Oxidation and Ammoxidation of Propylene over Bismuth Molybdate Cat a Iyst James 1. Callahan, Robert K. Grasselli, Ernest C. Milberger, and H. Arthur Strecker Research and Deuelopment Department, The Standard Oil Co (Ohio), Clewland, Ohio 44128
I
CALLAHAN
GRASSELLI
JAMESL. CALLAHANis Manuper for Chemicals and Monomrs Synthesis Research for The Standard Oil Co. of Ohio. Dr. Callahan has twenty years of research in refining processes, catalytic chemistry, and chemicals manufacture. His entire’ professional career has been with The Standard Oil Co. of Ohio. Dr. Callahon received his B.S. from BoldwinWnlloce College and his M.S. and Ph.D. from Case Western Resew University. ROBERTK. GRASSELLI is a Senior Research Associate in the area of petrochemical heterogeneous catalysis with The Standard Oil Co. (Ohio), where he began his professiond career in 1952. In addition to research in chemicals and mommer manufacture, his interests included the fields of liquid thermal diffusion and fuel cells. He receiued the A.B. in chemistry from Haruard Uniwrsity (1952) and his M.S. (1955) and Ph.D. (1959) fmm Case Western Reserue Uniuersity. Dr. Gmselli is a member of the American Chemical Society, Catalysis Society, Huruard Association of Chemists, and Sigma Xi.
134
Ind. Eng. Chern. Prod. Res. Develop., Vol. 9, No. 2, 1970
MILBERGER
STRECKER
ERNESTC. MILBERGER receiued his A.B. and M.A. in chemistry from Uniwrsity of Missouri in 1941-1942; his Ph.D. from Western Reserve Uniwrsity in 1957. He was associated with The Texas Company, Beacon, N . Y., from 1942 to 1946; since 1946, he has been with The Standard Oil Co. (Ohio). Research intere.vts with Sohio included work on liquid thermal diffusion, grease and industrial lubricants, and bench scale process research and catalysis in the petrockmical field. He was a member of research teams who were awarded scrolls as “Modern Pioneer in C r e a t i ~Industry” for development of the Sohio Acrylonitrile Pmcess by N A M , and an I-R-100 award for compounded uranium catalyst, Catalyst-21, He is a member of the American Chemical Society andsigma
xi.
H. ARWURSTRECKER received a B.A. in chemistry from Cornell Unisrsity in 1940 and his Ph.D. from Cornell in 1948. He was associated with OSRD High Explosives Division from 1942 to 1945. From 1948 to date, he has worked in the fild of petmleum and chemical catalysis with The Standard Oil Co. (Ohio). He is a member of the American Chemical Society, Catalysis Society, Society of Applied Spectroscopy, and Sigma Xi.
Research on the catalytic oxidation and ammoxidation of propylene is described which led to the development of commercial processes for the production of acrolein and acrylonitrile. The initial research comprised a search for regenerable metal oxide ”oxidants” of high activity and selectivity. This line of attack culminated in the discovery of a direct air oxidation catalyst based on bismuth molybdate, which gives high single pass conversion of propylene using near stoichiometric ratio of feeds. Conclusions are drawn from literature as well as from unpublished data in summarizing some of the important aspects of propylene oxidation and ammoxidation chemistry over bismuth molybdate catalysts.
THE
past ten years have seen a large amount of research and commercial activity in the vapor phase catalytic oxidation of light olefins. The production of acrylonitrile by the oxidation of propylene and ammonia has become a major factor in meeting the world’s requirements (Veatch et al., 1960). Prior to the propylene-ammonia process, acrylonitrile was made commercially by two processes. One of the earliest processes used ethylene cyanohydrin made by combining ethylene oxide with HCN. The dehydration to acrylonitrile occurred readily and cleanly. This process has not been used for the commercial production of acrylonitrile since 1966. Between 1953 and 1960, the acetylene-HCN process gradually supplanted the cyanohydrin process. This reaction is generally conducted in the liquid phase with copper salts as catalyst. One feature of this route is the production of by-product acetylenic impurities which are difficult to separate from the acrylonitrile. However, such impurities must be removed to the parts-per-million concentration range to avoid trouble in the polymerization of the acrylonitrile monomer. Although this process is still practiced commercially, it accounts for only a minor amount of the total acrylonitrile production. I n addition to the propylene-based acrylonitrile process, efficient olefin oxidation processes leading to methacrylonitrile, acrolein, acrylic acid, butadiene, and isoprene are now either commercialized or in advanced states of development. Sohio’s contributions to these developments lie mainly in the area of propylene and isobutylene oxidation, specifically, in processes for the production of acrylonitrile, methacrylonitrile, and acrolein.
