3.2. Catalyst deactivation mechanisms

3.2.1. Deactivation by thermal degradation and sintering

Temperature has become an increasingly important factor for deactivation of three-way catalysts due to the fact that the converter is installed near the engine to confirm the efficient purification of hydrocarbons. Most of the emissions are formed during cold start, and during the low temperature operation of the catalyst, as mentioned earlier. Since the pre-converter is installed near the engine, the temperature inside the converter is higher than in the main converter due to higher temperature of exhaust gas. Thus, catalytic materials have to work even at temperatures higher than 1000°C (Koltsakis & Stamatelos 1997, Becker & Watson 1998). Thermal degradation of a three-way catalyst begins in the temperature area of 800°–900°C, or even at lower temperatures, depending on the materials used. It is a physical process leading to a catalyst deactivation at high temperatures because of the loss of catalytic surface area due to crystal growth of the catalytic phase, the loss of washcoat area due to a collapse of pore structure, and/or chemical transformations of catalytic phases to non-catalytic phases. The first two processes are typically referred to as sintering, and the third as the solid-solid phase transition at high temperatures. (Butt & Petersen 1988, Somorjai 1994, Bartholomew 2001, Moulijn et al. 2001)

In Fig. 7, the deactivation mechanisms of a three-way catalyst are illustrated. These mechanisms are examined in detail in the following paragraphs and references to Fig. 7 are given concurrently. Sintering, as illustrated in Figures 7C and 7D, is the loss of catalyst’s active surface due to crystal growth of either the bulk material or the active phase. In the case of supported metal catalysts, reduction of the active surface area is provoked via agglomeration and coalescence of small metal crystallites into larger ones (Gunter et al. 1997, Stakheev & Kustov 1999). Two different models have been proposed for sintering i.e., the atomic migration and the crystallite migration models. As such, sintering occurs either due to metal atoms migrating from one crystallite to another via the surface or gas phase by diminishing small crystallites in size and increasing the larger ones (atomic migration model). Or sintering can occur via migration of the crystallites along the surface, followed by collision and coalescence of two crystallites (crystallite migration model) (Forzatti & Lietti 1999, Bartholomew 2001). Figure 7C presents a schematic representation of atomic migration and crystallite migration models.

As mentioned earlier, sintering on supported metal catalysts involves complex physical and chemical phenomena that make the understanding of the mechanistic aspects of sintering difficult. Experimental observations have shown that sintering is strongly temperature-dependent (Bartholomew 2001, Mouljin et al. 2001), but is also affected by the surrounding gas atmosphere (Forzatti & Lietti 1999). The rate of sintering increases exponentially with temperature and, for example, the sintering of precious metals becomes significant above 600°C. The underlying mechanism of sintering of small metal particles is the surface diffusion, or at higher temperatures, the mobility of larger agglomerates. The so-called Hüttig and Tamman temperatures indicate the temperature at which sintering starts. The following semi-empirical relations for Hüttig and Tamman temperatures are more commonly used (Mouljin et al. 2001):

Equation 2.

Equation 3.

Temperature at which the solid phase becomes mobile depends on several factors such as texture, size and morphology. For instance, highly porous γ -alumina is much more sensitive to sintering than non-porous α-alumina. (Mouljin et al. 2001)

Sintering processes at high temperatures are also affected by atmosphere, as expressed earlier. Supported metal catalysts sinter relatively rapidly under an oxidizing atmosphere, however the process is more slow under reducing and inert atmospheres (Wanke & Flynn 1975). Sintering is also generally accelerated, e.g. in the presence of water vapor (Mowery et al. 1999, Bartholomew 2001). In addition to temperature, atmosphere and time, the sintering rate is also dependent on several other factors, such as precious metal loading and washcoat composition. The presence of specific additives is known to reduce the sintering of a catalyst: BaO, CeO₂, La₂O₃ and ZrO₂ improve the stability of γ -Al₂O₃ towards sintering in the presence of high H₂O content in the exhaust gas. (Heck & Farrauto 1997)

Solid-solid phase transitions, as presented in Fig. 7D, can be viewed as an extreme form of sintering occurring at very high temperatures and leading to the transformation of one crystalline phase into another. Phase transformations typically occur in the bulk washcoat, e.g. aluminium oxide has many phases from the porous γ -Al₂O₃ to non-porous α-Al₂O₃, which is the most stable phase of alumina. The phase transformations of Al₂O₃ (boehmite), as shown in Fig. 8, beco1me significant at high temperatures and remarkably decrease the surface area of the catalysts (Hayes & Kolaczkowski 1997, Forzatti & Lietti 1999). On supported metal catalysts, the incorporation of the metal into the washcoat can be observed, i.e. the reaction of Rh₂O₃ with alumina to form inactive Rh-aluminate at high-temperatures, especially in lean conditions. (Shelef & Graham 1994)

