1WIUSTHIi.I i\L) EhGIhEERING CHEMISTRl
1132
with terpenes generally, the terpenemaleic anhydride resiii. tend to form peroxides which catalyze the polymerization of drying oils and thereby hasten film formation. In any discussion on resins of the alkyd type or the base> from which such resins are made, the question of the upe of rosin always enters. For that reason we may also consider the reaction of rosin with maleic anhydride. It is customary to differentiate between the terpenes and the terpene acidi such as abietic acid. I n the diene synthesis no such distinction exists. The abietic acid molecule was shown by Ruzicka to react with maleic anhydride to give a compound with the following structure: CH3,
VOL. 28, NO. 10
This fact has been utilized industrially by preparing resins from rosin and maleic anhydride. The addition product between methyl abietate and maleic anhydride is unique in being extremely stable to heat. It may be refluxed a t atmospheric pressure with only a small amount of decomposition. There seems to be no liberation of either maleic or succinic anhydrides, and, after refluxing for 30 minutes, the product iiiay be recovered in substantially pure form by a single crystallization. When rosin is reacted with maleic anhydride, there is a marked increase in melting point and the product 1,ehaves in general like a tribasic acid.
Literature Cited
/
Diels and Alder, Ann., 460, 98 (1927). Hurd, IND.ENG. CHEM.,26, 53 (1934): Hurd arid Hallman, J . Am. Chem. SOC.,55, 699 (1933); 56, 447 (1934). ( 3 ) Xienle, IND. ENG.CHEM.,22, 590 (1930). (4) Koch, Dissertation, Christian Albrecht University, Kid, 1932 (5) Littmann, J. Am. Chem. SOC.,57, 586 (1935). RECEIVED September 2, 1936
Polymer Distribution in 1
on such resins as polystyrene and polyisobutylerie; the copolymer of vinyl chloride and vinyl acetate can also be separated into fractions of different solubility.
Vinyl Ester
Molecular Weight Determination by Cryoscopic Method
Resins S. D. DOUGLAS AND W. N. STOOPS Carbide and Carbon Chemicals Corporation, South Charleston, W. Va.
Conventional methods of determining molecular weight, \rich as osmotic pressure or freezing point measurements, are risually applied only to low polymers with molecular weights of only a few thousand. The results of cryoscopic molecular weight determinations on four such Iow-polymer fractions are given in Table I. Fractions 1 and 2 were prepared by extracting the low polymer from one lot of vinyl chlorideacetate resin and fractions 3 and 4 from another. They were carefully purified by repeated precipitation. Attempts to
S
TSTHETIC resins are made up of molecules which differ widely in their molecular weight or degree of polymerization. The physical properties of synthetic resins vary in R general way with degree of polymerization, at least until they become highly polymerized. A determination of the polymer distribution, or percentage composition based on degree of polymerization, of a resin would obviously b P useful both in studying its physical properties and the polymerization reaction by which it was made. The ultracentrifuge affords the only method known a t present for obtaining this information directly, but the equipment is too coniplicated and expensive for general use. 111 rase the resin is soluble and there are appreciable differences of solubility among the various polymer bands, it can be separated into fractions of differing average molecular weight by extraction with appropriate solvents or by fractional precipitation from solution. This proredure was wed by Staudinger and others
FIGCRE1. POLYMER DISTRIBUTION CURVESFOR Two TYPESOF VINYL CHLORIDE-ACETATE RESIN
IXDUSTHIAL AND ENGINEEKING CHEklISTRP
OCTOBER, 1936
1153
obtain molecular weights by this method on more highly polymerized samples were not successful. Dioxane was used as a solvent for the freezing point measurements. It was first refluxed over sodium, distilled, and then fractionally crystallized several times. The purified material had a freezing point of 11.78" C. and a molecular freezing point depression of 4.54" C., determined with benzil. The freezing points were measured with a Beckmann thermometer graduated in 0.002' C. and could be determined to 0.0005~c.
