Theory and practice of the formulation of heterogeneous catalysts

Chrmicsl E#tgi#vring Sckncr Printed in Great Britain.

Vol. 36. No. 9. pp. 1431-1445.

1981

REVIEW

ARTICLE

NUMBER

2

THEORY AND PRACTICE OF THE FORMULATION OF HETEROGENEOUS CATALYSTS S. P. S. ANDREW Imperial Chemical Industries Limited, Agricultural Division, Billingham, Cleveland, England Abstract-This review has been written from the viewpoint of the formulator of industrial catalysts with the objective of explaining the practice but giving only such general theories and speculations as the author has found to be of real utility. All the important industrial aspects of formulation are covered from the initial choice of catalytic species, through the processes of fabrication of the catalyst to its final reductioo in the reactor. INTRODUCTION

Outside of the vast and confusing patent literature there is little publicly available information on the formulation of industrial heterogeneous catalysts. The only text book that the writer has come across being Catalyst Techno&y, editor I. P. Mukhlyonov, Mir, Moscow 1976. The very considerable scientific literature on catalysts is for the most part concerned with investigations into and speculations on the theory of catalysis, when written by chemists, and algebraic and computational exercises on diffusional and beat flow effects in catalyst particles and reactor optimisation, when written by chemical engineers. This review does not touch on either of these areas in any depth as an attempt is made to present a balanced picture of catalyst formulation. CHOOSING THE ACTIVE SPECIES

The first stage in catalyst formulation is the choice of the active catalytic species. This choice is invariably made by experiment guided by knowledge of the catalytic properties of substances. Experiment is necessary as the general catalytic properties of metals, metal oxides etc. are not so well defined that the precise active species can be selected from the literature alone. Past knowledge is a guide to the most fruitful areas for experimental investigations. In order to use this knowledge it is necessary to consider the reaction which it is desired to catalyse in some detail and to decide which of the bonds in the reagent molecules it is hoped will be ruptured so as to bring about the reaction. First, however, it is as well to check that the thermodynamic maximum possible yield of the reaction is adequate, and whilst doing this to check what yields of undesired products would be possible if the catalyst was somewhat unselective. With very many reactions it is evident that there are an embarrassingly great number of thermodynamically possible products formed either from the reagents taken singly, or from the reagents together, or from a reagent and a product or by reaction together of products. In organic systems, carbon is usually one of these undesired possible products. It is therefore necessary to list also

bonds which it is not desirable to acitvate. The choice of active species is then made bearing the negative criterion as wetl as the positive criterion in mind, Table 1 shows a typical list of the types of bond activation usually associated with different species. Depending on the species and on the composition of the reagents and their partial pressures and temperatures, the catalytic species can either, in use, be in the elemental (usually metallic) form, or with more oxidising conditions it could exist as an oxide or a sulphide or halide etc. The process of bond activation required in catalysis requires a reversible chemical reaction to take place between the reagent species and the solid catalytic surface. This is termed”chemisorption”. For this to take place the catalytic species must be, thermodynamically, operating close to a readily reversible change between the reacted and unreacted state. Thus metallic iron is a good catalytic species for synthesising NH3 from N2 and Hz because, under synthesis conditions, it operates close (but not too close) to the formation of solid iron nitride, thereby rupturing the N-N bond, and also the formation of iron hydride, thereby rupturing the H-H bond. Ruthenium, osmium and molybdenum are also good catalysts. Palladium, platinum, rhodium and iridium, though also excellent catalysts for the activation of the H-H bond are no good for rupturing the N-N bond (see Table I). Ammonia synthesis catalyst must therefore be chosen from species that are present in both the first and fourth lines in Table 1. With ammonia synthesis, at the temperature at which these catalysts are reasonably active ( > 350°C). ammonia is thermodynamically the only likely product. The situation with methanol synthesis from CO and H2 is more complex. Not only is C a possible product, but also CH4, higher alcohols, and higher paraffins. It is therefore necessary to seek selectivity by first ensuring that CO and H-H are activated but the bond between C and 0 is not activated otherwise CH, and paraffins, waxes and even C could be formed. The choice is therefore between those species falling in line 1 of Table 1, H-activation, provided they also fall in the line for CO activation and do not fall in the line for CEactivation.

those

1431

S. P. S. ANDREW

1432

Table 1 Selection of

catalytic

species

TiVATlON QUIRED

STATE

OF CATALYST

H-

METAL,

CXIDE,SULPHIDE

3=

METAL

ce

METAL

Fe,Ru,Os,

NE

METAL

Fe,Ru,Os,Mo

w,

51

SULPHIOE

Mo,W

Cc,Ni,Cu,Fe,Sn,Zn,V

C ct

-

2O,OH_

n+ HC.!

OR

HIGH Pd.

OXIDE

PI,

ACTIVITY

Pi.Rh,Ru, Pd,

CARBONATE

K, Na

CHLORIDE

Cu.Zn,

Mn,Co.

ACTtVlTY

MEDlUM

,r

Mn,Fe,Ni,Cu,W,Ag,Cr,

Co,zn,V,Mo

C”

Ag,NI,Fe,V,

TI

Rh

MO,Sb,Cr,

Ni,Co t-m,U

CQ Hg,Ag

OXIDE

OR

HYDROXIDE

W,q

OXIDE

OR

IiALlOE

(Si

V,Ca,Th,Mg

OXIDE

OR

CHLORIDE

SIAP,AI

AtaP,

Cr,W)

B,AI,Tl,Hg,Zn,

SlAC

(~t,Sn,~n,zr,B)

OXIDES

HALIDES

ZCZC’ METAL -c:c-CEC-

OR

OXIDE

Pd,Pl,

SALT5

Ztl

H%CU,A9

co

METAL

OR

OXIDE

Pt ,Cu,Pd,

502

METAL

OR

OXIDE

PI ,v

UGNE

Co,Ni,Fe,Ir,W,Mo,Cr,Cu

Rh,Ru

>

OXIDE

I

Zn,Co,Fe,Mn.Ag

lr

Fe,AC

AP,

So far it has been assumed that a single species must be used for carrying out the reaction. This is, indeed, desirable but not always essential, as certain reaction intermediates can, after activation on one species, spill over onto another cataIytic species. The hydrogen atom H is well known for this ability. In general, however, larger surface (chemisorbed) species are inadequately mobile. Two or even more catalytic reactions may be carried out in a single reactor and indeed in a single catalyst particle by incorporating more than one catalytic species. In these cases, however, molecules pass from one catalytic surface to the other through the pores and not as chemisorbed species on the surface. The combination of a hydrogenation activity (line I of Table 1) with a hydrocarbon cracking activity (activation of H’) is found in hydrocracking catalysts where naphthenes are hydrogenated, using a precious metal such as ptatinum which is dispersed as small crystals over a cracking catalyst such as a silicalumina compound. When employing more than one active species in a formulation the overall result however becomes a little more unpredictable as even the structure of the catalytic species itself is open to great uncertainty. If two metals are employed which are mutually soluble then the formulator is uncertain to what extent he has the two present unmixed or mixed. Furthermore, when mixed, he is often uncertain which metal atoms are concentrated on the surface under reaction conditions and if so to what extent and where and what effect does this have on activity and selectivity? Fortunately for industry a process of trial and error in formulation can arrive at a satisfactory catalyst even though a further 50 years’ experiments might be required to find out why it is

Mg,

MgAI,Mg

Si,

Co At2

satisfactory. Even when the two catalytic components are in no way miscible or apparently capable of mutual reaction on the bulk scale, as with platinum metal crystals on a ceramic oxide component, it is quite possible that some interaction occurs which alters the catalytic properties of the very small platinum crystals (5%1dia.) due to electron transfer between the metal and the underlying oxide. DEVICES

TO ENHANCE

CATALYTIC

SEL.ECTIVlTY

If adequate selectivity for the desired reaction has not been attained by adjustment of solid catalytic species there are a number of chemical devices that on some occasions prove beneficial. These are shown, together with commercial examples of their use, in Fig. 1. Three

Fig. 1. Typical devices to enhance catalytic selectivity.

