Download pdf - Reaction’phase, and/or (iii) chemical transformations of cat-alytic phases to non-catalytic phases. The ﬁrst two processes are typically referred to as “sintering”. The third

Catalyst deactivation is the focus of the current lecture

Reaction kinetics

Aspects of heat & mass transport

Catalyst Deactivation

Fluid flow

Developing models for catalytic reactors

C.H. Bartholomew / Applied Catalysis A: General 212 (2001) 17–60 19

Fig. 1. Conceptual model of poisoning by sulfur atoms of a metalsurface during ethylene hydrogenation.

available for catalysis. Thus, poisoning has operationalmeaning; that is, whether a species acts as a poisondepends upon its adsorption strength relative to theother species competing for catalytic sites. For exam-ple, oxygen can be a reactant in partial oxidation ofethylene to ethylene oxide on a silver catalyst and apoison in hydrogenation of ethylene on nickel. In ad-dition to physically blocking of adsorption sites, ad-sorbed poisons may induce changes in the electronicor geometric structure of the surface [16,20].

Mechanisms by which a poison may affect cat-alytic activity are multifold as illustrated by a con-ceptual two-dimensional model of sulfur poisoningof ethylene hydrogenation on a metal surface shownin Fig. 1. To begin with, a strongly adsorbed atom ofsulfur physically blocks at least one three- or four-foldadsorption/reaction site (projecting into three dimen-sions) and three or four topside sites on the metalsurface. Second, by virtue of its strong chemicalbond, it electronically modifies its nearest neighbormetal atoms and possibly its next nearest neighboratoms, thereby modifying their abilities to adsorband/or dissociate reactant molecules (in this case H2and ethylene molecules), although these effects do notextend beyond about 5 a.u. [20]. A third effect maybe the restructuring of the surface by the strongly ad-sorbed poison, possibly causing dramatic changes in

Table 2Common poisons classified according to chemical structure

Chemical type Examples Type of interaction with metals

Groups VA and VIA N, P, As, Sb, O, S, Se, Te Through s- and p-orbitals; shielded structures are less toxicGroup VIIA F, Cl, Br, I Through s- and p-orbitals; formation of volatile halidesToxic heavy metals and ions As, Pb, Hg, Bi, Sn, Zn, Cd, Cu, Fe Occupy d-orbitals; may form alloysMolecules which adsorb with

multiple bondsCO, NO, HCN, benzene, acetylene,other unsaturated hydrocarbons

Chemisorption through multiple bonds and back bonding

catalytic properties, especially for reactions sensitiveto surface structure. In addition, the adsorbed poisonblocks access of adsorbed reactants to each other (afourth effect) and finally prevents or slows the surfacediffusion of adsorbed reactants (a fifth effect).

Catalyst poisons can be classified according to theirchemical makeup, selectivity for active sites and thetypes of reactions poisoned. Table 2 lists four groupsof catalyst poisons classified according to chemi-cal origin and their type of interaction with metals.It should be emphasized that interactions of groupVA–VIIIA elements with catalytic metal phases de-pend on the oxidation state of the former, i.e. howmany electron pairs are available for bonding and thedegree of shielding of the sulfur ion by ligands [15].Thus, the order of decreasing toxicity for poisoning ofa given metal by different sulfur species is H2S, SO2,SO4

2!, i.e. in the order of increased shielding byoxygen. Toxicity increases with increasing atomic ormolecular size and electronegativity, but decreases ifthe poison can be gasified by O2, H2O or H2 present inthe reactant stream [20]; for example, adsorbed carboncan be gasified by O2 to CO or CO2 or by H2 to CH4.

Table 3 lists a number of common poisons forselected catalysts in important representative reac-tions. It is apparent that organic bases (e.g. amines) andammonia are common poisons for acidic solids suchas silica-alumina and zeolites in cracking and hydroc-racking reactions while sulfur- and arsenic-containingcompounds are typical poisons for metals in hy-drogenation, dehydrogenation and steam reformingreactions. Metal compounds (e.g. Ni, Pb, V and Zn)are poisons in automotive emissions control, catalyticcracking and hydrotreating. Acetylene is a poison forethylene oxidation, while asphaltenes are poisons inhydrotreating of petroleum residual.

