ENGINEERING. DESIGN, A N D PROCESS DEVELOPMENT would have a high separation factor but a very low selective adsorption. Other desirable characteristics of an adsorbent for large scale commercial application are high durability, high bulk density, Ion. pressure drop, low cost, and long life.
Acknowledgment The authors gratefully acknow-ledgcthe assistance of Harriet G. Jones, Robert L. James, and W. B. 31.Faulconer in carrying out the experimental ~ o r and k calculations reported in this paper.
literature Cited (1)
Brunauer, S., Emmett, P. H., and Teller, E., J . Am. Chmn. SOC., 60, 309 (1938).
Carslaw, H. S., and Jaeger, J. C., "Conduction of Heat in Solids," Oxford University Press, London, 1947. (3) Day, D. T., Proc. Am. Phil. Soc., 36, 112 (1897). (4) De Vault, D., 6.Am. Chem. SOC.,65, 532 (1943). ( 5 ) Dreisler, P. F., Jr., and Wilhelm, R. H., IND.EXG.CHEM.,45, (2)
1219 (1953).
(6) Eagle, S., and Scott, J. W., Ibid., 42, 1279 (1950). (7) Emmett. P. H.. and De R i t t . T. W.. J . Ana. Ckem. Soc., 6 5 , 1253
( 8 ) Gaffney, B. J., and Drew, T. B., IND.Eso. CHEW.,42, 1120 (1950). (9) Geddes, R. L., Trans. A m . I n s t . Chenz. Engrs., 42, 88 (1946). (10) Gurmitsch, L. G., J . Russ. Phys.-Chem. Soc., 47, 805 (1935). (11) Harper, J. I., Ohen, J. L., and Shuman, F. R., Chem. ETLQ. Progi., 48, 276 (1952); Shuman, F. R., and Brace, D. G . , P e t r o l e u m Eng., 25, No. 4, C-9 (1953). (12) Hirschler, A. E., and Amon, S.,IND.ENG. CHEY., 39, 1585 (1947). R., U. S. Patent 2 , 4 4 1 , 5 7 2 ( M a y (13) Hirschler. A . E., and Lipkin, 18, 1948). (14) Lipkin, M. R., Ibid.,2,398,101 (April 9, 1946). (15) AIair, R. J., and Forziati, -4. E'., .I. Research Sntl. Bur. S t a n d a r d s , 32, 151, 165 (1944); hlair, B. J.. Ibid., 34, 435 (1945); Mair, B. J., and Whit,e. J. O., Oil Gas J . , 34, No. 18, p 29 (1935). (16) hlair, B. J., Westhaver, J. TV., and Romini, F. D., IKD.ENG. CHEY.,42, 1279 (1950). (17) Rescorla, A. R., Ottenweller, J. H., and Freeman, R. S.,A n d . Chem., 20, 196 (1948). (18). Ries, H. E., "Advances in Catalysis," Vol. IV, p. 128, ilcadeinic Press, Kew York, 1952. (19) Shull, C. G., Elkin, P. B., and Roess, L. C., J . Ana. Chem. SOC., 70, 1410 (1948). ( 2 0 ) U'ilkie, C. R., Chent. Eng. Piogr., 45, 218 (1949). (21) Williams, A . M., M e d . K . Ventenslcapsahd. Nobel-imt., 2 , No. 27 (1913). R E C E I V Efor D review July 29, 1954.
(1943) .,
ACCEPTEDDecember 1, 1934.
Thermal Diffusion Separations of Organic liquid Mixtures CHARLES R. BEGEMAN AND PAUL 1. CRAMER Research laboratories Division, General Mofors C o p . , Defroif 2 , Mich.
The separation of organic liquid mixtures b y liquid-phase thermal diffusion in Clusius-Dickel type apparatus is described. Four columns are used in this work and their relative separating effectiveness is evaluated using a cetane-Decalin test mixture. The effect of time, temperature gradient, and viscosity on the approach to steady state i s discussed. Batchwise fractionation of some low molecular weight binary and ternary mixtures is illustrated. Mineral oils are fractionated batchwise and on a continuous flow basis, and the distribution of molecular types in the products is measured. The concentration b y thermal diffusion of lubricating oil additives is illustrated.