Theory and practice of the formulation of heterogeneous catalysts of these devices use the addition of a species which is chemisorbed on the catalytic species, but does not react with the reagents or products, to modify the catalyst surface. Volatile species such as HCl or HzS may be used in very low concentrations so as to chemisorb preferentially on certain parts of the catalyst surface, and thereby possibly enhancing selectivity without excessive loss in activity. Or, like CO, which is employed during the selective hydrogenation of CH=CH in CH2=CHz over a palladium catalyst, such species may, by chemisorption, displace from the catalyst surface a species which otherwise would have chemisorbed and then have reacted (ethylene). Inert optically active molecules may be chemisorbed on a catalytic surface with the objective of leaving “optically active” (i.e. handed) holes into which non optically active reagents could slot, combining on the catalyst to give an optically active product. More crudely, the catalytic species can be set in a molecular sized box, such as the cage of a zeolite, which limits the access to it of molecules and limits its ability to form large molecules. Zeolitic cages are used for these purposes in petroleum reforming and in the synthesis of hydrocarbons from CO and Hz_ Simple considerations such as the above in a given instance may lead, when formulating a new catalytic system, to experiment to show whether a superficially plausible device may be of benefit. POISONS

Of the devices listed above several fall into the class of catalyst poisons used in judicious small amounts. There is a general rule in heterogeneous catalysis (i.e. a rule which is usually correct!) that the more active the catalyst, the less selective, hence the use of small amounts of poisons. Or some occasions even great amounts of poisons may be present in the gas stream, they may even be products of the reaction such as HIS when hydrodesulphurising oil. In this case a poisoned catalyst is the active species and the most active poisoned species must be chosen. Less obvious examples of poisons are listed in Fig. 2. The surface active, but non-catalytic, metal or ion is one variety. Thus when formulating a catalyst for steam reforming methane, where it is derived to activate Cwso as to form CO from CH, by reaction with steam, a nickel catalyst is very effective. Quite low concentrations of copper in the nickel will, however deactivate it, as copper is surface active relative to nickel (lower melting point and there-

Fig. 2. Different types of catalyst poisons.

1433

fore lower surface tension) and furthermore copper does not activate Cm. Similarly low amounts of Ca in Co304 catalyst for ammonia oxidation (which operates at SOOOC) diffuses to the surface and deactivates the catalyst. Pb in Fea04 catalyst for the reaction CO + HzO+ COz + H2 is a similar poison. Another form of “poison” is a species which accelerates the formation of high molecular weight “polymer” or “coke” to such an extent that the catalyst is rapidly deactivated by the formation of a covering of this by-product. Fe in Cu methanol synthesis catalyst or Fe in %-AI cracking catalyst are examples. HCl and l-J20 have a more pernicious effect than merely acting as temporary poisons, in as much as the chemisorbed species that they form are mobile and by diffusing across the surfaces or through the gas phase from small crystals to large crystals result in a rapid growth in the average crystal size and hence a loss in total surface area and in catalytic activity per gram of catalytic species. This loss is, of course, irreversible even when the poison has been removed. ACTIVE SITES

It has often been estimated that only a small fraction of the surface of catalytic species is active and hence has arisen the concept of “active sites”. These have been hypothesised to be caused by dislocations, crystal steps, corners etc. in metals as well as by impurity atoms in metals and oxides. With metal catalysts the formulator normally ignores this aspect as there is no practical way in which he can affect these sites other than through the bulk composition of the metal crystal&es, the support on which these crystallites are dispersed, and the temperature, reagent composition and pressure. Industrial metal catalysts and some oxide catalysts almost invariably operate at a temperature at which surface restructuring of the metal crystals fairly rapidly takes place, such that the surface defects and active sites are maintained in a dynamic steady state appropriate to their environment. For this reason the creation of extra defects in the metal crystals of a catalyst by, for insubjecting it to nuclear radiation prior to stance, employing the catalyst have only a temporary effect. There is one class of catalysts where the above does not apply, ceramic acid catalysts such as silica-alumina gel and its crystalline zeolitic form. Because these catalysts are employed for their H’ exchange ability at a temperature which is relatively low compared with their melting points, defects, in the form of Al” ions in place of Si4’ ions, which are created during the preparation of the catalyst can last throughout its life and continue to act as sites of catalytic activity. The activity of the catalysts can thus be correlated with the density of sites measured prior to their use by titration methods. PORES In order to secure a high activity per unit volume of solid catalyst it is necessary to produce this catalyst in a very fine state of subdivision, crystals from 5 to 5008, diameter are common in industrial catalysts. As these crystals must be contacted with the the reagent gas or

1434

S.

P. S. ANDREW

liquid and then separated from the product, they must be aggregated into particles of from 50 p to 10 mm in size and these particles must be porous to allow diffusional ingress of the reagents and egress of the products. The size distribution of the voids or pores in a catalyst particle is dependent on the processes of preparation of the catalyst. Roughly three sorts of pore may exist-see Fig. 3. Smallest-falling in the range 2-g A, are pores that are formed in a crystal lattice by some type of phase change involving a loss in volume. The dehydration of chromium hydroxide, for example, produces such shrinkage holes but these are too small to be of use in catalysis. The regular passages and cages in zeolites are sufficiently large to admit usefully sized reagent molecules and to allow the egress of useful product molecules. The next larger range of pores falls between 30 and 200 A and is formed by the spaces between particles of a very fine precipitate of catalytic species and support which usually range in the 20-500~ crystal diameters. The final stage of catalyst forming may involve compaction of a powder made from aggregates of these precipitates in, for instance, a tabletting machine. The spaces left between the crushed together aggregates usually falls in the range 500-2oooA. These size ranges are compared with the boundary between bulk gaseous diffusion and Knudsen diffusion in the pores in Fig. 3. Catalysts operating in gases at pressures below 5 bar and containing internal porosity made by precipitation almost invariably fall in the Knudsen regime in which the gaseous diffusion mean free paths are restricted by the closeness of the catalyst particles. When operating above 50 bar, little restriction of the mean free path occurs. The extent to which diffusion restricts egress of the reaction product is particularly important if good selectivity is required of an intermediate product in a series of consecutive reactions. The gas phase oxidation of an organic compound is a typical example where over oxidation leads to carbon monoxide. Under these circumstances catalysts having little or even no internal porosity are used. For example-silver needles for the

