REVIEW Adsorption, chemisorption, and catalysis - Springer Link

Chemical Papers 68 (12) 1625–1638 (2014) DOI: 10.2478/s11696-014-0624-9

REVIEW

Adsorption, chemisorption, and catalysis a,b

a VUCHT b Faculty

Milan Králik*

a.s., Areál Duslo a.s., 927 03 Šaľa, Slovakia

of Chemical and Food Technology, Slovak University of Technology in Bratislava, Radlinského 9, 812 37 Bratislava, Slovakia Received 16 January 2014; Revised 14 June 2014; Accepted 25 June 2014

Dedicated to the memory of professor Elemír Kossaczký A short history of the relationships among adsorption, chemisorption, and catalysis with solid catalysts is reviewed. A special focus is on the development of quality and descriptions accuracy using computers, both for the modeling of elementary physical phenomena and adsorption, as well as for the solution of more complex problems like quantum chemical approach to chemisorption, kinetics over solid catalysts, and reactor systems. Modern approaches to the characterization of solid catalysts from the adsorption–desorption data based mainly on n-layer adsorption and non-linear three parameter BET isotherm regarding the volume of micropores as one of the parameters are demonstrated. Instrumentation techniques like infrared spectroscopy or NMR techniques for the analysis of the strength of component chemisorption are mentioned. As for the kinetics, a vague capability of the Langmuir–Hinshelwood–Hougen–Watson models to describe a reaction system in more complicated cases, e.g. bimolecular surface reactions, is discussed. In this context, the simplest model with a minimum number of parameters is advised. To estimate the most realistic values, intrinsic reaction kinetic and mass transport phenomena are taken into account. Usefulness of quantum mechanistic models for better understanding of the catalytic phenomena and more efficient design of catalysts are outlined. c 2014 Institute of Chemistry, Slovak Academy of Sciences Keywords: adsorption, chemisorption, catalysis, infrared spectroscopy, quantum chemical models, reaction kinetics, mass transport

Introduction The phenomenon of catalysis is very old. Already Aristotle (384–322 BC) wrote about active substances (catalysts) and passive substances (reactants) undergoing transformation (Schwab, 1981). Observations in the 17th and 18th century provided Berzelius (1835) with a basis to define: “A catalyst is any substance (including light) that directly alters the rate of a chemical reaction without entering into the net chemical reaction itself”. It is worthy to note that by this definition, a catalyst might increase or decrease the reaction rate. However, in today’s definition, only substances increasing the reaction rate are denoted as catalysts.

A modern and still valid definition was introduced by Ostwald (1896) who defined a catalyst as a substance which accelerates a chemical reaction and it is not consumed in the course of this reaction. Very important were his thermodynamic considerations resulting in the conclusion that a catalyst does not change the chemical equilibrium of the reaction. Experimental results and theoretical investigations allow describing a catalytic process with a solid catalyst as a sequence of actions: – chemisorption of at least one of reaction components; – surface reaction of a chemisorbed component either with other chemisorbed component or with a

*Corresponding author, e-mail: [email protected], [email protected]

1626

M. Králik/Chemical Papers 68 (12) 1625–1638 (2014)

component from the fluid phase surrounding the solid catalyst; – desorption of products. Adsorption and chemisorption of reaction species are unavoidable in a heterogeneously catalyzed process and, in addition, all steps are accompanied by the diffusion of reactants and products as well (Wisniak, 2010; Satterfield, 1980). The presented paper summarizes data on adsorption, chemisorption, and catalysis obtained in laboratories at the Slovak University of Technology, as well as relevant important data from other authors. Due to the enormous number of papers dealing with these topics, this paper cannot be considered as a full review, but an attempt to suggest suitable procedures for the study of heterogeneous catalysts and of the adsorption/chemisorption effects on their performance.