The concept was further augmented by several potential advantages over conventional air oxidation. The effluent stream from the reactor would not be diluted with large amounts of nitrogen which would complicate separation of the intermediate products and the recovery of unconverted hydrocarbon feed for recycle. Further, it was reasoned that nonselective radical chain hydrocarbon oxidation in the vapor phase would be avoided because there would be no molecular oxygen in the reactor or its effluent system. With a large number of variable valence metal oxides to choose from and a large measure of optimism, there should be a good chance of discovering a t least one oxidant with a satisfactory combination of activity and selectivity for partial hydrocarbon oxidation. Early Oxidant Exploration
Early experiments were conducted in simple fluidized bed reactors of about 1-inch internal diameter. The oxidant charge was contacted with air a t reaction temperature to assure its complete oxidation; the flow of ,air was stopped, the reactor purged with nitrogen, and then a light hydrocarbon introduced, usually accompanied with diluent nitrogen or steam. Oxygenated products were recovered from the effluent stream by water scrubbers, carbon dioxide was determined by absorption in caustic solution, and fixed gas composition was analyzed by mass spec-
The Original Concept
Research on the vapor phase oxidation of hydrocarbons was initiated at Sohio in the 1950’s. I t began with the concept (Lewis et al., 1949) that the lattice oxygen of a reducible metal oxide would serve as a more versatile and useful oxidizing agent for hydrocarbons than would molecular oxygen. A process as shown in Figure 1 was visualized in which a light hydrocarbon would be contacted at elevated temperature with a fluidized mass of supported, variable valence metal oxide in its higher oxidation state. The hydrocarbon would be partially oxidized to intermediate oxygenated products, and the fluidized oxidant would be reduced. The oxidant would be restored to its higher oxidation state by contacting with air a t elevated temperature in a fluidized regeneration vessel. On a continuous basis, partially reduced oxidant would be withdrawn from the reactor, regenerated in a separate vessel, and returned to the reactor.
OXIDATION PRODUCTS
HYDROCARBON
FEED
-yLIFT I
Figure 1. Oxidant process Ind. Eng. Chem. Prod. Res. Develop., Vol. 9, No. 2, 1970
135
Table I. Definitions of Conversions and Yields
(All on carbon atom basis) per Cent total - gram atoms of carbon converted x 100 conversion gram atoms of carbon fed Per cent yield = Per cent selectivity
gram atoms of carbon to useful products x 100 gram atoms of carbon converted
- gram atoms of carbon to specific product x 100 gram atoms of carbon to useful products
Useful products are taken as the sum of oxygenated and/or nitrogen containing compounds other than oxides of carbon.
trometry. The duration of experiments was usually limited to 15 to 30 minutes. At the completion of an experiment, the hydrocarbon flow was stopped, the reactor purged with nitrogen, and the oxidant reoxidized with air. A variety of analytical techniques were employed: titration for organic acids, derivative formation (such as 2,4-dinitrophenylhydrazones for carbonyls)-followed by infrared analyses of the derivatives to determine the relative amounts and nature of carbonyl compounds. These laborious techniques were eventually displaced by gas chromatography, permitting much more facile product analyses. Because a single molecule of converted hydrocarbon can lead t o more than one molecule of products through fragmentation, the conversions and yields are defined on a carbon ztom basis as indicated in Table I. The original oxidants were prepared by incorporation of variable valence oxides with high surface area refractory supports such as silica gel using both coprecipitation and impregnation methods. These oxides gave a wide range of hydrocarbon oxidation activities and a t least a modest range of selectivities for useful product formation. Thus, oxidants with active components of, for example, C r O a , V205,and MnOz were found t o vary widely in both activity and selectivity for a variety of hydrocarbon types. Typical results of these initial experiments are given in Table 11. Hydrocarbon total conversions ranged from less than 1% up to about 10% while selectivity for conversion t o intermediate oxygenated products ranged from less than 1% to as high as 80%. The decision was made a t that point to restrict the exploration largely to propylene oxidation with acrolein as the desired product. At the time, propylene was an
over-abundant refinery by-product and seemed to represent an ideal petrochemical feedstock from the standpoint of cost and availability. Acrolein was seen as a very versatile intermediate which could be used t o synthesize any number of large volume end products'and which had potential as a monomer in its own right. I n selecting oxidant systems for propylene oxidation, our working hypotheses assumed that a correlation would exist between the metal oxygen bond strengths of the bulk oxides used to prepare a series of oxidants and the activity-selectivity relationship. The structural and energy states peculiar to the surface, as opposed to bulk properties, would really determine the course of hydrocarbon oxidation. Solid state diffusion rates would be important in determining the extent to which oxygen of the bulk could participate in the surface reactions. While reasonably good thermodynamic and structural data existed for bulk oxides, similar data for their surface states did not exist. The search for improved oxidants then was guided by experiments to test working hypotheses and not by major reliance on primary scientific data. This search for improved oxidant performance led us from simple metal oxide systems to more complex oxide salts and mixed oxides. An interesting example is afforded by a silicasupported V?O:-K20 oxidant. As shown in Figure 2 vanadium pentoxide, alone, supported on silica gave propylene conversions to oxidized products but with little selectivity to acrolein. A mixture
1
0
'
0 z 0
I
'loo "C
4Y)
ATM PRESS 6 SEC APP CONT
TIME
' c > 4
co c
u A W v) W
0
05 K z O / V ~ O J MOLE R A T I O
10
Fipre 2. Effect of K?O/V205 ratio on
C1 conversion a n d acrolein selectivity
Table II. Typical Oxidant Screening Results
136
Run Length, Min
Conversion
150-190
10
8.9
1 5 8 M n 0 2 on silica
396
4
5.7
Propylene
9% V,Oi on Si02-K2S04
460
5
0.7
Toluene
9% VZOS o n Si02-K2S04
404
9
8.0
o-Xylene
9%
404
9
10.0
Hydrocarbon
Solid Oxidant
Propylene
14.5% CrOl on silica
Propylene
V205
on Si02-K2S04
Temp.,
O
Ind. Eng. Chem. Prod. Res. Develop., Vol. 9, No. 2, 1970
C
YO O h
Yield, Intermediates
15.7 (Acetone-propionald.) 46.5 (Acetone-propionald.) 80 (Acrolein) 33.5 (Benza1d.-maleic acid) 12.0 (Phthalic anhydride-o-phthalald.)
