SEPARATION OF n-HEXANE FROM A

SOLUTION IN BENZENE BY ADSORPTION

ON MOLECULAR SIEVE 5 A E P H R A I M K E H A T A N D Z U S l A R O S E N K R A N Z

Llefiarfment of Chemicul Engintering, Y'echnion -Israel Inrtitute of 7'rchriology, E f a q a , Israe

Adsorption of n-hexane from a 6 mole 70 solution in benzene, on Molecular Sieve 5A, was studied at tem- peratures of 150" to 250' C., pressures of 40 to 150 p.s.i.g., and Reynolds numbers of 2.2 to 47. Adsorp- tion capacity and length of the mass transfer zone are reported. Decrease of over-all mass transfer co- efficients with increased flow rate is attributed to the requirement of orientation of the adsorbed molecules at the surface openings of the molecular sieve.

E\V economic industrial separation operations based on adsorption were made possible by the advent of moleciilar

sieves. Molecular sieves have high specificity for particular molecular sizes, high capacity a t low concentrations, and stability a t high temperatures.

I n small-scale operations, the adsorbent is stripped by heating. T h e time required for heating and cooling the adsorbent and the heat exchange facilities required make an adsorption cycle. using desorption by heating, unattractive for industrial use.

'The stability and long life of molecular sieves a t high tern- peratiires made possible a pressure cycle. where adsorption takes place under pressure and desorption is accomplished by reducing the pressure. T h e whole cycle is run a t temperatures over 200' C. This results in a loss of adsorptive capacity; but the increased rates of adsorption. and particularly desorp- tion, make u p for the loss of capacity.

A number of cornpanies offer adsorption separation processes utilizing a pressure cycle for a variety of systems ( 7 , 8-70), However, except for dehumidification processes (77), no design data are available in the literature on separation processes utilizing molecular sieves.

T h e primary ob,ject of this work was to obtain data that will be useful for the design of an adsorption system for the separa- tion of straight-chain aliphatic hydrocarbons from aromatics. For convenience in analysis n-hexane was separated from a solution of n-hexane in benzene by adsorption on Molecular Sieve 5A, which adsorbs n-hexane and does not adsorb benzene (76). Another object was to study the mechanism of adsorp- tion bv molecular sieves.

Adsorption is inherently a batch operation.

Experimental Procedure

T h e feed solution was pressurized with nitrogen in a 3.5- liter stainless steel container, and was passed through a capillary flo\vmeter and a flow regulator (Moore Products, Model 63RD-1), through an evaporator coil and the adsorption tube (Figure 1). T h e evaporator coil was a coil of '/l-inch copper titbing 1.0 meter long. T h e adsorption tube was a 3/a-inch (7.2-mm. i.d.) copper tubing 3.68 meters long, coiled into a 3-inch coil. about 0.5 meter long. T h e coil contained 110 grams of Molecular Sieve 5A in the form of 1,'i6-inch graniiles. The void fraction was 0.632. Three thermocouples soldered to the coil \\.ere spaced a t the beginning, middle, and end of the adsorption tube.

T h e two coils were held in a constant temperature bath made from a vertical 3-inch steel pipe, 1.20 meters long, filled with paraffin, around which resistance \\tire was wound. T h e

paraffin c o d d be heated up to 400' C. At steady state? the temperatiire {vas controlled Lvithin 1" C. and the difference between the readings of the three thermocouples was less than l o c.

Another coil of ';'d-inch copper tubing \vas used to cool the paraffin. A condenser made from 0.30 meter of a finned l l g - inch copper tubing, in a water bath. \\'as used to condense the effliient.

I n operation, the thermostar was heated to the desired temperatiire and the controller \vas set. T h e system \vas pressiirized \vith nitrogen a t the pressirre selected for the run. Approximately 1 hour was required to achieve a constant tem- perature along the adsorption tube. T h e nitrogen connection to the evaporator was closed and all the valves in the feed and analysis lines were opened. T h e flow rate depended on the relative opening of the valve in the flow regulator and the valve a t the end of the system, and could be set only approui- mately for a predetermined flo\v rate. T h e nitrogen initi,illy present in the system was expelled through the analysis line, ahead of the benzene.

