INDUSTRIAL AND ENGI.VEERING CHEMISTRY
July, 1931
Acknowledgment
791
Literature Cited
The writers wish to acknowledge the help of R. D. Jones and Francis Scofield in securing the data for Table 111, and also the kindness of the Archer-Daniels-Midland and Wm. 0. Goodrich companies for permission to publish these results.
(1) Eibner. "Das Oeltrocken," Berlin, 1931. 53 (1931), (2) Long and ,-hatawmy, IND, ENG. CHEM,, (3) Long, gittelberger, Scott, and Egge, x d . , ai, 950 (1929). (4) Morrell and Marks. J . Oil C O h 4 7 Chem. Assocn., 12,183 (1929).
Viscosity Increase and Gelation in Phenolic Resin Varnish Cooking' V. H. T u r k i n g t o n , R. C . Shuey, and W. H. Butler BAKEIJTE CORPORATION, BLOOMFIELD, N. J.
HE changes that occur in the physical properties of China wood oil when it is heated alone or with various resins are of great importance in varnish making. However, published numerical data on which to base the choice of proper temperatures and times for securing desired results have been rather meager. It is common practice to determine the rates of gelation of different lots of China wood oil by the Brown or Worstall tests (S), after which the variations in treatment which will probably be necessary are found by experience, final determination of the end point being made by drip, string, or pill formation of a sample withdrawn during the heating. The advent of synthetic resins has brought many surprises to the varnish maker because of the suggested departures from old practices during cooking. Therefore, methods of determining by experiment the optimum conditions of treatment of tung oil with natural and synthetic resins are desirable. For the purpose of showing these effects on the oil bodying, the following two methods were developed in separate laboratories and are presented here, together with discussion of the two sets of data:
T
METHODOF TEsTmG-The oil bath is heated to the required temperature, which is then adjusted and held constant by means of a pilot burner fitted with a needle valve. The test tube containing 1.5 grams of the sample, with a glass rod inserted in the tube, is then immersed in the oil bath and allowed to remain there until the tung oil forms a gel strong enough for the test tube to be lifted out of the apparatus with the glass rod. The time required is determined by raising the glass rod a t short intervals after thickening has commenced. At 200" C. there is a gradual thickening, for about 15minutes, before the gel structure is strong enough to support the test
--cone
(1). A development of the Brown test (3),which gives comparative numerical values t o the restraining or accelerating effects of the resin on the final gelation of tung oil. Variations produced a t various temperatures and concentrations are shown. (2) Use of the DeVilbiss viscometer ( 2 ) for recording the increase in viscosity of tung oil with natural and synthetic resins during heating a t a constant temperature b y placing the viscometer in the kettle during cooking. The numerical data were plotted for comparison and later gas check tests ( 4 ) were made upon the varnishes a t different stages in the cooking. Gelation Tests
APPARATUS-The apparatus is a modification of that commonly used for the Brown test. The following changes are made: A bath of mineral oil is used in place of cottonseed or soy-bean oil, the total volume of mineral oil is increased, the bath is fitted with a stirrer, and the unit is insulated with asbestos. The size of the test tube is reduced from 15 cm. x 16 mm. to 7.62 cm. X 9.51 mm., and the sample from a volume of 5 cc. to 1.5 grams. These changes are made in order to obtain a more uniform temperature of the oil bath and sample. With this modified apparatus it is easily possible to attain the desired temperature quickly and to hold this temperature constant within * 1" C. PREPARATION OF SAMPLE--A sample of approximately 100 grams is prepared by weighing in a 250-cc. Erlenmeyer flask the resin and the tung oil; it is heated to 180" C. in 15 minutes and then cooled to room temperature and stoppered. The temperature and time are arbitrarily chosen as being sufficient to effect solution of high-melting resins in the tung oil. Received April 9, 1931
tube, but a t 225" C. and above the gel structure forms rapidly. A thin film of oxidized oil forms on the surface of the oil and this can be easily broken by a slight movement of the glass rod. D A T A - T ~ curves ~ were plotted from the average of two or more determinations made with each sample. The results agreed within 1 minute a t temperatures to 250" C., and within 30 seconds or 1-ss at higher temperatures. Table I
INDUSTRIAL AND ENGINEERING CHEMISTRY
792
2000 c. Min. 265
RESIN TUNG OIL Parts Parts Tung oil only Bakelite XR-254:
Table I-Gelation Data 2250 c. 237' C. Min. Sec. Min Ser. 86 30 55 0
.