of VzOj-KzO on silica in amounts corresponding to potassium vanadate (mole ratio 1 to 1) gave essentially no oxidation of propylene. Thus, under the conditions of the reaction, potassium vanadate is essentially irreducible by propylene while vanadium pentoxide is a very active oxidant. As potassium oxide is added to V2Os in increasing amounts up to the stoichiometric value for potassium vanadate formation, the activity of the resultant oxidant is observed to decrease continuously with an accompanying enhancement of selectivity for acrolein as a partial oxidation product. The best compromise between activity and selectivity for converting propylene to acrolein was an oxidant of approximate composition V205 0.35 K,O. An Oxidant Process for Converting Propylene to Acrolein
With the emphasis now largely di,rected toward a propylene to acrolein process, the continuing search for oxidants of greater selectivity eventually led to the discovery of a series of highly efficient molybdates. Particularly effective oxidants were the silica-supported salts or complex oxide mixtures of molybdic oxide or phosphomolybdic acid with copper, antimony, tin, and bismuth. The oxidant process is illustrated by the results obtained with two different oxidants, using data obtained from a continuous advancement unit. The results from a copper phosphomolybdate oxidant are shown in Figure 3. The conversion and yield values were obtained under steady-state conditions. Acrolein constitutes approximately 70% of the useful products, the remainder being various carbonyl compounds and oxides. Both per pass conversion to acrolein and the yield of useful products decrease with increasing oxidant-topropylene weight ratio. The performance of the bismuth-phosphomolybdate (BiPM) oxidant was strikingly different as shown in Figure 4. Under steady-state conditions both conversion to acrolein and yield of useful products increase with increasing oxidant-to-propylene weight ratio instead of decreasing as was the case in the Cu-phosphomolybdate system. Approximately 80% of the useful product was acrolein. Optimum bismuth-phosphomolybdate oxidant perfor-
1 1 7 % C u 3 1 P 0 4 ) 2 ~31 6 %
CuMo04
Catalytic Oxidation of Propylene
During the course of the oxidant work, the more promising oxidants had also routinely been checked as direct air oxidation catalysts-i.e., gaseous oxygen is, in this case, present in the reactor vessel simultaneously with the hydrocarbon feed. This had been done with both the V 2 0 5 - K 2 0and the copper phosphomolybdate oxidants, and both functioned poorly as catalysts, giving near complete oxidation of propylene to waste products. A test of the BiPM oxidant as a catalyst for the air oxidation of propylene revealed it to be a highly active and selective direct air oxidation catalyst as shown in Table 111. Even the initial tests indicated that more than 50% of the propylene could be converted t o acrolein in a single pass (Veatch et al., 1961). This discovery
-I
1
56 7 % Si02
TEMP i 460'C ATM PRESSURE 6 SEC A P P CONT
q 4
mance is obtained when it is maintained a t or near its highest state of oxidation. The great difficulty with such a process as far as commercial practicality is concerned is the very large weight ratio of solid oxidant to hydrocarbon required for high conversion and selectivity. If it is assumed that, in the hydrocarbon oxidation reaction, only the molybdenum in the oxidant is reduced from +6 to +4 in oxidation state, then there is a theoretical requirement of about 53 pounds of oxidant per pound of propylene feed. Actually, as was evident from the experimental data, only a very small fraction of the potential oxidizing capacity of the oxidant could be efficiently used. T o maintain the equilibrium oxidation state of the oxidant a t a level necessary to obtain both high reactivity and selectivity, it was necessary to consider oxidant-topropylene weight ratios of 300 or more. Steady improvement in oxidant composition and in technique had resulted in increased selective oxidation of propylene to acrolein of from less than 1% conversion to over 20%. From this standpoint, the oxidant system appeared to be of commercial utility. However, practical engineering aspects shadowed the picture. Apparently, the necessity of maintaining a high oxidation state of the oxidant to achieve both high conversion and selectivity was the key for the final solution.