T
HERNIAL diffusion separation of organic liquid mixtures
has become of increased interest in recent years, and numerous reference to both analytical and industrial applications have appeared in the literature (9, 11, 12, 14, 17). For the most part these investigations have involved the use of rather conventional Clusius-Dickel type apparatus ( 5 ) . I n attempts to increase the net production rate of separated products, various other investigators have utilized increased gravitational force (6, l a ) , moving walls ( I C ) , or barriers to prevent remixing of separated components ( 4 ) . The work that is described in this paper was also done with columns of ordinary Clusius-Dickel type construction. This equipment is relatively easy to construct and simple to operate. The present data is divided into five principal parts: 1. A description of apparatus and general fractionation procedure 2. The effect of temperature difference and time on degree of separation of a binary test mixture 3. The batchwise fractionation of some low molecular %-eight binary mixtures 2Q2
4. The fractionation of lubricating oils on a batchwise and on a continuous flow basis 5 . The concentration of metal-containing additives in lube oils Apparatus and Procedure Four thermal diffusion columns have been made, all of concentric tube arrangement, three of Rhich are fabricated of stainless steel and one of glass. The important dimensions have been listed in Table I. Column 1 is equipped with drawoff ports a t 10% intervals along the vertical axis so that the entire column contents may be removed by draining each section separately starting a t the top, thereby reducing contamination between fractions. The outer wall is heated by means of a helical resistance winding of 15.6 ohms of 20-gage Chrome1 A wire, wound with a pitch of l j p inch and having one extra turn a t the top end and ll/z extra turns a t the bottom. Hot wall temperature is measured hy six thermocouples soldered to the outer wall at intervals along its length. Cooling water enters the inner tube a t the bottom and exits a t the top. A high flow rate of coolant
INDUSTRIAL AND ENGINEERING CHEMISTRY
Vol. 47,No. 2
HYDROCARBON SEPARATIONS Column 4 is also designcd for continuous flow operation and can be operated with either 35 pounds per square inch gage steam or with controlled temperature hot water on the hot wall. For continuous flow operation the process liquid flows by gravity from a constant-head feed reservoir into the column at the center feed port and is withdrawn continuously, or automatically in small amounts of frequent intervals from each end. A detailed description of flow control devices for this application is being prepared. A photograph of the columns used in this work is shown in Figure 1.
Effect of Time and Temperature Difference on Separation I n order to study the effect of time and temperature gradient across the slit on separating effectiveness, the test mixture used should have a minimum volume change due to mixing, reasonably low vapor pressure, and be susceptible to easy analysis, preferably by refractive index. The components should be of similar and low viscosity and be modestly difficult t o separate by the method in question. By a process of trial and error, the system cetaneDecalin was selected as approximating these requirements. Cetane was purchased from the D u Pont Co. and had a refractive index of 1.43249% and a density of 0.7735;' [reported (1) 1.4325%,
1.4800
Figure 1.
Thermal diffusion columns 1.4700
is used in order t o keep the temperature rise a t a minimum, and thermocouples soldered to the water lines close t o the column are used to indicate water-in and water-out temperatures. This column is very similar t o the one described by Jones and Milberger
(IS).
9 1.4600 Table 1.
Thermal Diffusion Columns for Liquids
Small Tube, Outside Col- Length, Diam., Annulus, Ft. Material Inches Inches Heating Medium umn 1 5 Stainless 0 50 0.015 Electrical resistance elesteel ment 15.6 ohm, 20gage Chrome1 A wire I/! inch pitch helicai wmding 2 10 Stainless 1 00 0.032 35 lb./sq. inch gage steel saturated steam 2.00 0.032 35 Ib./sq. inch gage 3 11 Stainless steel saturated steam 4 8 Borosilicate 0 63 0.031 Hot water or 35 lb./sq. glass inch gage steam
2 X W
n z 1.4500
c
0
9 LL
w
a
1.4400
Columns 2 and 3 are designed for continuous flow operation, and each is equipped with a single feed port located a t the vertical center of the column and drawoff ports a t each end; 35 pounds per square inch gage saturated steam enters at the top of the innermost tube a t a rate controlled by a small valve in the steam exit line. This valve is located only a short distance from the bottom of the column. Thermocouples located in the steam lines a t the top and bottom of the column indicate the same temperature when only a small excess of steam is discharged with the condensate. Under these conditions, with oil in the annular space, 1.9 gallons per hour of condensate is collected from column 2, which corresponds to approximately 14,900 B.t.u. per hour (4.4 kw.). February 1955