250' C. Min. Sec. 34 50
Vol. 23, No. 7
*
275' C. Min. Sec. 14 30
300' C. Min. Sec. 8
0
61 9
30 45 27 50
~
I
I
1 2 1 4 1 8 Ester gum: 1 2 1 4 Congo: 1 2 1 4 Bakelite XR-254 and ester gum: 1 2 Bakelite Ester gum 75 25 50 50 20 80 10 90 Bakelite XR-254 and ester gum: 1 4 Bakelite Ester gum 75 25 50 50 80 20 10 90 Rosin: 1 2 1 4 Limed rosin: 1 2 1 4
93
39 41 44
22 52 50
40 27 23 27
25 32 45 55
33 12 11 9
30 45 2
30
0
83 59
45 37
52 37
35 30
29 18
42 17
95 85
0 0
67 53
5 37
44 32
40 45
22 15
35 42
73 89 111 103
30
51 61 73
30 1.5 30
30 34 47 41
12 22 30 52
15 19 23 21
30 10 57 45
29 30 34 37
37 5 30 55
13 13 16 16
47 53 52
182 207 227
70 76
15
257 220
116 87
217 198
68
74 73 94 93
30 0
0 30
15 0 30
4s 49
shows the average of these results. Table I1 shows the properties of the resins. Table 11-Roperties RESIN
ACID
NUMBER
of Resins SPECIFIC GRAVITY
MELTING POINTa
c.
Ester gum 4.8 1.090 87.5 Run Congob 67.4 1.0584 101 Bakelite XR-254 92.0 1.220 100 Rosin 165 80 Limed rosin 85 105 Modification of the A. S. T . M. ball and ring method. b Congo resin used in these tests was run to 610' F. (321' C.), held at this temperature 2 minutes, and then poured out t o cool.
The curves plotted from the results (Figures 1 to 3) with various types of resins, using 1 part of resin to 2 parts of tung oil. show: (1) The net effect of Bakelite resin XR-254 is that of acceleration, whereas Congo and ester gum retard final gelation of tung oil. (2) The accelerating effect of the phenolic resin is greatest a t 200' C., and a t this temperature Congo also has a slight accelerating effect on the final gelation of tung oil. (3) Tung oil with ester gum shows no tendency t o change to the solid phase within 5 hours at 300' C . ; however, with 10 per cent of Bakelite resin XR-254 the tung oil reaches its final gelation in 28 minutes a t 300" C. This shows markedly the effect of a low percentage of this phenolic resin in accelerating the final gelation of tung oil in the presence of a resin, such as ester gum, which itself actually prevents gelation a t 300" C.
I n Figures 4 and 5 are shown the effects of the same resins with tung oil, using 1 part of resin to 4 parts of tung oil. Comparison of these curves with Figures 1,2, and 3 shows: (1) The general effect of the resins on the final gelation of tung oil remains in the same order as in Figures 1, 2, and 3. (2) As the resins are diluted further with tung oil, they have a decreasing, restraining, or accelerating effect on the final gelation of tung oil, but the general direction is that of acceleration.
Figures 6 and 7 show the effects of Bakelite resin XR-254 in comparison with the effects of mixtures of this phenolic resin and ester gum on the final gelation of tung oil. I n Figure 6, using 1 part of resin to 2 parts of tung oil, the results show: (1) The final gelation of tung oil is retarded as the Bakelite
37 37
50
6
5
5 hrs.; no gelation 23
30
28
15
3 hrs.; no gelation at 250°, 275", or 300' C. 54 30 33 34 46
0
3 hrs.; n? gelation at 250°, 275', or 300' C. 53 20 47 0 75
0
0
resin XR-254 is decreased and the ester gum increased, the curve for the phenolic resin being on the extreme left, and the curve for the ester gum on the extreme right, with mixtures of these resins in between. (2) The mixture containing 10 per cent of Bakelite resin XR-254 is faster than the 20 per cent mixture, but the writers can offer no explanation for this phenomenon a t present. A similar break was observed in a curve plotted from the results of hardness tests of fdms (7) taken from a similar series of varnishes. (3) .In Figure 7, as also in Figures 4 and 5, the resin with increasing dilutions of tung oil is shown to exert a lesser influence on the restraining or accelerating of the final tung-oil gelation, and the general direction is that of acceleration.
