tensiometer, is thus not a completely reliable guide as to which surface active agent will allow evaporation of the water in the membrane with retention of the desalination and physical properties of the membrane. A second effect coulld be operating-namely, “coating” of the cellulose acetate so that the interface of the pore water and the cellulose acetate is changed, thus reducing the capillary shrinkage forces on the cellulose acetate. This could be part of the reason why glycerol worked so well and FC-170 did not (Figure 1). There must be some adsorption of the material by the cellulose acetate; many of the surface active agents tested appear capable of filling this role. T h a t the combination of glycerol and a surface active agent--e.g., Sterox DJ-can yield better results than either alone may be attributable to the coating action of the glycerol supplementing the surface tension effect of the surface active agent. A third effect should also be considered as part of the drying-Le., the surface active agent may act as a humectant. Conclusions
Water can be evaporated from modified cellulose acetate membranes with no loss in desalination or physical properties by soaking the membrane in a surface active agent before drying. Surface active agents representative of all major types can apparently be used successfully. Dry membranes have several potential advantages which could make the process worthy of larger scale study and experimentation.
Literature Cited
Bartell, F. E., Ray, B. R., J . A m . Chem. Soc. 74, 778 (1952). Harkins, W. D., Jordan, H. F., J . A m . Chem. SOC.52, 1751 (1930). Loeb, S., Sourirajan, S., Aduan. Chem. Ser. No. 38, 117 (1963). Lonsdale, H. K., Merten, U., Riley, R. L., J . A p p l . Polymer Sci. 9, 1341 (1965). Manjikian, S., Loeb, S., McCutchan, J. W., Proceedings of the First International Symposium on Water Desalination, Washington, D. C., Oct. 3-9, 1965, Vol. 2, p. 159. Merten, U., Ind. E r g , Chem. Fundamentals 2, 229 (1963). Riley, R . L., Gardner, J. O., Merten, E., Science 143, 801 (1964). Riley, R. L., Merten, U., Gardner, J. O., Desalination 1, 30 (1966). Sourirajan, S., Govindan, T. S., Proceedings of the First International Symposium on Water Desalination, Washington, D. C., Oct. 3-9, 1965, Vol. 1, p. 251. Vos, K. D., Burris, F. O., Jr., Riley, R . L., J . A p p l . Polymer Sci. 10, 825 (1966a). Vos, K. D., Hatcher, A . P., Merten, U., IND.ENG.CHEM. PROD. RES. DEVELOP.5 , 211 (196613). Washburn, E . W., Ed., “International Critical Tables of Numerical Data,” Vol. IV, McGraw-Hill, New York, 1928. Zisman, W. A., Aduan. Chem. Ser. No. 43, 1 (1964). Zisman, W. A., “Adhesion and Cohesion, Proceedings,” P. Weiss, Ed., p. 176, Elsevier, Xew York, 1962. RECEIVED for review June 26, 1968 ACCEPTED January 4, 1969
Acknowledgment
The authors thank .R. L. Riley for helpful discussions.
Work sponsored by the U. S. Department of the Interior, Office of Saline Water, Contract KO.14-01-001-767.
GLYOXAL FROM OZONOLYSIS OF BENZENE WILLIAM
P .
K E A V E N E Y ,
R A Y M O N D
V .
RUSH,
Central Research Laboratories, Interchemical Corp., Clifton, N . J .
AND
JAMES
J.
PAPPAS
07015
Ozone reacts with benzene in low-molecular-weight carboxylic acids, or mixed solvents incorporating such acids, at a reasonable rate, with very little absorption by the solvent. Subsequent reduction by dimethyl sulfide (DMS) and gravimetric analysis with 2,4-dinitrophenylhydrazine (DNPH) have demonstrated yields of glyoxal up to 73% based on absorbed ozone. Tests for several possible by-products wer’e all negative. Various solvent systems, temperatures, conversions, catalysts, and reducing agents were utilized, and the effects of each variation noted. The ozonolysis of cyclooctatetraene w a s briefly investigated and related to that of benzene.