Adsorption According to definition (Ruthven, 1984), adsorption is a process that occurs when a gas or liquid solute (adsorptive) accumulates on the surface of a solid or a liquid (adsorbent), forming a molecular, ionic, or atomic film. Species (molecules, atoms, ions) of the adsorptive adhered to the adsorbent are called adsorbate. The formed film of adsorbate is usually a multilayer for the gas phase adsorption and a monolayer for the liquid phase adsorption. Physical adsorption is enabled by the van der Waals forces between molecules and/or atoms on the adsorbent surface and adsorbate, whereas adsorbate resembles a condensed liquid form of the adsorptive. Adsorption from the gas phase very often involves condensation of the adsorptive in pores of the adsorbent. For quantitative evaluation of the adsorbent capability, adsorption isotherms are applied. They are expressed as a function obtained at a constant temperature expressing an equilibrium amount of the adsorptive adsorbed at a certain partial pressure or concentration (generally activity) of the adsorptive in the gas or liquid (generally bulk) phase. Adsorption isotherms start at zero activity up to a maximum; for a gas phase it is a pressure related to the saturated one at a temperature at which the adsorption isotherm is determined. To obtain reliable experimental adsorption–desorption data, maximum outgassing of the adsorbent and sufficient time for reaching an equilibrium between the adsorptive and the adsorbate are unavoidable conditions (Lecloux, 1981; Rouquerol et al., 1994; Sing et al., 1985; Ruthven, 1984; Thomas & Thomas, 1997). Six types of isotherms reflecting the type of a porous material are considered generally (Fig. 1, IUPAC classification, Sing et al. (1985)). Texts in individual charts of Fig. 1 denote the main feature of the adsorption process. Terms: micropores (diameter < 2 nm), mesopores (diameter = 2–50 nm), and macrop-

Fig. 1. Typical adsorption isotherms (IUPAC classification (Sing et al., 1985)). Abscissas and ordinates express relative pressure (x = p/ps ) and adsorbed amount (vA /(cm3 g−1 )), respectively.

ores (diameter > 50 nm) are distinguished (Sing et al., 1985; Hudec, 2012). When the adsorbent has a high volume of micropores (type I) even at low activity of the adsorptive, a large amount of it is accumulated in the adsorbent. On the contrary, materials without micropores accumulate material proportionally to the activity of the adsorptive. Usually, the strength of the adsorption of the second and next layers of the adsorbate differs from that of the first layer. If the adhesion of the first layer is weak, then adsorbate has to cover the surface first at a relatively large activity of the adsorptive and then, the formation of the second and next layers is performed only at a little increase of the adsorptive activity (type III). Type IV represents adsorbents with mesopores in which capillary condensation occurs (Lecloux, 1981). Type V is similar to type IV but for the week adsorption of the adsorptive. Type VI comprises multilayered adsorption, and/or adsorbents with different dimensions of mesopores. Adsorption processes accompanied with large condensation of the adsorptive are typical with hysteresis presented by desorption isotherms. Due to capillary condensation of the adsorptive, desorption proceeds at lower activity than adsorption (Lecloux, 1981; Sing et al., 1985; Hudec, 2012; Thomas & Thomas, 1997). Description of adsorption isotherms Attempts to describe adsorption isotherms have


led to various models. The Langmuir adsorption model is based on the assumptions that: – all sites on the adsorbent surface possess equal affinity for the adsorbate; – interactions among adsorbate molecules are negligible; – the adsorption proceeds with formation of only one layer of the adsorbate on the adsorbent surface (the so called: monolayer adsorption) : va,La = VLa

KLa x 1 + KLa x

(1)

where va,La /(cm3 g−1 ) is the adsorbed amount calculated according to the Langmuir adsorption isotherm, VLa /(cm3 g−1 ) the monolayer (maximum) adsorption capacity, KLa the equilibrium constant in the Langmuir adsorption isotherm, and x the relative pressure (x = p/ps , where p and ps are adsorptive absolute and saturated pressure at given temperature, respectively). For a low activity of the adsorptive (x → 0), the Langmuir isotherm is reduced to the form of the Henry isotherm: va,He = VHe x (2) where va,He /(cm3 g−1 ) is the adsorbed amount calculated according to the Henry isotherm and VHe / (cm3 g−1 ) the maximum adsorption capacity. As it can be seen from the comparison of Eqs. (1) and (2), coefficient VHe is constant only in a low range of adsorptive activity starting from zero. If exponential distribution of the adsorbent active sites with respect to the adsorption heat is assumed, it is possible to derive an empirical adsorption isotherm (Freundlich, 1932): va,Fr = VFr xkFr