At the appearance of the first drop out of the condenser the time \vas recorded and a I-ml. sample was taken a t intervals from ' / 2 to 20 minutes. 'The total floiv during the run was 100 to 350 ml. The samples Ivere analyzed by a thermostated Abbi: refractometer, n i t h an acciiracy of 0.05% mole n-hexane in benzene. (The original benzene used was C.P. grade and the n-hexane purity was better than 99%',.) The adsorption stage \vas terminated when the ratio of efffuent to feed con- centration was over 0.98. Desorption was achieved by flowing nitrogen through the system for 10 minutes and then increasing the temperature to 300' C. and decreasing the pressure to 0.1 p.s.i.g. for 30 minutes, \vith the help of a vacuum pump. A glass trap Ivith acetone and dry ice protected the oil in the vacuum piimp from contamination by the effluent. This

coolma

P

Figure 1 . Schematic diagram of apparatus

V O L . 4 NO. 2 A P R I L 1 9 6 5 217

I

. 250°C.150R~lG 0 250OC. 100 PSJG A 250OC. 70 PSlG

0 200'C. 100 RSIG

- - - I I I 1 I I

I 2 5 IO 20 ! Reynolds number

Adsorptive capacities of Molecular Sieve 5 A for Figure 2. n-hexane

desorption method did not permit a check of the material balance of the adsorbed material. as the holdup included the material in the evaporator and the condenser.

The range of flo\c rates was 0.4 to 5.6 ml. per minute \cith an accuracy of 0.1 ml. per minute. Runs \there the flo\c rate \vas not kept constant \cere discontinued.

The pressure \cas read by Bourdon gages before the evapora- tor tubing and after the adsorption tube. 'The accuracy of the gages \cas k 2 . S p.s.i.g. and the pressure tvas taken as the average of the t\co pressure gages. In the lo\c pressure runs the pressure drop a t high flow rates was a considerable frac- tion of the inlet pressure, \chich is probably reflected in the greater scatter of these data . T h e lo\v ratio of adsorption tubing to granule size (approximately 4.5) could not be avoided. Tubing of larger diameter did not have enough heat transfer area to keep the temperature inside the adsorption tubing constant, and the use of molecular sieve powder resulted in an excessive pressure drop. Since granules \vi11 be used in an industrial operation. the use of po\vder \vas also undesirable because of scale-up problems.

Results

All runs were made \cith one feed concentration-6 mole % n-hexane in benzene-and one adsorbent bed. T h e same adsorbent gave results reproducible Lvithin 5% after more than 200 hours of adsorption operation.

Runs were made at 150', 200'. and 250' C. and 40, 70. 100, and 150 p.s.i.g. (In Figures 2 to 4 only the runs made at 250' C. and 40, 70, 100j and 150 p.s.i.g., and a t 100 13.s.i.g. and I j @ O , 200' and 250' C. are shomm.) T h e range of flo\c rates \cas 390 to 7260 kg.,'hr. sq. meter. corresponding to Reynolds numbers of 2.2 to 47.

The experimental data were in the form of concentration- time curves. The form of these curves was a horizontal line of zero concentration up to the breakthrough point and an S- shaped curve thereafter.

T h e total time for a run, te. was defined from the appearance of the first drop to the time the effluent concentration \cas

I I I I I I

- 250oc.150 RSIG. 0 250OC. 100 PSI6 A 250OC. 70 RSJG.

-v 25OOC.40 RSJG; 0 20O0C.1O0~SIG

r?

E E Y A 150°C.100 RSlG

0)

= I E ii

/ - - - ). - - e -

0.5 -

Reynolds number

Figure 3. Length of mass transfer zone

0.98 of the feed concentration. The breakthrough time, 16: was defined from the appearance of the first drop to the time the effluent concentration was 0.02 of the feed concentration. The total amount adsorbed per unit mass of adsorbent, Q T . was calculated by graphical integration of the effluent con- centration-time curves.

where TI' is the benzene mass flow rate, E O the mass of ad- sorbent. Yo the feed mass fraction (kilograms of n-hexane per kilogram of benzene), and Y the mass fraction of n-hexane in the effluent, a t any time.

