Adsorption and Separation of Hydrocarbons on Mordenite Zeolites Paul E. Eberly, Jr. Esso Research Laboratories, Humble Oil & Rejining Co., Baton Rouge, La. 70821
The adsorption properties of c6-c9 hydrocarbons were investigated on several samples of hydrogen mordenite differing in degree of alumina deficiency. Conventional mordenite with a SiOz/AIz03 ratio of 12 [HM( 1 2 ) ] preferentially adsorbs n-nonane with respect to 2,2-dimethylheptane. Adsorption rates are slow. Alumina-deficient mordenite, HM(93), has lower diffusion resistance and shows no preference for the normal paraffin. With aromatics and paraffins of the same molecular weight, HM(93) exhibits no preference for the aromatic. Thus, with toluene/n-octane and toluene/n-heptane mixtures, the paraffin was preferentially adsorbed. These results are in contrast to those obtained with silica gel. Other adsorption and separation results are presented and discussed with respect to mordenite crystal structure.
M o r d e n i t e is a crystalline aluminosilicate whose structure has been firmly established by X-ray diffraction techniques (Meier, 1961). Also, nearly all of the alumina in the zeolite can be removed by extraction with mineral acids (Dubinin et al., 1968; Eberly and Kimberlin, 1970; Kranich et al., 1970). This removal of alumina induces critical changes in the nature of the acid sites and in adsorption and catalytic properties. A recent report (Eberly et al., 1971) showed that the total acidity, as measured by NHs adsorption techniques, and the concentration of OH groups decrease progressively as the alumina is removed. In regard to adsorption, the acid-extracted mordenites generally shoLv higher adsorptive capacity and less diffusion resistance for hydrocarbon molecules. However, there has not been universal agreement. Because of mordenite's unidimensional pore structure, small amounts of ifnpurities can exert enormous effects on the adsorption capacity and rates of adsorption. Consequently, differences in performance reported by various iiives tigators are not unexpected. For example, Kranich et al. (1970) reported that on their samples the diffusivity of cumene decreased with decreasing alumina content. Our studies, however, show that alumina-deficient mordenites exhibit considerably less diffusion resistance for cumene, n-decane, and decalin molecules (Beecher et al., 1968; Eberly and Kimberlin, 1970; Eberly et al., 1971). For catalytic purposes, the removal of alumina creates effects which appzar to depend on the nature of the reaction and size of the reactant molecule. Thus, previous investigations have shown that the cracking of cumene and hydrocracking of n-decane and decalin are much greater with a mordenite having a silica/alumina ratio of 64 than one with a ratio of 12 (Beecher et al., 1968; Eberly and Kimberlin, 1970). With n-pentane hydroisomerization, however, the activity decreased as the alumina was removed (Eberly et al., 1971). For some reactions, maxima in activity were observed a t intermediate levels of Si02/A1203 ratio (Weller and Bauer, 1969; Kranich et al., 1970). A review of mordenite catalysts has recently been given by Burbidge et al. (1970). The present investigation is concerned with the adsorption properties of Cs-C9 hydrocarbons on several samples of hydrogen rnordenite differing in degree of alumina deficiency.