SINCE the end
of World War 11, glyoxal has evolved from a laboratory chemical to a large-scale commercial product which has expanded to a reported 1967 capacity of some 180,000,000 pounds. I t s most prominent use has been as a key component in crease-resistant formulations for textiles, notably with urea and formaldehyde. The industrial synthesis of glyoxal involves catalytic oxygenation of ethylene glycol, giving as the ultimate salable product a concentrated aqueous glyoxal solution containing
unreacted glycol, formaldehyde, and formic and glycolic acids as the main impurities (Bohmfalk et al., 1951). The ozonolysis of benzene has long been recognized as leading to glyoxal, as mentioned in Long’s review (1940). The lower cost of benzene as a starting material uis a uis ethylene glycol encouraged investigation of this application of ozonization as a synthetic route to glyoxal. The reaction of benzene with ozonized oxygen has been reported only rarely in the chemical literature of this VOL. 8 NO. 1 M A R C H 1 9 6 9
89
century. Harries and Weiss (1904) found that benzene, when ozonized neat and then hydrolyzed with hot water, gave about 2 equivalents of glyoxal (isolated as the osazone) and no hydrogen peroxide. The authors surmised that the third mole equivalent of glyoxal reacted with hydrogen peroxide during hydrolysis to give carbon dioxide and water. The ozonolysis of benzene has more recently been studied from a kinetic point of view (Bernatek et al.. 1967; Kakagawa et al., 1960; Sixma et al., 1951). The reaction rate increased when benzene was replaced by its alkylated homologs, upon going to solvents of higher dielectric constant, and with the addition of Lewis acids, confirming the expected electrophilicity of ozone attack upon the aromatic ring. Aluminum chloride-catalyzed ozonolysis of benzene in nitromethane, followed by hydrolysis and p-nitrophenylhydrazine determination, gave 14 to 155 yields of glyoxal over a range of conversions. An attempt to optimize the yield of glyoxal had to take a number of factors into account. The solvents utilized were exclusively participating, for both ease of workup and safety reasons. Harries and Weiss (1904) experienced several explosions when ozoning benzene neat; the only major difference between their system and those involving nonparticipating solvents would probably be one of concentration. An additional attribute looked for in the solvent was exceptional stability toward ozone, since benzene itself was expected to ozonize only with difficulty. Because a batch reactor was used, a moderately slow gas flow was employed to minimize entrainment. A reductive workup was utilized, since this technique was assumed to be milder and more specific than the hydrolytic workup of previous workers. Glyoxal, which solvates and polymerizes with great facility and is rarely isolated as the monomer, was assayed through formation of the bis-2,4dinitrophenylhydrazone. Experimental
A sample gas stream introducing 12 to 13 mg. of 0.3 per minute was obtained by charging pure dry oxygen to a Welsbach T-816 Ozonator set a t 86 volts and 0.2 liter per minute. The “conventional” apparatus was Ace Mini-Lab Assembly Xo. 10104; a subsequent modification, caused by chronic difficulties with ozone leakage, consisted of replacing the hollow stirrer with a coarse-porosity filter stick fitted a t the proper height with a 19/38 joint. Ozone was determined iodometrically. Benzene and all solvents were reagent grade and were used directly. Temperatures were controlled (to -1” C.)by cold water, ice-methanol, dry ice-carbon tetrachloride, or dry ice-acetone, depending on the temperature range desired. Adequate shielding was used as a precaution, but the ozonolysis and reduction steps were completely uneventful from a safety standpoint. Procedure. The modified apparatus was used, with the reactor connected to the terminal K I wash bottle through two West condensers in series. The charge consisted of 6.28 grams (80 mmoles) of benzene, 25 ml. of acetic acid, and 55 ml. of nitromethane. The inlet gas, containing 12.15 mg. of ozone per minute, was introduced to this solution, held at -5’C.. for exactly 3 hours. Of the 2186 mg. of 0 , added, 360 mg. passed through unreacted, so that 1826 mg. (83.5‘; efficiency), or 38 mmoles. reacted in some fashion. I’mmediately after dissolved ozone had been fiushed from the solution with nitrogen (the ozone absorption calculation includes this operation), 5 90