Poisoning selectivity is illustrated in Fig. 2, a plotof activity (the reaction rate normalized to initial rate)

26 C.H. Bartholomew / Applied Catalysis A: General 212 (2001) 17–60

Boudart’s more general classification of structure-sensitive and structure-insensitive catalytic reactions.In coke-sensitive reactions, unreactive coke is de-posited on active sites leading to activity decline, whilein coke-insensitive reactions, relatively reactive cokeprecursors formed on active sites are readily removedby hydrogen (or other gasifying agents). Examples ofcoke-sensitive reactions include catalytic cracking andhydrogenolysis; on the other hand, Fischer–Tropschsynthesis, catalytic reforming and methanol synthesisare examples of coke-insensitive reactions. On the ba-sis of this classification Menon [49] reasoned that thestructure and location of a coke are more importantthan its quantity in affecting catalytic activity.

Consistent with Menon’s classification, it is alsogenerally observed that not only structure and loca-tion of coke vary but also its mechanism of formationvaries with catalyst type, e.g. whether it is a metalor metal oxide (or sulfide, sulfides being similar tooxides). Because of these significant differences inmechanism, formation of carbon and coke is dis-cussed below separately for supported metals and formetal oxides and sulfides.

2.2.1. Carbon and coke formation on supportedmetal catalysts

Possible effects of fouling by carbon (or coke) onthe functioning of a supported metal catalyst are illus-trated in Fig. 9. Carbon may (1) chemisorb stronglyas a monolayer or physically adsorb in multilayersand in either case block access of reactants to metalsurface sites, (2) totally encapsulate a metal particleand thereby completely deactivate that particle, and(3) plug micro- and mesopores such that access ofreactants is denied to many crystallites inside thesepores. Finally, in extreme cases, strong carbon fila-ments may build-up in pores to the extent that theystress and fracture the support material, ultimatelycausing disintegration of catalyst pellets and pluggingof reactor voids.

Mechanisms of carbon deposition and coke forma-tion on metal catalysts from carbon monoxide andhydrocarbons [4,44–48] are illustrated in Figs. 10 and11. Different kinds of carbon and coke which varyin morphology and reactivity are formed in thesereactions (see Tables 6 and 7). For example, CO dis-sociates on metals to form C!, an adsorbed atomiccarbon; C! can react to C", a polymeric carbon

Fig. 9. Conceptual model of fouling, crystallite encapsulation andpore plugging of a supported metal catalyst due to carbon depo-sition.

film. The more reactive, amorphous forms of carbonsformed at low temperatures (e.g. C! and C") areconverted at high temperatures over a period of timeto less reactive, graphitic forms [47].

It should also be emphasized, that some forms ofcarbon result in loss of catalytic activity and some donot. For example, at low temperatures (<300–375!C)condensed polymer or "-carbon films and at high tem-peratures (>650!C) graphitic carbon films encapsulatethe metal surfaces of methanation and steam reform-ing catalysts [47]. Deactivation of steam reformingcatalysts at high reaction temperatures (500–900!C)may be caused by precipitation of atomic (carbidic)

Fig. 10. Formation, transformation and gasification of carbon onnickel (a, g, s refer to adsorbed, gaseous and solid states, respec-tively) [47].

28 C.H. Bartholomew / Applied Catalysis A: General 212 (2001) 17–60

Fig. 12. Electron micrograph of 14% Ni/Al2O3 having undergone extensive carbon deposition during CO disproportionation at 673 K,P CO = 4.55 kPa (magnification of 200,000; courtesy: BYU Catalysis Laboratory).

hydrogenation). In the intermediate temperature rangeof 375–650!C, carbon filaments (Fig. 12) are formedby precipitation of dissolved carbon at the rear sideof metal crystallites causing the metal particles togrow away from the support [44]. Filament growthceases when sufficient carbon accumulates on thefree surface to cause encapsulation by a carbon layer;however, encapsulation of the metal particles doesnot occur if H2/CO or H2O/hydrocarbon ratios aresufficiently high. Thus, carbon filaments sometimesformed in CO hydrogenation or steam reforming ofhydrocarbons would not necessarily cause a loss of

intrinsic catalyst activity unless they are formed insufficient quantities to cause plugging of the pores[47] or loss of metal occurs as the carbon fibersare removed during regeneration [52,53]. However,in practice, regions of carbon forming potential insteam reforming must be carefully avoided, sinceonce initiated, the rates of filamentous carbon for-mation are sufficiently high to cause catastrophicpore plugging and catalyst failure within a few hoursto days.