On the basis of these results, a number of separation experiments involving binary hydrocarbon mixtures were conducted. The adsorption and separation were contrasted with those exhibited by amorphous silica gel. Experimental
Materials. A list of the solids used in this study is given in Table I. N a mordenite was obtained from the Norton Chemical Co. It was exchanged with NH4N03 solutions and subsequently calcined in air a t 1000°F to produce HhI(12). The aluminum-deficient mordenites HM(66) and HM(93) were prepared by extracting alumina from "(12) with 6h7 HC1 solution for times up t o 8 hr a t refluxing temperature ( ~ 2 1 2 ° F ) .The materials still exhibited the X-ray diffraction pattern expected for a highly crystalline mordenite. Silica gel was obtained from Davison Chemical Co. (Grade 408). Adsorption Measurements. A Cahn microbalance was used to measure the adsorption of pure hydrocarbons. About 20 mg of adsorbent was placed on one a r m of the balance. This was then evacuated and heated to 800°F to remove adsorbed impurities. The sample was then cooled to the desired temperature, usually 200"F, and small volumes of the hydrocarbon were pipetted into the system through a Hg-covered glass frit. Weight changes were continuously recorded. Pressures were read with a Model 145 precision pressure gage (Texas Instruments). Desorption rates were recorded by opening the balance to the vacuum pumping system. Separation Measurements. For separation studies, about 3 grams of the adsorbent was packed into a nominal 1/4-in. stainless steel column placed in a fluidized sand bath. The adsorbent was degassed with flowing helium a t 800°F and then the temperature lowered to 200°F for the separation experiment. The hydrocarbon mixture was introduced to the column by diverting the incoming helium flow through a saturator containing the liquid mixture maintained a t 65°F. I n most cases, the amount of hydrocarbon in the saturabor was greatly in excess of that volatilized during the experiment. Emergence of the hydrocarbon in the effluent stream was detected by a thermal conductivity cell and from the breakthrough curve, the Ind. Eng. Chem. Prod. Res. Develop., Vol. 10, No. 4, 1971
433
Table 1. List
of Solids
Designation
SiOz/AIz03 molar ratio
Surface area, M2/g
pv, cc/g
HM(12) H N (66) HM (93) Si02 gel
12 66 93 ..
540 602 602 696
0.27 0.49 0.37 0.45
adsorption capacity of the solid could be estimated. After breakthrough, samples of the effluent stream were periodically taken every few minutes and analyzed by a n F&hI gas chromatograph. The total composite was also trapped using liquid NB for additional analysis by mass spectrometry. Results
Single-Component Adsorption. The adsorption results of a number of hydrocarbons on hydrogen mordenites and silica gel are listed in Table 11. TI', represents the equilibrium adsorption capacity a t the indicated temperature and pressure. With H M ( l 2 ) , rates of adsorption are very slow and the capacities are enclosed in parentheses t o indicate t h a t it is questionable whether true equilibrium was reached in these experiments even though several hours 1Fere allowed for equilibration. With the CS hydrocarbons, HM(12) exhibits a marked preference for the normal paraffin. Adsorption and desorption rate curves are plotted in Figure 1. After 1 hr the capacity for n-nonane is nearly four times that of 2,2-dimethylheptane illustrating the molecular sieving action of HM(12). Alumina-deficient samples do not have this property. With HM(66) and HU(93), equivalent amounts of the two Cs isomers are adsorbed as shown in Table I1 and Figure 2 . Other improved properties of the
Table
II. Hydrocarbon Adsorption Results
Solid
Hydrocarbon
H M (12)
Toluene n-Octane n-Konane 2,2-Dimethylheptane Toluene n-Nonane 2,2-Dimethylheptane n-Hexane Benzene n-Hept ane Toluene n-Octane n-Nonane 2,2-Dimethylheptane Toluene n-Hept'ane
HiCI (66)
HM (93)
Si02 gel
alumina-deficient mordenites are rapid attainment of adsorption equilibrium and comparative ease of desorption. More extensive studies on HM(93) indicate that the solid is rather insensitive t o the type of hydrocarbon molecule of a given molecular weight. Thus, nearly equivalent capacities are achieved for n-hexane and benzene. The same is true for n-heptane and toluene. Figure 3 contains the desorption rate curves for these hydrocarbons on HM(93). V7 represents the amount of material remaining on the solid at any time t and W eis the initial equilibrium capacity. The differences in desorption rates appear t o be solely attributable to molecular weight rather than molecular type. This behavior differs from that observed with pure silica gel. The pores in silica gel are nonuniform and are significantly larger than the uniform "molecularly sized" pores in mordenite. Consequently, silica gel does not have as high a n adsorption capacity at low pressures. This can be seen from the toluene isotherm data in Table 11. At 200°F, the hydrocarbons are too easily removed from silica gel to permit distinctions to be made among the various hydrocarbons. At 78"F, desorption rates comparable to those on "(93) are obtained and these are shonn in Figure 4. The relative ease of removal of the hydrocarbons is quite different from that on HM(93). Toluene is most tightly held t o the surface. This is followed, in turn, by n-octane and n-heptane. Separation Experiments. 011the basis of the desorption rates of pure compounds, it can be predicted that aluminadeficient mordenite will preferentially retain n-octane from its mixture with toluene and consequently affect the preferential adsorption of a paraffin from a n aromatic. SiOz gel would do the opposite. This is substantiated by the results shown in Figure 5 and tabulated in Table 111. Toluene has the higher vapor pressure, and on this basis, one would expect n-octane t o be preferentially adsorbed.