L"
6
25
2 (L 0
a 2o
A
w LL
15
2
izol 0
0 LL
z El
n
Y
w -
z
5 s
0
IO
>
.e TIME
0
70
I"
50
1 ;
E
140
I5
J
w
IO
>
z 5
I
z I
I
1
I
OXIDANT/PROPYLENE WT RATIO
Figure 3. Performance of Cu-phosphomolybdate oxidant in continuous advancement reactor
Figure 4. Performance of Bi-phosphomolybdate in continuous advancement reactor
oxidant
Ind. Eng. Chem. Prod. Res. Develop., Vol. 9 , No. 2, 1970
137
56
roughly equivalent to operation of the two-vessel process under conditions in which the oxidant solids circulation rate is allowed to go to a very high value. The air-to-hydrocarbon ratio used in the direct oxidation scheme is clearly in the explosive range (U. S.Bur. Mines, 1966). Fortunately, there is no opportunity for detonation or uncontrolled combustion in the reactor because of the very efficient heat transfer properties of the fluidized catalyst and the quenching effect of the large amount of catalyst surface on the vapor phase free radicals required to propagate combustion. These process characteristics permit one to achieve high single-pass conversions of both the hydrocarbon and oxygen, minimizing the uneconomic dilution effects which would be caused by requiring a substantial excess of either reactant.
16
Catalytic Arnrnoxidation of Propylene to Acrylonitrile
Table 111. Catalytic Air Oxidation of Propylene to Acrolein
(Veatch et al., 1961) Conditions Catalyst: 50% BigPMol?Oj2on silica 450 Atmospheric 1.6
Temperature, C Pressure Contact time, sec Feed mole ratio Propylene Air Water
1
9.3 3.5
Results per pass conversion to: Acrolein Acetaldehyde Acetic acid Acrylic acid ,f Carbon oxides
1
20.5 Total
92.5
represented a substantial gain in efficiency and simplicity over an oxidant scheme requiring almost prohibitively large solid circulation rates. I t is interesting to compare the performance characteristics of bismuth molybdate as a direct air oxidation catalyst and as an oxidant. As an oxidant, it is characterized by display of a maximum in both activity and selectivity a t or very near its highest state of oxidation. As a direct air oxidation catalyst, it displays high activity and selectivity as long as a t least a small amount of unreacted oxygen is maintained in the reactor effluent. Removal of a sample of catalyst from the reactor under these reaction conditions confirms that the catalyst is maintained in a high state of oxidation. When the airto-propylene ratio is increased, the oxygen concentration rises in the effluent but propylene conversion and product distribution remain essentially unchanged. If the air-tohydrocarbon ratio is reduced to the point a t which the oxygen concentration in the effluent falls to zero, one observes a decay in both conversion and selectivity with simultaneous reduction of the catalyst to a lower oxidation state. Increasing the air ratio again causes the catalyst to regain its higher oxidation state with a return t o normal conversion and selectivity levels. These reaction characteristics point to an over-all process for the direct oxidation which can be broken down into two major chemical processes. I n one, the oxidized form of the catalyst acts as an oxidant in converting the hydrocarbon and is reduced. I n the other, the reduced catalyst is regenerated to the higher oxidation state by molecular oxygen. T o account for this behavior, the kinetic rate of reoxidation of reduced catalyst must be very much greater than the rate a t which the propylene can reduce the oxidized form of the catalyst. Thus, only a very small amount of excess oxygen is required in the reaction mixture to keep the catalyst in a high state of oxidation. The direct oxidation process may be thought of as a form of the original oxidant process scheme. The difference is that a single vessel is employed simultaneously as both the reactor and regenerator and functions so as to provide a constant high oxidation state oxidant. This should be 138
Ind. Eng. Chern. Prod. Res. Develop., Vol. 9, No. 2, 1970
With the discovery of a satisfactory catalyst and process for the conversion of propylene to acrolein, the research effort was expanded to include work on the further conversion of acrolein to such marketable chemicals as acrylic acid and glycerol and to find additional oxidative conversion applications for the best newly found oxidants and catalysts. An experiment was run in which a mixture of air, propylene, and ammonia was passed through a fluidized bed of bismuth phosphomolybdate catalyst. The conditions and results are given in Table IV (Idol, 1959; Veatch et al., 1962). Reaction temperature was 470" C., pressure a t atmospheric and superficial contact time 9.1 seconds. The ratio of reactants was 1.1 moles of NH3 and 13.1 moles of air per mole of propylene. T h e surprising result was that almost total conversion of the propylene occurred with 65% of the propylene fed converted to acrylonitrile. Smaller amounts of by-product acetonitrile, hydrogen cyanide, and oxides of