methods such as the ultracentrifuge. However, because of the comparatively small range over which the viscosity equation is being applied in the case of vinyl chloride-acetate resins, it would probably give substantially correct values. The viscosity measurements were made in a modified Ostwald viscometer ( 5 ) . It was so constructed that dipropyl ketone, the solvent used, had a running time of over 200 seconds a t 20 O C., the temperature for all measurements. The resin solutions were kept a t constant temperature for a t least 8 hours and were filtered through sintered glass before measuring. Individual viscosity measurements could be checked to 0.2 per cent, and the molecular weight values to TABLEI. CRYOSCOPIC MOLECULAR WEIGHTDETERMINATIOSS about 3 per cent. The molecular weights calculated from the ON Vrsl-i, CHLORIDE-ACETATE RESIN viscosity equation were found to be independent of concentraFreezing tion up t o specific viscosities of about 0.35. Methyl amyl Point Mol. Fraction Resin Dioxane Devression Weight Knr X 10-4 ketone was also used as a solvent, and the relative viscosities Grams Grains of the resin solutions were found to be the same in both 1 0.545 49.45 0.036 1400 6.1 'olvents. 0.063 1600 5.3 1.004 44.15 2 3 4
-
0,889 1.072 0.957 1.832 1.07i 1.390 1,583
18.50 48.63 44.81 44.87 43.80 44,82 4 4 . ?P
0.021 0.027 0.0165 0.0325 0.015 0.0185 0.026
4000 3700 5900 5700 7500 7600 6300
3.3 3.6 3.4 3.6 3.3 3.3 4.0
__
-
Determination of Polymer Distribution
__
Molecular Weight Determination from Viscosity Measurements The cryoscopic method is satisfactory for determining molecular weight on only a small percentage of the polymer present in most vinyl chloride-acetate resins. Tiny1 compounds are generally believed t o polymerize to form linear macromolecules. Staudinger ( 2 ) found the following enipirical relation between the molecular weight of such linear or threadlike molecules and the viscosity of their dilute jolutions: N s p = CKmM where N s p = N r - 1 Nr = relative viscosity, or viscosity of solution divided by that of solvent C = concentration, baae moles/liter K m = a constant M = mol. weight "Base molecular weight" nieans the molecular weight of the recurring unit in the polymer, such as 62.5 for polyvinyl chloride and 86.0 for polyvinyl acetate. I n applying thib equation to the vinyl chloride-acetate co-polymer, a base molecular weight was calculated from the analysis of the resin irhich is usually about 87 per cent vinyl chloride. The constant Km must be determined from some independent measurement of molecular weight. The Km values given in Table I were calculated by this equation from the cryoscopic molecular weights and the viscosities of dilute solutions of the fractions. The larger Jralues for Km found for fraction 1 may be due to the fact that it was more difficult to purify than the others. On precipitation, the resin tended to come out of solution as a soft gum rather than as a powder. However, Staudinger (4) reports an increase in the Kni values for polystyrene a t low molecular Tveights and attributes it t o an increasing effect of the end groups as the molecules become smaller. The value of 3.4 X for Km is considerably larger than those that Staudinger (3) reports for styrene and respectively. A detervinyl acetate, 1.8 and 3.6 X mination of this constant for polyvinyl chloride would be desirable as a check but is not yet available. Kraemer and Van Xatta (1) investigated the applicability of Staudinger's viscosity equation for determining the molecular weights of high polymers and report fairly good agreement between viscosity and chemical determinations of niolecular weight in the range below 25,000. The determinations reported here have not been checked by any independent
Although an average molecular weight can be obtained on any sample of resin by the viscosity method, it tells little about the polymer distribution of the resin unless it has first been carefully fractionated so that molecular weights can be determined on each fraction. Such a fractionation is much too laborious for use as a routine test; hence an attempt was made to develop a fractional precipitation procedure in which the amount of resin precipitated could be determined from a total solids measurement on the supernatant liquor without actually having t o remove the precipitated fraction from the solution. Acetone was selected as a solvent and an acetonewater mixture as precipitant, although other solvents and precipitants could be used. The procedure was simply to dissolve the resin sample, add a certain amount of precipitant, and withdraw a 10-cc. sample of the supernatant liquor from which the amount of resin still in solution could be determined by evaporating to dryness. I n case the precipitated resin did not settle readily, a portion of the liquor was centrifuged and the sample taken from it. More precipitant was then added, and another sample taken. The exact procedure for securing the five fractions given in Table I1 was to dissolve 15 grams of reyin in 900 grams of dry acetone and precipitate by adding successively 86.0, 29.0, 30.0, and 80.0 cc. of a 50-50 (by volume) acetone-water mixture. The progress of the precipitation could be followed in this way and check results obtained if the procedure was carried out in a closed vessel a t constant temperature and with thorough stirring. TABLE11. EFFECTO F RESIN CONCENTRhTION O N POLYMER DISTRIBUTIOX DETERMINATIONS Hesin Concn.,
R
10.0 5.25 2.70 1.64 1.10
Mol. weight range
I
Per Cent Precipitated Fraction 1 Fraction 2 Fraction 3 Fraction 4 Fraction 5
?