Figures 8 and 9 show the effects of varying proportions of Bakelite resin XR-254 on the final gelation of tung oil. (1) In Figure 8 it is noticeable that this phenolic resin accelerates the final gelation of tung oil, in proportions of 1 t o 2, 1 to 4, and 1 to 8. (2) At 225" C. the accelerating effects increase as the resin increases, and there is a reversal of this effect a t 275' C. (3) In Figure 9 this resin with tung oil in the proportion of 1 to 2 has a slight retarding action on the final gelation of tung oil a t 300" C., and this restraining effect becomes very noticeable when the ratio is 1 part of resin to 1 of tung oil.
Tests were also made with rosin and limed rosin, but the curves not included here. These resins with tung oil in the proportion of 1 to 2 will hold and prevent tung oil from forming a solid gel a t temperatures of 250°, 275", and 300" C. within 3 hours. On further dilution with tung oil and in the proportion of 1 to 4, they have more of a restraining effect than ester gum on the solidification of the tung oil. However, the tung oil forms a solid gel a t 250°, 275", and 300" C. There have been varying opinions as to the optimum temperature to be used in making varnishes from tung oil with different resins, especially rosin and ester gum. The bodying of the tung oil is generally terminated just short of the final gelation point, and in order to keep it from going to the final insoluble stage it is checked with a portion of the gum, a specially prepared oil, or a thinner. A brief comparison of the time-temperature curves on the final solidification of the tung oil with the resins tested in various dilutions with tung oil will indicate why it is necessary to use a high temperature in order to obtain gel structure and body with rosin and ester gum, and why lower temperatures are advisable with Congo and the Bakelite resin XR-254.
July, 1931
INDUSTRIAL AND ENGINEERING CHEMISTRY
793
794
INDUSTRIAL AND ENGINEERING CHEMISTRY
Viscosity Increase of Tung Oil with Different Resins and Varying Oil Length
APPARATUS--A heavy copper restaurant stew pan, 20 cm. wide, 20 cm. high, bottom 0.4 cm. thick; a single-burner kitchen hot plate drilled and adjusted to give a clear flame on either very low heat or full blast; armored thermometer; watch or other time piece showing seconds; a DeVilbiss viscometer operated on 110 volts, 60-cycle alternating current; pint or half-pint cans for samples withdrawn to be cooled or diluted for gas-check tests; block-tin collapsible tubes, 2.54 x 15.3 cm., for preservation of cooled samples out of contact with air. METHOD-The time of heating to the boiling temperature of 232" C. (450' F.) was arbitrarily chosen a t 25 minutes as a
Vol. 23, No. 7
rate of heating easily reproduced in the factory. Practice soon enabled the operator to adjust his burner so that this temperature was generally reached within 5 minutes of the desired time and maintained within 2' C. of this temperature thereafter. This temperature was found to give ample spread to the readings without consuming an excessive amount of time. A total batch weight of 3000 grams was chosen as comfortably filling the kettle and allowing for generous sampling without danger of the level falling too low for accurate operation of the viscometer. I n all cooks an amount of cobalt linoleate paste equivalent to 0.2586 gram of cobalt was added for each 1000 grams of oil, in order to insure a uniform amount of drier in all the samples for the later comparisons. No effect of this slight addition could be observed in the timeviscosity curve. At frequent intervals the viscometer was inserted in the kettle and, after the voltage had been adjusted to 100, the current consumption was noted. The taking of instantaneous readings in this manner a t the temperature of the cook is a great advantage, because it does not introduce errors incident to removal of a sample to be dissolved or cooled for external determination. These intervals were 5 or 10 minutes in the early stages of the cook, but were reduced to 1 minute or less toward the end, when the viscosity increased rapidly. The experiment was generally terminated a t the point of incipient solidification-that is, beyond the normal cooking viscosity. At longer intervals samples for gas-check tests were withdrawn and, after closing, the cans were placed in running water to cool. As soon as they were cool enough to handle, a collapsible tube was filled with the material, care being taken that no air was allowed to remain in the tube during closing. CALCULATIONS AND CORRECTIONS-with the instrument is furnished a graph converting the milliamperes current
July, 1931
INDUSTRIAL A N D ENGINEERING CHEMISTRY
795
dispersion in the diluted oil to delay or stop gel-structure aggregation.