I
1.4300
0
I
20 40 DISTANCE FROM
I
I----
60 80 TOP, PER CENT
I 100
Figure 2. Separation of cetane-Decalin test mixture in column 1, showing distribution of components for different temperature gradients for a constant residence time
INDUSTRIAL AND ENGINEERING CHEMISTRY
203
ENGINEERING, DESIGN, AND PROCESS DEVELOPMENT 0.77345i0]. Decalin was Eastman Kodak white label and was a mixture of cis and trans isomers, ng = 1.47726, dZ5 = 0.8861. Both components were used without further purification. Figure 2 shows a series of curves illustrating the distribution of components in column 1 used for this study for various levels of AT, for a residence time of 17 hours. Separate fractionations of the individual components are also shown, which reveal a partial separation of cis- and trans-Decalin isomers. Several important facts are brought out in this plot. It shows the large influence of AT on the time required to establish a steady state distribution of components. Also, the steep composition gradient toward the ends of the apparatus during the initial period of separation is evident. This effect has been predicted from theory by Debye (6). Slightly purer terminal concentrations of components are produced by the AT-47" C. condition over that of the AT-47" C. condition. This is probably due to the turnaround of liquid streams a t the ends of the column, which is greater a t the higher temperature differential. This turnaround must be counteracted by the thermal diffusion separation effect and represents one of the limiting factors of a gravitational type column for batch operation. The effect of time on separation has been investigated by a series of experiments conducted a t constant AT. These results are shown in Figure 3 for AT-97' C. The rapid change in terminal compositions immediately after imposing the thermal gradient is also evident. For the cetane-Decalin system a steady state is
reached after about 8 hours in this column. Similar families of curveB were obtained for AT-14', 28", and 39" C. The separation effect produced by various combinations of AT and time have been evaluated from the average compositions existing above and below the feed line a t the center of the column. Such average compositions have been measured by mechanically
Column 5' x 0 5 ' x 0 015"
Figure 4.
Time required to reach steady state as function of thermal gradient across slit A1
and A2 are areas from Figures 2 and 3
COLUMN 5' X 0 5 X 0.015" CETANE 50 VOL % DECALIN 50 VOL.%
IO
8
c-------___
---a
6
A,
17 HOURS
4
2
0 2 4 A 2
6
E
\ -
2600
4600
(*,.,e
l0,60@
\
-\
IC
Figure
I 1,4300 ----0
AT on separation effect
1 20
DISTANCE
40
60
FROM
80
too
TOP, PER CENT
Figure 3. Separation of cetane-Decalin mixture in column 1 I &owing dish+bution of components for several residence times
204
5. Effect of
integrating the shaded areas A I and Az for families of ourves such as those shown in Figures 2 and 3. I n experiments conducted a t constant AT-28", 39', or 97" c., separation increased nearly linearly with time until a steady state condition was approached. These d a b are illustrated in Figure 4. I n a similar manner, as shown in Figure 5, the magnitude of separation at constant time
INDUSTRIAL AND ENGINEERING CHEMISTRY
Vol. 47, No. 2
HYDROCARBON SEPARATIONS 1.5200
Table 11.
1.5100
Density. Components D 5" Toluene0.8669 methyloyol~hexane~ 0.7984
1.5000 0'
2 1.4900 X 0 W
1.4800 W
E
2 1.4700 a
VISCOSITY. IOOOF.
LL W
a 1.4600
0
Figure 6.
20 40 60 80 DISTANCE FROM TOP, PER CENT
too
Effect of viscosity on time to reach steady state
Column dimenriontr 5 ft. X 0.5 inch X 0.015 inch
is approximately a function of (AT)*. I n his theoretical exposition of the thermogravitational separation method (Clusius and Dickel column) deGroot (7) predicts that for a column equipped with reservoirs a t each end, the approach to a steady state concentration of species in such reservoirs would be a function of time X (AT)*,where the average temperature is constant. Since there are no physical reservoirs attached to the ends of this apparatus, these results may not properly be interpreted in light of the deGroot theory. Nevertheless, the separation obtained follows the course predicted for the reservoired column. The fact that greater separation effect is obtained a t higher values of
IO0 80
z
2 u 0
Iz w
60
0
a
y
40
&I
3 P
COLUMN:5' X&'X
0.015'
AT u 48. C. 2o
0
20 DISTANCE
40
60
FROM TOP,
80
100
PER CENT
Figure 7. Separation of cetane-cyclohexane blends
February 1955
Vol.