0 0 0 5 6 20,000-16,000
57 46 35 28 26 16,000 12,000
25 29 31 30 23 12,0008,000
13
1s
24
2:3 21
8,0004,000
5 7 10 14 24
Belov 4,000
The principal source of error was the mutual solubility effect of neighboring polymer bands. The data in Table I1 show this effect over the range of concentration between 10 and 1 per cent resin; the fractional precipitation procedure and hence the amount of precipitant present were exactly the same in each case. The results show that low polymer is precipitated with the high polymer in concentrated solutions; a slight solvent action is also exerted, as shown by the appearance of a small first fraction in the dilute solutions. Although there is still some change in polymer distribution with resin
INDUSTRIAL Ah D ENGI \ EERI 2 (; CHE 11IST K\I
1151
concentration in the nio-t tlllute duti(iiis, It n a< iountl dewable to carry out the aiialysia 111 tliiq concentration range. n ith the limiting value determined by t' e amount of resin necei5ary to give reproducible total-solids determinations. This optimum concentration n a s found to be about 1.G per cent A 50-50 acetone-water mixture appears to be the moqt satisfactory precipitant. K i t h inixtiires containing a inuch larger percentage of acetone, the dilution is so great a hen the last fraction is precipitated that it is difficult to determine accurately. With less acetone, the amount of precipitant required for the various fractions becomes so siiiall that the experimental error is increased. I n order to determine the actual iiiolecular weight liantls corresponding to a particular fractionation procedure, the polymer composition was determined in detail for tn o re4ns by means of a repeated extraction process in which the resins were each split up into about twenty-five fractions Each fraction was isolated and its molecular weight determined by a riscosity measureinent. JTith this information it was possible to determine the polymer distribution of these resins within about 5 per cent for each band of 4000 in molecular weight, based on the composition of the resin as a XT hole. The amounts of precipitant used in the precipitation procedure were then adjusted to give this polymer distribution in five fractions of 4000 spread each for one of these resins (number 1, Table 111). The same procedure gives an analysiq for resin 2 differing from that obtained by the reperited eytraction proce+- onlu in the first two fraction. TABLE 111. POLYMER DISTRIBTTIOS D.Y~.L Resin S o . 1 (calcd.) 1 iexutl.) 2 icaicd.) Z (exgtl.1 , Blen 1 (calcd) Blend 1 (exptl.) Blend 2 (calcd.) Blend 2 (exptl.) Blend 3 (calcd.) Blend 3 (exptl.) Blend 4 (calcd.) Blend 4 (exptl.)
Fraction 1 Fraction 2 Fraction 3 Frartioii 1 Fiaction 5
%
%
%
%
%
6 5 40 31 21 22 17 10 36 33
25
29 30 14 15 23 23 29 29 16 16 5 10
22 22 12 10 16 16 21 21 11 12 55
18 14
13
8
28
30 38 27 28 26 :33 30 :34
8 10
52
1 6 13
11
Y #
5
19 90
The degree to which it is possible to calculate the polynier distribution of a blend from that of its components is also illustrated in Table 111. I n general, repeated determinations on the same resin give results differing by not more than 3 per cent for corresponding fractions, based on the composition of the resin as a whole. The calculated and experimental value:: for blends usually agree within 5 per cent; where the difference is more, it is almost always compensated in the next fraction rather than distributed over all the fractions. The greatest difference between calculated and experimental TABLE Iv.