Active precious metals are well-known catalysts for the exhaust gas purification. The sintering behaviour of Pt, Pd and Rh under different ageing atmospheres is considered in Table 5. Among the active metals, rhodium is known to be the most sensitive metal towards sintering at high temperatures under the exhaust gas conditions. This leads to poor activity, especially in the reduction of NO_X (Taylor 1993). The use of a bimetallic catalyst, such as Pd-Rh or Pt-Rh, gives a better catalytic activity at high temperatures. The operation conditions (rich or lean) also affect the sintering of active metals; for example, ageing atmosphere and the oscillation between oxidizing and reducing atmospheres can accelerate deactivation. (Butt & Petersen 1988, Koltsakis & Stamatelos 1997)

Redispersion is an opposite process to sintering. During redispersion, many complex phenomena take place, the particle sizes decrease and surface areas increase. In particular, the interaction between oxygen and precious metals may lead to the formation of species that are mobile on the surface and reverse the process of agglomeration. Sintering is normally physical rather than chemical in nature and, therefore, the magnitudes of thermal activation are quite different. Typical activation energies for sintering may be twice or even three times lower than those associated with the chemical processes in poisoning or coke formation. Furthermore, ageing time is important because it correlates both with sintering and redispersion. (Flynn et al. 1975, Butt & Petersen 1988, Gunter et al. 1997)

The kinetics of a catalyst deactivation is a function of temperature, time, pressure and the concentrations of different substances. The change in catalytic activity can be an effect of one or several of the previously mentioned processes. For sintering, the kinetics can be derived from active metal surface area versus time measurements at constant temperature. A number of researchers, Wanke & Flynn (1975), Bartholomew (1984) and Fuentes & Gamas (1991), have attempted to correlate sintering kinetics of supported metal catalysts. The following simple correlation of sintering kinetics can be applied:

Equation 4.

where

D is the metal dispersion (or metal surface area),

D₀ is the initial metal dispersion (or initial metal surface area),

k is the kinetic rate constant for sintering, and

n the sintering order.

It has been found that the use of this equation (Eq. 4) leads to values of k varying with sintering time, and hence with dispersion. Recently, a more sophisticated expression for sintering kinetics has been proposed by Bartholomew (1997) and Fuentes & Salinas-Rodriguez (1997), which takes into account the asymptotic approach (by adding the term –D_eq/D₀) observed in the typical dispersion vs. time curves (as will be presented in Fig. 26):

Equation 5.

where

D is the metal dispersion (or metal surface area),

D₀ is the initial metal dispersion (or initial metal surface area),

D_eq is the final dispersion (when the asympototic approach is achieved),

k is the kinetic rate constant for sintering, and

n the sintering order.

Equation (5) can be applied in a quantitative comparison regarding the effect of temperature, time and atmosphere on the sintering rate of supported metal catalysts. (Bartholomew 2001)

3.2.2. Deactivation by poisoning

The activity of a three-way catalyst reduces gradually when the unwanted, harmful components of fuels and lubricants, or other impurities, are accumulated on the catalyst’s surface and slowly poison the catalyst (Koltsakis & Stamatelos 1997). Poisoning is defined as a loss of catalytic activity due to the chemisorption of impurities on the active sites of the catalyst. Usually, a distinction is made between poisons and inhibitors. Poisons are substances that interact very strongly and irreversibly with the active sites, whereas the adsorption of inhibitors on the catalyst surface is weak and reversible. In the latter case, the catalytic activity can be at least partly restored by regeneration. This irreversible/reversible or permanent/temporary nature of deactivation and the regeneration possibility of a catalyst are the main differences between poisoning and inhibition (Butt & Petersen 1988, Forzatti & Lietti 1999). However, the distinction between permanent and temporary poisoning is not always so clear, since strong poisons at low temperatures may be less harmful in high-temperature applications (Moulijn et al. 2001). Catalyst poisons can also be classified as selective or non-selective. The description of a poison as selective or non-selective is related to the nature of the surface and the degree of interaction of the poison with the surface. A poison can also be selective in one reaction, but not in another. (Butt & Petersen 1988)

Poisoning of a three-way catalyst as a result of the accumulation of impurities on the active sites (see Fig. 7B) is typically a slow and irreversible phenomenon. The accumulation of poisons on the active sites blocks the access of reactants to these active sites (Butt & Petersen 1988). As a result of poisoning, the catalytic activity may be decreased without affecting the selectivity, but often selectivity is also changed since some of the active sites are deactivated while others are practically unaffected. In some cases, depending on the adsorbed poison, the poisoned catalyst can be regenerated and its activity can be at least partly restored (Angelidis & Sklavounos 1995, Forzatti & Lietti 1999, Rokosz et al. 2001). However, the poisoned three-way catalyst can hardly be regenerated and, therefore, the best method to reduce poisoning is to decrease the amount of poisons in the fuel and lubrication oils to more acceptable levels.