(3)

where va,Fr /(cm3 g−1 ) is the adsorbed amount calculated according to the Freundlich adsorption isotherm, VFr /(cm3 g−1 ) the maximum adsorption capacity, and kFr a coefficient. Maximum adsorption capacity VFr is usually positive and generally not an integer. Despite the fact that Freundlich isotherm, Eq. (3), lacks the required linear behavior in the Henry’s law region, data on heterogeneous adsorbents over a wide range of the adsorptive concentrations can be correlated successfully (LeVan et al., 1997). The Temkin isotherm, firstly introduced as a footnote in the paper of Frumkin and Shlygin (1935), considers a linear decrease in the binding energies adsorbate–adsorbent, which leads to the equation: va,Tem =

RT ln(ATem x) bTem

(4)

where va,Tem /(cm3 g−1 ) is the adsorbed amount calculated according to the Temkin isotherm, R

1627

(= 8.314 J K−1 mol−1 ) represents the universal gas constant, T/K the absolute temperature, bTem a coefficient, and ATem the Temkin binding coefficient. Langmuir, Freundlich, and Temkin isotherms give a very simplified idea of the adsorption mechanism. More realistic is the Dubinin–Radushkevich (DR) isotherm model that is based on the assumptions (Lecloux, 1981) that: – the Polanyi potential expresses the surface working function (Somorjai, 1994); – the adsorption proceeds only in micropores (type I adsorption model, Fig. 1). va,DR = VDR exp −D [ln(x)]2 for x > 0 (5) where va,DR /(cm3 g−1 ) is the adsorbed amount calculated according to the Dubinin–Radushkevich isotherm, VDR /(cm3 g−1 ) the maximum adsorption capacity, and D a coefficient. The most frequently used models for the descriptions of type II and IV isotherms are BET isotherms, whereas the assumptions are: – adsorption sites form a regular array on the adsorbent surface and the adsorption enthalpy (∆Hads ) for the first monolayer is constant; – no interaction between the neighboring adsorbed molecules (adsorbate); – adsorption of the second and next layers is not limited; – formation of the second and next layers proceeds with the same enthalpy as that of liquefaction, (∆Hliq ) Vm,N x CN (1 − x) 1 + (CN − 1)x E1 − EL CN = exp RT

va,BET =

(6)

(7)

where va,BET /(cm3 g−1 ) is the adsorbed amount calculated according to the BET isotherm, Vm,N / (cm3 g−1 ) is the monolayer adsorption capacity, E1 /(J mol−1 ) the activation energy for the adsorption in the first layer, and EL /(J mol−1 ) is the activation energy for the adsorption in the second and next layers. Activation energies E1 and EL are considered to be proportional to the enthalpy of condensation (∆Hads ) and liquefaction (∆Hliq ). In case of micropores (type I isotherms), there are generally three possibilities of employing the BET model: – Concept of the volume of micropores, M-BET, being filled firstly and then multilayer BET adsorption continues (Schneider, 1995; Hudec, 2012) va,M-BET = VM-BET + va,BET

(8)

where va,M-BET /(cm3 g−1 ) is the adsorbed amount calculated according to the M-BET isotherm and

1628


Vm,BET /(cm3 g−1 ) the volume of micropores filled before the start of the BET type adsorption. – Utilization of standard isotherms (Lecloux, 1981; Hudec, 2012). The most commonly applied is the equation of Harkins and Jurra (1944): tHJ = 0.1

13.99 log(x) + 0.034

0.5

x ∈ (0.05, 0.5) (9) t < 1 nm

where tHJ /nm is the average (statistical) thickness of the adsorbate layer on the surface. Lecloux (1981) introduced a more general isotherm reflecting the difference between the adsorption strength (enthalpy) of the first and the next layers expressed by constant CN in Eq. (7). However, this type of generalized isotherm is not used frequently in practice. – Linear combination of DR and BET isotherms: va,DR−BET = va,DR + va,BET

(10)