The rate at which deactivation occurs for a givencatalyst and reaction depends greatly on reaction

Deactivation is an observable phenomenon; it is easy to conceptualise at least some modes of deactivation


Fig. 16. Two conceptual models for crystallite growth due tosintering by (A) atomic migration or (B) crystallite migration.

growth of the catalytic phase, (ii) loss of supportarea due to support collapse and of catalytic surfacearea due to pore collapse on crystallites of the activephase, and/or (iii) chemical transformations of cat-alytic phases to non-catalytic phases. The first twoprocesses are typically referred to as “sintering”. Thethird is discussed in the next section under solid–solidreactions. Sintering processes generally take placeat high reaction temperatures (e.g. >500!C) and aregenerally accelerated by the presence of water vapor.

Most of the previous sintering and redispersionwork has focused on supported metals. Experimentaland theoretical studies of sintering and redispersionof supported metals published before 1997 have beenreviewed fairly extensively [8,80–89]. Three principalmechanisms of metal crystallite growth have been ad-vanced: (1) crystallite migration, (2) atomic migration,and (3) (at very high temperatures) vapor transport.The processes of crystallite and atomic migration areillustrated in Fig. 16. Crystallite migration involvesthe migration of entire crystallites over the supportsurface followed by collision and coalescence. Atomicmigration involves detachment of metal atoms fromcrystallites, migration of these atoms over the supportsurface and ultimately, capture by larger crystallites.Redispersion, the reverse of crystallite growth in thepresence of O2 and/or Cl2, may involve (1) formationof volatile metal oxide or metal chloride complexeswhich attach to the support and are subsequently de-composed to small crystallites upon reduction and/or(2) formation of oxide particles or films that breakinto small crystallites during subsequent reduction.

There has been some controversy in the literatureregarding which mechanism of sintering (or redisper-sion) operates at a given set of conditions. However,each of the three sintering mechanisms (and two

dispersion mechanisms) is a simplification whichignores the possibility that all mechanisms may occursimultaneously and may be coupled with each otherthrough complex physicochemical processes includ-ing the following: (1) dissociation and emission ofmetal atoms or metal-containing molecules frommetal crystallites, (2) adsorption and trapping of metalatoms or metal-containing molecules on the supportsurface, (3) diffusion of metal atoms, metal-containingmolecules and/or metal crystallites across supportsurfaces, (4) metal or metal oxide particle spreading,(5) support surface wetting by metal or metal oxideparticles, (6) metal particle nucleation, (7) coales-cence of, or bridging between, two metal particles,(8) capture of atoms or molecules by metal particles,(9) liquid formation, (10) metal volatilization throughvolatile compound formation, (11) splitting of crys-tallites in O2 atmosphere due to formation of oxidesof a different specific volume, and (12) metal atomvaporization. Depending upon reaction or redispersionconditions, a few or all of these processes may be im-portant; thus, the complexity of sintering/redispersionprocesses is emphasized.

In general, sintering processes are kinetically slow(at moderate reaction temperatures) and irreversibleor difficult to reverse. Thus, sintering is more easilyprevented than cured.

2.3.2. Factors affecting metal particle growth andredispersion in supported metals

Temperature, atmosphere, metal type, metal disper-sion, promoters/impurities and support surface area,texture and porosity, are the principal parameters af-fecting rates of sintering and redispersion (see Table 8,[8,85–89]). Sintering rates increase exponentially withtemperature. Metals sinter relatively rapidly in oxygenand relatively slowly in hydrogen, although dependingupon the support, metal redispersion can be facilitatedby exposure at high temperature (e.g. 500–550!C forPt/Al2O3) to oxygen and chlorine followed by reduc-tion. Water vapor also increases the sintering rate ofsupported metals.