Catalytic converters are poisoned by the impurities in fuel and lubrication oil, or by shavings from the exhaust tailpipe. Even the low levels of impurities are enough to cover the active sites and decrease the performance of a catalytic converter. It follows that the analysis of poisoned catalysts may be complicated since the content of poison of a fully deactivated catalyst can be as low as 0.1 wt-% or even less (Forzatti & Lietti 1999). Lead (Pb), sulfur (S), phosphorus (P), zinc (Zn), calcium (Ca), and magnesium (Mg) compounds are typical catalyst poisons (Liu & Park 1993, Culley et al. 1996). Earlier, the effects of lead (Pb) had been studied extremely carefully (Williamson et al. 1979a, Williamson et al. 1979b, Monroe 1980). The catalytic converters were known to loose already their effectiveness after 10 refills with leaded gasoline (Heck & Farrauto 1996). Nowadays, mostly due to the use of unleaded or low Pb concentrations in gasoline, the role of lead as a catalyst poison is far less significant than in the past.

Fuel (gasoline) contains sulfur in small amounts. New requirements for a low sulfur content (<50 ppm) in the gasoline fuel are introduced together with Euro IV (Directive 98/70/EC). Sulfur clearly affects, and often very quickly, the efficiency and oxygen storage capacity (OSC) of the catalyst. Sulfur poisoning can lead to the formation of new inactive compounds on the catalyst’s surface and also to the morphological changes in the catalyst. Fast poisoning by sulfur can be to some extent reversible and the poisoned catalyst can be regenerated (Yu & Shaw 1998). Beck & Sommers (1995) have shown that the impact of sulfur on vehicle-aged catalysts was irreversible at temperatures below 650°C, but the original activity could be restored at higher temperatures. However, it should be noted that although the purification efficiency is recovered, the oxygen storage capacity is not (Beck et al. 1997b). During the combustion processes in the engine, fuel sulfur oxidizes to SO₂ and SO₃. These compounds adsorb on the precious metal sites on the catalyst’s surface at low temperatures (below 300°C) and react with alumina to form aluminium sulfates that reduce the active surface of washcoat and deactivate the catalyst. The air-to-fuel ratio also affects the behaviour of sulfur. In lean conditions, SO₂ is stored on cerium, and in rich conditions both SO₂ and SO₃ are reduced to form hydrogen sulfide (H₂S). Three-way catalysts are known to loose their activity more in oxidizing (lean) than in reducing (rich) conditions in the presence of sulfur compounds. The problem of sulfur poisoning also appears to be more significant at low temperatures, whereas at elevated temperatures (T>1000°C), the adsorption of sulfur species is almost absent. (Butt & Petersen 1988, Heck & Farrauto 1996)

Phosphorus (P), zinc (Zn), calcium (Ca), and magnesium (Mg) compounds are typical impurities in the lubrication oils. These substances and/or their compounds accumulate on the catalyst’s surface and they can be regarded as notable as fuel poisons. The considerable amounts of phosphorus, zinc, calcium, and/or magnesium are normally observed on the surface of an aged catalyst after years of driving. Several studies of the deactivation of a three-way catalyst by phosphorus, calcium and zinc compounds have been published. (Williamson et al. 1984, Williamson et al. 1985, Inoue et al. 1992, Liu & Park 1993, Culley et al. 1996)

Zinc dialkyldithiophosphate (ZDP), a typical oil additive, is a common source of phosphorus and zinc. Several studies have shown that the individual effects of P and Zn on deactivation are small compared to that of the combined effect of P and Zn. At low exhaust temperatures in particular the formation of zinc pyrophosphate (Zn₂P₂O₇) decreases the catalytic activity (Williamson et al. 1984, Williamson et al. 1985). The phosphorus contamination is observed either as an overlayer of Zn, Ca and Mg phosphates (M₃(PO₄)₂, M= Zn, Ca or Mg), or as aluminium phosphate (AlPO₄) within the washcoat (Liu & Park 1993, Ueda et al. 1994). Recently, cerium has also been observed to form cerium phosphates, CePO₄ and/or Ce(PO₃)₃ (Rokosz et al. 2001). Phosphates form a film layer on the catalyst surface that covers the precious metals in the porous washcoat and prevents contact between the catalyst and the surrounding gas atmosphere. (Brett et al. 1989, Heck & Farrauto 1996)