Both Eqs. (8) and (10) have physical meaning but the volumes of micropores differ. Horniakova et al. (2001) and Kralik et al. (2008) showed that the volume of micropores assessed by Eq. (8) is lower than that obtained from Eq. (10). Due to the simpler concept of the physical meaning, the Schneider modification (Eq. (8)) is more advised (Schneider, 1995). Surprisingly, for the description of microporous adsorbents, the n-layer BET isotherm showed good interpretation of experimental data (Horniakova et al., 2001; Kralik et al., 2008). Vm,nL CnL x 1 − (n + 1)xn + nx(n+1) (1 − x) 1 − x + CnL x(1 − xn ) (11) where va,nL-BET /(cm3 g−1 ) is the adsorbed amount calculated according to the n-layer BET isotherm, CnL the constant for the n-layer adsorption, and n the maximum number of adsorbate layers (statistically, it can have a non-integer positive value). Besides the BET isotherm with micropores and the n-layer BET isotherm, all others mentioned above are usually applied in their linearized form. Estimation of relevant parameters is commonly obtained by treating data with a standard SW built in apparatus for adsorption–desorption data measurement. It is worthy to stress that a non-linear treatment of data provides more reliable information. Fig. 2 shows a comparison of experimental data for a microporous material (mordenite type zeolite with the Si/Al modulus equal to 78) and calculated values obtained using different isotherms. As it can be seen, just the n-layer BET isotherm was proved to be the best one for such a description. Another frequently appearing mistake is the interpretation of results using the linearized form of the va,nL-BET =

Fig. 2. Experimental adsorption isotherm (1) and various model isotherms; parameters estimated using nonlinearized forms and non-linear regression: BET (2), modified BET isotherm, considering micropores (3), combination of BET and Dubinin–Radushkevich isotherms (4), – n-layer BET isotherm (n = 1.4) (5). Further details can be found in the paper by Horniakova et al. (2001).

BET isotherm to treat experimental data for microporous materials possessing a negative value of the C-BET constant. As implied from Eq. (7), this constant cannot be negative and such a situation indicates a strong non-suitability of the multilayer BET model (Hudec, 2012). Employment of the n-layer isotherm for microporous materials is also recommended by Rothenberg (2008). Effect of the condensation in meso- and macropores is used for the calculation of characteristic dimensions of these pores using the modified Kelvin equation (Lecloux, 1981; Hudec, 2012): rP =

bK + tliq,M ln (1/x)

(12)

where rP /nm is the radius of a pore and bK is a constant. For nitrogen, as the adsorptive, the thickness of the monolayer is tliq,M = tliq,N2 = 0.354 nm. Barrett et al. (1951) extended this method by introducing a different thickness of the monolayer adsorbate to Eq. (12). This is called the BJH method and belongs to the standard procedure of isotherm data treatment. Lecloux (1981) applied the Broeckhoff and de Boer’s approach (modified Kelvin equation) based on adsorption thermodynamics. In spite of the significantly wider characterization of porous adsorbents compared to the simple Eq. (12), this procedure is not frequently applied as it requires very precise experimental data, and it is not commonly included in the software of standard adsorption equipments.


Databases of measured isotherms and power of computers allow using methods of statistical mechanics. Adsorptive–adsorbent (G–S) and adsorbate– adsorbent (L–S) interaction phenomena at a given temperature and pressure are able to describe pore size distribution and pore geometry. Such models allow calculating density profiles and the adsorbed amount can be derived. Obtained results are “calibrated” against real isotherm data of non-porous materials (Jagiello & Thommes, 2004). There are many other modifications of adsorption models described in detail by LeVan et al. (1997) and Wisniak (2010). A good comparison of the applicability of individual isotherms including statistical evaluation has been performed by Foo and Hameed (2008). D˛abrowski (2001) in his broad review (640 references) discusses the history and suitability of individual isotherms, as well as their applicability for the solution of practical adsorption problems and relationships with heterogeneous catalysis. From the catalytic point of view it is necessary to distinguish between models used for the description of solid catalytic materials and models suitable for the description of a catalytic reaction. For the latter, exclusively the Langmuir approach is applied. Accessibility of adsorption/catalytic sites can be increased by the increase of macropores radius (rP > 25 nm), which is characterized more precisely by other methods than the adsorption/desorption isotherms. The most common is mercury porosimetry, which allows distinguishing pores with the radius of 1.6 nm at pressures up to 400 MPa (Hudec, 2012). Of course, mechanical strength of the material has to be sufficient to ensure no breakage under such pressures. Microscopic techniques are also useful in the characterization of macropores (Lecloux, 1981). A good assessment of methods for the characterization of microporous solid materials can be found in the book by Che and Vedrin (2012), chapter prepared by Llewellyn et al. (2012). Besides the adsorption methods, evaluation of micropores and mesopores using combination of results obtained by Density Functional Theory and Gas Chromatographic Mass Spectroscopy measurements (DFT/GCMS) with subsequent isotherm reconstruction have proved to be useful for research as well as technological purposes. Dynamic features of adsorption are also discussed, whereas Pulsed Field Gradient Nuclear Magnetic Resonance (PFG NMR), Quasi-Elastic Neutron Scattering, Zero Length Column (ZLC), and Optical Impedance Spectroscopy (OIS, observations of probe molecules by IR) are outlined. Gas (vapor) phase adsorption The gas (vapor) phase adsorption is usually a multilayer process. Depending on the adsorbent properties (mainly microporosity), the adsorbent–adsorbate