%-Octane Toluene
434 Ind. Eng. Chem. Prod. Res. Develop., Vol. 10, No. 4, 1971
Temp,
P,
We
O F
mm
mrnol/g
200 200 200 200 200 200 200 200 200 200 200 200 200 200 78 78 78 78 200 200 200 200 200 200 200
0.77 0.43 0.41 0.46 0.73 0.41 0.44 0.72 1.06 0.64 0.77 0.43 0.40 0.46 0.73 0.61 1.10 0.64 0.96 1.60 2.24 2.67 3.26 4.27 5.15
(0.49) (0.29) (0.38) (0.11) 0.31 0.29 0.31 0.34 0.39 0.38 0.44 0.37 0.41 0.43 1.37 0.81 1.02 1.22 0.12 0.18 0.24 0.27 0.33 0.37 0.41
Table 111. Separation of Hydrocarbon Mixtures at 200°F Vapor pressure of pure components Mixture, mol
70
S i 0 2 gel
@ 2OO0F,
ff
mm
Distillation
Ads. capacity mmol/g
345 275 650 345 1140 650 1140 1060 1780 1140
1.25
1.17
1.88
96. 57y0 Toluene 3.43% %-Octane 11.58% %-Heptane 88.42y0 Toluene 95. %yoBenzene 4.72% %-Heptane 95.85% Benzene 4 . 15y0 Cyclohexane 6.86% Hexene-1 93, 14y0 Benzene
HM(931
Pref. ads. compound
Ads. capacity, mmol/g
Pref. ads. compound
Toluene
0.78
n-Octane
...
...
0.72
n-Heptane
1.75
1.28
n-Hept ane
0.95
n-Heptane
1.08
...
0.95
Cyclohexane
1.56
1.28
1.08
Hexene-1
... Benzene
TIRE, R l N
Adsorption of
Figure 1. 2OO0F
C9 hydrocarbons
on HM(12) at TIME, WIN.
The circles and squares refer to the adsorption o f n-nonane and 2,2dimethylheptane, respectively
ADSORPTION PATE AT 0 . 4 MM.
0.05
a
I IO
I zo
I
I
I
I
I
40
so
o
IO
TIME.
Figure 2. 200°F
Adsorption of
The symbols, o,*,O,O,V, refer to the desorption of n-hexane, benzene, n-heptane, toluene, and n-octane, respectively
DESORPTION RATE U M E R V A C U u l
30
I zo
I
I
I
30
40
50
I
nin.
Cs hydrocarbons
Figure 3. Desorption of hydrocarbons from HM(93) at 200°F under vacuum
on HM(93) at
The circles and squares refer to the adsorption of n-nonane and 2,2dimethylheptane, respectively
This is the case with HM(93) which is relatively insensitive to the type of molecule. SiOz gel, like most other adsorbents, shows a preference for aromatics. With a n-heptaneltoluene feed (Table 111), HlI(93) still shows a preference for the normal paraffin, although the separation is not so effective as with the n-octaneltoluene feed. Similar experiments with the tolueneln-paraffin mixtures were conducted on HM(12) but no separations were achieved. Results on a benzeneln-heptane mixture are given in Figure 6. I n this case, silica gel and HhI(93) behave in a similar manner in that both prefer to adsorb the n-heptane. Apparently, the larger difference in volatility offsets the inherent preference of silica gel for aromatic compounds. I n regard t o cycloparaffins, HRI (93) will preferentially adsorb cyclohexane lvith respect to benzene as seen in Table 111. Another example of the dissimilarity between HM(93) and silica gel is seen by the results on the separation of a hexene-llbenzene mixture (Figure 7). Silica gel preferentially Ind. Eng. Chem. Prod. Res. Develop., Vol. 10, No. 4, 1971
435
100
r
0
0.02
l a i
004
I
I
I
I
0.05
0.08
0.12
0.11
1
GRMS EFFLUL'(1lGRW SOL10
Figure 6. Separation of a 95.28% benzene-4.72% heptane mixture at 200°F
n-
The circles and squares refer to experiments on HM(93) and silica gel, respectively
100
90 80
70
E T I M . 113.