PHYSICAL PROPERTIES OF I'IXYL
Original Resin 10,600 Mol. weight 85.9 Tiny1 chloride content, % 1.526 Refractive index Brinnell hardness No. a t 2.5-kg. (5.5-1h.) ioad 14.2 TT-ater absorption a t 60' C. (140' F,),% gain in n-eight: 18 hr. 0.39 1.30 168 hr. Heat distortion point: 58 c. F. 136.4 Yiscosity of 2OY0 soln. in methyl isobutyl ketone, 190 centipoises Plasticity a t 120° C. (248O F.) b y Scott piastometer, 76 Tensile %. strength, lb./sq. in. 9,025 Impact strength ft.-lh./piece 0.25 Fatigue r e s i s d c e a t 2400 lb./sq. in., cycles 26,000 Modulus of elasticity lb./sq. in. 400,000 Modulus of rupture, lb./sq. in. 13,300
I
101' 28,
\o,
\-slues occurs when two resins of very different cnmposition tire mixed-for instance, resins with average niolecular weights of 17,000 and 6000 (blend 4, Table 111). The tn-o .-eparate peaks which should appear in the polymer distribution curve tend to be pushed together. S o method has been found which entirely avoids this difficulty, but in general t'he agreement is satisfactory. Obviously 110 method hasetl 011 such alight differences of solubility can give extreme accuracy, but the general picture presented by such an analysis is believed to be suhstantially correct. Polymer distribution data for two reFin4 are shown graphically in Figure 1 and illustrate the type of polymer tlistrihution usually found in vinyl chloride-acetate resin. The higher the average degree of polyiiierinatirrii in a resin, the less is t,he difference in solubility between neighboring polymer bands, with the result that its polyriier distribut'ion cannot be obtained in much detail. Polyvinyl chloride and polystyrene are usually so highly polyinerized that 7 5 per cent or more of the resin appears as a single high fraction which cannot b e fiirther separated on the basis of difference in solubility. Coniiiiercial grades of poly.\-inyl acetate usual1.v have a coiisiderably lower arerage degree of polyriierization, and this resin can be anslyzed by a proredurr similar to that described.
Dependence of Physical Properties on Molecular Weight K t l i a niethod available for tleteriiiiriing t,he polymer distribution of the resin in considerable detail, it becomes important to know how the physical propert'ies vary with molecular lyeight. I n order to obtain this information, a quantity of resin was separated into six fractions by a fractional precipitation procedure. It v-as dissolved in acetone, and approximately one-third was precipitated by the addition of isopropanol. This precipitate was filtered out. and enough inore isopropanol was added to precipitate another third of the resin. The three fractions so obtained were then each tlivided in half hy the same method, so that six apprcrsimately equal fractions were obtained from the original resin. They \yere then each redissolved and trimmed by fractional precipitation to narron- the polymer bands further. However. because of the slight difference in solubility betwern polymer honiologs, there is undoubtedly some owrlapping iii the six fractions. The inolecular iyeights and physical prol)ertie%or' the*[: six fractions are given in Table IT. Four of the properties listed-namely, heat distortion, vater absorption. refractive index, and Brinnell hardness-are practically iiiilependent of niolecular weight for fractions with a molecular weight above 5000. The heat distortion of the original resin is 1" C. less t,lian that) of any of the fractions, and the wat'er alworption of
CHLORIDE-*kCET.iTE
RESIXF R . ~ C T I O X ~ _.
Fraction
B
C
15,800 86.4 1.530 13.9
14,600 86.6 1.526 13.5
13,200 85.7 1.524 13.2
11,100 83.1 1.523 12.7
7,500 84.8 1.523 13.2
0.27 0.44
0.40 0.47
0.37 0.78
0.31 0.56
0.29 0.54
0.34 0.86
63
66 150.8
63.5 146.3
6% 143.6
145 8" 9,180 0.31 360,000 403,000 13,500
15 94 6,680 0.16 18,000 394,000 9,100
A
10
64 147.2
65 149
Gel 39 9,955 0.36 940,000 418,000 14,400
Gel 52 9,745 0.33 1,000,000 407,000 14,600
D
145.4 370
66 9,630 0.32 1,000,000 401,000 13,000
E
F .i.800 83.9 1.524 Fractured
24
Could not mold 1,120 Could not mold Could not mold 350,000 1,470
OCTOBEH, 1936
INDUSTRIAI, AhD ENGINEERIYG CHEI1IISTRI-
tlie original resin about doiible that of the fractions. This difference in heat distortion and v-ater absorption vias due to the presence in the original resin of 10 to 13 per cent of very low polymer, ranging in molecular weight from 1000 to 4000, 15-hich n-as lost in the fractionation procedure. The presence of impurities such as cntalyst or monomer T v O U k ~also affect these properties, h i t these impurities are not likely to be present in ,sufficient quantity. The other physical properties listed show a pronounced variation JT-ith molecular weight. Tensile strength, impact strength. fatigue resistance, and the moduli of rupture and elasticity all show the same SfJrt of variation-that is, a rapid increase between molecular weights of 5000 and 8000 followed by a much more gradual increase above 8000. The average molecular weight of the highest fraction in Table I1 is 15,800, but fractions which show an average molecular \%*eight as high as 19,000 have been separated from other resins. Such frnctions show >very little change in physical properties
1155
froni fraction- A and B, except t'hat the plasticity is definitely higher. To summarize, these data show a rapid increase in mechanical strength with increasing niolecular weight u p to about' 8000, f o l l o ~ e dby a more gradual increase in the higher niolecular JTeight range. Similarly the heat distortion increases rapidly and the wat,er absorption decreases rapidly with increase in molecular weight up to about 4000; then they change more gradually.