The total amount adsorbed was in the range of 0.15 to 0.6 of the static equilibrium concentration (9: 20).

In an industrial operation the adsorption stage would be terminated at, or before. the breakthrough time and only the amount adsorbed per unit mass of adsorbent to the break- through point. Q B , need be considered here. This value was defined as the adsorptive capacity and was calculated by :

The adsorptive capacity as a function of pressure. temperature. and Reynolds number is given in Figure 2.

T h e Reynolds number was calculated by the Leva formula for a fixed bed of cylindrical particles (74. The viscosities and densities of pure benzene were used. T h e effect of pres- sure on viscosity \\.as estimated by the method of Grunberg and Nissan (78).

The length of the mass transfer zone. Z,. was defined as the length of column required to increase the effluent concen- tration from 0.02 to 0.98 of the feed concentration. I t \vas calculated by assliming that a t the breakthrough point the mass

218 l & E C P R O C E S S D E S I G N A N D D E V E L O P M E N T

so

c

'r 20 0

0 X 1

I -

20 -

2 5 IO 20 50 Reynolds number

Figure 4. Mass transfer coefficients

transfer zone has still to travel its o\cn length, and that the amount adsorbed a t any point is proportional to the distance traversed by the mass transfer zone

z, = z 1 - -- ( 3 where Z is the length of the adsorption tube. T h e results are given in Figure 3.

Over-all mass transfer coefficients. kga, were calculated by the method of Ilougen and hlarshall (5, 7 7 ) , neglecting the small change of the vapor flow rate as the adsorption proceeds. Point values of kga \cere calculated from the experimental concentration-time curves, a t Y,'l'o intervals of 0.1, and averaged. T h e spread of the point values around the average value of kga was less than 2076. From the average value of kga a smooth time-concentration curve could be calculated. which fitted the experimental curve closely. The average values of kga. calculated by this method, are given in Figure 4.

Discussion of Results

T h e adsorption isotherms for n-hexane on Molecular Sieve 5A converge a t high pressures, a t the maximum sieve capacity of 11 kg. of n-hexane per 100 kg. of adsorbent (9). At higher temperatures, higher pressures are required to approach that value. This value can be considered as the limiting capacity at zrro flo~c rate. At increased Reynolds numbers the adsorp- tive capacity is decreased (Figure 2). Pressure and tempera- ture affect the adsorptive capaciry and the adsorption isotherms in a similar manner. At any Reynolds number. increased preswre and decreased temperature increase the adsorptive capacity. ,At lo\vrr pressures. 40 to 100 p.s.i.g.. the pressure effect is less significant than a t 100 to 150 p.s.i.g. and the ef- fect of temperature decreases a t highrr Reynolds numbers.

'The best operating conditions for an indrrstrial operation \$-i l l he determined by an economic balance, birt some indication of these conditions can be obtained from Figure 2.

T h e rate of desorption increases with increased temperatiire. In order to decrease the time of the desorption part of the operating cycle? it is desirable to operate a t the Iiighest prac- tical temperature, At higher flo\v rate5 the drcrease of ad- sorptive capacity with increased temperature is sinal1 and operation a t 250' C . \vi11 he reasonable.

T h e cost of the adsorption vessel \vith its diixiliaries \vi11 increase wit11 increased pressure. Since the adsorptive capac- ity does not increase much \ \ i th pressurr l i p to 100 p.s.i,g,. the optimum operation pressiirr \ \dl probably b r foiind bet\\-rrn 100 and 150 p.s.i.g.. if an economic balance is made o f opvrat- ing in the range of 40 to 150 i1.s.i.g.

Increased flo\v rate reduces the time of the adsoiption part of the cycle more than it rediicrs the capacity of the adsorbent. At 2.50' C.. doubling the flo\v rate reduces the hrrakthrough time by 4: and the adsorption capacity by 2. I t \\ill not bc advantageous, hon.ever. to reduce the adsorption time milch belo\e the desorption time. The desorption iinie is relatively independent of the capacity, as it is deterniinrd by the timc required to pump d o \ m the system to a lo \ \ prtwure. \vhich is determined by the size of the adsorptive pellets and shape of the adsorption bed.