I

0. I

I I

Pressure, Fig. 3. Mean free

I

I IO

100

oxidation of methanol to formaldehyde for the oxidation of ammonia to NO.

or platinum wire

CBOOSINGTHE SUFTORT

The continuous formation and decomposition of surface compounds formed between metal atoms and the reagents during catalysis on metals results in an effective mobility of the metal which rapidly results in very small crystals fusing together with a consequent loss of surface area and activity per gram. Provided the crystals are in contact, this sintering process proceeds at a rate which is determined by the crystal size and the closeness of the temperature to the melting point, or the decomposition temperature of the solid if that is lower. Figure 4 gives a rough indication of the effect and is based on data for sintering such materials as copper, alumina and carbon. Thus a compact of copper crystals having a melting point at lOg2”C, if held at 250°C for 6 months, however small they were initially, will be at least 1 p in size at the end of the period. However alumina, melting point 203O’C, might still be in crystals no bigger than 20 A. If the substance suffers a decomposition, for instance 2C030d+ 6CoO + 02, then the temperature which should be taken is the decomposition temperature in using Fig. 4. The function of a support is, quite simply, to enable a large particle or tablet of catalyst to be made composed of very small readily sinterable crystals of catalyst which are stopped from coalescing by being separated from each other by the support component. The support component clearly has to be much more resistant to sintering than the catalytic species and neither react with nor be wetted by the catalytic species. Typically refractory oxides such as alumina, silica and magnesia are used as supports. Two methods of utilising a support are common. Firstly, a support may be used as a form of refractory surface on which the catalyst crystals are dispersed: rather like putting small drops of mercury on a table top. Secondly, the support may be used in the form of a more finely divided refractory crystal than the catalytic species positioned between particles of this

I I.000

bar

paths and pore dimensions in typical heterogeneous catalysts.

Fig. 4. Minimum crystal sizes for sintered compacts.

Theory and practice of the formulation of heterogeneous catalysts

species so that they are kept out of contact-rather like dust which can be used to stop a heap of mercury drops form coalescing. With the first method, the volume of support is always considerably greater than the volume of catalytic species, consequently this method is normally employed with precious metal catalysts. The second method, where the volume of catalyst normally exceeds that of the support, is employed usually with non-precious metal catalysts. Combinations of the two methods are also employed, as the use of the first method has the additional advantage that the catalyst particle has a continuous support structure which can be made mechanically very stmng and being made of inert species remains so during changing chemical conditions in the reactor (eg changes from oxidising to reducing). A very large pore continuous support can therefore give strength to the particle whilst small metal crystals at relatively high volumetric concentration are stabilised against sintering by packing them together with a very fine refractory support component, into the large pore structure. A typical catalyst using this technique is nickel hydrogenation catalyst powder where the coarse pore structure is made from diatomaceous earth and the fine support is silica gel. Before leaving the subject of the support, it must be emphasised that no support material is completely inert and the effect of the support on the course of the reaction chemistry must be considered. It must further be remembered that the use of a support may well introduce into the catalytic system more chemical components than at first expected. Thus alumina frequently contains sodium oxide which may be catalytically active. An example of the undesirable effects of components that may be present in a support on the course of methanol synthesis from CO + CO2 + HP over a copper catalyst is shown in Fig. 5. Quite small amounts of some of these components have suprisingly large effects on selectivity and life. STAGES IN PREPARATION OF SILICA-ALUMINA

Fig. 5. Bad effects of support components in low temperature methanol synthesis catalysts. silica and alumina-for instance a solution of sodium silicate mixed with a solution of aluminium sulphate plus sulphuric acid. If a crystalline zeolite is required then a additional component, normally an organic compound which is soluble in water is often also added. This compound affects the type of crystal which subsequently appears if the amorphous precipitate is allowed to ripen by holding it a temperature (usually near 100°C or above) for many hours. This additional compound sometimes has so large a molecule, for example the tetrapropyl ammonium cation, that it is trapped in the cages of the forming zeolite crystal and in these circumstances is supposed to fulfil the role of a template around which the crystal forms, thereby determining which of many possible zeolites having the same chemical composition crystallises. The choice of this “template” compound is entirely empirical and no very obvious relation exists between its molecular form and that of the zeolite cage it promotes.

CATALYSB

Having above discussed briefly the fundamentals of choice of the catalytic species and of the support the remainder of this review will be concerned with the practice of industrial catalyst preparation. The silicaalumina catalysts are an unusual class in as much as was indicated above, no support is needed and active sites are built into the catalyst during preparation and remain structurally present (even though deactivated possibly by coke coverage) for relatively long periods of catalyst use. Even here, however, restructuring of the catalyst lattice occurs in time, particularly if accelerated by the combined effects of high temperature and high steam partial pressure. The general technique for preparing silica alumina catalysts is shown in Fig. 6. If an amorphous lattice is required then the steps enclosed in double lines are ommitted. If a crystalline product is required (a zeolite) then they are included. The first stage of the preparation involves the rapid precipitation of an amorphous hydrated silica alumina gel by mixing aqueous alkaline and acid solutions containing

H

Fig. 6.

zeolite

Typical preparationof SiAl catalyst or of zeolite

1436

S. P. S.

After precipitation, if an amorphous product is required, or precipitation and ageing if a zeolite is required, the solid is filtered off, dried and, if containing trapped “template” molecules, is then subjected to gentle oxidising conditions which burn away the template molecules. Subsequently a thorough wash, including an acid wash, removes the bound sodium ions. The product, when finally dried, is the active catalytic H’ exchanging amorphous silica-alumina or crystalline zeolitic catalyst much used in petroleum refining for cracking and also for isomerisation duties. The presence of aluminium ions replacing silicon ions in the structure is an essential feature of the generation of the H’ activity. The ratio of Al/Si in the catalyst can be varied very greatly-ertain zeolites can even be made with no aluminium. Experiment determines which is the best ratio for any particular duty. Other cations, for example Fe+ and etc, can be incorporated into the lattice. THEPREPARATION

OF PREFORMED SUPPORT

When the catalytic species is dispersed on the surface of a support (the first method of using a support described above), it is usual to prepare the support separately as a preformed particle or tablet before inserting the catalytic species. In industrial practice this support is normally made from refractory oxides though occasionally active carbon is used. The preparative technique for the refractory oxide supports is that of making a fired ceramic except that an initial stage of making the “clay” is often required; particularly if a very high area but low porosity support is required. This first stage is essentially a precipitation of the hydroxides of the support material-for example aluminium hydroxide gel or magnesium hydroxide. This precipitation is brought about often by mixing an aqueous solution of the nitrate with sodium carbonate or of the sulphate with ammonium carbonate. Thorough washing is usually required to remove residual sodium in the first case or sulpbate in the second. Precipitation using a mixture of the nitrate and ammonium carbonate would eliminate such washing problems but would introduce a different hazard! The formation of ultra fine precipitates is an ill-understood phenomenon. Generally however, it appears that the lower the solubility of the precipitated material-the finer the precipitate, and the more complex the mixture being simultaneously precipitated--the finer the precipitate. Long ageing is, of course, undersirable when producing very fine precipitates but flocculation is essential otherwise filtration and washing is virtually impossible. Experimental investigation of the effects of pH, temperature, ageing time and precipitation solution strengths and compositions is always required in order to optimise the filtration and washing operations. The wet paste from the filters is dried and, where necessary, formed into desired shapes by extrusion or tabletting (see later in this review) and then given strength by firing. The fusing operation may roughly be pictured for a single component such as MgO or Al203 by reference to Fig. 4. High area alumina supports receive only a relatively low temperature firing, 600700“C, whereas low area supports are fired at tem-

