Sunteți pe pagina 1din 35

2.

Heterogeneous Catalysts

Heterogeneous catalysts are more often applied than homogeneous catalyst, despite the latter
being sometimes more active and selective than their solid, heterogeneous counterpart. This is
due to the ease of separation (and often regeneration as well) of the heterogeneous catalysts.

In heterogeneous catalysis the catalytic reaction takes place on the surface of the solid. The
catalytic activity is proportional to the surface area of the catalytically active phase in the catalyst.
It is therefore desired to expose create an as large as possible surface per unit mass (SA) of
catalyst to increase the activity of the catalyst. The surface area is directly related to the density of
the material, if a non-porous solid (e.g. platinum or iron carbide) is the catalyst. Take for instance
a sphere with a diameter dsphere:,

area of sphere area of sphere   d2sphere 6


SA    
mass of sphere   volume of sphere  3   dsphere
 d
6 sphere

The identical relationship can be derived for cubical crystallites. Small crystallites in the range of a
few nanometers are therefore required to obtain a high surface area per unit mass of catalyst
(larger than 100 m2/g). Crystallites with a diameter of a few micrometers have a thousand fold
smaller surface area per unit mass compared to crystallites with a diameter of a few nanometer.

It is therefore clear that small crystallites are needed for high catalytic activity, if a non-porous
solid is used as the catalytically active material (also the use of materials with an intrinsic pore
system, such as molecular sieves, require the use of small crystallites, for other reasons – internal
mass and heat transport limitations: see chapter 3). They can however not be utilised directly in
technical processes, since the pressure drop increases dramatically with decreasing particle size.
(see Ergun equation:


dp

u

1   
 150 
1      1.75    u  )
dz dparticle  3  dparticle
gas 
 
Very small particles (e.g. with a diameter of a few nanometer) will cause large pressure drops,
which make the direct utilisation of these particles in large technical scale reactors unfeasible.

The need for small crystallites (with a high surface area) and the requirement of larger particles in
technical reactors means that a catalyst particle must either comprise of an agglomeration of
small crystallites or small crystallites must be brought onto a support material. The agglomeration
of small crystallites will create a porous particle with the pores located in the intracrystalline
space. Alternatively, the nano-crystalline catalyst can be brought onto a porous support (see
Figure 2.1).

2.1 Porosity
Particles consist of spaces filled with crystallites (solid) and void spaces. The porosity of materials
() is defined as the volume of the void relative to the total volume of a particle. The void space in
a particle is the volume taken in by the particle minus the volume of the solid.

14
Figure 2.1: Catalyst particles as:
Left: agglomeration of catalyst crystals (crystals of zeolite Ti-Beta [1]
Right: nano-crystalline cobalt on Aerosil [2]

Vsolid
Vv oid Vparticle  Vsolid mparticle  particle
   1  1
Vparticle Vparticle Vparticle  solid
mparticle

The mass of a particle equals the mass of the solid in the particle, since the void space does not
have any mass. The porosity of particles can thus be determined knowing the density of a particle
and the density of the solid.

solid space

void space
mercury helium
pycnometry pycnometry
Figure 2.2: Determination of catalyst particle volume using mercury pycnometry and the
volume occupied by solids using helium pycnometry

The volume of a particle (or its density) and the volume of the solid phase in the particle (or its
density) can be determined experimentally using pycnometry. Mercury at atmospheric pressure
will not enter pores with a diameter less than ca. 5 m. If the pores within a catalyst particle are
smaller than 5 m, mercury at atmospheric pressure can be used to determine the volume of a

15
particle. For that purpose catalyst is added in a volumetric flask and the flask is filled up with
mercury. The volume mercury added must be noted, since the volume of the particles is given by:
Vparticle  Vf lask  Vmercury added  Vmercury displaced

mparticle mparticle
The particle density is then given by: particle  
Vparticle Vmercury displaced

Similarly, the volume occupied by the solid in the particle can be determined using helium. Helium
will enter all pores, but cannot enter the spaces filled with solid.
Vsolid  Vf lask  Vhelium added  Vhelium displaced

The density of the solid in the catalyst particle is thus given by:
mparticle mparticle
 solid  
Vsolid Vhelium displaced

The volume occupied by the solid can be estimated even in the absence of helium pycnometry
data knowing the mass composition of the various solid phases present in the catalyst particle
and the density of the individual phases (e.g. from the Handbook of Physics and Chemistry). For
instance, if the mass composition of a catalyst is 30 wt.-% Co on SiO2 (Co = 8.9 g/cm3 and
SiO2(quartz) = 2.6 g/cm3) the volume of the solid phase will be 0.3 cm3/g.