1629

interaction and process conditions (temperature, pressure, activity of the adsorptive), at least one of the isotherm models listed above is suitable for the description. Sorption of linear alkanes on molecular sieves was measured by Bobok et al. (1970a), whereas a precise apparatus set up was found to be crucial for obtaining proper data. Due to the low rate of calculation devices in the past, analogue models were used to simulate process dynamics (Ilavsky et al., 1970). Besides reliable experimental techniques, an optimal plan of experiments also contributes to gaining good adsorption data (Bobok et al., 1982). Kossaczký and Bobok (1974) described a numerical solution of diffusion and adsorption of pentane inside a particle of a molecular sieve Calsit 5A. Adiabatic equilibrium desorption of carbon dioxide from molecular sieves (fixed bed Calsit 5A) by a nitrogen stream can be considered as a continuation of this work (Kossaczký et al., 1979; Bobok et al. 1979). For the description of the adsorption equilibrium, these authors used the modified Henry’s relationship. Bobok et al. (1970b) showed non-suitability of the Langmuir isotherm, as well as that the Dubinin assumptions that the density of the adsorbed phase does not change with the temperature is only a rough approximation for the system comprising the pentane–molecular sieve Calsit A. They suggested a modification by transforming equilibrium data to another value of temperature, different than that at which the isotherms were measured. Bobok et al. (1975) applied the actual form of the Freundlich isotherm for the description of equilibrium data of the heptanes adsorption on molecular sieve Calsit 5A. Chovancová et al. (1986) studied the adsorption dynamics of ethyl acetate, butyl alcohol, toluene, and vapors of industrial solvents on activated carbon, whereas the breakthrough curves served as a basis for the design of industrial adsorbers. Kossaczký et al. (1986) described nonisothermal adsorption of carbon dioxide on the molecular sieve Calsit 5 by a system of partial differential equations for mass and heat balance. The partial differential equations were transformed into ordinary differential equations (ODE) applying reverse difference instead of derivation with respect to length. The fourth order Runge– Kutta method was applied for the solution of the obtained ODE. Estimation of heat and mass transport parameters by experimental data fitting allowed designing an adsorber system. Dynamics of adsorption and values of the Knudsen diffusion coefficients for heptane in a molecular sieve NaY particle were estimated by Bobok et al. (1989). A very simple linear model was applied for the description of the adsorption equilibrium, partial differential equations expressing flow and adsorption phenomena in a packed column and the transport in a particle

1630


were replaced by relationships emerging from momentum analysis. Besedová and Bobok (1995) and Besedová et al. (2004) analyzed competitive adsorption of cumene and acetone on activated carbon. Equilibrium of the adsorption was expressed by isothermal and nonisothermal Langmuir models considering one or two types of adsorption centers; the latter proved to be more accurate. Solution of ecological problems, as well as demonstration of adsorption of non-polar adsorptive on non-polar adsorbents were reported by Bobok and Besedová (2001), who investigated the diffusion of chlorinated hydrocarbons in particles of activated carbon. The same research group (Bobok et al., 2004) developed a corrected diffusion model involving pores filling for the adsorption of ethanol vapors on activated carbon. Demands on the calculation extent describing the adsorption dynamics can be significantly decreased using efficient approximations; e.g. Rajniak et al. (1982) and Rajniak (1985) analyzed one-component sorption in a single adsorbent particle considering lineardriving-force approximation. The one-point collocation method proved to be very successful for the calculation of adsorbate diffusion. Desorption from a porous material can be more realistically described by models based on the percolation theory than by the simple Kelvin equation. In derived models, adsorption–desorption hysteresis is ascribed to blocking of liquid adsorptive discharge from larger pores by precedential emptying of smaller pores surrounding the larger ones (Rajniak & Yang, 1993, 1996, Rajniak et al., 1999, Šoóš & Rajniak, 2001). Generally it is considered that adsorption is not connected with strong adsorbate–adsorbent interactions. However, another general statement can be added: that molecules (adsorptive) of a nature similar to that of the adsorbent are adsorbed preferentially in comparison with those of not similar nature. Thus non-polar substances, e.g. hydrocarbons, are better adsorbed on non-polar activated carbon, and polar substances, e.g. amines, are better adsorbed on silica gel and zeolites. Probably the most relevant industrial example is Pressure Swing Adsorption (PSA). Chemisorption of more polar CO2 is stronger on the surface of an acidic form of zeolite (including entire micropores) than on non-polar H2 . When pressure is decreased, H2 is released firstly but CO2 only when a further decrease in the overall pressure is achieved (Ruthven, 1984; Chang et al., 2004). Liquid phase adsorption As mentioned above, the liquid phase adsorption is a typical monolayer process (LeVan et al., 1997). Therefore, suitable models for its description are the Langmuir, Freundlich, and Henry models, but also the Temkin and BET models are preferred. Báleš

