetonitrile. CONVENTIONAL METHOD FOR PURIFYING ACETONITRILE
The conventional method for the purification of acetonitrile generally involved some kind of pretreatment to remove acetic acid-for example, shaking with aqueous potassium hydroxide, solid potassium carbonate, or aluminafollowed by repeated distillations from phosphorus pentoxide until the residue was no longer colored (orange or black). Certain authors included an additional preliminary drying operation (shaking with calcium chloride, magnesium sulfate, etc.) and an additional (final) distillation of the product, either alone or with added potassium carbonate or barium oxide. Repeated distillation of acetonitrile from phosphorus pentoxide has a major VOL 34, NO. 9, AUGUST 1962
11 39