AT,
50 50
53
Time, Hr. 18
0.7025 0.6919
50 50
67
16
n-Heptane0.6838 2,2-dimethyl~entane~0.6738
50 50
61
18
Cyclohexane0.7786 2,2,3-trimethylbutane 0 . 6901
50 50
53
2,2-Dimethylpentane0.6738 2,2-dimethylbutanem 0.6492
50 50
2,2-Dimethylpentane3,l-dimethylpentane
0.6738 0.6933
2,2,3-Trimethylbutaneethanola
0,6901 0.7893
50
Ethylbenzene2-ethoxyethanol
0.8670 0.9311
n-Octane2,2,4-trimethylpentane5
0
W
Separation of Binary Mixtures by Liquid Thermal Diffusion %
C.
Final Composition Top Bottom 10% 10% 52.5 47.5 47.5 52.5 67 33
38.6 61.5
36
64
37 63
20
43 57
58 42
22
20
50.7 49.3
49.3 50.7
50 50
50
20
62 38
30 70
60
44
16
48.5 51.5
51.8 48.2
58.5 41.5
72
I6
85 15
16 84
1-Propanol0.8035 76 75 18 0.9982 24 water These pairs involve forgotten effect phenomenon.
>99
50 50
state cwnditions. binary mixtures, the addition of a third component t o such a mixture usually effects some separation among all three comContinuous Flow and Batchwise Oil Fractionations ponents. I n the case of a benzene-cyclohexane blend, addition Columns 2 and 3, rquipprd N ith suitable flow control devices, of acetone, ethanol, or methanol causes the benzene and cyclohave been used to fractionate lubricating oils on a continuous hexane to be separated, although the fractionation effect is small. 8 shows the distribution Figure patterns for benzene and cyclohexane after the polar comTable 111. Continuous Flow Fractionation of White Oil pound had been removed by washing with water. I n these (Column 2, AT-122' C.; feed rate 13 nil /hour) experiments, benzene, cycloParaffinic Ai,. Viscosity, S.U.B. yjscoqity A v . Mol. Carbonb, Rings/ hexane, and the third comn" ; u"," Wt.5 70 Xoleculeb 100' F. 210O F. Index ponent were each present in Original oil 1 4802 0,8785 388 55 4 2 6 32 1 51 30 79 Top fractions equal parts by volume. Tem45 76 108 384 64.6 2.0 179.8 0,8587 First pass 1.4718 perature difference and time 1.5 125.1 42.5 127 362 70.4 0.8433 Second pass 1.4658 39.82 145 346 74.8 1.1 91 95 Third pass 1.4609 0.8316 were 20" C. and 16 hours, reBottom fractions spectively, which combination First pass 1.4889 0.8999 399 47.5 3 3 748.9 63.90 42 -6 Second pass 1.4955 0.9166 397 41.4 4.0 1688 80.3 is too small for the components Third pass 1.4993 0.9257 405 36.2 96.2 - 42 4.2 2992 to have reached a steady state a Measured by ebullioscopic method, using benzene. distribution. 6 Calculated by method of Tadema and Leendertse ( 1 5 ) . deGroot ( 7 ) has shown that 206
INDUSTRIAL AND ENGINEERING CHEMISTRY
Vol. 47, No. 2
HYDROCARBON SEPARATIONS
0
2
I
NUMEER
3
3
I
2
0
OF FRACTIONATIONS
Figure 10. White oil composition change with successive refractionation of top and bottom streams
I
I
I
I
20
40
60
80
t
of 321 S.U.S. and a viscosity index of 79, was fractionated in this manner and the end fractions separately refractionated in the same equipment. After the second such refractionations the 100' F. viscosities of the top and bottom fractions were 91.95 and 2992 S.U.S., and viscosity indexes 145 and -42, respectively. Other physical properties are summarized in Table 111. Figure 10 illustrates the changes in oil composition, a s reflected in the percentage of paraffinic carbon and ring number of top and bottom fractions, with successive retractionations. In a similar manner a solvent-refined, mid-continent lube oil having a 100' F. viscosity of 525 S.U.S. was successively refractionated to produce a top fraction that had a viscosity index of 157 and a 210' F. viscosity of 46 seconds. Such a viscosityviscosity index combination is usually associated with blends of very low molecular weight neutrals plus viscosity index improver, whereas for the case of the thermal diffusion product these properties result from the very high paraffinic carbon content of this oil. The molecular weight is only slightly lower than the feed stock from which it was prepared. Physical properties of these fractions are listed in Table IV. I n addition to information concerning