et al. (1983) studied adsorption equilibria of phenol, p-cresol, and p-nitroaniline on activated carbon in aqueous solutions using the Henry and Langmuir isotherms. Langmuir modification for a binary mixture gave a very good description of simultaneous adsorption of p-cresol and p-nitroaniline. Liquid phase adsorption is widely applied mainly for “cleaning”, removing of non-desired species from a mixture of components. A brief review of wastewater treatment processes is offered by Grassi et al. (2012), who described individual adsorbents (activated carbon, clays, minerals, low cost adsorbents) and their capability to treat wastewaters with emerging pollutants. Adsorption (ion-exchange) equilibrium using ion exchange materials is a specific problem (LeVan et al., 1997). A nice example of the isolation of noble metals from the spent ceramic automotive catalysts is given by Shen et al. (2010). Besides the description of the equilibrium (Rh(III) chloride complexes with the anionic exchange resin Diaion WA21J), these authors also presented a simplified kinetic model for the adsorption process. Dada et al. (2012) investigated the sorption of Zn2+ on phosphoric acid modified rice husk and compared the application of Langmuir, Freundlich, Temkin, and Dubinin–Radushkevich isotherms for the equilibrium description. In accordance with the assumption of one layer adsorption from the liquid phase, the Langmuir isotherm exhibited the best description of the studied process; the coefficient of determination (R2 ) reached the value of 0.99. Melo et al. (2013) studied adsorption equilibria of Cu2+ , Zn2+ , and Cd2+ on EDTA-functionalized silica spheres, which can be considered as a functionalized adsorbent. Similarly to the previous authors, the best results were obtained by fitting with the Langmuir isotherm model, and, again, it fully corresponded with the availability of adsorption centers for one layer.

Chemisorption Similarities and differences between adsorption and chemisorption Differences between adsorption and chemisorption are listed in Table 1. It is worthy to stress that despite similarities of a chemical compound and an adsorbate– adsorbent complex, in the majority of chemisorption processes, adsorbate–adsorbent bonds are significantly weaker than those in typical chemical compounds. As discussed bellow in the theory on catalysis, there is an optimal strength of chemisorption (“the volcano curve theory”) for a proper course of the catalytic process.

Catalysis Catalysis is a phenomenon in which a substance (catalyst) enhances the rate of a chemical reaction


1631

Table 1. Differences between physisorption and chemisorption Property

Physical adsorption

Chemisorption

Type of bonding forces

Van der Wals

Similar to a chemical bond

Adsorption heat

Low, 10–40 kJ mol−1

High, 20–400 kJ mol−1

Chemical change of adsorptive

None

Formation of a surface compound

Reversibility

Fully reversible, i.e. desorption of adsorbate occurs by decreasing the activity of the adsorptive in the fluid surrounding the surface

The process is irreversible; “desorbed compounds” are different from the adsorbed ones

Activation energy

Very low (close to zero)

High, similar to a chemical reaction

Effect of temperature

Negative

In some extent of temperatures positive; so called activated adsorption

Specificity of adsorbate–adsorbent interactions

Very low

High

Formation of multilayers

Yes, in gas phase adsorption usually accompanied by liquefaction in microand mesopores