ANDREW

peratures of up to 1200°C. The latter are no problem as the product is very strong. The production of high.area single component supports with reasonable strength is much more difficult and various devices, such as impregnating a weak support with nitric acid or with nitrate and subsequent light firing to “glue” together the structure may be employed. The addition of small amounts of other oxides which act as a sinter aid increases strength but at the possible disadvantage of complicating the catalytic properties of the support adversely. A typical additive is a small amount of Ti02 in A1203. THE PREPARATION OF IMF’REGNATFQ CATALYSTS

Performed supports are normally made into industrial catalysts by impregnating them with a solution of the catalytic species. A number of, types of such catalysts may be distinguished-see Fig. 7. Precious metal catalysts, for reasons of cost, always have a low metal content. This metal is dispersed in very fine crystals (5-15 A) either uniformly .on the surface of the support throughout its volume, or it is distributed only over a small depth into the particle, thereby forming an outer shell of catalyst. The choice of which type is employed depends on the nature of the desired and undesired catalytic reactions. If the desired product can suffer further degrading reactions over the catalytic species then it is usual to employ impregnation in the outer shell only, in order to eliminate the high build up of product concentration which could occur particularly with fast reactions as a result of a diffusive resistance inside a large particle of a catalyst which was active throughout its volume. A typical shell type catalyst is that used for the selective hydrogenation of acetylene in an ethylene stream. Here the catalytic species is metallic palladium dispersed in a very thin zone close to the outer surface of an alumina tablet. A typical example of a catalyst where the precious metal, for instance platinum, is dispersed throughout the volume is petroleum hydrocracking catalyst where the support also has isomerisation or cracking activity. Non precious metal catalysts with a high content of catalytic species on the support are invariably prepared by impregnation. Typical examples are nickel methane steam reforming catalysts which contain Z&30% nickel

Fig. 7.Differing types of impregnated catalysts.

Theory

and practice

of the formulation of heterogeneous catalysts

supported on preformed ceramic rings. A significantly different type of catalyst which can be prepared by impregnation is the cobalt molybdenum sulphide catalyst for hydrodesulphurisation of hydrocarbons. In this type, the active species, molybdenum sulphide, exists as a coating on top of the cobalt sulphide, which modifies its properties, and the cobalt sulphide itself coats the high area alumina support. The much higher degree of chemical interaction between the catalytic species and the support compared with the interaction between a reduced metal and a ceramic support such as alumina results in this coating phenomenon. Reduced metal, having little interaction with the support, does not “wet” it and therefore exists in discrete crystals, which, when small and in appropriate gas compositions and temperatures, hop around on the surface of the support being so poorly anchored. Another type of catalyst which can be formed by impregnation is that where the catalytic species is a liquid at operating temperature and the support merely acts like a sponge. Molten salt catalysts, such as Vz05K$O, salt mix supported on a silica support for SO2 oxidation or CuCll supported on alumina for oxychlorination of ethylene are examples. Control of the depth of impregnation is primarily by chemical means-see Fig. 8. The objective is to control the intensity of chemisorption of soluble salts of the impregnating solution on the support. With a weak SOIUtion of, for instance, a metal chloride, chemisorption of the metal ions on the ceramic surface is by exchange: M’Cl- + SOH+SOM

f H’Cl-.

By introducing a competitive ion, for instance NH; or an amine ion, the chemisorption of the metal ions is reduced and, as the impregnating solution advances into the support particle, when the dry particle is immersed in the impregnating bath, the metal ions penetrate further before being chemisorbed and removed from the solution. The chemisorption equilibrium can also be pushed to the left in the above equation by addition of acid (e.g. HCl) to the impregnating solution. Any free alkali in the

II

Limited

Metal

I

penetration

1437

support also has a marked effect in reducing the penetration of metal salts so that pre-impregnation of the support with alkali followed by drying and then impregnation with a metal salt is a method for producing very shallow depths of penetration. Limited impregnation is usually effected either by dipping the dry support into a bath of solution or by spraying solution onto the dry support whilst it is being tumbled so as to expose its surface. In the former case the amount of metal left on the support is determined by the voidage of the support multiplied by the strength of the solution and in the latter by the quantity of liquid sprayed multiplied by the strength of the solution. When complete penetration is required, dipping is normally used. Usually the support is preformed to its final particle shape before impregnation, however with some formulations the support is impregnated whilst in a fine powder state and then the catalyst is formed subsequently. THE PREPARATION

OF CO-PRECIPITATED

CATALYSTS

A particularly effective method of securing intimate mixture of support and catalytic species and one which is often used with high content non precious metal catalyst is co-precipitation of both support precursor and catalytic species precursor; for instance, the co-precipitation of nickel carbonate and aluminium hydroxide from aqueous mixed nitrates by a solution of sodium carbonate. It is important during this operation that the intimate mixing of the nickel and aluminium components when in solution is not entirely lost by mismanaging the precipitation. In particular the alkaline solution must never be added to excess of the metal salt solution otherwise precipitation will proceed in a sequential manner as the pH of the mixture increases and the precipitating components will precipitate separately. No subsequent agitation, however vigorous, is capable of effecting the desired intimacy of mixing once this has been lost. Figure 9 illustrates the sequential precipitation process by taking as an example the preparation of a copper on alumina catalyst by adding sodium carbonate solution sIowly to mixed copper aluminium nitrate solution. As precipitation proceeds and the pH rises, aluminium hydroxide precipitates first followed

11

Complete

I

By strong chemisorption and

short

chsmisorption due to strong

1Diwng 1 Fig. 8. Control of depth of impregnation.