consumption a t 100 volts directly to centipoises, a t specific gravity 1.OO. The readings, after conversion to centipoises, were corrected for the specific gravity of the cook a t room temperature, thus putting all cooks on an equivalent basis irrespective of resin content or specific gravity of components. No correction factor for specific gravity change from room temperature to 232" C. was used, as the change was assumed to be approximately the same in all cases. A correction of this sort would add about 25 per cent to all viscosity values. The formula used for specific-gravity correction was Viscosity value a t 232" C. = approximate absolute viscosity (Specific gravity a t 25" C.)z
The plotting of these values against the time elapsed since reaching 232" C. gave curves approximately logarithmic, but with indications of a break generally around 25 centipoises. New graphs were made on semi-logarithmic paper, thus plotting the logarithm of the viscosity against time. I n these graphs an attempt was made to consider the change truly logarithmic and draw the portions in as few segments of straight lines as possible. The maximum number of type segments found was four, but all were never found in one graph. They were designated as follows:
(t) Region in which viscosity remained stationary with lapse of time (static). (2) Gradual increase in viscosity (assumed t o be thickening without gel formation). (3). Rapid increase in viscosity (assumed to be formation of gel dispersed in the oil). (4) In short-oil varnishes a retardation of the rate of viscosity rise, apparently due to dilution by the resin, or t o sufficient
From a glance a t the logarithmic graphs such a separation may seem rather daring and the assumptions unwarranted, but allowing the widest possible latitude to the choice of locations of these lines would effect but little change in an arithmetic plotting of such lines. That is, in the region of very low accuracy (around 10 centipoises) the presence of a horizontal (static) line would not be discernible on the arithmetic plotting. It is probable that there is no static delay of bodying, but that the increase is not truly logarithmic or the instrument not sensitive enough to show the small differences. However, the so-called thickening-to-gelation break is certainly present and quite significant. The arithmetic graphs follow the curves indicated by the logarithmic graphs to show the departure of the observed points from a truly logarithmic curve. The slope of the logarithmic straight lines was calculated and stated in minutes required to produce a tenfold increase in viscosity. Also the location of the intercepts of the lines was recorded in elapsed time from and a t 232" C. with corresponding viscosities. Comparison of Data by Two Methods
For comparison with the gelation figures obtained by the first method, the time required to reach a viscosity of 200 centipoises is given. This is about as far as any regular production cook would be carried for 300 centipoises or more would be the last obtainable reading before complete gelation. Even if the 300-centipoise value were given (it was not reached in most cases), actual agreement would not be expected, for the character of the gel formed and the amount that attaches
Vol. 23, No. 7
INDUSTRIAL AND ENGINEERING CHEMISTRY
796
itself to the rotating disk cause wide variations in readings as soon as visible gel structure forms. I n choosing a temperature for making the comparisons, three boilings of the same oil were made-at 260' C. (500" F.) , 232' C. (450' F.), and 204.2' C. (400'F.). Theseareshown iniFigure 10 on arithmetical coordinates and in Figure 10-L with the viscosity on logarithmic ordinates, to show the method of locating the breaks in the curves. The values are given in Table 111. It was decided to use the temperature of 232' C. on the major portion of this work because it allowed
of Viscosity Increase of Tung Oil with Different
T a b l e IV-Rate
Resins (Temperature, 232' C.) TIME FOR
RESIN
(a)
Rosin Ester gum Limed rosin Bakelite XR-254 Tung oil only
(b)
of Viscosity Increase of T u n g Oil a t Different Temperatures TIMEFOR
TOTAL
TENFOLD TIME
TIME
VISCOSITY
AT-TEMP. INTERCEPT INCREASE TO REACH
POINT
AT
80 68.5 73 40.5 54
1 PART RESIN
27 30 28 27 (6)
Rosin Ester gum Limed rosin Bakelite XR-254 Tung oil only
Table 111-Rate
TIME
1 PART RESIN TO 4 PARTS TUNG OIL
26 25 21 23 25
Ester gum Congo Bakelite XR-254 Tung oil only
ample time for taking readings, even during the gelation period, without requiring too long a total time for the cook. For the sake of brevity no tabulation of individual readings is given, since the observations can be read directly from the graphs within the accuracy of experimental error.