'*
Figure 4. Desorption of hydrocarbons from SiOz gel at 7 8 ° F under vacuum I?, V,
The symbols, respectively
0,
and
refer to n-heptane, n-octane, and toluene,
a
60
50 40
30 20
h
lo 0
0.02
0.04
0.01
0.W
0.10
Figure 7. Separation ,of a 93.14% hexene-1 mixture at 200°F
100 98
0.12
0.14
benzene-6.86%
96
The circles and squares refer to experiments on HM(93) and silica gel, respectively
94 92 90
2
88 86
84 82
I' 0
I
I
I
I
0.02
0.04
0.06
0.08
I 0.10
I
I 0.12
0.14
0.11
GRAMS E F F L U t N l l t R R n SOLI0
Figure 5. Separation of a 96.57% octane mixture at 200°F
toluene-3.43%
n-
The circles and squares refer to experiments on H M ( 9 3 ) and silica gel, respectively
adsorbs the benzene. With mordenite, the opposite is true. I n the latter case, there was an insufficient amount of liquid in the saturator t o permit breakthrough t o feed concentration. All the hexene-1 was adsorbed by HM(93) and only pure benzene issued from the column. After the run, the adsorbent could be easily regenerated by desorption with water vapor and heating t o elevate temperature. This indicates that polymerization of hexene-1 was not extensive. Discussion
Since the chemical composition of silica gel and aluminumdeficient mordenite, HM(93), are nearly the same, th8 differences in adsorption behavior must be attributed to differences in crystal and pore structure. The surface of amorphous silica gel is contained among pores of a wide variety of sizes generally much larger than those in mordenite. Hydrocarbon molecules are considerably smaller than the pore diameter and, consequently, wall effects are of a different nature than those of mordenite. The pore structure 436
Ind. Eng. Chem. Prod. Res. Develop., Vol. 10, No. 4, 1971
of mordenite consists of parallel tubes each having an approximately elliptical opening with a major and minor diameter of 6.95 A and 5.81 %I, respectively (Meier, 1961). There are smaller side pockets perpendicular t o the main tubes which can only adsorb molecules smaller than nbutane (Barrer and Peterson, 1964). These side channels have restrictions which essentially prohibit motion of molecules from one main tube to the other. The experimental observations of this investigation show that among hydrocarbons of a given molecular weight the alumina-deficient mordenite does not have a preferential affinity for aromatic molecules. This is different from silica gel and most other adsorbents and could be owing to both thermodynamic and kinetic factors. I t is possible that because of the presence of side openings in the pore channels, planar aromatic molecules cannot interact as strongly with the mordenite surface as would otherwise be expected. This is supported by previous work on conventional hydrogen mordenite which showed that heats of adsorption for benzene, cyclohexane, and n-heptane were 12.5, 13.5, and 16.0 kcal/ mol, respectively (Eberly, 1963). From a kinetic viewpoint, it is possible that the nonplanar molecules may be able to enter the side pockets to a degree which retards their motion through a pore relative t o that of the planar aromatic. I n any event, it is suggested that the presence of the side pockets is responsible for the adsorption behavior reported here. Acknowledgment
The author expresses his sincere appreciation to Dorothy Webb for her excellent experimental assistance.
literature Cited
Barrer, R. &I., Peterson, D. L., Proc. Roy. Soc. ( A ) , 280, 466 (1964). Beecher, R., Voorhies, A., Jr., Eberly, P. E., Jr., Znd. Eng. Chem. Prod. Res. Develop., 7, 203 (1968). Burbidge, B. W., Keen, I. M., Eyles, 11. K., International Conference on Molecular Sieve Zeolites, Worcester, Mass., 1970. Dubinin, M. M., Federova, G. M., Plavnik, G. M., Piguzova, L. I., Prokofeva, E. N., Zzv. Akad. Aauk SSSR, Ser. Khim, 11, 2429 (1968). Eberly, P. E., Jr., J.Phys. Chem., 67,2404 (1963). Eberly, P. E., Jr., Kimberlin, C. N., Jr., Ind. Eng. Chem. Prod. Res. Devehp., 9, 335 (1970).