The pore volume can be determined knowing the volume occupied by the solid and the volume
taken in by the particle. The pores in the particle are the void spaces in the catalyst particle. The
pore volume is typically given per unit mass of catalyst. Thus the pore volume is given by:
Vv oid Vparticle  Vsolid Vmercury displaced  Vhelium displaced 1 1
Vpore     
mparticle mparticle mparticle particle  solid

2.1.1 Mercury porosimetry


Mercury is a non-wetting fluid at room temperature for most porous materials. At low pressure
mercury cannot penetrate small pores due to the high surface tension. Mercury can, however, be
forced into the pores by applying pressure [3]. At any given pressure, the force applied to enter a
pore with a radius r is resisted by the surface tension. The force balance thus reads:
2
  rpore  papplied  2    rpore  cos   
 2  cos   
rpore 
p applied

Mercury porosimetry can thus be applied to determine the pore radius distribution for a given
catalyst knowing the surface tension of Hg (0.485 N/m) and the contact angle (140 o). One should
however be aware of the pitfalls in mercury porosimetry, such as the assumption of (non-
connected) cylindrical pores [4]. For a detailed analysis of modern pore trends in mercury
porosimetry the student is referred to [5,6] and references therein. The major pitfall is the
assumption of non-connected pores. Figure 2.4 shows the presence of an ink-bottle type of pore.
The pore will only be filled if mercury can penetrate the bottle-neck. This leads then to the
interpretation that a large pore volume is associated with a smaller diameter than actually is the
present.

16
rpore

rneck

rbottle

Hg  2  cos   
p
 2  cos   
p
 2  cos   
rbottle rneck rneck

no filling of neck; instantaneous filling


no access to bottle of neck and bottle
thus no Hg in bottle
Figure 2.3: Hg- penetration Figure 2.4: Filling of an ink-bottle
in a pore type of pore

Example: Determining pore radius distribution from Hg-porosimetry


Pirard et al. [7] measured the volume intruded by mercury as a function of the applied pressure on silica
xerogels. They obtained following data (taken from Figure 1 out of the article):

p, MPa 0.02 0.1 0.49 1.49 4.5 13.8 24.1 42 73.2 127.8
3
Vintruded, cm /g 0 0.1 0.3 0.5 0.9 1.43 1.73 2.48 2.58 2.63

With the dependency of the pore radius on the applied pressure:


 2  cos     2  cos140   0.485 0.743
rpore   
papplied papplied papplied
2
with applied pressure in N/m and the rpore in m.

p, MPa 0.02 0.1 0.49 1.49 4.5 13.8 24.1 42 73.2 127.8
rpore, m 37.153 7.431 1.516 0.499 0.165 0.054 0.031 0.018 0.010 0.006
3
Vintruded, cm /g 0 0.1 0.3 0.5 0.9 1.43 1.73 2.48 2.58 2.63

The plot of the intruded volume as a function of the pore radius is a cumulative plot, i.e. the intruded volume
3
belonging to a pressure of 42 MPa (1.73 cm /g) corresponds to the penetration of all pores with a pore
radius of larger than 0.018 m. A better view of the differential pore radius distribution is obtained by
differentiating the integral plot of the intruded volume as a function of the pore radius on a logarithmic scale.
This is obtained plotting the differential (dVintruded/dlog(rpore) or Vintruded/log(rpore). This is most easily done
by taking two consecutive data points, e.g.

p, MPa 0.49 1.49


rpore, m 1.516 0.499

17
3
log(ripore), cm /g 0.18 -0.30
3
Vintruded, cm /g 0.3 0.5

Vint ruded 0.5  0.3 0.2


    0.41

 log rpore   .30  0.18 0.48

This differential belongs to an average pore radius of 1 m.

dVintruded/dlog(rpore)
3

0
0.001 0.01 0.1 1 10 100
rpore, m

Integral (left) and differential (right) plot of intruded volume as a function of the pore radius

The pore radius distribution can in principle also be determined using other liquids. However, the
surface energy of other liquids is typically much lower than that of mercury (see for instance
reference [8]).