TOTAL
TENFOLD AT TEMP. POINT TIME VISCOSITYTO REACH AT WHICH TO INTERCEPT INCREASE VISCOSITY VARNISH REACH THICKENINGDURING OF PASSEDGAS 232' c. TO GELATION GELATION200 CP. CHECK TEST Min. C p . Min. Min. Min. Cp.
94 65 43 65
26 22 23 22.5 26
108 69 80 35 54
95 82 99 58 66
95 82 99 15 66
182 203 200 10 203
123 86 10 77
278 194 13 221
95 92 204 10 66
20 175 161 15 203
TO 4 PARTS TUNG OIL
47 25 31 30
1 PART RESIN TO
25 25 21 24 25
17.3 13.5 27.5 17 13.5
28 18 21 33 26
36 23.5 25 14
a
117 87 63 76
PARTS TUNG OIL
36 23 77
27.5
13.5
139 93 155 57 66
Run Congo has less effect than either of the rosin products, but the net effect is that of restraint. Bakelite XR-254 decreases the thickening time noticeably (about as much as ester gum increases it) and then exerts a slight restraining effect during gelation, leaving, however, a net result of acceleration, which seems to be a property unique to the phenolic resins. Increasing the ratio to 1 part resin to 2 parts of oil, (approximately a 25-gallon varnish) gave similar results except for an increase in spread due to the higher percentage of resin. These values are shown in Table IV (c) and Figures 13 and 13-L. The two extremes of resin performance with change in concentration are shown in Table V and Figures 14-L, 15, and 15-L. I n the case of Bakelite XR-254 variation from straight-oil to a short-oil varnish shows progressive diminution of the thickening time followed by increasing restraint of the gelation rate, so that, with the exception of the 1:l varnish, the net result is always acceleration. The probable
TO WHICH VARNISH REACH THICKENINGDURING VISCOSITY PASSEDGASTEMP, TEMP. TO GELATION GELATIONOF 200 CP. CHECK TEST a C. Min. Min. Cp. Min. Min. Cp. 260 23 23 23 50 sec. 24 24 ZOO+ 232 36 74 25 22 min. 92 94 214+ 26 210 23 63 min. 268 269 216+ 204.2
For the purpose of comparing the bodying time of oil with different resins, cooks were made using 4 parts of another tung oil to 1 part of each of the various resins. These varnishes are shown in Table IV (a) and Figures 11 and 11-L, respectively, for logarithmic and arithmetic plotting. For the purpose of including Congo, a graph previously made with a different lot of oil is shown in Table IV (b) and Figures 12 and 12-L. This series is of a lower order of accuracy than the remainder of the work. The makers of the viscometer do not advise its use with the present disk for values below 10 centipoises, and in this series the bottom bearing of the motor was not freshly lubricated. I n the rest of the work this was done, with a consequent improvement in accuracy, especially in the low values. Although rosin and limed rosin appear to be almost equally efficient in holding or restraining the oil, it is also apparent that the effect of the liming has been largely manifested in restraining the gelation, slowing it from 13.5 minutes to 27.5 minutes for a tenfold increase with only 7 minutes' change in the time of thickening. Ester gum, on the other hand, exerts its effect on the thickening, with less change in the gelation rate, as most varnish makers are aware.
significance of this acceleration of the thickening will be discussed later. It is apparent that the various concentration curves tend to converge to a point somewhere above the highest reading. The maximum departure from the time of the oil at 200 centipoises is only 9 minutes, or 14 per cent. In the 1:l varnish there is no upward break and, although the curve starts out logarithmic, retardation of a new order sets in after 53 minutes and the time required for a tenfold increase changes from 44 to 78 minutes. For the lack of a better term this is called "dilution retardation." This break causes the curve to cut all the others and is the only case of net delay.
INDUSTRIAL A N D ENGINEERING CHEMISTRY
July, 1931
Table V-Rate of Change in Viscosity of Tung Oil w i t h Bakelite Resin XR-254 and Rosin and Variation in Oil Length (Temperature, 232' C.) TIMEFOR TOTAL TENFOLD TIME VISCOSITY AT TEMP. INCREASE TO REACH RESIN- TO INTERCEPT VISCOSITY OIL REACH THICKENINQDURING RATIO 232O C. TO GELATION GELATIONOF 200 CP. Min. Min. Cp. Min. Min. TIME
Tung oil onlv
25
54
26
50
POINTAT WHICH VARNISHPASSED GAS CHECKTEST Min. Cg.