l & E C PRODUCT RESEARCH A N D DEVELOPMENT
ml. of dimethyl sulfide (DMS), almost a 1 0 0 5 excess, was added dropwise below 10°C. After the solution had warmed to room temperature (KI-starch paper gave a negative test a t this point), it was diluted t o exactly 250 ml. with methanol. A 20-ml. aliquot was removed and added to 450 ml. of saturated aqueous 2,4-dinitrophenylhydrazine hydrochloride. Heating, overnight digestion, filtration, methanol-washing, and drying gave 0.91 gram of glyoxal bis-dinitrophenylhydrazone (identification through infrared and mixture melting point). This figure normalizes to 27.2 mmoles, equivalent to a ’71.65 yield based on ozone. Attempted Detection of By-products. The D N P H filtrate was reheated before being diluted to twice its volume with water. Kothing precipitated, indicating the absence of a significant amount of glyoxylic acid. Next, 25 ml. of the product solution was stripped of solvent by heating on the steam bath under a nitrogen stream. The residue was dissolved in 13 ml. of water, and the solution treated with 2 ml. of saturated aqueous calcium hydroxide. No precipitate or cloudiness was observed, and oxalic acid was thus shown to be completely absent (Dorland and Hibbert, 1940). Three drops each of the product solution and 2 N hydrochloric acid were treated with powdered magnesium until no more gas evolution occurred. Then 4 ml. of 1 2 N sulfuric acid and a little chromotropic acid were added, and the mixture was heated in a hot water bath for 15 minutes. KOcolor appeared, although subsequent addition of a drop of formalin resulted in an indigo-colored solution. Therefore, formic acid was not a by-product (Feigl, 1960); carbon monoxide (for test, see Feigl, 1960) was also contraindicated when the exit gas stream was sampled with phosphomolybdic acid. Ozonolysis of Benzene. YIELDBASEDON BENZENE. The conventional apparatus was used, with the addition of a dry ice-acetone-filled condenser between reactor and K I trap. Ozonized oxygen was added under the usual conditions for 2.5 hours a t 15.C. to 80 mmoles of benzene in 80 ml. of acetic acid. After nitrogen flushing, the solution was reduced by slow addition of 3.5 ml. of DMS. The yellowish solution was transferred t o a 100-ml. volumetric flask and diluted to the mark with acetic acid, and a 5-ml. sample was removed and added to an excess of D N P H solution. The resultant yield of the osazone was 0.46 gram, normalized to 22 mmoles of total glyoxal, derivable from the ozonolysis of 7.33 mmoles (0.57 gram) of benzene. The condensed material was meanwhile allowed to thaw into a tared flask: total condensate, 2.60 grams. Subjection of chlorobenzene-enriched samples of both the condensate and the product solution to vapor-phase chromatography revealed that 4.17 grams of benzene remained in the reactor and 1.01 grams had been trapped on the condenser. This left unaccounted 1.10 grams from the initial charge. Since only 0.57 gram of benzene accounted for the total glyoxal, the yield of glyoxal based on consumed benzene was 5 2 ‘ ~ . Ozonolysis of Cyclooctatetraene. A sample of this material purchased from a laboratory supply house was vacuumdistilled: Vapor-phase chromatography of the yellow liquid c Twenty millimoles (2.08 showed only about l C impurity. grams) was dissolved in 27 ml. of acetic acid and 53 ml. of propionic acid, and the solution cooled to -35°C. The ozonator was set a t 140 volts and 0.5 liter per minute, producing 3i.9 mg. of 0, per minute in the sample stream.