2.1.2 Capillary condensation tfilm


Condensation of a compound in a pore can occur in pores at pressures
much lower than the vapour pressure of that compound. It might be thought that
condensation in a pore occurs through the build-up of a liquid film on the walls
of a pore. However, a rigorous thermodynamic treatment of this problem
[9,10] shows that the successive growth of the film thickness with increasing
pressure does not occur. Initially the walls of the pore are wetted (yielding a
decrease in the surface energy of the system). Raising the pressure further will
only marginally increase the thickness of the film. At a certain pressure, the
pores suddenly become filled.
rpore

Thermodynamic proof for the lack of continuous growth of a liquid film in a cylindrical
pore
Consider a cylindrical pore, whose walls are wetted with a liquid. The total number of moles of a compound
in the system is distributed over the liquid and the gas phase. For the Gibbs free energy of the system, the
interfacial energy has to be taken in:
Gsy stem  nG  GG  nL  GL  Gint erf acial with ntotal  nG  nL

 
Gsy stem  ntotal  nL  GG  nL  GL  Gint erf acial

18

Gsy stem  ntotal  GG  nL  GL  GG  Gint erf acial 
The number of moles in the liquid phase is related to the volume of the liquid phase and the molar volume
of the liquid:
V liquid
nL 
VL

The volume of the liquid, which is as a film on the pore wall is given by: Vliquid    rpore  t f ilm2  Lpore

The interfacial Gibbs free energy is given by the Gibbs free energy of the solid liquid interface and the
liquid-vapour interface. If the pore walls are fully wetted with the liquid, the interfacial Gibbs free energy of
the solid-liquid interface is given by:
Gint erf acial  SL  A  2    rpore  Lpore   SL

The interfacial Gibbs free energy for the liquid-vapour interface is given by:

Gint erf acial  LV  A  2    rpore  t f ilm  Lpore  LV 
Thus, the Gibbs free energy of a system comprising of a liquid film in a pore is given by:

Gsy stem  ntotal  GG 



  rpore  t f ilm 2  Lpore     
 GL  GG  2    rpore  t f ilm  Lpore  L  V
L
V
 2    rpore  Lpore   SL
At equilibrium the Gibbs free energy of the system is minimal. Therefore, the variation of the Gibbs free
energy of the system with the thickness of the liquid film will be zero:
 Gsy stem
  

2    rpore  t f ilm  Lpore   
 GL  GG  2    Lpore  L  V  0
 t f ilm  total L
 n ,p,T V

L
rpore  t f ilm   LL V  VG
G G 
The difference between the molar Gibbs free energy of a pure compound as a liquid, i.e. without any
interaction between the surface of the solid and the liquid, and the molar Gibbs free energy of that
compound as a vapour, is:
1. less than zero, if the pressure is larger than the vapour pressure of the compound
2. zero, if the pressure equals the vapour pressure of the compound
3. larger than zero, if the pressure is less than the vapour pressure of the compound

Capillary condensation takes place under conditions, where the liquid would normally not condense, i.e. the
pressure is lower than the vapour pressure of the liquid. Thus, the difference between the molar Gibbs free
energy of the liquid phase and the molar Gibbs free energy of the gas phase will be larger than zero. This
means that the film thickness will be larger than the radius of the pore, which corresponds to a physically
impossible situation. Thus, according to this model a film will not exist in the pores. It should however be
noted that a very thin film may exist due to van der Waals interaction between the gas phase molecules and
the surface.

The walls of the pores are initially slightly wetted and at a given pressure the pores will suddenly
become filled with liquid. This is caused by the difference in the pressure of the liquid covering the
wall and the pressure in the gas phase. The pressure of the liquid in a cylindrical pore of a radius r
and a film thickness t, will be:

19

p v apour  pliquid  L  V
r  t 
A vapour will condense out, when the Gibb’s free energy of the vapour phase becomes equal to
the Gibb’s free energy of the liquid phase taking into consideration the difference in the pressure:
Gv apour  Gliquid
V v apour  dp v apour  Vliquid  dpliquid

The molar volume of the vapour phase is given through the ideal gas law. Thus,
RT
 dpv apour  Vliquid  dpliquid
v apour
p

In a capillary the vapour will condense at a pressure p instead of at the saturation pressure p*.
The pressure of the liquid is then p*-p.

p
v apour V liquid p*  p liquid
 dp 
RT
  dp
p* p*

Solving this integral:


 p v apour  liquid
ln   V  p
 p*  RT
 

The difference in the pressure of the liquid phase is mainly due to the curvature of the liquid
surface, i.e.

p  L  V
r  t 

Thus condensation will take place, if


V liquid L  V
 
RT r  t 
p v apour  p  e

The emptying of pores starts from the formation of a meniscus at the end of the
pore:

The capillary pressure is given by:



pv apour  pliquid  2  L  V
r
The factor 2 comes into because the area of the meniscus is
proportional to r2 (and not r as in the case of pore filling). Developing the
argument now in the same manner as for the filling of the pore leads to the
statement that evaporation of the content of the pores will happen if the pressure
in the vapour phase, pvapour:
rpore
V liquid  L  V
 2 
p v apour  pe RT r

20
A compound will condense out earlier in capillary pore, due to the reduced pressure of the liquid.
This will result in the rapid condensation within a pore if a certain pressure is exceeded. The pore
will be emptied (evaporation of the liquid in the pore) at a pressure, which is not identical to the
pressure at which the pore was filled. The difference between the pressure of the liquid and the
pressure of the vapour is less for a completely filled pore than for a thin liquid layer on the wall of
the pore.