The adsorption capacity in the practical rangr of the olrrat- ing variables is 10 to 30% of the rnaxiriirim capacity of the adsorbent.

'This low capacity has one advantagr. I t is possible to operate the adsorption c)-cle adiabatically. For 3OYG of maximum capacity, assliming that all thr heat of adsorption is retained by the adsorbent. adsorbent heat capacity of 0.24 cal. per gram 'C. (75). heat of adsorption of 12 kcal. per mole (7 ) : the temperature rise of the adsorbent \\ill be only about 20' C. This heat \vi11 be recovered in thr desorption part of the cycle.

'The length of the mass tranbfer zone for the same range of operating variables as in Figrirr 2 is givrn in Figure 3. T h e length of this zone determines the minimum length of the adsorption bed needed to get any separation.

T h e length of the mass transfer zone is relatively insensitive to temperatures in the range of 150' to 250' C. and to pressiircs a t high flow rates. The length of the mass transfer zone is considerable. about 1.5 to 2 meters a t a Reynolds niimber of 10. and increases rip to 3 meters a t higher flow rates. 'This puts a severe limitation on the flo\v rate, as a short bed is dc- sired for faster desorption.

T h e effect of bed length was not studied in this Lvork. Su t t e r and Burnet (77) have sho\vn: for adsorption of \vater on molec- ular sieves, that the length of the mass transfer zone is in- dependent of the length of the adsorbent coliimn.

Camp- bell et al. (6) have sho\vn that for adsorption of n-hexane on silica gel. the total amount adsorbed increased \\ith inlet concentration up to a concentration of 1 pound of n-hexane per 100 pounds of inlet gas. and was not affected hy highrr concentrations. bfolecular sieves can he exprcted to sho\v a similar independence of adsorptive capacity on inlet concen- tration. above feed concentrations of 1 to 2'j: of n-hrxanr.

The mass transfer coefficients were calciilated by the merhod of Hoiigen and hiarshall ( 7 7 ) . This method lrimps all thc resistances to mass transfer. a t the surface of the pellets. be- ticeen the pellets, inside the pellets. and other effects. into one over-all coefficient. I t assiimes a linear adsorption eqiiilibriiim relationship. HoIvever. in the calciilation of the mass transfrr coefficient it is not necessary to iisr the eqtration of the rqriilib-

T h e feed concentration was not varied in this \vork.

V O L . 4 NO. 2 A P R I L 1 9 6 5 2 1 9

rium line, and the problem of fitting a “best” straight line through the equilibrium curve is bypassed.

T h e mass transfer coefficient increases with increased tem- perature. Above 200’ C. the effect of temperature is negli- gible. T h e low values of the mass transfer coefficient a t 150’ C. and 100 p.s.i.g. may be caused by the presence of a liquid phase. Mass transfer coefficient increases with pressure. This effect is less pronounced a t low flow rates. T h e effect of flow rate on the mass transfer coefficient is unusual. Increased flow rate decreases the mass transfer coefficient. Generally, increased flow rate decreases the diffusional boundary layer, decreasing the resistance to mass transfer.

In some cases, particularly where mass transfer is controlled by diffusion within particles ( a ) , the mass transfer rate is insensi- tive to flow rate. Decrease of mass transfer coefficients with increased flow rate, in fluidized beds, was reported by Kettern- ing, Manderfield, and Smith (72) for desorption of water. and by Resnick and White (79) for evaporation of naphthalene a t low Reynolds numbers. However. there is no parallel be- tween their studies and this work

T h e proposed explanation of this phenomenon is based on the special properties of molecular sieves. Molecular Sieve 5A contains large cavities. regular in shape, having a free diameter of 11.8 A. Each cavity has six windows of free diameter of nominally 4.2 A. ( 2 ) , and effectively about 5 A. T h e diameter of normal paraffins is about 4.9 A. ( 2 ) . By stretching out the paraffin chain, spiraling motion, or accom- modation of the positions of the hydrogen atoms to the concave parts of the surface, the molecules can enter through the surface openings. T h e n-hexane molecules have to be properly orientated in order to get through the surface openings. Increased flow rate increases the number of collisions of other molecules with a molecule about to enter, or partly through, a surface opening, and decreases the residence time of the mole- cules a t the openings. T h e cross section a molecule presents parallel to the opening is much smaller than the average cross section it presents normal to the opening. Therefore, an increased frequency of collisions with other molecules, due to increased flow rate, hinders the movement of a molecule through a n opening.