Fig. 9. Variation in catalyst composition with acidity precipitation for a copper alumina catalyst.

during

1438

S. P. S.

by copper carbonate. Figure 9 also indicates the hazard of adding excess alkali, as re-sohttion of the precipitate occurs, again in sequential manner. The objective is to produce a flocculated multi-component precipitate in which the catalytic metal containing microcrystals are each surrounded by support precursor crystals and both these sets of crystals are very small, preferably less than 5OA. Furthermore it is desirable to effect this with the maximum ratio of metal precursor crystals to support precursor crystals. If it assumed that an effective separation of the metal crystals requires a “monolayer’ coverage of each by the support crystals, then this requirement defines a geometrical relation between the ratio of the crystal size of the metal crystals to that of the support crystals and the ratio of the volume of metal to that of the support material. As metal crystals which are in contact in the final catalyst rapidly unite when the catalyst is operating, by sintering together, the attainment of the above geometrical relation is one of the objectives of the precipitation process. In practice, the structure is not as regular as implied above and the proportionality constant between ratio of sintering component crystal size (the metal) divided by the support crystal size to the ratio of volume of sintering component (the metal) divided by the volume of support component in the catalyst has been obtained empirically by reference to a number of good industrial formulations. The relation so obtained is shown in Fig. 10 and is an attainable objective in many circumstances. It is the reviewer’s personal belief that the microgeometrical relations in supported precipitated catalysts are a key to the attainment and maintenance of a high exposed surface area of the catalytic species. Unfortunately this is a particular difficult area to observe directly as the particles of the support component (dSUPPORT) in Fig. 10 have to be very small. With metal crystals (dSINTER) of 100 A in size-which is readily observable by X-ray diffraction line broadening or by electron microscopy, and a high metal content catalyst containing, say 4 times the volume of metal to that of support, the support particles must only be about 5 A in size which is difficult

CSINTE.

(-1 “SUPPORT Fig.

Volume of metal Volume of ossocioted

ANDREW

to observe directly and is obtained by deduction from surface areas measured by physical absorption. The desired structure described above can be produced either by precipitating the metal component precursor and the refractory support component precursor as separate particles or in the form of a compound which is subsequently decomposed. In the first case the pH of the precipitate is important as it influences the flocculation process and hence the distribution of the two types of pL:rticle relative to each other and also the filterability and washability of the precipitate. Ideally, opposite charges should be formed on the two types of particles so that the floes consist of a neutral mass composed of oppositely charged support precursor particles surrounding catalyst precursor crystals. In the second case, ageing of the precipitate at an appropriate pH and, usually, an elevated temperature is necessary in order to promote the crystallisation of the compound phase-a typical example is a basic carbonate of the form MO, A1203b OH,COw where MO is a metal oxide (e.g. CuO, NiO). The process by which comparatively large crystals of compound are broken down into a supported metal will be described later. Thorough washing of the precipitate, preferably with de-ionised water is usually required in order to remove unwanted components, for example sodium ions, from the precipitate. Chloride ions are particularly deleterious to catalyst life and should not be used (e.g. as metal chlorides) in catalysts which are expected to operate in chloride free atmospheres at elevated temperatures otherwise sintering is accellerated. Following washing, the precipitate is dried and calcined to convert it from hydroxides and carbonates to oxides. This calcination must be carefully controlled, both with respect to temperature and with respect to degree of completion as it affects both the ease of reduction of the metal oxide to metal and the ease of forming the material into a tablet. This subject will therefore be discussed when outlining the theory and practice of catalyst particle forming and of reduction. The relation between metal particle size and support particle size given in Fig. 10 sets a lower Iimit to the metal particle size that will exist in the catalyst after reduction. If the metal precursor crystals (e.g. the metal carbonate or hydroxide) produced during precipitation contain more metal than corresponds to this lower limit size then the reduced and stabilised metal crystals will be appropriately larger than the lower limit. There is thus no virtue in increasing the support volume/metal volume ratio excessively in the hope that this must result in small metal particles. This point is illustrated in Fig. 11 which shows how the reduced and stabilised metal crystals decrease in size as the support fraction of the mixture increase (the concave part of Fig. 11 marked “sintering limit”) until the metal size reaches a limiting smallness set by the size of the metal precursor precipitate.

support

10. Size of metal crystals stabilised by well dispersed refractory support.

THE CREATION OF FORM

Unformed

catalyst

or support

mixtures

are shaped

Theory and practice of the formulation of heterogeneous catalysts

%

Fig. 1I.

Specific

Metal

in

catalyst

metal surface of a metal-metal

oxide catalyst

after sintering.

usually by one of five methods: Extrusion, wet pressing, dry tabletting, granulation, or casting. The first tHio essentially are similar from the physico-chemical standpoint. The powder to be formed is mixed with water and usually, a plasticising and probably also an extrusion aid, to give a mouldable clay like body. This is then shaped either by extrusion through a die or pressing between male and female moulds. Usually catalyst mixes of powdered oxides made by precipitation do not by themselves have the necessary mouldability and plasticity even when the optimum water content is added. They behave rather like wet sand. In order to give plasticity the particles must be given a lubricating layer of a clay, such as bentonite and the water made more rigid by dissolving starch in it. Figure 12 shows how an aluminium trihydrate powder mixed with 5% bentonite is plasticised by addition of starch. The much reduced sensitivity of yield strength to water content as a result of adding the starch results in good plastic properties suitable for feeding an extruder. Uniformity of feed composition, particularly water

1439

content, and absence of pockets of air is essential when operating an extruder, as the material must be weak enough to extrude without the generation of excessive pressures behind the die and strong enough when it emerges to retain its form even when sliced off with a knife. Pockets of air are normally removed by passing the extruder feed through a vacuum chamber situated half way along the extruder. The product is normally quite strong after drying (because of the starch) but must be fired to give it real strength in use when the starch decomposes at the catalyst operating temperature. Normally this firing is effected in a carefully controlled oxidising atmosphere, to remove the starch residues, and to form a ceramic bond between the components, a task which is aided by any bentonite in the formulation. The use of such additives as bentonite which remain in the catalyst mixture, however, in some circumstances can have deleterious effects on the catalyst performance by promoting undesired reactions. Dry tabletting of a catalyst powder obviates the necessity for such additives and also the necessity for producing strength by firing. Tabletting is however, much more expensive, as each individual tablet is separately formed however large or small. The production of small catalyst tablets is, therefore, particulady expensive per cubic meter of product, The production of strong particles by tabletting requires that the particles of tablet feed are subjected to a reasonably uniform triaxial (e.g. hydraulic) pressure and that under this pressure they crush together to form a dense bed which then welds at points of contact. At the tabletting stresses capable of being generated by normal machines the requirement of welding requires that a sufficient fraction of the material has adequate ductility, a quantity neither tabulated nor easy to measure in these circumstances. However a very crude relationship exists between ductility, melting point, elastic modulus and Mob’s scale of hardness. This is shown in Fig. 13. Bearing in mind that the tabletting machine has steel punches-even if hard faced-it is clearly necessary that the material to be tabletted should more ductile and in general “softer” than the dies otherwise they will

Moh’s

scale

of 4

5

6

7

Elastic

modulus

e

9

IO

107

lo=

105

Fig. 12. Typical variation of yield strength with water content and plasticiser content for an alumina trihydrate +5% bentonite mixture.

hardness

kg/cm’

Fig. 13. The relation between pelletability, melting point.

crystal strength and

S.