Eberly, P. E., Jr., Kimberlin, C. N., Jr., Voorhies, A., Jr., paper presented a t Second North American Meeting of the Catalysis Society, Houston, Tex., February 24-26, 1971. Kranich, W.L., Ma, Y. H., Sand, L. B., Weiss, A. H., Zwiebel, I., International Conference on Molecular Sieve Zeolites, Worcester, Mass., 1970. Meier, w, z, Krist., 115, 439 (1961), Weller, S. W., Bauer, J. M., "Studies of the Catalytic and Chemical Properties of Acid-Extracted hIordenite," Preprint, 62 Ann. Meeting A.I.Ch.E., Washington, D.C., 1969. RECEIVED for review March 12, 1971 ACCEPTED July 12, 1971
Ammonium Potassium Polyphosphate Fertilizer Bench-Scale Production Charles A. Hodgel and David R. Boylan Engineering Research Institute, Iowa State University, Ames, Iowa 60010
The technical feasibility of producing an ammonium potassium polyphosphate fertilizer was studied with bench-scale processing equipment. The kinetics of the reaction between potassium chloride and phosphoric acid was experimentally investigated, and the constants in the Arrhenius equation were determined for the forward portion of the reaction. A monopotassium orthophosphate-phosphoric acid melt was ammoniated, and the variables of temperature, pressure, residence time, and ammonia-to-phosphoric acid feed ratio were studied. A specially designed reactor was constructed for the ammoniation reaction. Analysis of the product ranged up to 8-26-17 (NPK) during optimum operating conditions. Experimental results showed that the degree of ammoniation of the final product increased with increasing pressure and residence time; however, the degree of ammoniation decreased with increasing temperature in the range of 177-232OC. The optimum ammoniation conditions for a melt of K:P mole ratio equal to 0.50 were 200-210°C with a pressure in excess of 40 psi and a residence time above 4 min.
I n the past 10 years bhe fertilizer industry has undergone considerable change. The general trend in the industry has been toward large-capacity single-stream processing units for all three of the primary plant food nutrients. Also, the coiicentration of single-plant nutrient fertilizers has increased steadily through the years. I n addition, mixed fertilizers have shown a substantial increase in total plant food analysis. According to published data (TVA, 1969), the average nutrient analysis (N, P205,and K20) of mixed fertilizers containing two or more primary nutrients has risen from 23.1y0 in 1950 to 38.2% in 1968 and is expected t o be slightly above 43% by 1975. Demand has been increasing in the liquid fertilizer industry for phosphates in the polyphosphate form for formuating higher analysis nonprecipitating liquid fertilizers. In view of this trend, ammonium potassium polyphosphate is of considerable interest, being one of the highest analysis fertilizers known. When transportation, marketing, and agronomics are considered, such a fertilizer could be quite 1 Present address, Division of Chemical Development, Tennessee Valley Authority, Muscle Shoals, Ala. 35660. To whom correspondence should be addressed.
advantageous compared with present three-component fertilizers. Ammonium potassium polyphosphate can be made by reacting potassium chloride with an excess of phosphoric acid and then neutralizing the resulting potassium phosphate and excess phosphoric acid with ammonia. Bench-scale work reported in this paper was undertaken to determine the technical feasibility of such a process. Previous Work
A novel method of producing a complete concentrated fertilizer was tested on a small pilot-plant scale and reported as early as 1925 (Ross, 1925). This fertilizer contained 5,6% N, 24.7% P, and 15.6% K or 5,6% T\;, 56.6% PzOS,and 18.8% K 2 0 . The following main reactions were reported: KCl
+ 2H3P04
+
KHzPOI
+
+ HClt
This process involved two steps. I n the first, potassium chloride was reacted with two or more equivalent amounts of concentrated phosphoric acid a t 25OoC. Three products were obtained : volatilized hydrogen chloride, a liquid solution of Ind. Eng. Chem. Prod. Res. Develop., Vol. 10, No.
4, 1971 437