Literature Cited (1) Rraerner and Van N a t t a , J.P h y s . Chem., 36, 3175 (1932). (2) Staudinger, "Hochmolekulare organische Verbindungen," p. 56, Berlin, J. Springer, 1932. (3) Ibid., p. 135.
(4) Ibid.,g . 179. ( 5 ) Willihnganr, McCluer, Fenske. and McGrew, IND. ESG. C H E h f . , Anal. Ed., 6, 231 (1934). RECEIVED September 2 , 1936.
Polymerization of Vinyl Acetate K . G . BLAIKIE AND R. N . CROZIER Shawinigan Chemicals Limited, Shawinigan Falls, Quebec, Canada
S
I S C E the discovery of vinyl acetate in 1912 (9, 16) numerous papers on its preparation (3, 8) and polymerization (3) have appeared. The present paper is an outcome of work carried out with a view to placing on a commercial basis the manufacture of polymerized vinyl acetate. It has been found that the factors affecting the polymerization and viscosity of the resulting polymer are impurities, temperature, solvents, catalyst, and percentage conversion. The polyvinyl acetate resins described are characterized by a number calculated from the outflow time, through 811 Odnald viscometer, of a standard solution in benzene:
T7 = = )SI = Ti = Db = TB = I'b
viscosity of polymer, centipoises absolute viscosity of benzene at 20" C. density of solution of polymer at 20" C. outflow time of solution at 20" C., seconds density of benzene at 20' C., seconds outflow time of benzene at 20" C., seconds
Impurities Polymerizations involving chain formation are extremely sensitive to promoters and inhibitors. Tiny1 acetate is no exception; thus it is necessary, especially for commercial production, t o pay particular attention to the purity of the ingredients as well as to the standardization of the conditions under which polymerization takes place. A good grade of commercial vinyl acetate contains over 99.9 per cent ester; the only impurities are traces of water, acetic acid, and acetaldehyde. The first two do not have much effect but acetaldehyde influences the viscosity t o a marked degree. The experimental method throughout has been to heat the material under reflux or in sealed glass tubes. Concordant results are obtained provided the time of reaction and the temperature are comparable. The effect of acetaldehyde is shown in the following table:
60 40 0.0II 4.5 66-70
Vinyl acetate, parts Benzene, parts Catalyst, part Reflux period a t 73-78' C . , IIOUE 70conversion Acetaldehyde
Viscosity
% 20
0.1 0.2 0.6
15
6
I n addition t(J these impurities, certaiii materials which actively inhibit the polynierization can be picked up by the ester and solvent unless precautions are taken. Chief among them are sulfur and copper compounds. Thus solvents must not, contain sulfur compounds, and tanks and pipe lines must not be assembled Fith gaskets containing sulfur. I n order to prevent polymerization in stills and columns, copper equipment is used : but as it is imperative that none of this material shall appear in the finished product, vapor lines and condensers should be constructed of aluminum or stainless steel. Cold Vinyl acetate, containing less than 0.03 per cent acetic acid, does not corrode iron, so that t>hismetal can be used for storage t,anks.
Temperature An increase in the temperature of polymerization has been shown to lower the viscosity of the product. The following table provides further evidence of this: Vinvl acetate, parts 90% butyl acetate, parts Catalyst, part Temp. Conversion
=
c.
70 75 80
70 30 0.07 Viscosity
% 68 70 65
23
19.5 16.5
Solvents The specific effect of variuu.' -01vents has been noted by several observers ( 1 1 , I J ) . Tables I and I1 a h o r clearly that not only the particular soh-ent, but also its concenhas a marked effect on the
