Fiacco, A. V., McCormick, G. P., “Nonlinear Programming. Sequential Unconstrained Minimization Techniques,’’ pp. 60-1, Wiley, New York, 1968. Fletcher, R., Powell, M. J. D., Computer J . 6, 163-8 (1963-4). Goldfarb, D., Lapidus, L., Ind. Eng. Chem. Fundam. 7, 142-51 (1968). Griffith, R. E., Stewart, R. A., Management Sei. 7, 37992 (1961). Gue, R. L., Thomas, M. E., “Mathematical Methods in Operations Research,” Macmillan, Xew York, 1968. Gurel, O., Lapidus, L., Ind. Eng. Chem. Fundam. 7, 61722 (1968). Hooke, R. J., Jeeves, T. A., J . Assoc. Comp. Mach. 8, 212-29 (1961). Keefer, D. L., Gottfried, B. S., A . I . I . E . Trans., in press, 1970.
Lee, E. S., A.1.Ch.E. J . 15,393-400 (1969). Pearson, J. D., Computer J . 12, 171-8 (1969). Prabhakar. T., Ind. Eng. Chem. Fundam. 7, 626-32 (1968). Rudd, D. F., Watson, C. C., “Strategy of Process Engineering.” pp. 251-73, 282-8, Wiley, New York. 1968. Weisman, J., Wood, C. F., Rivlin, L., Chem. Eng. Progr. S y n p . Ser. 61, 50-63 (1965). Wilde, D. J., Chem. Eng. Progr. 61, 86-93 (1965). Wilde, D. J., Beightler. C. S., “Foundations of Optimization,” Prentice-Hall, Englewood Cliffs, N. J.. 1967.
RECEIVED for review September 22, 1969 ACCEPTED May 21, 1970
Influences of Catalyst Formulation and Poisoning on the Activity and Die-off of l o w Tempera tu re Shift Catalysts John S. Campbell Agricultural Division, Imperial Chemical Industries, Ltd., Billingham, Teesside, England The use of copper catalysts for the low temperature shift reaction has become much more widespread in the last decade in hydrogen and ammonia synthesis gas production.
This has led to a general awareness among plant operators of the problem of catalyst die-off. Die-off i s shown to be the consequence of two factors: thermal sintering, which can be reduced and, in some cases, eliminated by correct catalyst formulation methods; a n d poisoning b y small concentrations of impurities such as sulfur and chlorine, carried in the gas stream. The importance of the method of catalyst manufacture on subsequent activity and stability i s outlined; with a well-formulated catalyst, not maltreated, poisoning
i s the maior cause of loss of activity on the commercial scale. The reaction kinetics, using a commercially available catalyst, have been examined and a design equation
is proposed. The activity of the catalyst, in the absence of a diffusion limitation, is directly proportional to the copper area and initial activity increases with increasing copper content. The use of hydrogen in steam for reduction of the catalyst i s discussed.
THE
great majority of recent ammonia and hydrogen plants employ the low temperature water gas shift reaction in conjunction with methanation for carbon monoxide removal. This combination has come about for economic reasons and is advantageous because, a t the reaction temperature obtained with a commercial iron oxide-chromia catalyst, the reaction is equilibrium-limited. Since there is no volume change and the reaction is exothermic, the equilibrium conversion of carbon monoxide can be increased only by decreasing the temperature. The variation of the equilibrium concentration of carbon monoxide with temperature for typical shift exit conditions is shown in Figure 1. The widespread use of low temperature shift catalysts during the last decade [although they have been known since the later twenties (Larson, 1931a,b; 1947)] has led 588
Ind. Eng. Chem. Process Des. Develop., Vol.