No

whereas the substance itself is not consumed during the reaction (Ostwald, 1896). Historically (Berkman et al., 1940; Rothenberg, 2008), three groups of catalytic systems are distinguished: – homogeneous, i.e. reaction components and catalytic moieties are in the same phase, for example hydrolysis of esters by mineral acids; – heterogeneous, i.e. reaction components and catalytic moieties are in different phases; the most common are reactants in gas or liquid and catalysts in solid phase, e.g. synthesis of ammonia from nitrogen and hydrogen over iron catalysts; – biocatalysis or enzymatic catalysis. To these, the next two groups should be added: – heterogenized catalysis, i.e. a catalytically active moiety is anchored to the surface or molecule, e.g. ionized liquid of the second phase; a typical example is the acid based catalysis performed by anionically modified polymers (Rothenberg, 2008); – electrocatalysis, i.e. employing some mechanisms which contribute to the rate of half-cell reactions on electrode surfaces (Koper, 2009). In accordance with the adsorption and chemisorption phenomena, the following parts of this paper are devoted mainly to heterogeneous catalysis. Chemisorption and catalysis Heterogeneous catalysis with solid catalysts is connected with adsorption and mainly chemisorption directly or indirectly involved in all classical concepts of catalysis (Schwab, 1981):

Fig. 3. Dependence of the reaction rate on the strength of adsorption, the so called volcano plot introduced by Sabatier. Dissociative adsorption as a rate determining step increases with the concentration of reactants, the rate equation of positive order. Desorption as the rate limiting step reveals the negative order of the rate equation (van Santen, 2010).

– active centers, i.e. points on a catalyst surface at which at least one of the reactants is chemisorbed; – intermediate compounds formed as a result of the interaction of reactants with active centers; – complexes and clusters as a form of intermediates formed on catalytic centers, usually polynuclear, more than one atom of a catalytic center is involved in the interaction with the substrate; this concept is also the basis for the quantum mechanical treatment of chemisorption; – multiplet hypothesis and geometric factor, i.e. arrangement of active centers has to lead to proper chemisorption of reactants, e.g. benzene parallel chemisorbed on (111) faces of transition metals; – volcano curve (Fig. 3), i.e. the strength of

1632


Table 2. Activation of methane on different metals and surfaces (van Santen, 2010) Metal and surface configuration

Ea /(kJ mol−1 )

d-character/% (Bond, 1957)

Ru (0001) Ru (1120) Rh (111) Rh step Rh kink Pd (111) Pd step Pd kink

76 56 67 32 20 66 38 41

50 50 50 50 50 46 46 46

chemisorption has to be sufficient to modify electronic structure of the substrate but not too strong to be close to the compound formed by the catalytic center; this is characterized by the dependence of the reaction rate on the chemisorption heat; – band theory, introduced by Schwab in 1940; either an electron from the reactant enters the catalytic center (usually a defect on a metal crystal) or electrons can be donated by a reactant to the catalytic centers; this theory was in 1951 extended also for semiconductors. However, the acceptance of chemisorption as a key factor in heterogeneous catalysis was not simple (Robertson, 1983). The Faraday’s paper introducing the adsorption theory to the Royal Society in 1834 was discussed for a long time and a final proof of its validity was demonstrated by Langmuir in 1906. Afterwards, chemisorption as a key factor in heterogeneous catalysis was generally accepted. Similarly to the classification of reactions to i) acid–base, ii) redox, and iii) coordination, the catalysts can be distinguished in the same way. A typical example of solid acid catalysts are zeolites, which are used either with a neutral component, support, as a pure acid (e.g. for cracking in petrochemistry), or in combination with other catalytic substances like transition metals as multifunctional catalysts in hydrocracking (Satterfield, 1980). Catalysis by transition metals has its peculiar features which have attracted attention of researchers from the very beginning (Berkman et al., 1947). Transition metals are used as catalytic substances for oxidation, reduction, and combined, e.g. reductive alkylation, processes. A very important feature of transition metals as catalysts is the occupation of electron orbitals, especially d-orbitals. Pauling (1949) introduced the so called d-character parameter, which is directly related to the chemisorption and catalytic activity. Electrons from d-orbitals are easily movable (transition metals are good conductors) and they can interact with the adsorbate. Similarly, electrons from the adsorbate can interact (enter) with the d-orbitals of transition metals (Somorjai, 1994). As reported in Table 2, strength and extent of chemisorption on the metal surface depends on the type of the metal and on the defects on the metal surface. Metal crystal-