The poisoning of a catalyst is clearly dependent on the phosphorus level in the lubrication oil (Brett et al. 1989, Culley et al. 1996). The use of calcium or magnesium containing oil additives can decrease the harmful effects of phosphorus (Culley et al. 1996). Calcium and magnesium sulfonates form Ca and Mg phosphates and thus prevent the accumulation of phosphorus on the catalyst’s surface (Ueda et al. 1994). Similar observations have also been made in the case of zinc compounds (Monroe 1980). The largest contaminant levels are typically observed in the front edge of the catalyst (Culley et al. 1996, Beck et al. 1997a, Beck et al. 1997b). Experimental observations have also shown that even small amounts of these compounds are high enough to decrease the performance of a catalytic converter. (Joy et al. 1985)

Active metal catalysts are preferred in the controlling of the exhaust gas emissions, because they are less liable to sulfur poisoning than metal oxide catalysts (Shelef et al. 1978), as reported above. Precious metals have different types of resistance against poisoning. Palladium is more sensitive than platinum and rhodium to chemical deactivation, in particular to poisoning by sulfur and lead (Taylor 1993, Lox & Engler 1997). Currently, the use of Pd catalysts is possible because of the rapid decrease in fuel lead content, as discussed previously. The additives used and the chemical composition of the washcoat has an effect on the sulfur behaviour in the catalyst. In particular additives, which play a significant role in Pd-only catalysts. (Heck & Farrauto 1996)

Driving conditions also affect the catalyst’s chemical deactivation. Especially in Nordic countries, where the cold weather and urban driving keep the catalyst’s temperature low during a long time period. This accelerates the catalyst’s chemical ageing, because the unburned soot and particles adsorb on the active material. The stability against thermal and chemical deactivation can be improved by a proper choice of the catalyst material. In addition, the placement of the active material in separate washcoat layers improves the durability. (Laurikko 1994, Laurikko 1995)

3.2.3. Other relevant mechanisms of deactivation

There are other essential forms of the deactivation of three-way catalysts. For example, pore blockage, encapsulation of metal particles, volatilization of active compounds, fouling or coke formation and metal-metal or metal-washcoat interactions, which will be briefly discussed below.

According to Graham et al. (1999), high temperature ageing may result in deep encapsulation of sintered precious metal particles (see Fig. 7D) as the surface area of the washcoat decreases. This is a serious type of deactivation because of its permanent nature. The encapsulated metal particles cannot participate in catalysis since they are inaccessible to gas phase molecules. Furthermore, support can interact with the metal catalyst also by the support-induced changes observed in the metal particle morphology, by the formation of specific active sites on the metal-support interface and by the charging of metal particles. (Hu et al. 1998)

Fouling covers all phenomena where the surface is covered with a deposit, e.g. with combustion residues such as soot or with mechanical wear. Coke formation is the most widely known form of fouling (it is even used as a synonym for fouling). Coke formation is not very clearly defined. There are probably as many mechanisms of coke formation as there are reactions and catalysts where this phenomenon is encountered. During the coke formation, carbonaceous residues cover the active surface sites (see Fig. 7A), and decrease the active surface area. First, this blocks out the active compounds to reach the surface sites, and second, the amount of coke might be so large that carbon deposits block the internal pores in the catalyst. In many cases, hydrocarbons and aromatic materials are primarily responsible for coke formation. Among these other deactivation mechanisms, pore blocking is probably one of the most important mechanisms. Pore blocking is often connected to coke formation, and when the amount of coke is high on the catalyst’s surface, it may be possible for the coke itself to block off the pore structure. (Butt & Petersen 1988, Mouljin et al. 2001)

At high temperatures, catalysts may suffer from the loss of active phase through volatilization. Metal loss through direct volatilization is generally an insignificant route of the catalyst deactivation. By contrast, metal loss through the formation of volatile compounds is important over a wide range of conditions (Bartholomew 2001). Large amounts of catalytic materials can be transported to either substrate where they can react, or into the gas phase where they are lost in the effluent gas stream. High volatility limits the selection of otherwise useful catalytic materials, e.g. the oxides of Pt, Pd and Rh formed during the reaction cycles are not as volatile as the other noble metal oxides, such as RuO₂, OsO₄ and Ir₂O₃. (Cotton & Wilkinson 1988)