Normalized dispersion (percentage of metal expo-sed at any time divided by the initial percentageexposed) versus time data in Fig. 17 show that attemperatures of 650!C or higher, rates of metal sur-face area loss (measured by hydrogen chemisorption)due to sintering of Ni/silica in hydrogen atmosphere

Sintering

Fouling


[117–119] observed surface etching, enhanced sinter-ing, and dramatic surface restructuring of Pt thin filmsto faceted particles during ethylene oxidation over arelatively narrow temperature range (500–700!C). Thesubstantially higher rate of sintering and restructuringin O2/C2H4 relative to that in non-reactive atmo-spheres was attributed to the interaction of free radi-cals such as HO2, formed homogeneously in the gasphase, with the metal surface to form metastable mo-bile intermediates. Etching of Pt-Rh gauze in a H2/O2mixture under the same conditions as Pt surfaces(600!C, N2/O2/H2 = 90/7.5/2.5) was reported byHess and Phillips [120]. A significant weight loss was

Fig. 20. (a) SEM of Pt-Rh gauze after etching in N2/O2/H2 = 90/7.5/2.5 at 875 K for 45 h [120]. (b) SEM of Pt-Rh gauze after use inproduction of HCN; magnification: 1000" (photographs courtesy of Dr. Ted Koch at Du Pont).

observed in a laminar flow reactor with little change insurface roughness, while in an impinging jet reactor,there was little weight loss, but substantial restructur-ing of the surface to particle-like structures, 1–10 !min diameter; these particles were found to have thesame Pt-Rh composition as the original gauze. Thenodular structures of about 10 !m diameter formedin these experiments are strikingly similar to thoseobserved on Pt-Rh gauze after use in production ofHCN at 1100!C in 15% NH3, 13% CH4 and 72% air(see Fig. 20). Moreover, due to the high space veloc-ities during HCN production, turbulent, rather thanlaminar flow would be expected as in the impinging

SEM micrographs showing deactivation

Poisoning

(Bartholomew, 2001)

This is from my course notes a few years ago. In the previous slide, we saw representations of each of these modes of deactivation

130 S.T. Sie / Applied Catalysis A: General 212 (2001) 129–151

mechanisms of catalyst deactivation are operative andthese are coped with in different ways. Moreover,depending upon reaction conditions chosen or char-acteristics of the feedstock to be processed, the speedof catalyst deactivation can vary widely and conse-quently lead to different process configurational andoperational choices.

2. Nature of catalyst deactivation and remedialactions

Catalysts can lose their activity during a processfor a variety of reasons. Most common reasons arepoisoning or reaction inhibition by impurities in thefeed or by reaction byproducts, deposition of poly-meric material including “coke” on the catalyst asa result of side or consecutive reactions, and loss ofcatalyst dispersion by “sintering” of small particles ofthe active material. In addition, catalysts may becomedeactivated by loss of active components by leachingor vaporisation, or by changes in their porous texture.

Fig. 1. Flow scheme of BP’s paraffin isomerisation process, showing the dosing of organic chloride to compensate for catalyst deactivationby stripping of hydrogen chloride [5].

Changes in porous texture that can affect the perfor-mance of a catalyst are, for instance, loss of specificsurface area through sintering of the carrier or loss ofpermeability through plugging of pores.

An important distinction between the deactivationprocesses is whether they are reversible or irreversibleunder the conditions of the process. A reversible de-activation caused by leaching of active material fromthe catalyst in a continuous process can be copedwith by adding the leached product to the feed. At thecorrect dosing rate, the supply and leaching rates arein balance so that the net loss of active material fromthe catalyst is zero. An example of such a process isthe isomerisation of paraffins with a catalyst based onchlorided alumina. Loss of chlorine from the catalystas a result from some hydrolysis by moisture is coun-teracted by feeding hydrogen chloride or a hydrogenchloride generating compound to the feed to make upfor the hydrogen chloride lost, see Fig. 1.

An irreversibly deactivated catalyst has either to bediscarded or to be subjected to a reactivation treat-ment to restore its performance. The latter treatment

(Sie, 2001)

Various means and methods are used for overcoming catalyst deactivation. This is one of them.