Table IV. Bench-Scale Acrylonitrile Run
(50% Bi9PMo12052 - 50% Si02 catalyst) Conditions Reactor 1YJ-inch pipe fluid bed Catalyst charge Temperature Pressure, psig Contact time Moles feedihour Propylene Air
566 g 470" C Atmospheric 9.1 sec
1"
Propylene wt. hourly space velocity
0.255 3.355 0.282 0.0253
Results Mole % excess 02 coper pass conversion(C-atom basis) to : Acrylonitrile Acetonitrile Acrolein HCN
co
Con Total % carbon balance % unreacted NH1
2.2 65.2 4.0 0.09 4.1 5.5 16.8 95.69 95.6 8.2
Ratio 1 13.1 1.1
carbon were formed with only a trace of acrolein and other oxygenated hydrocarbons. A more recently developed uranium-antimony oxide catalyst gives conversion of propylene feed to acrylonitrile of about 80% under very similar run conditions (Callahan and Gertisser, 1965, 1967; Grasselli and Callahan, 1969). Reaction Kinetics and Mechanism
I n agreement with recent literature (Adams et al., 1964b; Dalin et al., 1962; Kolchin et al., 1964; Pasquon et al., 1967), our laboratory has found that the ammoxidation reaction over bismuth molybdate containing catalysts has a first-order dependence with respect to propylene and a zero-order dependence on both oxygen and ammonia when they are supplied in a t least stoichiometric amounts. Most workers have observed the capacity of the catalyst to act as a reactive oxygen sink, thus being able to function for some time in the absence of supplied vapor phase oxygen (Aykan, 1968; Batist et al., 1967; Sachtler and deBoer, 1965; Triflro et al., 1968). Assignment of a reactant role for lattice oxygen in the reaction mechanism has stimulated detailed studies of the crystalline phases of bismuth molybdate and other olefin oxidation and ammoxidation catalysts in an attempt to define the catalytically active structures and gain some knowledge of the detailed workings of an active site. I n the BiZOs-MoOj system, some eleven crystalline phases have been reported (Doyle and Forbes, 1965). However, only a few of these are important from the standpoint of olefin oxidation catalysis. Table V outlines reported crystalline phase data in the region of experimentally determined catalytic activity (Aykan, 1968; Bleijenberg et al., 1965; Boutry et al., 1969; Erman et al., 1964, 1966, 1968; Gelbshtein et al., 1964; Kolchin et al., 1964; Mekhtiev et al., 1965; Zemann, 1956). An a phase of composition Bi?03.3Mo03has been determined to be monoclinic with relatively good agreement between
workers on the values of the lattice parameters. A p phase of composition Bi203 2Mo03 has been characterized as tetragonal or orthorhombic. Finally, a y phase with a composition of Bi2O3.MoO3has been shown to be identical to the mineral koechlinite with an orthorhombic structure as originally determined by Zemann (1956). Recently, Erman (1968) has shown that there are three separate y phases within this structure, the concentrations of which are a function of temperature. Kolchin et al. (1964) determined the specific rate constants for the formation of the major products in both the oxidation and ammoxidation of propylene over the 01, p, and y phases of bismuth molybdate (Table VI). The CY and p phases have appreciable activity and selectivity in the conversion of propylene to acrolein. However, the y phase displayed very low activity. Roughly the same activity relationship between the various phases was displayed in the ammoxidation of propylene. The specific rate constants for the formation of coproduct HCN and C 0 2 are also given and are observed not to vary greatly between the catalytically active and phases. Significantly, the activation energies for the formation of both acrolein and acrylonitrile are reported to be almost identical a t 19 to 21 kcal per mole, and do not vary for the various bismuth molybdate phases. This is consistent with the view that there is a common rate limiting step in both reactions. The manner in which propylene intecacts with the catalyst in the over-all oxidation and ammoxidation reactions has been treated comprehensively in the literature. Adams and Jennings (1963, 1964a; Adams, 1965) and others (Sachtler, 1963; Sachtler and deBoer, 1965) as illustrated in Figure 5, have shown by means of isotopic labeling that the rate limiting step in the attack on propylene is the abstraction of the first hydrogen from the methyl group leading to a symmetrical allylic intermediate. This initial step is followed by a second abstraction of (Y
Table V. Crystallographic Data of Catalytically Active Bi-Molybdate Phases Phase Composition
Bi2O3.3MoOi
Density, G/Cm’ Crystal Syst.
Monoclinic
(a)
Monoclinic Monoclinic
Lattice Parameters (A)
a = 7.89, b = 11.70, c = 12.24 B = 116”12’ a = 7.85, b = 11.70, c = 12.25 i3 = 116”20’ a = 7.719, b = 11.516, c = 11.985, /3 = 115”25’
Calcd.