Theoretical ozone uptake was 3840 mg. (80 mmoles), corresponding to 101.3 minutes. Ozone addition was performed for exactly 2 hours. A precipitate was observed, increasing in quantity as ozonation proceeded. Hour 1
2
ivg 0 Added
in Trap
Mg A baorbed
2274 2274
179 886
2095 1388
Mg
The apparent ozone uptake was 3483 mg. (72.6 mmoles), or 90.75 of theory. An equivalent amount of D M S was added, after the nitrogen flush, a t -30" to -15°C. (the mixture had first been allowed to warm until all the solid had dissolved, and then recooled). The solution was then warmed to ambient temperature and diluted t o 100 ml. Addition of a 5-ml. aliquot to excess D S P H afforded 1.32 grams of the derivative, equivalent to an over-all total of 63.1 mmoles of glyoxal. This was a 79'; yield based on the total charge of cyclooctatetraene, 87'; based on the ozone consumed. Results and Discussion
Several potentially acceptable solvents were screened by passing ozonized oxygen through the neat liquid and ascertaining the quantity of ozone taken up by determining the amount entering a terminal trap containing aqueous potassium iodide. Solvent susceptibility t o ozone attack was methanol > neodecanoic acid > pelargonic acid >> propionic acid N acetic acid. Preliminary experiments in which low concentrations of benzene in these last two solvents (40 mmoles in 80 ml.) were ozonized a t 15' to 20°C. to ca. 4 5 5 completion (based on ozone uptake) and then reduced with excess D M S (Pappas et al., 1966) gave the following results: At 12 to 13 mg. of ozone per minute, 60 to 7O';c of the ozone introduced was absorbed; yields of g1:yoxal based on ozone were around 50%. Attempts to inclorporate water into the system, as either solvent (with and without emulsifying agent) or cosolvent with either a participating or nonparticipating partner, were notably unrewarding. Glyoxal yields were under 4 0 5 in all cases. The best solvents a:mong those tested were lipophobic, low-molecular-weight carboxylic acids; admixture with a low-molecular-weight nitroalkane constituted a slightly more favorable system (with respect to yield) than when the acid alone was used. Alcohols, which would produce neuLral systems in which the hydroperoxidic fragments might be less prone to degradation, were completely unsuitable because of their relative sensitivity toward ozone. The experimental sequence summarized in Table I constituted an attempt to estimate the effects of both temperature and degree of' conversion on the yield of glyoxal. In all experiments in this table, the modified apparatus was used (see Experimental). Not only was leakage eliminated, but the superior gas dispersion which resulted markedly improved the efficiency of ozone absorption. Allowing the ozonization to proceed to higher conversions did not significantly affect the yield until the process was pushed close to or beyond completion, indicating that product deterioration by ozone becomes prohibitive a t low benzene concentration. Lowering the reaction temperature diminished t'he reaction rate, as shown by the reduced ozonization etiiciency (corresponding reduction in thermal decomposition of ozone could have played a minor
Table I. Ozonolysis of Benzene" (c
Ozonolysis Mmoles Temp., Time, Benzene C. Hours 80
15
-a
-20
10 -5 15
1 2 3 3 4 5 1.5 3 4.5 1.5 3 3 5 4.67 4.5 4.5 3.33 4
Apparent i
,c
of
Introduced Ozone
Completiono Absorbed 6.0 11.9 18.2 18.5 24.6 30.0 8.2 15.8 24.3 6.5 12.5 12.9 20.5 46 51 101 102 123
92.3 92.2 90.5 92.0 91.6 92.7 84.4 83.5 83.7 70.7 64.0 66.0 64.6 20.1 21.1 42.6 58.3 57.8
C
Yield 51.1 60.3 60.0 59.4 58.7 57.5 63.3 71.6 71.7 73.5 70.0 69.8 71.0 62.5 63.9 38.5 35.3 28.9
'Soluent. 55 ml. of nitromethane and 25 ml. of acetic acid; I2 to 13 mg. o f ozone added per minute; reduction by DIMS. 'Assuming that each mole of absorbed ozone reacted exclusiwly uith one-third mole of benzene. Based on ozone.