XR-P64
BAKELITE
RESIN
26 22
13.5
66
66
15
A4
58
203+ 66
66
66 95 110 60
203+ lSZ+ 31 9
797
concentration finally forms a gel. He also states that during cooking the bromine number drops at first in a straight line, then somewhat more slowly, and finally remains constant. As long as the bromine number drops the viscosity rises but slowly, finally rising faster and faster. The rapid rise takes place when the bromine number reaches a minimum. The refractive index is also practically stationary during this rise.
7sa ROSIN
T u n g oil only 1:4
1:2 1:l
a
25
26 25 25
54 80 108 a
26 26
28 0
13.5 17.3 36 Gradual change 144 t o m
95
139 445+
Dilution retardation.
K'ow consider the case of increasing resin concentration at the other extreme-rosin. The 1:l varnish is again neglected, for in this case it would not gel a t all. Stated in percentage of time required by the oil, the 200-centipoise point was reached in 144 and 210 per cent in the 50- and 25-gallon varnishes, respectively. The thickening-to-gelation point was reached in 148 and 200 per cent of the oil time, showing distinctly how great a difference in treatment is required with a change in oil length. As a general rule it can be stated that, whereas other resins require noticeably different treatments at different oil length, the 100 per cent phenolic resin requires but little different treatment to reach a given viscosity from that demanded by the straight oil. This difference, however, is a reduction in temperature or time rather than an increase. There is a remarkable parallelism in the gelation slopes of the various resins at a given concentration. The most marked deviation from this is noticed as the effect of liming or saponification in the limed rosin.
Cutter ( 1 ) discusses the work of Slansky and Kohler ( 6 ) , and Ostwald (J), stating their conclusion that in drying oils untreated except for refining the quantity of disperse phase present must be small and of secondary importance. On the other hand, the presence of two phases in heat-treated oils is undoubted. These conclusions tend to confirm the assumption of the present writers that the lower viscosity portion of the curve probably represents chemical polymerization or condensation, while the later portion of rapid rise is physical aggregation or formation of gel dispersed in the other substances present. An excess of resin could thus act as a dispersing medium, and retardation would be expected. A very important difference in behavior of Bakelite resin varnishes as compared with varnishes made from other resins is shown in the last two columns of the tables. Whereas the oil alone 'or cooked with other resins (except in the very short oil ones) is subject to gas checking unless cooked to a high viscosity, the phenolic resin varnishes a t concentrations of 1:4 or shorter become gasproof before the thickening-togelation break in the curve is reached. This makes possible extremely short cooking when desired for making high gloss, very full bodied varnishes, which will build film thickness with a minimum number of coats. The writers are of the opinion that lack of gloss in long-oil varnishes has been due to the necessity for cooking to a high body to obtain gasproofness, with consequent gel formation. These short-cook varnishes have all the gloss common to short-oil varnishes and have no great tendency to thicken with addition of basic pigments, such as zinc oxide. Literature Cited
WOE(8) shows a graph of the bodying of linseed and tung oils. The tung oil, when replotted with the viscosity shown o n a logarithmic scale, indicates a similar break near 20 centipoises. He concludes that the gelation product of wood 'oil is not a single substance, and that the thickening must be considered as a reaction in which &n unknown product is formed (which may also be a polymerization product), which Is dissolved colloidally in unchanged oil and with increasing
Cutter, J. O., J . Oil Colour Chem. Assocn., 13, 66-80 (1930). Fawkes, C. E.,Paint, Oil Chem. Rev., BO, 90-1 (Oct. 23, 1930). Gardner, H.A.,"Physical and Chemical Examination of Paints, V a r nishes, Lacquers and Colors," pp. 430-2 (1927). Gardner, H.A,, Ibid., p. 466. Ostwald, Kolloid-Z.. 46, 136 (1928). Slansky, P.,and Kohler, L., Ibid., 46, 128-36 (1928). Turkington, V. H., Shuey, R. C., and Butler, W. H., IND. ENG.CHEY., 22, 1177 (1930). Wolff, H., Z . angew. Chem., 88, 729-32 (1924).