hydrocarbon types present in the top and bottom oil fractions obtained by continuous flow separations, a more systematic study was made by batch fractionation using column 1. Oils of so-called paraffinic and naphthenic type have been fractionated, and the distribution of components according t o carbon type has been calculated by the structural group analysis method, refractive index-density-molecular weight (n-D-M method) of Tadema and Leendertse (15). A typical pattern for an asphaltic base oil is illustrated in Figure 11. Percentage of paraffinic carbon ranged from approximately 64% in the top 0.1 of the column t o 9% in the bottom 0.1 of the column. Naphthenic carbon constitutes approximately 64% of the total in the bottom fraction and 27% of the total in the top fraction. Average aromatic and naphthenic ring numbers for these fractions indicate that the separation process does not differentiate sharply between saturated and unsaturated rings. This is in agreement with the observations of Broeder and van Nes ( 2 ) . By charging a n oil having a ring number of 4 with an average molecular weight of 310 it has been possible to obtain a 10% fraction in which more than 95% of the carbon is in ring structure. Whole Oil
0
4
To P
100
DISTANCE F R O M TOP. PER CENT
Figure 1 1. Carbon-type distribution in thermal diffusion fractions from asphaltic base oil separated in column 1 Ba
P
Mg
Ca
Zn
Bottom
flow basis. I n all such experiments, the feed was entered a t the Figure 12. Emission spectra of thermal diffusion oil fractions vertical center of the column, and the products were taken off containing inhibitor and dispersant additives on a 50% overhead, 50% bottoms split. Feed rate was approximately 13 ml. per hour for column 2 and 25 ml. per Table IV. Continuous Flow Thermal Diffusion Fractionation of Solvent Refined hour for column 3, which Mid-continent Oil corresponds t o 0.035 ml. per (Columns 3 and 2, AT-120' C.) square inch of heated surface Av. AroNaphParafAv. matic thenio finic Rings/ per hour; 280" F. saturated Mol. Carbon, Carbon, Carbon, Mole- V?E?!!LSU .& Viscosity % % cule IOO'F. 210° F. Index nzg Wt. % steam heats the inside tube, and tap water the outer Original oil 1.4816 0.8730 466 4 30 66 2.3 523 64.6 96 Top fractions First pass 1.4727 0.8539 53.5 119 458 2 27 71 2.0 270 tube. Secondpass 1.4667 0.8403 458 1.2 19.7 79.1 1.2 175 48.2 137 A highly refined white oil, Thirdpass 1.4628 0.8319 440 0.6 15.4 84.0 0.9 134 46.0 157 having a 100' F. viscosity
.~~
February 1955
INDUSTRIAL AND ENGINEERING CHEMISTRY
20 7
ENGINEERING, DESIGN, A N D PROCESS DEVELOPMENT Metal or Metalloid Containing Lube Additives
Literature Cited
Organic compounds consisting of metal salts or containing certain other heavy elements have been found t o be concentrated very strongly a t the bottom of the column. This includes such materials as zinc alkyldithiophosphates, calcium and barium sulfonates and other compounds commonly added to mineral oils to impart oxidation resistance, antiwear, or dispersing properties. Emission spectra of thermal diffusion fractions of an additive oil are illustrated in Figure 12. The actual concentration in the bottom fraction of elements zinc, phosphorus, calcium, or barium is more than 10,000 times their concentration in fraction 1. The process, therefore, may be applicable as a nondestructive means of concentrating certain compounds which may be present in trace amounts.
(1) Am. Petroleum Inst., Pittsburgh, Pa., Res. Proj. 44 (1950). ( 2 ) Broeder, J. J., and Nes, K. van, Proc. Third World Petroleum Conor., 6, 25 (1951). (3) Clusius, IC.,and Dickel, G., Naturwissenschajten, 26, 546 (1938).
Acknowledgment The authors wish t o express their thanks t o L. L. Withrow and J. M. Campbell of these laboratories for their continued support of this work; t o B. M. Johnson and S. G. Anderson for experimental assistance; and t o A. L. Jones, Standard Oil Co. (Ohio) for helpful advice and discussion.