role in this trend). Temperature decrease from 15" to -5.C. also resulted in markedly higher yields; this trend did not continue upon further decrease to -2OCC., and apparently an optimum for these conditions had been reached. The reasonable stability of the ozonolysis products under conditions where ozone was present only in low concentration was verified by results from the ozonolysis of cyclooctatetraene. This completely conjugated but nonaromatic hydrocarbon reacted with ozone much more readily than benzene; indeed, absorption was almost quantitative until over half of the tetraene had reacted. As a result, 12.5 mmoles ozonized to 7 5 5 completion a t l ? C . in acetic acid gave a 785 yield of glyoxal, while 20 mmoles ozonized to 91% completion a t -35" C. in acetic-propionic gave an 875 yield. Any alteration, therefore, which accelerates the rate of benzene ozonolysis (such as the improvement of gas dispersion) should be capable of increasing the yield of glyoxal. Since Sixma et al. (1951) had demonstrated catalysis by Lewis acids, the effect of these compounds in participating systems was investigated. Acetic acid was thoroughly unsuitable as a benzene ozonolysis solvent when aluminum chloride was present. The "catalyst" was largely insoluble and the rate of ozonolysis sharply decreased (from 60 to 25% of the introduced ozone was absorbed, compared to 805 in the absence of aluminum chloride), while the yield of glyoxal also dropped, from 535 without to 3 6 5 with aluminum chloride. Nitromethan? was then incorporated into the system in an attempt to improve catalyst solubility, and a homogeneous system was achieved with 70 ml. of nitromethane and 10 ml. of acetic acid. Again, however, both rate and yield were markedly depressed. Acetic acid apparently was acting as a Lewis base with aluminum chloride, forming HAl(OC0CH JCli. The aluminum atom, now electron-rich, should tend to associate with the positive end of the ozone dipole, deactivating VOL. 8 NO. 1 M A R C H 1 9 6 9
91
it toward electrophilic attack on benzene. Anhydrous ferric chloride was tried as a catalyst whose metal atom contains “open” orbitals. The rate of ozone absorption increased slightly, but the eventual yield of glyoxal was only 2OLC. Evidently the catalyst was involved in accelerating product degradation. Despite the convenience of DMS as a reductant for laboratory ozonolysis, it was realized that more economical or efficient alternatives might be available. Comparative experiments were performed whereby glyoxal bis-DNPH levels of samples ozonized in acetic acid were determined after DMS reduction, no reduction, or PtO>-catalyzed hydrogenation. Unreduced aliquots gave practically the same quantities of osazone as those treated with DMS, while those subjected to hydrogenation resulted in yields of this derivative only about one half as great as with the other techniques. However, while the base-catalyzed condensation of DMS-derived product with o-phenylenediamine (Dvornikoff, 1950) gave good yields of quinoxaline, none of this product could be obtained from unreduced samples. Evidently the still-intact peroxidic intermediates were readily convertible to the osazone by the strongly acid D N P H solution, but (after removal of solvent under reduced pressure) were destructively decomposed by the diamine-aqueous base system. The results obtained with hydrogenation could have been due to the wrong choice of catalyst, solvent, etc.; the different variables were never independently evaluated. There was significant decomposition of glyoxal and/ or a precursor under the conditions used; in agreement with this, catalytic reduction of glyoxal to ethylene glycol has been reported (Union Carbide Corp., 1965). Initially produced compounds I to 111 (Ac = acetyl; aldehyde groups represented as nonsolvated) probably undergo condensation to give materials such as IV, especially since such condensations are known to be acidcatalyzed (Keaveney et a1 , 1967). D M S is stable to peroxides (as contrasted with hydroperoxides). CHO L O