Y, No. 4,
1970
to a general awareness among plant operators of the problem of catalyst die-off. This paper shows that die-off is the consequence of two factors: thermal sintering, which can be reduced and in some cases eliminated by correct catalyst formulation methods, and poisoning, by small concentrations of impurities, such as sulfur and chlorine, carried in the gas stream. The importance of the method of catalyst manufacture in subsequent activity and stability is outlined. The structural problems that have to be taken into consideration prior to deciding on a method of formulation are discussed. With a well-formulated catalyst. provided it is not maltreated, poisoning is the major cause of loss of activity. The function of each of the components in IC1 and Katalco catalyst 52-1 (subject of catalyst and process patents and patent applications in many countries) has
"
COPPER AREA M% CATALYST 0 Cu ZnO Cu 2nO A1203 A CU AI2 0 3
Figure 2. Relationship between copper area and catalytic activity
200
300
400
TEMPERATUREOC C 0 2 = 13% H2 =40%
5
H20 = 33% Figure 1. Equilibrium carbon monoxide concentration for typical exit conditions a t different temperatures
been examined. I n the reduced state the catalyst consists of copper supported on zinc oxide and alumina; in the absence of a diffusion limitation, the activity of the catalyst is directly proportional t o the copper area and increases with increasing copper content. I n addition to the role of support, the zinc oxide in the catalyst appears to absorb sulfur, which prevents poisoning of the copper. The reaction kinetics, using a commercially available catalyst in pellet form, have been examined. The reaction appears to be strongly pore diffusion-controlled above ca. 200OC. This factor is taken into consideration in the design equation. I n some cases, where no natural gas or nitrogen is available, catalyst reduction has to be carried out with hydrogen, using steam as an inert carrier. In the past. this has led to problems because of loss of activity. Progress in overcoming this problem is discussed. Active Species
Although copper has been recognized as an essential ingredient of low temperature shift catalysts since the late twenties, it has always been used in conjunction with zinc oxide. This has led to the belief that copper is promoted by zinc oxide. Recent work with copper-alumina, copper-zinc oxide, and copper-zinc oxide-alumina catalysts of varying composition does not support this view, since specific activity was found to be proportional to copper area in both the presence and absence of zinc oxide (Figure 2). T o obtain these results, it was first necessary to develop a special method for measuring copper areas in the presence of zinc oxide. The chemisorption of oxygen gave the most satisfactory results and areas obtained in this way were in good agreement with those calculated from x-ray diffraction line-broadening experiments. As a result of this work, it is concluded that copper is the active species for the reaction. Further support for this conclusion can be found by an examination
of the initial activities of catalysts with similar supports containing different proportions of copper. With a copperalumina catalyst, initial activity was proportional to copper content; catalysts containing 44. 27, and 6 ' r copper oxide had activities (expressed in arbitrary units) of 7.5, 3.8, and 1.0, respectively. Choice of Support
Because copper is not a very refractory metal (melting point 1083°C) and thus sinters easily, it must be supported for catalytic purposes on a stable high-area support material. The efficiency of such a material depends largely on three factors: the inherent thermal stability of the material, the ability of the support to resist attack by the reaction atmosphere, and the geometry of the support relative to that of the supported species. Unless a support is carefully chosen to take account of all these factors, activity will be lost through sintering. Although a great many materials have been tried, in the case of low temperature shift catalyst, zinc oxide-alumina has given the best results. The rates of loss of activity of copper supported on different materials are shown in Figure 3; the experiments were carried out a t 230°C and with very pure gases to avoid complications caused by poisoning. Figure 3 shows that copper supported on different materials sinters a t widely different rates. Correct over-all chemical composition per se is not sufficient to ensure a stable catalyst, because one of the copper-zinc oxidealumina catalysts shown is obviously very stable, whereas the other, of apparently similar composition, is very much more prone to thermal sintering. Once a support material is found which combines good thermal stability with resistance to attack by the reaction atmosphere, a method must be devised to incorporate it with the copper in such a way as to prevent the copper from sintering. This is purely a geometrical effect and the stability of the catalyst depends almost entirely on the ability of the support particles to limit the degree of contact between the copper crystallites. I t is obvious, therefore, if the initial size of the copper crystallites is small, that the potential sinterability of the copper will depend on the average crystal size of the support and the amount of support relative to the amount of copper. A large proportion of support leading to a high degree of dispersion of the copper crystallites together with small support crystal size should inhibit sintering. Experimental work indicates a very approximate relationship of the type shown in Figure 4. Ind. Eng. Chem. Process Des. Develop., Vol. 9, No. 4, 1970
589
,
0
Figure 3. Thermal stability of different formulations
0h
Standard 52-1 (Cu: ZnO: AI203)
A Cu: ZnO,
1:2
Poorly formulated Cu: ZnO: compositon as 52-1
Method of Preparation