(4) Debye, P. J. W., U. S. Patent 2,567,765 (Sept. 11, 1951). ( 3 ) Debye, P. J . W., Ann. Physik, 36, 248 (1939). (6) Farber, M., and Libby, W., J. Chem. Phys., 8 , 965 (1940). ( 7 ) Groot, S. R. de, Physica, 9, 801 (1942). (8) Ibid., p. 923. (9) Haak. F. A , , and Xes, K. van, J Inst. Petroleum, 37, 245 (1951) (10) Horsley, L. H., Anal. Chem., 19, 508 (1947). (11) Jones, A, L., Petroleum Processzng, 6, 132 (1951). (12) Jones, A. L., and Foreman, R. TV., IND.ENG.CHEM,,44, 2249 (1962). (13) Jones, A. L., and Milbergor, E. C., Ibid., 45, 2689 (1953). (14) Kramers, H., and Broeder, J. J., A n a l . C h i m . A c t a , 2, 687 (1948). (15) Nes, K. van, and Westen, H. A. van, “Aspects of the Constitution of Mineral Oils,” Elsevier, Netherlands, 1951. (16) Niini, R., Suomen Kemistilehti, 26B, 42 (1953) (in English). (17) O’Donnell, G., A n a l . Chem., 23, 894 (1951). (18) Tilvis, E., SOC.Sci. Fennica, Commentationes Phya.-Math.. 13, 16 (1947) (in English). RECEIVED for review August 4, 1964.
ACCEPTED
November 2d, 1954.
Rotary and Packed Thermal Diffusion Fractionating Columns for Liquids LLOYD J. SULLIVAN, THOMAS C. RUPPEL, AND CHARLES 6. WILLINGHAM MeIIon Institute of Industrial Research, Pittsburgh 73, Pa.
The percentage separation of organic liquid mixtures b y thermal diffusion in a concentric tube fractionator, operated as a batch unit, is found to b e enhanced when the inner member i s rotated, or when the annular space is packed; glass wool is satisfactory as a packing material. Operated as each of three column types, a bench model apparatus gave the higher percentage separation when operated as a packed column. Construction details are given for the bench model apparatus, and for a 5-foot laboratory packed column having a 0.063-inch width b y 1.245-inch inside diameter annulus and a 200-ml. charging capacity. This laboratory packed column produces separations equivalent to a 5-foot Jones-Hughes open annulus column having a 0.01 2-inch width b y 0.625-inch inside diameter annulus and a 25-ml. charging capacity. The test mixtures used were n-hexadecane-decahydronaphthalene, n-hexadecane-1 -methylnaphthalene, and a urea nonadductible microcrystalline wax.
T
HE separation of fluids, particularly organic liquids, by thermal diffusion has recently been reported by several laboratories (4, 6-8). Furry and Jones ( 5 ) in a column theory have shown t h a t the separation obtainable varies as t h e inverse fourth power of the annular spacing, and t h a t equilibrium time of separation is approximately a n inverse seventh power function of the annulus. One type of apparatus used is the concentric tube column introduced by Jones and Hughes ( 7 ) . Jones-Hughes columns ”4 inch in diameter and 5 or 6 feet long with approximately a 0.01-inch annular space have charge capacities of 25 t o 30 ml. In order to separate larger amounts of material batchwise, several columns have been designed. These units retain the narrow open annulus of 0.01 t o 0.03 inch, and the columns have been modified by addition of reservoirs a t the ends ( 4 )or over the length of the unit (9). The time necessary for separation increases in proportion to the size and number of the reservoirs. 208
TWOconcentric tube batch fractionators without reservoir8 have been developed that will separate larger volumes of liquids within a reasonable time. One is a rotary thermal diffusion column, and the other is a column in which the annular space is packed. Glass wool is a satisfactory packing. A thermal diffusion unit with rotating members has been reported in the literature (11). However, it differs from the concentric tube columns in that it was a converted cream separator in which the heated and cooled cones were rotated to increaae the gravitational force field of the system. Debye (8,8 ) has reported and patented the separation of polymers by thermal diffusion. H e used a 10-cm. column packed with glass wool to show the enhanced separation. No other application of a similar column t o the separation of mixtures by thermal diffusion is known to the authors. The theory of separation of organic liquids by thermal diffusion is not as yet adequate t o predict the extent of separation
INDUSTRIAL A N D E N G I N E E R I N G C H E M I S T R Y
Vol. 47, No. 2