I
POH
/OAC OHC----CH
t O=H-CH=O-0
o=cH-~H-o-~ 0
0
V
H
I
H
I
0
II
O=C-O-C=O+HC-OH
+ CO
chloride (Briner and Wunenburger, 1929); the preliminary breakdown product of the ozone-acetylene initial adduct should also be V. Negative reactions, however, were obtained when sensitive spot tests (Feigl, 1960) were applied to ozonized (benzene-containing) solutions for formic acid and to the off-gas for carbon monoxide. Similarly, oxidation of the primary products was seemingly contraindicated by negative tests for oxalic and glyoxylic acids; carbon dioxide also could be detected. Thus no one- or two-carbon by-product was observed, and the exact nature of the side reaction(s) was not determined. Equal aliquots from a benzene ozonolysis were reduced immediately following the (5-hour) period of ozone addition, and 19 hours later. The quantities of D N P H derivative subsequently isolated were 1.52 and 0.60 gram, respectively, this decay process accounting for a drop in glyoxal yield of well over 50%. Literature Cited
EO0/cH-H\o*c
111
IV
Potassium iodide, a broader-based reductant than DMS, was therefore investigated. Comparative experiments revealed no improvement in glyoxal yield upon going from DMS to potassium iodide. This result was interpreted as indicating that I1 and/or I11 were present, even if only in minor equilibrium concentrations. Yield calculations were heretofore based on ozone absorption, but it was realized that similar values should be obtained using benzene consumption as the basis, if, over the conversion range covered, ozone reacted predominantly with benzene. Reductive ozonolysis was performed, using the conventional reactor and acetic acid as solvent, with off-gas volatiles trapped on a dry ice-acetone condenser. The amounts of unreacted benzene in the reactor and entrained benzene on the condenser were determined by vapor-phase chromatography. Comparison of the quantity of glyoxal bis-DNPH with that of the unaccounted 92
benzene resulted in yields of 48 to 52%, assuming 3 moles of glyoxal from 1 mole of benzene; thus the same values resulted from each basis of calculation (cf. above). Therefore, attempting to account for these yields largely from the standpoint of deviation of some ozone-benzene product from the desired reaction path would seem a valid approach. One plausible side reaction involves rearrangement of the short-lived zwitterion V to formic anhydride, which spontaneously decomposes t o formic acid and carbon monoxide. Both glyoxal and, to a lesser extent, formic acid were produced upon ozonolysis of acetylene in methyl
I & E C PRODUCT RESEARCH A N D DEVELOPMENT
Bernatek, E., Karlsen, E., Ledaal, T., Acta Chem. Scand. 21, 1229 (1967). Bohmfalk, J. F., McKamee, R . W., Barry, R . P., Ind. Eng. Chem. 43, 786 (1951). Briner, E., Wunenburger, R., Helu. Chim. Acta 12, 786 (1929). Dorland, R. M., Hibbert, H., Can. J . Res. 18B, 30 (1940). Dvornikoff, M. S . (to Monsanto Chemical ‘20.1, U.S. Patent 2,521,287 (Sept. 5, 1950); C A 44, 10740f (1950). Feigl, F., “Spot Tests in Organic Analysis,” 6th ed., pp. 346, 368, Elsevier, Amsterdam, 1960. Harries, C., Weiss, V., Ber. 37, 3431 (1904). Keaveney, W. P., Berger, M. G., Pappas, J. J., J . Org. Chem. 32, 1537 (1967). Long, L., Chem. Reu. 27, 437 (1940). Nakagawa, T. W., Andrews, L. J., Keefer, R . M., J . A m . Chem. SOC.82, 269 (1960). Pappas, J. J., Keaveney, W. P., Gancher, E., Berger, M., Tetrahedron Letters 1966, 4273. Sixma, F. L. J., Boer, H., Wibaut, J. P., Rec. Trau. Chim. 70, 1005 (1951). Union Carbide Corp., “General Chemistry of Glyoxal,” p.5, 1965. RECEIVED for review June 28. 1968 ACCEPTED October 21, 1968