1454
INDUSTRIAL AND ENGINEERING CHEMISTRY
(13) Riedel, L., 2. ges. KdZte-Ind., 48, 105-7 (1941); BuLZ. Intern. Inst. Refrig., 23, No.3, 1-5 (1942). (14) Russell, H., Golding, D. R. V., and Yost, D. M., J. Am. Chem. Soc., 66, 16-20 (1944). (15) Smith, D.C.,Brown, G. M., Nielsen, J. Rud, Smith, R. M., and Liang, C. Y., J. Chem. Phus., 20, 473 (1952). (16) Smith, D. C.,Alpert, M., Saunders, R. A., Brown, G. M., and Moran, N. B., Naval Research Lab., Rept. 3924, Feb. 1, 1952. (17) Smith, D.C., Nielsen, J. R., Berryman,L. H., Claassen, H. H., and Hudson, R. L., Naval Research Lab., Project NR019-120, Rept. 3567, Sept. 15. 1949.
Vol. 47, No. 7
(18) Soll, J. (to I. G . Farbenindustrie A,-G.), U. S.Patent 2,118,901 (May 31, 1938) [equivalent Ger. Patent 641,878 (1937)1. (19) Swarts, F., Compt. rend., 197, 1261-4 (1933). RECEIVED for review November 13, 1954. ACCEPTED February 24, 1955. Material supplementary to this article has been deposited as Document No. 4502 with the AD1 Auxiliary Publications Project, Photoduplication Service, Library of Congresa, Washington 25, D. C. A copy may be secured by citing the document number and b y remitting $5.00 for photoprints or $2.25 for 35-mm. microfilm. Advance payment is required. Make checks or money orders payable to Chief, Photoduplication Service, Library of Congress.
Vapor-Liquid Equilibria in Dilute Aqueous Solutions of Ethylene Oxide IC. E. MAcCORMACK AND J. H. B. CHENIER Applied Chemistry Division, National Research Council of Canada, Ottawa, Canada
P
REVIOUS work on the system ethylene oxide-water covers the extremes of dilute and concentrated solutions, but is not very consistent. Kaplan and Reformatskaya ( 7 ) measured the solubility of ethylene oxide a t 5", lo", and 20" C. over the range 2 t o 23 mole % ethylene oxide in the liquid phase, but did not make vapor phase analyses. Coles and Popper ( 3 ) and Kireev and Popov (9) employed the Othmer still method to determine the normal boiling points and phase compositions for the whole range of liquid concentrations. Over the range 10 to 60 mole % ethylene oxide (liquid), the Kireev and Popov boiling points lie from 2" to 4" higher than those of Coles and Popper and the vapor phase compositions are correspondingly much higher. Extrapolation of the Kaplan-Reformatskaya data a t 20" C. indicates this to be the normal boiling point a t 15 mole % ethylene oxide (liquid). For this composition Coles and Popper (3) give 22.5" C. and Kireev and Popov (9) give 25.5" C. by interpolation. The present interest lies chiefly in obtaining reliable solubility data a t low ethylene oxide concentrations. Information in this region is required for absorber design where low partial pressure ethylene oxide gas mixtures are encountered, as, for example, in the process for manufacture of ethylene oxide by silver-catalyzed air oxidation of ethylene, The results obtained have been successfully correlated with those of Coles and Popper and it is believed that accurate equilibria can now be calculated over the whole range of concentration a t temperatures from 5" to 30" C. Absorption a t higher temperatures would be undesirable commercially by virtue of reduced solubility and rapid hydrolysis of the ethylene oxide to glycol.
and the mass transferred was determined by weighing. The whole of the dual cell and constant-volume manometer, M , shown in Figure I, were thermostated in a water bath and stopcock A was opened t o allow the establishment of equilibrium while the liquid phase was agitated by means of a glass-covered ma netic stirrer, 8. 'faking pressure constancy as the criterion, equilibrium was established in 1 to 3 hours, though a t least 1hour and often much longer was allowed t o elapse after the onset of pressure constancy. Stopcock A was then closed and the vapor sample condensed into a previously weighed, evacuated ampoule, followed by reweighing to determine the total mass of the vapor phase. The efficiency of this condensation was checked by measuring the low residual pressure in the vapor cell and errors due to condensation of mercury va or were checked by reweighing the flamed evacuated ampoule. $he ampoule was then attached to a closed 300-cc. vessel containing a known amount of Luhatti reagent ( I f ) and shaken for 10 to 15 minutes, followed by back-titration of the residual hydrochloric acid, Blanks performed under these conditions with known amounts of ethylene oxide showed the absorption efficiency to be 98.9 j=0.4%.