Catalyst structure critically affects activity and lifetime and thus catalysts of apparently similar chemical composition performed differently. Experiments have shown that the method of precipitation has a major influence in determining the structure of the catalyst. The effect of pH on precipitation is shown in Figures 5 and 6. Clearly, the more acid the conditions, the bigger is the crystallite size which leads to a catalyst with low initial activity and consequently a poor life. The size of the fundamental catalyst particles produced under alkaline conditions is shown in Figure 6 (upper); these are much smaller than those produced under acid conditions (Figure 6 lower), giving a more active product. The direction from which the pH of precipitation is approached is also important, since this may change the order in which the different components are deposited. Figure 5 3 , shows that if precipitation is approached from a low pH, zinc is precipitated last; if' approached from a high pH, zinc is precipitated before copper. In addition to having an effect on the structure of the catalyst, and hence the rate of sintering, the pH
AI203,
same
of precipitation can also affect the surface area and pore volume, both of which are important parameters in determining activity. Under normal operating conditions, the reaction is strongly pore diffusion-controlled, under which condition the rate of reaction per unit volume of catalyst is given by the equation
v
O
r
140
120
V SINTER
(m) Figure 4. Stabilization of sinter size by high area supports
590
Ind. Eng. Chem. Process Des. Develop., Vol. 9,No. 4, 1970
Figure 5. Variation of catalyst structure and properties with pH of precipitation
10 9 8
7 6 3
9 8 7
6 5 4
3 I8
19
21
20
22
103/lnK
Figure 7.Variation of reaction rate with temperature
Figure 6 . Effect of pH of precipitation on catalyst particle size Upper. Alkali-precipitated lower. Acid-precipitated
To make the most effective use of the catalyst, the product of pore volume and surface area of active species must be a maximum; this is plotted in Figure 5,A. Kinetics
Results using commercial catalysts, obtained a t a constant total pressure of 13 atm and a fixed steam ratio of 0.5, a t temperatures over the range 180" to 170"C, with inlet dry gas compositions varying from 1 to 14 volume % carbon monoxide are shown in Figure 7. T h e results are the values of k relative to the value a t 270°C. The converter was charged with 125 ml of catalyst pellets and an appropriate high space velocity was chosen, so that results could be obtained under approximately differential conditions. The average converter gas composition and temperatures were calculated from the mean of the inlet and exit conditions. Before use, the catalyst, a supported CuO-ZnO formulation, was reduced with 2% hydrogen in nitrogen a t 230°C and atmospheric pressure. Over the range of compositions and temperatures studied, the reaction was found to be first-order with respect to carbon monoxide partial pressure, giving a rate equation of the form
The Arrhenius plot of the rate constants a t various temperatures shown in Figure 7 can he divided roughly
into two separate regions. At temperatures above 200" C the activation energy is 12 kcal per mole; below 200°C it is 28 kcal per mole. The approximate halving of the activation energy on moving from the low temperature to the high temperature region is consistent with a change from a chemically controlled reaction to a pore diffusionlimited reaction. This subject has heen treated theoretically by Thiele (1939) and Wheeler (1955), who have shown that under strongly pore diffusion-limited conditions the measured rate constant is almost proportional to the square root of the real rate constant, which corresponds to a halving of the activation energy. I t is unlikely that over a full temperature range the reaction is film diffusion-limited, since the activation energy would he rather lower than observed. A criterion for this effect is given by Wheeler (19551, who has pointed out that gas-film diffusion will have no limiting effect, provided that
u/10 =
VL
( Ma3PT
)l"
>k
(3)
Considering reaction a t the highest temperature (270" C), u/lO was calculated to be many times greater than k, and consequently the reaction would not be expected to he film diffusion-limited. Experiments have been carried out a t 240" C, not only by altering the proportion of gas allowed t o go through the preconverter but by changing the steam-dry gas ratio over the range 0.3 to 1.0 and by altering the total operating pressure. The results were analyzed by postulating a numher of reaction mechanisms, deriving the corresponding reaction-rate equations, and using a computer to fit the experimental data to these equations by a least-squares fitting technique. The most satisfactory equation was derived from a single adsorption-surface reaction-desorption model, assuming the surface reaction step and the pore diffusion of carbon monoxide to the surface to be rate-limiting. This equation is given as follows: Ind. Eng. Chem. Process Der. Develop., Vol. 9, No. 4, 1970 591
, pH 0
L1
-
'
PCO PH KPPCOPHO
I
The term involving the square root of total pressure is introduced to allow for the effect of pore diffusion of carbon monoxide; it is assumed that the diffusivity of carbon monoxide is inversely proportional to total pressure, Because a steam reformer was used for producing the reaction gas, there is a direct relationship between the hydrogen and carbon dioxide concentration:
[Hg] = A
+ [CO?]
Thus, using this relationship, the term in the rate equation involving either the hydrogen partial pressure of carbon dioxide partial pressure can be eliminated. This was borne out in practice, since it was found that two sets of values for the constants, depending on whether Kd or K , was zero, gave an almost identical fit with the experimental results. The values obtained for the constants a t 240°C relative to the values of K O are shown in Table I. I t is not possible with the available data t o distinguish between these two rate equations: Fit 1 has been used for all design work. Carbon dioxide might be expected to be more strongly adsorbed than hydrogen, in which case K , can be taken as zero. The temperature dependence of the constants has been examined over the range 180"to 280" C. At temperatures