Fig. 4. Reaction coordinates for non-catalyzed (the red curve) and catalyzed (the purple curve) reactions. Red and green curves describe situation without consideration of reaction intermediates. Transition states are indicated by the sign ‡, * indicates a catalytic site.

lites imperfections as kinks and steps are the most reactive catalytic sites. However, from the deactivation point of view, these sites are mostly attacked by reaction components. Smaller particles (with higher surface concentration of catalytic sites) are less stable than the bigger ones. Therefore, for technological purposes, a compromise between the catalytic activity and stability has to be considered. Kinetics and mass transport phenomena A modern concept of the reaction kinetics description was introduced by Eyring (1935) and by Evans and Polanyi (1935). This concept is based on the assumption that reactants form an activated complex which is subsequently transformed to products (Fig. 4). The theory of activated complexes can be extended to homogeneous as well as to heterogeneous catalytic processes. Let us consider a bimolecular reaction: k+

− → A+B − ← −− −− − − R+S

(13)

k−

If both reactants are chemisorbed on the catalysts surface (species A* and B*), then a reaction between

1633


these chemisorbed species proceeds. For simplicity, chemisorption of reactants (species A* and B*) is supposed to proceed simultaneously and after the chemical products (R* and S*) are released by desorption from the catalytic sites simultaneously. Energies of the chemisorbed species are lower than those of the nonchemisorbed ones. Chemisorption of species A* and B* proceeds via an activated chemisorbed complex (BA*)‡ which is subsequently transformed to the activated chemisorbed complex of products (RS*)‡ . This chemisorbed complex is split to chemisorbed species R* and S* and finally, products are desorbed from the surface. Two types of kinetic models are considered for heterogeneous catalytic processes (Berkman et al., 1940; Hougen & Watson, 1947; Thomas & Thomas, 1997; Satterfield, 1980): i) empirical and ii) mechanistic. The majority of empirical models is based on a power law model. For a reaction of nR reactants, the forward reaction can be written as: ξ˙V = kcat acat

nR

abi i

(14)

i=1

where ξ˙V /(mol m−3 s−1 ) is the reaction rate, kcat the rate constant (units depend on the number of reactants and stoichiometry), acat the activity of catalyst, usually expressed as mass concentration units (g m−3 ), ai the activity of reactant i, often expressed in molar concentration units (mol m−3 ), and bi is the stoichiometric coefficient for reactant i. In practice, activities ai are also expressed as partial pressures (pi ). The reaction rate can be also related to a unit mass of the catalyst or to the specific surface of the catalyst (Satterfield, 1980; Murzin & Salmi, 2005). An example of the power law application for the description of 4-aminodiphenyl amine reduction alkylation with acetone and hydrogen was presented by Králik et al. (1990). If the content of the catalyst expressed as a mass concentration is considered, the reaction rate of the reversible reaction is described by the relationship: ξ˙V = wcat (k+ cA cB − k− cR cS )

(15)

where wcat /(g m−3 ) is the mass concentration of catalyst, k+ /(m6 g−1 s−1 mol−1 ) and k− /(m6 g−1 s−1 mol−1 ) are the rate constants for the direct and back reactions, respectively, and ci /(mol m−3 ) is the molar concentration of component i. Evidently, the power law model, which can be derived from the mass action and/or activation complex theories, is not realistic for a surface reaction. A pioneer work on the heterogeneous catalytic reactions was presented by Hougen and Watson (1947), who used the Langmuir monolayer adsorption model (see Eq. (1)) and either a surface reaction of two chemisorbed species or a reaction of the chemisorbed

moiety and a component from the fluid surrounding the heterogeneous catalyst. This type of models is generally denoted as the Langmuir–Hinshelwood– Hougen–Watson (LHHW) models. Essential assumptions for the LHHW models are as follows: – all catalytic centers are of the same quality; – chemisorption on catalytic sites does not change the properties (energy) of the surrounding catalytic sites; – reaction consists of three elemental steps: a. chemisorption; b. surface reaction; c. desorption of products; – one of these steps is rate determining, others are in equilibrium; – reaction occurs either between chemisorbed species or one component is chemisorbed and the other one comes from the fluid; – activity close to the surface is equal to that in the fluid (in the interior of a porous catalyst). If the reactants are chemisorbed on active sites of different quality and the product and inert chemisorption are considered, Eq. (16) emerges: ξ˙V,i = kcat,i wcat N C

1+

j=1 νij