E
EXPERIMENTAL
A static method was chosen whereby very accurately known liquid phase compositions could be employed, and sufficiently large vapor samples were available for analysis with an accuracy within &lyo. Figure 1shows a diagram of the equilibrium cell. The cell consists of two parts, the liquid cell C (40 ml.), and the vapor cell, D (1 liter), separated b y a large hohow stopcock, A . Silicone grease was used on all stopcocks. The whole system was evacuated via tube E with stopcocks A and B open. Distilled water was then admitted from the previously weighed reservoir, R, with stopcock A closed. Residual droplets and vapor in the connecting tube were condensed into the liquid cell a t liquid air temperature and R was reweighed to determine the mass of mater transferred. Ethylene oxide was admitted t o the vapor cell via the evacuated tube, E,from a weighed ampoule. Wlth stopcock F closed, the residual line gas was recondensed into the ampoule
Figure 1.
Equilibrium cell
A knowledge of the vapor composition and its total mass made possible a calculation of the liquid phase composition from the known total mass of each component. Since the vapor phase mole fraction of ethylene oxide ( Y E ~ O ) is close to unity, the relative error in YEto would be approximately &OX%, thus placing a much larger relative error on the mole fraction of water. The residual vapor contained in the stem of the liquid cell has no effect on the vapor phase composition and a negligible effect on the computed value of the liquid composition by virtue of its extremely small relative mass
July 1955
INDUSTRIAL AND ENGINEERING CHEMISTRY
Usually the liquid sample was discarded after two further additions of ethylene oxide. The temperature was controlled within ~k0.05"C. and measured using a mercury-in-glass thermometer calibrated with an accuracy within *0.0la C. by the Division of Physics, National Research Council. Pressures were measured using a cathetometer which could be read to 0.001 em., density corrections being applied for that part of the mercury column kept at room temperature. However, limits are placed on the accuracy of the pressure measurements at 1 0 . 0 5 em., as no corrections were applied for meniscus and refraction errors involved in sighting one meniscus. through a glass-walled tank and I inch of water, or for hydrolysis.
Purity of Components. The ethylene oxide was manufactured by the Matheson Co. with a specified purity of 99.5%. This was condensed into a tube and distilled slowly a t dry ice temperature, the first and last thirds being discarded. The middle fraction was transferred to a storage flask, whence it was condensed into the weighing ampoule as required. Distilled water was thoroughly outgassed in the reservoir, allowing about 10% evaporation by boiling at room temperature before being weighed and attached to the equilibrium cell.
EQUIPMENT DESIGN DATA Required for absorber design where l o w partial pressure ethylene oxide gas mixtures occur
(where subscripts 1 and 2 refer to ethylene oxide and water, respectively, and x is the mole fraction of ethylene oxide in the liquid phase), that Raoult's law must hold for the solvent water. Equation 2 relates P (centimeters of mercury) and x (mole per cent ethylene oxide liquid).
P
= p;
*
50 P;
(HzO)
c.
60.89 0.6543
(2)
- pz
( 31
and
Solubility data were obtained a t 5O, loo, and 20" C., the upper limit being chosen to minimize errors due to hydrolysis of ethylene oxide. One experiment was attempted a t 40" C., using a 2.5 mole yoethylene oxide solution. The pressure fell almost linearly for 50 hours from approximately 38 to 28 em. of mercury. Assuming a first-order rate with concentration directly proportional to pressure, the velocity constant was calculated to be 1.2 X min.-l This is in fair agreement with the value 1.8 X min.-' given by Lichtenstein and Twigg ( I O ) and indicates that vapor pressure measurements above 30" C. are subject to large errors unless rapid equilibrium can be attained. A t 20" C. the velocity constant is given as 2.5 X 10+ and on the same assumptions the pressure drop should be approximately 0.02 em. per hour. The range of concentration employed was limited at 5' and 10" C. by solid hydrate formation. Interpolation from data given by Maass and Boomer ( 1 2 ) indicates the hydrate to be stable between 4.1 and 81 mole yo ethylene oxide a t 5" C. and between 8.7 and 22.0 mole yo ethylene oxide a t 10" C. It was found impractical to make measurements with solutions much more concentrated than the lower hydrate limit, because of formation of thick "ice" on the liquid surface on admitting ethylene oxide gas. I n any case it would not be practicable t o operate absorption columns in the presence of solid hydrate. It is possible, however, to extrapolate the present data up to this region. The vapor pressure of ethylene oxide was calculated from the equation given by Giauque and Gordon (4)and that for water was taken from steam tables (8). The values used (in centimeters of mercury) are as follows : P : (EtO)
+ kx
At50,100,and20oC.,khasthevalues3.17~O0.0~,4.00f0.05, and 5.98 0.15 em. of mercury, respectively. Since .Pl = p
RESULTS
1455
100
c.