1440

P. S. ANDREW

rapidly be destroyed mechanically if sufficiently high pressures are generated to produce strong tablets. Normally with catalysts given form by tabletting, a sufficient quantity of hydroxide or carbonate precursor is left in the tabletting feed (e.g. CuC03, Al(OH)s etc.) to act as a binder in a mixture which can contain untablettable material such as C&O, AlzOs. Excessive binder is, however, deleterious to tablet strength in use, as the COZ and Hz0 is stripped off leaving a much weakened tablet. With tabletting using a die with a cylindrical hole and one or two cylindrical punches compressing the powder charge, the desired uniform triaxial pressure is difficult to obtain. Indeed, if precautions are not taken the nonuniform stresses left in the tablet during compaction may be so great that the tablet breaks up spontaneously on ejection from the die. Part of the cause of this nonuniformity is the friction between the powder and the punch surface and die surface. By adding an extreme pressure lubricant, such as graphite flakes or calcium stearate, to the powder this effect is reduced. As the length/diameter ratio for the tablet is increased the necessity for good lubrication increases, as can be seen from Fig. 14. This shows the load on a stationary bottom

punch supporting powder being compacted by driving down the top punch for different Ild ratios with and without graphite addition. Tablets which are somewhat weakened by internal stresses left after tabletting, though not to the extent of fracturing may, if of appropriate

composition, be strengthened by an annealing operation (like glass bottles). Sometimes the catalyst formulator wishes to produce a strong tablet by tabletting and, at the same time, high porosity is required. In these circumstances it is important to use a feed having the minimum practical particle size as this usually gives the maximum tablet strength

for a given talbetting

pressure

and hence for a Other factors being constant, the tablet strength, measured, for instance as a stress imposed on the two flat faces at fracture (the axial crushing strength), is normally proportional to the tabletting pressure applied to these two faces. The,proportional constant typically falls between

given desired tablet density and voidage.

>

-0.01

IO Inltlal

strengths cannot readily be measured as above and they are laid horizontally on a flat surface and loaded to fracture using a cylindrical ram face whose axis is at

right angles to that of the tablet. The loading on the top is thus a point contact whereas the tablet is supported by a line contact. The fracture load per unit cross section thus measured (the maximum horizontal crushing stress, MHCS) is, it can be shown roughly double the tensile strength of the material. Typical tabletting pressures, raw tablet tensile strengths of calcined and reduced tablets are listed in Table 2 along with typical loads to which tablets may be subjected during converter charging (due for instance to dropping from a height) and during steady operation (due to bed weight and gas pressure drop). These figures indicate that tablets normally charged in the raw (oxide) state are adequately strong to resist the charging operation, but that if reduced they are liable to break up to some extent during discharging unless specially designed to be of greater strength. Granulation, the building up of spherical catalyst particles by layering as the particle rolls in a granulator (for instance an inclined dish rotating on its axis) usually requires a fine powder feed together with a spray of granulating liquid. The product balls have little strength unless a hydraulic setting cement is included in the mixture. Otherwise mixtures rather similar to extruder compositions can be used, starch may be added and salt solutions which can crystallise out on drying, for exam-

z”,“,‘s’,‘,‘R:“,”

(=“‘z

I

LOADS

I

2

of pellet shape and graphite lubricrication of computation of a cylindrical pellet.

& REDUCED

TABLETS

of catalyst

;I

STRENGTH

tablets

&;;oo

:-

2-b

:-

0’5

:-

0.2 - I’0

:-

0-

TABLET

STRENGTH iM.H.C.5.)

ON

CHARGiNG

on

{;F;;;rE

TABLET TENSILE (= ‘/2 M.H.C.S.)

TENSILE

I

(p)

size on tablet strength.

0.02 and 0.1, see Fig. 15. With domed cylindrical tablets,

CALCINED

Fig. 14. Influence uniformity

Particle size

Fig. IS. Influence of initial particle

RAW

0

averago

1000

Table 2. Typical loads on and strengths

1-1 I

I

100

DURING

-2

(TENSILE)

LOAD ON TABLETS IN BED DURING STEADY OPERATION (TENSILE)

o-2

Theory and practice of the formulation of heterogeneous catalysts ple, aluminium nitrate. The wet granule, after drying may then be given further strength as for extrudates, by burning out any starch and then firing to produce some sintering Spheres have the minimum outside surface area per unit volume of any shape and therefore are more subject to internal diffusion limitation than any other shape. They can, however, be made very strong, and they are resistant to attrition and flow readily. Finally, casting is a method of forming catalysts which has a number of applications. The gelation of silica, silica alumina or alumina in the form of a spherical particle whilst, as it were, being held in the “mould” of an aqueous drop “suspended” in oil is a form of casting which has considerable use. Molten mixed oxide catalyst precursors, such as the Fea04, A120s, I&O etc mixture for ammonia synthesis may be either cast in the form of spheres, by breaking up a jet of liquid with cooling stream of water or air or by casting the liquid into a mould made of metal or even of catalyst powderthough these are considerable practical difficulties with an oxide melt at over 16OPC ! CALCINATtON ANDCATALYWREDUCTION Catalyst and support precursors such as hydroxides, carbonates and nitrates are converted first to oxides then subsequently (and usually in the plant reactor) the catalytic metal oxides are reduced to metals or sulphided to metal sulphides. These solid state transformations can create strength or produce weakness in the catalyst pieces. They can increase or decrease surface area. They are often essential to the creation of activity but if performed without due regard to the kinetics and equilibria of solid state transformations the final activity may be very much less than the optimum. Many solid oxides are to some extent mutually soluble-quite a few even form compounds. The extent to which very intimately mixed dry particles of hydroxides, carbonates or nitrates of support processors and metal precursors react together, if such is thermodynamically favourable, is a function of their particle size, the temperature and the presence of “minerahsers”, ie ions such as OH- and Cl-, which by entering into the structure of the components render their diffusive interaction faster. It must be remembered that this interaction has, in part, to be reversed during the process of catalyst reduction as metal ions of the catalyst metal must be extracted from, for instance a ceramic compound, (e.g. Ni” from NiMgOz) and then reduced to give reduced metal catalyst and that this operation must take place in a reasonably short time (a few hours or less) in a specified environment (often the plant reactor) at a temperature which is not to exceed a certain value (set often by metallurgical considerations of the plant vessels). The precise velocity of these solid state reactions is clearly, in the present state of knowledge, very much a matter for experiment. A rough guide in deciding the relation between calcination temperatures and reduction temperatures is that the former should never substantially exceed the latter when compound formation between metal oxide and support oxide is likely. and preferably, be below the latter. The guide is based on the principle