Figure 2.5 shows the pressure at which benzene will condense in a pore with a radius r (assuming
that the film thickness can be neglected) and the pressure at which liquid benzene in a pore will
evaporate at a temperature of 293 K. Capillary condensation plays a major role in the filling of
sub-micron pores. In the range of a few micrometer, the capillary condensation effect becomes
marginal.

80
filling of pores
Pressure, mm Hg

60

40

emptying of pores
20

0
1.E-09 1.E-08 1.E-07 1.E-06 1.E-05
Pore radius, m

Figure 2.5: Pressure at which benzene in the vapour phase will condense and can evaporated
in a pore with a radius r at 293 K

Thus, a fluid in narrow pores can have quite different properties from those of bulk fluids. This can
only occur, if the attraction forces of the wall significantly alter the phase behaviour. For example
the critical temperature of a substance can change in a porous solid [11].

2.1.3 Classification of porous systems


Porous solids consist of a network of different, sometimes interconnected pores. Pores can be
classified in terms of their accessibility to molecules in the fluid phase. An open pore is a pore,
which is connected to the external surface of a solid and molecules can pass through these pores.
Closed pores are a void space within the solid with no access to the external surface and are
therefore isolated. So-called transport pores connect different parts of the external surface to the
inner microporosity. Blind pores are connected to transport pores, but are themselves not
connected to other pores or to the external surface.

21
The pores can be classified according to their size [12]. The IUPAC-classification of a pore
system is:
Micro-pores dpore< 2 nm
Meso-pores 2 nm<dpore<50 nm
Macro-pores dpore>50 nm

2.2 Surfaces
Surfaces are typically thought of as the termination of the bulk of a solid. However, it should be
recognized that some materials, such as molecular sieves (and thus also zeolites), have an
intrinsic pore system, i.e. the pores form a part of the crystal structure of the material. Figure 2.6
shows the crystal structure of the zeolite MFI (also known as ZSM-5). The crystal structure
contains pores as an intrinsic part of the crystal structure. The size of these pores is in the order
of the size of molecules (and hence the name molecular sieves). Furthermore, it can be noticed
that every atom in this material is a surface atom, i.e. every atom is located in a pore and ever
atom is a surface atom. The crystal is terminated by hydroxyl groups on the external surface of
the crystal.

Figure 2.6: Crystal structure of MFI (yellow: T-atom (Si or Al); red: oxygen).

The termination of the crystal structure of non-porous materials will result in a surface. The atoms
on a surface have typically less neighbouring atoms than the atoms in the bulk of a material (in a
fcc or hcp –structure each atom in the bulk has 12 nearest neighbours; in a bcc-structure each
atom in the bulk has 8 nearest neighbours). The reduction in the number of neighbouring atoms
(or coordination number of surface atoms) leads to a higher energy of the surface atoms than that
of the atoms in the bulk of the structure. The surface energy, , can be viewed as the difference
between the energy of the atoms in the surface and the energy of those atoms in the bulk of the
structure per unit area [13-15].

22
body centred cubic face centred cubic hexagonal closed packed
(bcc) (fcc) (hcp)

Figure 2.7: Common crystal structures

Surfaces are created by the termination of a plane and are commonly denoted with the vector
normal to the plane. Thus, the termination of the crystal with a (100) plane exposes the plane with
the vectors [001] and [010] to the surface. Not all surfaces appear flat and smooth. Some planes
(e.g. (210) plane of a bcc structure – see Figure 2.8) can be viewed consisting of terraces and
steps.

Surfaces (bcc): (100) (110) (210))

Surfaces (fcc): (100) (110) (111)

Figure 2.8: Surfaces of a bcc (top) and fcc (bottom) structure

The formation of a surface is energetically not favoured. Furthermore, a large number of planes
can be exposed in principal, but surfaces tend to equilibrate to a low energy state, i.e. surfaces
containing atoms with a high coordination will typically be preferred. In equilibrium the system
(crystal or set of crystals) tends to reduce the surface energy by the formation of specific crystal
shapes (Wulff plot) [16], by reconstruction of the surface [17], or by crystal growth (sintering). The
contribution of the surface energy increases dramatically with a decrease in the crystallite size
[18]. This is attributed to a dual effect, viz. the increase in the surface area of small crystallites
relative to the volume of the crystals and the decrease in the number of neighbouring atoms for
surface atoms with decreasing crystallite size (see Figure 2.9). The prevalence of sites with a
particular coordination number as a function of the lattice structure, morphology of the crystals,
and the total number of atoms for perfect crystals has been reported by van Hardeveld and
Hartog [19].