Pyc.
Lit.
2
Space Group
Cited
Scheelite like 5.86
5.99
4
6.197
6.14
4
6
d.
Tetragonal Orthorhombic
a = 11.80, c = 5.40 a = 10.79, b = 11.89, c = 11.86
6.61
6.39
P.
/
b
4
e
8
L I
Orthorhombic (Mineral koechlinite) Orthorhombic Orthorhombic ?(low t.) Orthorhombic y”(meta st.) Orthorhombic y ’ (hi temp.) Tetragonal (hi temp.)
a = 5.50, b = 16.16, c = 5.49 a = 5.502, b = 16.213, c = 5.483 a = 5.50, b = 16.215, c = 5.485
8.281 8.25
8.26
8.0
4 4
a = 11.90, b = 11.90, (c = 5.45)
7.85
7.8
6
a = 15.91, b = 15.80, c = 17.19
7.45
7.3
32
a = 3.95, c = 17.21
h
4
Isomorphous with LarMo06
u
?.
,
‘Mekhtiev et al., 1965. bErman et al., 1964. ‘Aykan, 1968. dBleijenberg et al., 1965. ‘Boutry et al., 1969. ’Kolchin et al., 1964. Zemann, 1956. ’ Erman and Galperin, 1968. ’ Blasse, 1966.
’Erman and Galperin, 1966.
Ind. Eng. Chem. Prod. Res. Develop., Vol. 9 , No. 2, 1970
139
H H H HC-C -C D
Table VI. Specific Rate Constants of Bismuth-Molybdates
1st H Abstraction
(Kolchin et al., 1964) klA, min-' Phose Composition
x 'OP Acrolein COZ
Bi*0?.3MoO~ 5.29
1.85
ky2b, min-' Selec-
m
HC;H,-CCDZ H, - CH A\ \
XIO'
tivity
AN
COP
HCN
Selec tivity
74
14.1
5.24
2.50
65
(4
H
H
H
A H HC=C-CD
C-C=CD
1.38
87
16.6
3.16
2.18
76
0.5
52
2.3
1.14
1.13
50
H H HC=C-C
0.55
(Y 1
H
.
I
c
+
I
(a) Bi?O?.MoOl
H
2nd E Abstraction
H
Bi207.2MoO? 9.4
H
P H C-C=CD
~crolein 3rd H Abstraction
Measured at 475" C, 0.8 sec. * Measured at 480° C, 0.75 sec. Figure 5. Mechanism of propylene oxidation and ammoxidation (Allylic intermediate formation) 7 = k D / k H = Relative probability between D and H abstraction (Isotope discrimination effect)
hydrogen from either end of the allylic intermediate. The oxidative reaction is known to proceed from this stage in a series of fast steps, which have not been elucidated, leading t o the final product, acrolein or acrylonitrile, depending upon whether ammonia is absent or present in the feed. I n the course of our own work, kinetic measurements were made up to high degrees of conversion in both fixed and fluidized catalyst beds. Differential reactors, designed with draw-off ports so that a small portion of the reaction products could be withdrawn for analysis after fixed reaction times, were employed. A schematic drawing of the experimental equipment is shown in Figure 6. The fluidized bed reactor contained sieve trays, which result in a more nearly plug flow of the gas through the reactor allowing more accurate control of residence time. Composition of the bismuth-phosphomolybdate catalysts used in both reactors was 24.1% Bi, 14.8% Mo, 0.4% P, 23.4% Si, and the balance oxygen. A " l f i inch x % c , inch cylindrical pellet was used in the fixed bed, and a microspheroidal form was used in the fluidized bed. A typical set of data determined in this way for the propylene ammoxidation reaction, using the above described catalyst in a fixed bed, is given in Figure 7 . Conversion to acrylonitrile increases with reaction time as expected for a reaction with a first-order dependence on propylene concentration. Conversion to acrolein was observed t o be very low a t all contact times with no more than 1%being observed a t any sampling port. Propylene oxidation and acrolein ammoxidation kinetics were measured in a similar manner. The rate constants determined from these data are summarized in Table VII. The rates observed for the ammoxidation of propylene are very similar for both fixed bed and fluid bed, indicating no effect for catalyst particle size, and that diffusion is not limiting the reaction rate. The rate of reaction of propylene is also nearly the same for both the oxidation and ammoxidation reactions, consistent with a common rate.limiting step for both reactions. The reaction rate of acrolein in the ammoxidation reaction is nearly twice that of propylene, indicating a different rate controlling step in this reaction. I n considering the role of intermediates in the reaction, we were concerned that the rate of reaction of acrolein a t low concentrations in the presence of a relatively high propylene concentration might be different than that 140
Ind. Eng. Chem. Prod. Res. Develop., Vol. 9,No. 2, 1970
Numbers 1 & 2
=
Relotive probability of preceding species to react along indicated poth
A - AMMOXIDATION REACTOR
- CHROMATOGRAPH-ELECTROMETER - ABSORBERS D - MARIOTTE B O T T L E ( 2 LITER) E - GRADUATED CYLINDER B
C
Figure 6. Schematic of experimental setup
[
O A C. R Y L O N I T R l L E 0 ACETONITRILE 0 ACROLEIN
PROPYLENE : A 1 R : N H j : HzO i FIXED BED T E M P :430'C
I:
1O:l 2 : 4
Lo v)
4
IO CL Y
5
0
6
12
I8
24