1441

that the velocity of decomposing by reduction the structure which was built during calcination must be adequately rapid and that the “solidity” of the structure built during catcination (i.e. crystal size and perfection of solid solutions or compounds) is determined by the temperatures of calcination, the higher the temperature the more “solid” the structure. More “solid” structures require higher reduction temperatures to break them down. Examples of the use of this guide are shown in Fig. 16. The two left hand columns show respectively the decomposition temperatures at atmosphere pressure of a number of hydroxides and carbonates as measured, for thermal analysis or therinstance, by differential mogravimetric analysis using relatively large pieces of filter cake of the pure components. The right hand column gives typical catalyst reduction temperatures for different supported metal catalysts. These reduction temperatures on the whole equate to the lower timit of temperature obtaining in the plant reactor during its steady operation. When the pieces of filter cake contain more than one component (as they invariably do) the decomposition of the more stable component is assisted by the decomposition of the less stable as, for instance, the CO2 of the carbonates is stripped off by the steam evolved from the decomposing hydroxides. A substantial fraction of the copper carbonate that might be used in preparing a low temperature water gas shift catalyst can thus be lost at a calcination temperature of 300°C due to the decomposition of the zinc hydroxide. The difficulty which would arise if a low temperature methanation catalyst (operating at say 320°C) was to be prepared from a nickel carbonates magnesium hydroxide co-precipitate can be seen, as the decomposition of the mixture during calcination would require temperatures substantially (at least 15O’C) above the reduction temperature, resulting in the formation of nickel magnesia spine1 which would be very slow to reduce leading to poor catalyst activity through lack of free nickel metal. The reduction process is the last stage of catalyst preparation with many supported non-precious metal and some metal oxide catalysts. The fact that this stage often occurs in the plant reactor should not disguise the fact that this is the stage where the final solid state trans-

Fig. 16. Calcination and reduction temperatures Cu, Zn, Fe, Ni, Fe, Al, Mg containing catalysts.

1442

S. P. S. ANDREW

formation occurs and where the catalyst comes into being. With precious metal catalysts and also certain nonprecious metal catalysts this stage is performed before the catalyst is charged to the plant which therefore receives active catalyst. Some reduction reactions are exothermic, in which case careful control of the reduction gas mix compositions is essential if unacceptably high catalyst temperatures are to be eliminated. These temperatures would result in catalyst sintering. Normally, as reduction usually takes place at around the final catalyst operating temperature and temperature rises of more than say 50°C are to be avoided, the reduction process takes place in an inert gas containing a low percentage of reductant-for example 1% HZ, the possible temperature rise is then limited by the heat capacity of the inert gas. With endothermic reductions, the above precautions are unnecessary. Other factors may be important, particularly the water vapour content of the reducing gas. The ratio of the water vapour partial pressure to the hydrogen partial pressure (pHZ0/pH3 during the reduction process is very important when reducing compounds or solids solutions of metal oxide and support oxide, as this quantity has a marked effect on the fineness of subdivision of the resultant metal plus refractory oxide structure. The ammonia synthesis catalyst which, before reduction, may be considered as a solid solution of alumina in Fes04 is the most studied example of this phenomenon. The effect of p~&p~~ may possibly be understood by envisaging the solid state transformation which gives rise to the free support oxide during the reduction process. The following analysis is basically a hypothesis as the necessary investigation of the reduction process as it proceeds under a very high resolution electron microscope has not yet been done. The un-reduced crystals may be considered as a solid which is reduced by means of a reduction front moving into the crystal leaving a porous mixture of metal crystallites and support oxide behind as the oxygen is removed as water. The support ions (for instance A13’ in the ammonia synthesis catalyst) presumably migrate in the reducing oxide being driven by supersaturation gradients to nucleation sites where they deposit to form AIz03. The distance over which this migration can take place is a function of the time available which is determined by the velocity and thickness of the reduction front. This distance determines the number of nucleation sites triggered and hence the density and size of the alumina crystallites produced. A simple algebraic analysis generally similar to that which has been applied to the solid state precipitation of iron carbide from a solid solution in carbon due to the movement of a cold front into a hot crystal suggests the relation shown against the top diagram in Fig. 17. Iron crystallites then form between the alumina crystallites and, over two or three days, those in contact sinter together until all that remain are isolated from each other by alumina crystallites (bottom diagram in Fig. 17). When this has occurred the geometric relation indicated earlier in Fig. 10 applies so that the ratio of the metallic iron crystallite size to that of

of support

Nuclei

phase

Partlolly reduced catdlyat

All metal separated

crystals by support

Stabilised reduced

catalyst

Fig. 17. Three stages in reduction

of a metal-supporl

compound.

the refractory alumina crystallite size is given by the equation at the bottom right of Fig. 17. Combining that equation with one at the top of Fig. 17 gives:

a

“1 metal ] (reduction velocity)-1’2. support

[ vol

Experiment shows that the rate of reduction of several metal oxides is roughly proportional to pHZ/pHlo at a given temperature: which Hence Da(vol metal/v01 support) (pH20/pHz)“2 explains the dependence of metal crystallite size and hence catalyst surface area on the ratio pH20/pH2 during reduction. By fitting data from a number of catalysts and in particular ammonia synthesis a value has been given to the proportionality constant and the resultant relation is shown in Fig. 18. Clearly this figure should be looked on merely as a guide or an informed guess in the absence of experimental data on any particular system rather than as a firm prediction. The quantitative science of these complex phase transformations does not permit anything more substantial at present. Adequately low partial pressures of water vapour are normally maintained during reduction by operating with a high velocity of reducing gas and maintaining the rate of reduction low

Pn,

Fig.

/ PHZO

18. Size of metal crystals produced support compound.

by reducing

metal-

Theory and practice of the formulation of heterogeneous catalysts

by keeping the catalyst bed temperature low. Reduction of metallic catalysts is an autocatalytic phenomenon as hydrogen atoms formed by chemisorption on the reduced metal spill over into the unreduced zone producing further reduction. This phenomenon can be utilised to formuiate catalysts capable of being reduced at lower temperatures than otherwise by incorporating very small amounts of easily reducible metal into the mixture, for example, platinum or palladium into a non-precious metal formulation. These small additions after they have accelerated the reduction process then dissolve in the main catalytic metal species, and, not being surface active, are effectively lost from the catalytic system.

TEE FABRICATION OF MATRIX CATALYSTS

Most industrial heterogeneous catalysts are used in the form of powder or of tablets, rings, spheres, extrudates or lumps. With fast reactions and the desire for low reagent pressure drops through the catalyst bed there has long been a call for more complex structures such as are employed in compact heat exchangers etc. As most catalyst structures are either ceramic or metal-ceramic the fabrication of such structures can be expensive and they are only used in special circumstances. Matrix catalysts almost invariably use a pre-shaped macrostructure to form the matrix to which the catalyst mixture is attached. There are two types-those in which the pre-shaped structure is ceramic and those where it is a metal-see Fig. 19. The ceramic structures may be fabricated by extruding a ceramic paste through a complex die which produces a large “sausage”-perhaps about 15 cm dia. having a very large number of parallel holes throughout its length, the holes being separated by thin ceramic walls perhaps only 0.5 mm thick. Alternatively they can be fabricated by forming flat and corrugated sheets of ceramic paste,

Dip in high

orea

ceramic

suspension

I Dry

ond tire I

Fig. 19. Fabrication of matrix catalysts.