Catalyst based on chlorided alumina is lost due to hydrolysis

S.T. Sie / Applied Catalysis A: General 212 (2001) 129–151 131

Fig. 2. Periodic catalyst reactivation during catalytic hydrodewaxing of Arabian heavy way distillate of 55–65!F pour point [6].

Fig. 3. Viable process technologies as determined by rapidity of catalyst deactivation.

Knowledge of the mode(s) of deactivation is essential in order to determine the best way to overcome, or ‘reverse’ the effects of it

(Sie, 2001)

In this example, the deposited polymeric material is removed from the catalyst by periodically increasing the reactor temperature

132 S.T. Sie / Applied Catalysis A: General 212 (2001) 129–151

is termed regeneration or rejuvenation. Rejuvenationis a term commonly used for a treatment which canbe carried out under conditions that are rather similaror not principally different from the operating condi-tions. For instance, polymeric material that had beendeposited on a catalyst under relatively mild condi-tions of hydroprocessing may still be sufficiently reac-tive to be hydrocracked under somewhat more severeconditions. Thus, the deposited polymeric materialmay be removed in the form of cracked fragmentsat somewhat increased temperature and/or increasedhydrogen rate (hydrogen stripping). An example ofsuch a catalyst rejuvenation is shown in Fig. 2.

When the deactivation process is irreversible andreactivation requires conditions that are incompatiblewith those of the main process, the catalyst has eitherto be regenerated in a separate step or it must bedisposed off. Common examples of such irreversibledeactivation processes are deposition of “coke” onthe catalyst and sintering of metals. “Coke” depositedon the catalysts at high temperatures in the form ofa carbon-rich polyaromatic material is generally toounreactive to be hydrocracked in a hydrogen strippingstep and has to be removed by oxidation to carbonoxides. Sintering of active materials can be undoneby redispersing, e.g. sintered platinum on aluminamay be redispersed by treatment with chlorine. Thesedecoking as well as redispersion steps are carriedout in an oxidative atmosphere, which is incompati-ble with the reductive atmosphere of a hydrocarbonconversion process.

If the regeneration option is not chosen, spent cat-alyst has to be disposed off. In catalyst disposal, thechoice between dumping and working up of spentcatalysts for recovery of valuable materials is dictatedby the economics of materials recovery and does notaffect the main process other than via the catalystreplacement costs. Regeneration of a catalyst can becarried out either ex-situ or in-situ. In the formercase, the regeneration may be carried out in a separatefacility at a different location from the main processand may even be performed as a service by an outsidecompany. In the case of in-situ regeneration, the re-generation facilities are much more an integral part ofthe process installation. The regeneration proceduremay even be carried out in the same reactor as used inthe main process, but with the latter temporarily out ofregular service during the regeneration campaign (this

mode of operation is often termed semi-regenerativeoperation). Another possibility is to install more thanone reactor, with each reactor being alternately op-erated in a process mode or in a regeneration mode(swing operation). Yet another possibility is not tokeep a batch of catalyst in the same vessel all ofthe time, but to circulate catalysts between reactionand regeneration vessels (continuous regeneration).This implies that the catalysts must be able to movein and out of a reactor, which is made possible bymoving-bed or fluid-bed technology.

3. Rate of catalyst deactivation

In batch processes, the rate of catalyst deactivationdetermines catalyst consumption and is in generalonly a factor of economic importance. Aside from the

Fig. 4. Compensating for catalyst deactivation in catalytic reform-ing by increasing the reactor temperature during the run [7].

•  Sometimes, we want to maintain a constant conversion; in this example, reactor temperature had to be steadily rising in order to ensure that.

•  This is a reforming reactor; so this must be very common in occurrence!

The costs associated with overcoming the adverse effects of deactivation will be significant!

(Sie, 2001)

Catalyst deactivation is important and the study of it a crucial component of catalyst science and technology…

•  In the next few lectures, we will •  Model deactivation •  Incorporate deactivation in simple reactor models •  Perform simulations to determine the impact of deactivation on

outlet conversions •  Perform calculations to determine rate laws for deactivation from

laboratory data •  Discuss important ways in which deactivation can be dealt with (at

least temporarily negated) & •  A couple of important industrial reactor designs for deactivating

catalytic reactions