30
36
4 2
48
5 4
T I M E , SECONDS
Figure 7 . Ammoxidation of propylene over Bi-phosphomolybdate catalyst
Table VII. Reaction Rate Constants
Table VIII. Acrolein Reaction Rate
Catalyst Type Pellet Propylene
"liked stock 1 Temperature, O C 425 Rate constant, sec-' Feed 0.34 Acrylonitrile 0.23 Acrolein 0.002 Acetonitrile 0.03 Acetaldehyde > hp, then h,/hl will express the ratio of the product formed by consecutive reaction to that formed
A
\"+/'
0.015%
Feed
Acrolein
Acrylonitrile
Acrolein CT-' + 0.015% acrolein"
0.47 0.45 =t0.07
0.38 0.34 =k 0.03
425
0.32
+
Rote Constant k, Sec-'
1
Microspheroidal, 20- to 150- p diameter.
Propylene
Feed composition = 1 mole acrolein of C?acrolein"/ 10 mole air/ lmole N H ?
A~rolein-2-C'~, specific activity = 0.831 mc/mmole.
by the parallel reaction. However, k l and h, cannot be determined independently and only the sum k l + hl can be measured. T o calculate the formation rate curves for acrylonitrile and acrolein using the equations of Figure 8, and to assess the relative importance of the parallel and consecutive reaction in acrylonitrile formation, the rate constants of Table VI1 are used. Since it is impossible to determine independently kl and k p in the presence of ammonia, the value of 0.2 sec-' (Table VII, column 2 ) which is the relative first-order formation rate constant for acrylonitrile with propylene as feed has been used for the sum h l + hp in the calculation. Similarly, a value of hl = 0.4 (Table VII, column 4), the relative first-order formation rate constant for acrylonitrile with acrolein as feed, has been used. Figure 9 shows the calculated formation rate curves for acrylonitrile with the experimental data points entered. The agreement between the calculated curve and the experimental points is very good, with deviation occurring only a t very high conversion. To calculate the curve for the formation of acrolein in the presence of ammonia, values of hl or h~ must be assigned in the equality k l + hz = 0.2 sec-'. Figure 10 shows calculated curves using values of h? = 0.02 sec-' and 0.005 sec-'. A value of h z in the range of 0.005 sec-' is required to fit the experimental points. Therefore, it follows that the ratio of kp to hl is approximately 0.025,
k , + k 2 i 0 2 SEC'I
6ot
i 501
k):
FIXED BED TEMPERATURE
i
0 4 SEC-'
4 3 0 'C
40
Acrylonitrile
301
Acrolein
8
Figure 8. Kinetics of propylene ammoxidation over Bi-phosphomolybdate catalyst
6
I 2
I8
24
30
36
42
48
54
60
T I M E , SECONDS
+ kz)t
A
=
Ane-(k'
B
=
[kpAn/(kl
C
=
Ao-(A
+
kp
+ B)
- kg)][e-k3f -e-(k1
+
k2)f
1
Figure 9. Comparison of calculated rate curve for acrylonitrile with experimental data points for parallel-consecutive reaction Ind. Eng. Chern. Prod. Res. Develop., Vol. 9,No. 2, 1970
141
h,
i
hz:
0 2 SEC-'
k3 = 0 4 SEC-'
$ 65 1 A
FIXED BED TEMPERATURE
:4 3 0
'C
LY V
a
'-
/-
TIME, SECONDS
Figure 10. Comparison of calculated rate curves for acrolein with experimental data points for parallel-consecutive reaction indicating that acrylonitrile is formed largely ( > 90%) by a mechanism not involving acrolein as an isolatable vapor phase intermediate. Conclusions
Variable valence metal oxides display a wide range of activities and selectivities as olefin oxidants. Mixed metal oxides can be employed to secure unique activity and selectivity properties not displayed by the individual oxide components. The conversions demonstrated by both the oxidants and catalysts investigated were found to be strongly- dependent on oxidation state. For bismuth molybdate, a high state of oxidation gives the best conversion properties. Kinetic measurements under acrylonitrile synthesis conditions follow a first-order dependence of reaction rate on propylene concentration and essentially zero-order dependence on both oxygen and ammonia concentration when these reactants are provided in a t least slight excess. A common rate limiting step involving abstraction of a hydrogen from the methyl group of propylene appears to be operative in both the oxidation and ammoxidation reaction. Our kinetic measurements are best fit by a mechanism in which propylene is converted directly into acrylonitrile following the initial attack of propylene on the catalyst surface, rather than via a mechanism involving acrolein as an isolatable vapor phase intermediate. Acknowledgment
The permission of The Standard Oil Co. (Ohio) to publish this work is gratefully acknowledged. Literature Cited
Adams, C. R., Proc. Int. Congr. Catal. 3rd. Amsterdam, 1964, 1, 240-9 (1965). Adams, C. R., Jennings, T. J., J . Catal. 2, 63-8 (1963).