74.37 0.9211
200
c.
109 2 1.753
Table I presents the phase compositions and pressures corrected for mercury density-temperature variations. I n Figure 2 total pressure, P, is plotted against liquid phase composition. Here it can be seen that up to 5 mole yoethylene oxide in the liquid phase, the total pressure is a linear function of the composition. This means that the partial pressures, p , for each component must also be linear with composition and Henry's law is thus applicable to ethylene oxide over this range. As the vapor pressures are much below 1 atmosphere, the vapor may be assumed ideal and it can easily be shown from the Duhem-Margules equation in the form
pz = p ; [1 - (2/100)1
(4)
substitution of Equations 2 and 4 into 3 gives pl = x [ k
-
(p2/100)1
(5)
Within the limits of accuracy obtained, Equation 5 may neglect the term in pg and the partial pressures for ethylene oxide may be calculated using the above values of k for the Henry's law constant. In Figure 3 the experimental values of ethylene oxide partial pressures are plotted as a function of the liquid phase composition (log axes), the straight lines up to 5 mole yoethylene oxide being calculated from Equation 5 and the k values derived. At 5" and 10" C. the dashed lines indicate the extent of hydrate formation. For comparison purposes some data of Kaplan and Reformatskaya (7) are included in both Figure 1 and 2, showing poor agreement. However, these data should not be used without first correcting the ethylene oxide partial pressures calculated by Kaplan and Reformatskaya. Numerical checking will show that these authors derived p , from the relation pl = P
- p;(z/loO)
instead of p1
=
P - p,O[l - ( X / l O O ) ]
Table I. Experimental Phase Compositions and Pressures Et0 Temp.,
c. 5
20
mole % 0.711 1.504 3.010 4.644
mole % 75,l 90.8 93.9 96.0
p = yP
2.08 4.99 9.55 14.9
Total P, Cm. Hg 2.77 5.50 10.17 15.57
1.325 2,309 2.609 3.472 4.531 5.014 5.678 7.604 8.192 10.096 11.105 14.186
81.1 89.3 90.7 92.5
7.53 13.6 15.1 21.6 27.9 29.8 34.1 43.2 46.0 54.3 57.2 67.0
9.28 15.22 16.70 23.36 29.52 31.73 36.06 45.09 47.43 55.48 59.07 68.68
Z,
u>
94.4
94.1 94.6 95.9 97 0 97.9 96.8 97.6
INDUSTRIAL AND ENGINEERING CHEMISTRY
1456 MOLE
O
'
i
x
E ~ O (LIP)
i
- 30
-25 -20
Vol. 47, No. 7
the corrected data of Coles and Popper ( 3 ) based on Van Laar solutions. These values therefore involve appreciable errors if used to calculate dilute solution equilibria. Furthermore, in recent correspondence with these authors, it has been agreed that the corrections applied for the variation of y1 with Tare erroneous by virtue of the mathematical sign employed for the heat of solution, L1. The latter is given by Bichowski and Rossini (9) as Q = - A H , = -L1 = 1500 cal. per mole, where A H s is the increase in heat content when 1 mole of liquid ethylene oxide dissolves in a large volume of water according to the reaction
Et0 (liq.) = , Et0 (solution)
(7)
The original work quoted by Bichowski and Rossini ( 2 ) is due to Berthelot (I), who states the heat of solution a t 13" C. as heat evolved (exothermic) in dissolving 1mole of liquid ethylene oxide in 160 moles of water. Thus AH8 or L1is negative.
LOO
- IO
I I lI l l l
70 60 50 m 100 00
eo
-
-
=
5 ! 1
2
3
I 4
6
l
l
6
7
1
1 8
9
1
1 1
0
1
1 1
0,
w
I 1
40-
1020-
w
2
MOLE Y. E t 0 ( L I P )
Figure 2. Total pressure-composition curves
0 0
0
Present work Coles and Popper, 20' C. Kaplan and Reformatskaya, 20" C.
where symbols previously defined are used. Table I1 lists the corrected values along with those published. For convenience, vapor phase compositions are plotted us. liquid phase compositions in Figure 4.
Table 11. EtO,
~~l~ % (Liquid) 2.53 4.08 5.62 7.27 12.31 15.40 22.99
p , 5'
Corr.
p , 10" C . , Mm. Hg p , 20°
Pub.
Corr.
Pub.
Corr.
C.,Mm.Hg Pub. 190.4 291.8 371.1 471.7 670.2
DISCUSSION
It is a common practice to attempt to smooth activity coefficient data by fitting them to equations of the Van Laar type derived by integration of the Gibbs-Duhem equation. For interpolation purposes these equations may be good, but they are inadequate for extrapolation to dilute solutions (say