The thermodynamics of volatilization and thermodynamic equilibrium calculations are useful in the evaluation of the volatility of metals and metal oxides in order to assess which materials are stable over long periods at high temperatures (Forzatti & Lietti 1999). Thermodynamic equilibrium calculations of the oxidation/reduction behaviour of palladium have shown that phase stability in a Pd/PdO system changes as a function of temperature and oxygen partial pressure. The lower the pressure and the higher the temperature are, the more likely is the Pd phase (Ribeiro et al. 1994). In Table 6, vapour pressures of Pt, Pd and Rh as metals and metal oxides are given at a temperature of 800°C in air (Shinjoh et al. 1991). The vapour pressure increases with temperature, and it is also strongly dependent on the composition of the surrounding atmosphere, i.e. Pd is volatile at temperatures around 850°C and above, depending on the surrounding environment. (Bartholomew 2001)

As can be seen in Table 6, the orders of vapour pressures of active metals and their oxides are as follows (Shinjoh et al. 1991):

Metals: Pd > > Pt > Rh

Oxides: Pt > Rh >> Pd

Hence the vapour pressure of metallic Pd is clearly several magnitudes higher than the vapour pressures of Pt and Rh, while as oxides, the situation is the reverse.

As an example, in Figures 9A and 9B, thermodynamics equilibrium curves of Pd and Rh are presented respectively. According to the thermodynamics, Pd is easily oxidized at room temperature to PdO and it reduces to metallic Pd in the temperature range of 500°–1200°C. The formation and decomposition of PdO occurs as follows:

Equation 6.

The most stable oxidation state of Pd is +2 and the formation of PdO is kinetically restricted at low temperatures. As shown in Fig. 9A, the metallic Pd is totally volatilised in a 5% O₂/N₂ atmosphere at a temperature of 1400°C, and the increased amount of oxygen in the gas phase moves the reduction curve of Pd to higher temperatures. According to Farrauto et al. (1992), two kinds of palladium oxides, PdO_X-Pd and PdO, and metallic Pd have been observed on the surface supported on γ -alumina. PdO supported on pure alumina is known to decompose in two steps to metallic palladium in air at a temperature above 800°C. Instead, the re-oxidation of metallic Pd to PdO and PdOX species during the cooling process is very slow at temperatures 550°–650°C, a temperature range at which PdO is the thermodynamically favoured phase. Temperatures above 800°C convert all PdO_X and PdO to metallic Pd and subsequent cooling again leads to redispersed PdO/Al₂O₃ and PdO_X-Pd/Al₂O₃ phases. Therefore, there is a window of a few hundred degrees in which the catalyst could be in the form of Pd metal or PdO. This hysteresis-like behaviour is strongly dependent on the surroundings of Pd/PdO phases, especially the chemistry of washcoat material and stabilisers. (Farrauto et al. 1992, Farrauto et al. 1995, Datye et al. 2000)

Studies of the stabilization of rhodium oxide phases supported on γ -alumina have shown several thermodynamically stable bulk rhodium oxide phases within the ageing temperatures of 500°–1050°C. According to Weng-Sieh et al. (1998), ageing in air below 650°C results in the formation of highly dispersed rhodium oxide – RhO₂, and above 650°C large particles of Rh₂O₃ are observed together with smaller particles of RhO₂. The observed low thermal stability and catalytic activity of rhodium under oxidizing conditions has been attributed to the interaction of Rh with the alumina support and the diffusion of rhodium into the bulk of alumina at high temperatures (Yao et al. 1977, Hu et al. 1998). However, the nature of rhodium oxides formed during the ageings in air is still unclear and it does not necessarily coincide with that expected on the basis of bulk-phase thermodynamics (Weng-Sieh et al. 1998). In Fig. 9B, the thermodynamical equilibrium curves of Rh are presented in a 5 % O₂/N₂ atmosphere. At room temperature, the most stable oxidation state of Rh is +3, and Rh has many oxidation states, as can be seen in Fig. 9B. The oxidation/reduction of Rh occurs as follows:

Equation 7.

At normal operation temperatures of the catalytic converter, Rh is in the form of Rh₂O₃, if the oxidation of Rh is kinetically favoured. The oxygen content in the exhaust gas strongly affects the formation of Rh oxides; the higher the amount of oxygen, the higher the transition temperature is.