1443

again only about 0.5 mm thick and then rolling them up together rather like a Swissroll. In both cases the damp ceramic paste matrix is then dried and fired to give it strength. Both these operations need to be accomplished with great care otherwise differential shrinkage will occur between diierent parts of the structure leading to cracking and mechanical failure. Large ceramic structures such as these, which may be several tens of centimeters in size, are subject to thermal shock failure if suddenly heated or cooled unless they are made of a ceramic having a very low volumetric coefficient of expansion. This additional requirement effectively limits the ceramic to a single species, cordierite, a magnesium alumina silicate. When thermal shock is absent other ceramics such as alumina are used. Metal structures may be made by assembling preformed metal sheets into a matrix, for instance by winding corrugated and flat sheets together, Swissroll fashion. This method of construction is used in a current type of automobile exhaust oxidation catalyst. Both ceramic and metal structures, though they have a high external surface area are unsuitable as anti-sintering supports for metal catalysts. Both catalytic species and supporting species must, therefore, be coated onto the structure. This is usually effected by dipping the structure into a suspension of a high area ceramic plus some binding salts and, drying and firing the coating at sufficient temperature to cause it to adhere to the underlying structure. When employing a metal structure the surface preparation of the metal before coating is most important (as with all ceramic enamelling processes) if good adhesion of the coating is to be secured. In general the metal surface must be non greasy and have a thin but strongly adherent oxide layer. The above automobile exhaust catalyst employs Fecralloy, a steel containing chromium, aluminium and yttrium which forms a strongly adherent oxide layer which resists further oxidation of the metal and forms a strong base onto which an aluminia support can be fixed. Following coating with the high surface area ceramic, the catalytic species, usually a metal is impregnated from solution into the coat. Frequently the ceramic coats have to be rather thin otherwise they detach from the metal readily when subjected to thermal shock. Thus it may not be practical to make a supported catalyst having a high weight of catalytic metal deposited in the ceramic. In order to obtain good activity, the catalytic metal should therefore have a high intrinsic activity and this type of support is often used with precious metal catalysts. The chief use that has been found for matrix catalysts is in the treatment of automobile exhaust gases where low pressure drop, low thermal inerta and the ability firmly to grip a matrix block when it is subject to the vibration of the exhaust system give them an advantage over more conventional shapes. They are also of use in various gaseous effluent treatment systems where low pressure drop is valuable. A non-matrix use of similar technology for coating a catalyst mixture on a metal structure is the catalytic oven liner which maintains domestic ovens clean by absorbing and then catalytically oxidising fat splashes on the walls.

1444 SOME

S. P. S. LESS USUAL

METHODS

OF CATALYST

ANDREW

PREPARATION

A number of less commonly used methods of catalyst preparation are shown in Fig. 20. The list is not exclusive. In particular the use of homogeneus organo-metallic catalysts made heterogeneous by chemically anchoring them on a high area support and the vaguely related technique of anchoring enzymes on supports are ignored as these methods are of a different class to the subject of this review. Of the methods listed in Fig. 20, the Raney method is frequently used for preparing a nickel hydrogenation catalyst for laboratory organic preparations. The porous nickel produced has a high surface area and is hence very active at low temperatures, it is not, however, particularly stable and would be unsuitable for commercial use at higher temperatures as there is little provision in the formulation for stopping nickel crystal coalescence and hence loss of area and activity. The hydrolysis of urea in a heated aqueous solution can be used as a method of generating in situ alkalinity in an otherwise inaccessible place. This phenomenon may be employed, for instance, to cause the precipitation of hydroxides uniformly throughout the pores of a support which has been impregnated with a solution of metal salts plus urea. For instance, a high strength low surface area alpha alumina ring could be impregnated with a solution containing a mixture of aluminium and nickel nitrates plus urea, the hydrolysis of the urea would then precipitate in the pores an intimate mixture of nickel hydroxide and aluminium hydroxide which, by calcination and reduction could be converted into a high strength high area nickel catalyst. The use of organo-metallic compounds which are decomposed in a reducing atmosphere to form a carbon supported metallic catalyst has the virtue that the route can be perfectly free from water (an accelerator of particle sintering with oxides). The disadvantage is that the method is somewhat costly, requiring, for instance, the synthesis of metal ketenides or at the least oxalates, more seriously, the carcitrates etc, and, perhaps bonaceous support is often unstable and disappears due to oxidation or hydrogenation when the catalyst is in use.

This method therefore is more of a laboratory curiosity than industrial process at present. The use of sputtering or ion impregnation techniques to coat the outer regions of a ceramic powder or coating with very fairly divided metal particles also receives episodic attention. The technique employs a high vacuum and the individual atoms of the metal can be deposited on the surface of the powder, which is tumbled to expose it to the metal beam, with a wide range of energies. High energy beams result in penetration of the metal atoms below the surface of the ceramic crystallites where they are buried and catalyticaNy ineffective. Low energies are therefore usually employed. Confinement of the metallisation to the outside visible surface of the catalyst particles results in metal contents of the catalyst being very low and activity per unit volume of reactor correspondingly low. Coupled with the high cost of catalyst fabrications this fact renders this method generally un-

attractive. Condensation of metal oxides and refractory oxides from the vapour phase due to sudden quenching of a very high temperature gas stream can produce very finely divided powders which could be used as a catalyst precursor. The gas stream can be prepared for instance by passing a cold stream plus a feed of catalyst precusors in solution through a plasma arc which vaporises all the catalyst components. Quenching can be effected by adding cold gas. Flocculation of the product in a residence chamber is desirable before collection by electrostatic precipitation. No industrial catalysts are so far made by this method which is clearly expensive, it is furthermore not easy to produce crystals of the desired

smallness in size (i.e. less than SOA) and to retain them. THE STATE

OF

THESCIENCE

The aim of this review is to indicate, in a brief space, the main factors which a practical formulator of industrial heterogeneous catalysts must bear in mind. Success is achieved by optimising what are often a series of conflicting requirements. This optimisation, though aided by a knowledge of the relevant theories of the catalytic process and theories of the solid state trans-

I

Tablet

Fig. 20. Some less usual methods of catalyst preparation.

Theory and practice of the formulation of heterogeneous catalysts formation occasioned during catalyst preparation, is not crucially dependent on such theories which primarily only shorten the experimental programme and suggest further lines of investigation. The complexity of the phenomena taking place, together with the fact that they frequently have opposing influences, means that their prediction by numerical means is neither a commercially realistic approach nor even scientifically possible in our present state of knowledge. This is not to argue that pure empiricism is the best course, for no pocket is sufficiently deep to be able to fund the preparation of the infinite variety of formulations that are possible. Rather, what is required, is an intelligent assessment of the relative importance of the factors which are likely to be of greatest importance in the preparation of an heterogeneous catalyst and a willingness to discard from experimental consideration those factors judged to be of secondary importance. Experimental formulations are prepared on this basis and

CES Vol. M. No. 9-B

1445

tested under as realistic conditions as can conviently be set up. From the results so obtained the formulations are modified in an attempt to correct observed imperusing general observations of the type fections, This process of review. listed in this formulation + testing+ modification to theory of formulation is repeated again and again until either the desired commercial target is reached or it is judged that the laws of nature do not permit it to be reached with the facilities that can be pressed into service. The subject of laboratory reactors and the testing of heterogeneous catalysts has not been touched on in this review, neither has the subject of the measurement of kinetics or the design of full scale reactors. Nevertheless the reviewer believes that engineers concerned with such measurements and their interpretation for full scale design would benefit from an understanding of the complexities of catalysts otherwise they are liable to build numerical castles on the most unsound of foundations.