142
Ind. Eng. Chern. Prod. Res. Develop., Vol. 9, No. 2, 1970
Adams, C. R., Jennings, T. J., J . Catal. 3, 549-58 (1964a). Adams, C. R., Voge, H. H., Morgan, C. Z., Armstrong, W. E., J . Catal. 3, 379-86 (1964b). Aykan, K., J . Catal. 12, 281-90 (1968). Batist, Ph. A., Kapteijns, C. J., Lippens, B. C., Schuit, G. C. A., J . Catal. 7, 33-49 (1967). Blasse, G., J . Inorg. Nucl. Chem. 28, 1124-5 (1966). Bleijenberg, A. C. A. M., Lippens, B. C., Schuit, G. C. A., J . Catal. 4, 581-85 (1965). Boutry, P., Montarnal, P., Wrzyszcz, J., J . Catal. 13, 75-82 (1969). Callahan, J. L., Gertisser, B. (to the Standard Oil Co.), U. S. Patent 3,198,750 (Aug. 3, 1965). Callahan, J. L., Gertisser, B. (to the Standard Oil Co.), U. S. Patent 3,308,151 (March 7, 1967). Dalin, M. A., Lobkina, V. V., Abaev, G. N., Serebryakov, B. R., Plaksunova, S. L., Dokl. Ahad. Nauk, S S S R 145, No. 5, 1058-60 (1962). Doyle, W. P., Forbes, F., J . I m r g . Nucl. Chem. 27, 127180 (1965). Erman, L. Ya., Galperin, E. L., Kolchin, I. K., Dobrzhanskii, G. F., Chernyshev, K . S., Zh. Neorg. Khim. 9, 2174-8 (1964). Erman, L. Ya., Galperin, E. L., Zh. Neorg. Khim. 11, 221 (1966). Erman, L. Ya., Galperin, E . L., Zh. Neorg. K h i m . 13, 927-31 (1968). Gelbshtein, A. I., Stroeva, S. S., Kulkova, N . V., Bakshi, Yu. M., Lapidus, V. L., Vasileva, I. B., Sevastyanov, N . G., Neftekhimiya 4, 906-15 (1964). Grasselli, R . K., Callahan, J. L., J . Catal. 14, 93-103 (1969). Idol, J. D. (to the Standard Oil Co.) U. S. Patent 2,904,580 (September 15, 1959). Kolchin, I. K., Bobkov, S. S., Margolis, L. Ya., Neftekhimiya 4, (2), 301-7 (1964). Lewis, W. K., Gilliland, E. R., Reed, W. A., Ind. Eng. Chem. 41, 1227-37 (1949). Mekhtiev, K. M., Gamidov, R . S., Mamedov, Kh. S., Belov, N. V., Dokl. Akad. N a u k , SSSR 162, 563-4 (1965). Pasquon, I., Trifiro, F., Centola, P., Chim. Ind. (Milan) 49, 1151 (1967). Sachtler, W. M. H., Rec. Trau. Chim. 82, 243-5 (1963). Sachtler, W. M. H., deBoer, N . H., Proc. Int. Congr. Catal., 3rd Amsterdam, 1964, 1, 252-63 (1965). Trifiro, F., Centola, P., Pasquon, I., Jiru, P., Int. Congr. Catal., 4th Moscow, Preprint No. 18, (1968). U. S. Bur. Mines, Bull. No. 627 (1966). Veatch, F., Callahan, J. L., Idol, J. D., Milberger, E. C., Chem. Eng. Progr. 56, No. 10, 65-7 (1960). Veatch, F., Callahan, J. L., Milberger, E . C., Foreman, R. W., Proc. Int. Congr. Catal., 2nd Paris, 1960 2, 264754 (1961). Veatch, F., Callahan, J. L., Idol, J. D., Milberger, E . C., Hydrocarbon Process Petrol. Refiner 41, 11, 187-90 (November, 1962). Zemann, J., Heidelberg. Beitr. Mineral. Petrog. M. 5 , 139-45 (1956). RECEIVED for review October 17, 1969 ACCEPTED March 12, 1970 Division of Petroleum Chemistry, 158th Meeting, ACS, Xew York, September, 1969.