T h e requirement for proper orientation of a molecule a t a surface opening of Type A Molecular Sieve was also found by KvitkovskiI and Sergienko (73) from energy considerations.

T h e decrease in rates of adsorption of normal hydrocarbons on Molecular Sieve 5A with increased chain length, reported by Barrer and Ibbitson (3) and Schwartz and Brasseaux (21), also supports this explanation.

Nomenclature

kga = mass transfer coefficient, hr.-l Q B = adsorptive capacity, kg. of n-hexane adsorbed up to

breakthrough time per 100 kg. of adsorbent QT = total amount adsorbed, kg. of n-hexane per 100 kg. of

adsorbent t b = breakthrough time, min. te = total time of adsorption, min. W = mass flow rate, kg./min. w o = mass of adsorbent, kg. Y = mass fraction of n-hexane in effluent, kg. of n-hexane

YO = mass fraction of n-hexane in feed, kg. of n-hexane per kg.

2 = length of adsorption tube, meters 2, = length of mass transfer zone: meters

per kg. ofbenzene

of benzene

Literature Cited

(1) Bacon, K. H., Henke, A. M., Petrol. Re jner 40, No. 4: 109

(2) Barrer, R. M., Brit. Chem. Eng. 4, 267 (1959). (3) Barrer, R. M., Ibbitson, D., Trans. Faraday Soc. 40, 195 (1944). (4) Beaton, R. H.. Furnas, C. C., Znd. Eng. Chem. 33, 1500 (1941). (5) Bird, R. B., Steward. LV. E., Lightfoot, E. N., “Transport

(6) Campbell, J. M., Ashford, F. E., Needham: R . B., Reid,

(7) Eberly, P. E., J . Phys. Chem. 66, 812 (1962). (8) Franz, W. F., Christensen, E. R., May: J. E., Hess, 13. V.:

(9) Griesmer, G. J., Rhodes, H. B., Kiyonaga, K., Zbid.. 39, No. 6 ,

(1961).

Phenomena,” p. 702, LViley. New York. 1960.

L. S., Petrol. ReJiner 42, No. 12, 89 (1963).

Petrol. ReJiner 38, No. 4, 125 (1959).

125 (1960). (10) Henke,’A. M., Stauffer, H. C., Carr, N. L., Zbzd., 41, No. 5,

(11) Hougen, 0. A,, Marshall. LV. R., Chem. Eng. Progr. 43, 197 161 (1962).

11947). (12) Kgtterning, K. N., Manderfield, E. L., Smith, J. M.. Ibid., 46. 139 (1950).

(13) ’Kvitkbvski!. L. N., Sergienko, S. R.. Proc. Acad. Scz. L’SSR,

(14) Leva, M., “Fluidization,“ p. 45, McGraw-Hill, New York; Phys. Chem. Sectton 147 (1-6), 890 (1962).

1 o<o .,”,. (15) Linde Co., “General Information on Linde Molecular Sieves.

Types 4A and 5A,” Form 8605-B. (1 6) Linde Co., “Molecular Sieves. Hydrocarbon Data Sheets,”

Form 9692-C (1959). (17) ”utter, J. I., Burnet, G., A.I.Ch.E. J . 9, 202 (1963). (18) Reid, R. C., Sherwood, T. K., “Properties of Gases and

(19) Resnick, W., LVhite, R. R., Chem. Eng. Progr. 45, 377 (1949). (20) Rosenkranz, Z., M. S. thesis, Israel Institute of Technology,

(21) Schwartz, R. D., Brasseaux, D. J., Anal. Chem. 29, 1022

Liquids,” p. 198, McGraw-Hill, New York, 1958.

Haifa, Israel, 1963.

(1 957).

RECEIVED for review May 5; 1964 ACCEPTED October 20, 1964

220 I&EC PROCESS D E S I G N AND DEVELOPMENT