23
Average number of "broken
9

bonds" per surface atom


tetrahedron
cube
octahedron
rhombic dodecahedron
cubo-octahedron
6

3
10 100 1000 10000 100000
Total number of atoms in crystallite

Figure 2.9: Average number of missing bonds for a surface atom in a fcc-structure as a
function of the total number of atoms in a crystallite calculated using the statistics
as published by van Hardeveld and Hartog [19,20]

2.2.1 Dispersion
Catalysis takes place on the surface of the catalytically active material. The effectiveness of a
catalytically active material is thus a function of the number of atoms on the surface relative to the
total number of atoms present in the material. This is called the dispersion. The dispersion is a
function of the crystal size of the catalytically active material and the structure of catalytically
active material. We can consider a single crystal with a spherical geometry. The number of atoms
at the surface of this single crystal is given by the surface area and the number of atoms per unit
surface area, X. The latter is a function of the crystal structure and the exposed plane of the
catalytically active material. The number of metal atoms per unit surface area for a variety of
materials is given in Table 2.1. The number of metal atoms can principally be calculated from the
crystal structure, but it is a function of the crystal planes exposed on the surface. When a number
of crystal planes are exposed, an average number of metal atoms per unit surface area can be
estimated. This value may however become size dependent.

Table 2.1: Number of metal atoms per nm2 [21]


Fe Co Ni
16.3 14.61 15.4
Ru Rh Pd Ag
16.3 13.3 12.5 11.5
Ir Pt Au
13.0 12.5 11.5
1
:Data from [22]

24
The total number of metal atoms is given by the volume of the single crystal and the molar density
of the material, which can be expressed as a function of the mass density, , and the molar mass,
M. Thus, the dispersion can be related to the (average) diameter of the crystal of the catalytically
active material.
2
number of atoms at surface   dcry stal X
dispersion  
total number of atoms in crystal  3 
 dcry stal
6 M
6 M X
dispersion 
dcry stal 

Substituting the known values for e.g. cobalt leads to:


g atoms 1 mol
6  58.9 
 14.6   
2 23
dispersion 
mol nm 6.02  10 atom
3
g cm
dcry stal 8.9   10 21 
cm3 nm3
dispersion% 
96
dcry stalnm
Thus, 9.6% of all atoms in the crystal are at the surface for a 10 nm crystal, and the fraction of
metal atoms at the surface relative to the total number of metal atoms will decrease with
increasing crystal size.

2.3 Adsorption on surfaces


Reactants have to interact with the catalyst surface for a reaction to take place. Surfaces can be
unsaturated, since the surface is a termination from the bulk structure. Hence, a surface may
reconstruct itself, and the surface structure may differ significantly from that of the bulk of the
solid.

There is always an attractive force between surfaces and molecules in the fluid phase. The
attraction can be based on physical attraction (physisorption) or the formation of a chemical bond
(chemisorption).

Physisorption is based on the weak interaction due to the van der Waals forces between the
surface and molecules in the fluid phase. The van der Waals’ attraction is based on the attraction
between objects with a dipole moment (Keesom force), between objects with a dipole moment
and objects which can be polarized (Debye force) and the attraction between two polarisable
objects (London force). It can be derived that the attractive energy is proportional to r -6. Close
proximity of the objects lead to repulsion, and the repulsive energy is can be modelled to be
proportional to r-12. Thus, the physical attraction between objects (e.g. surfaces and molecules in
the gas phase) can be modelled using the Lennard-Jones potential:
A B
E 
6
r r 12
Physisorption occurs on all type of solids, if the fluid phase molecules are at a temperature below
their critical temperature. The process is very similar to a condensation process, and it is therefore
a low temperature process. However, only a thin film can be formed, since the continuous growth
of the liquid film is thermodynamically not allowed. However, physisorption may lead to the
formation of multi-layer (but the number of layers will not go to infinity). Physisorption is a non-

25
activated process (i.e. it has a low activation energy), which is exothermic, since the change in
entropy upon the formation of a liquid is less than zero. The heat of the physisorption process is
similar to the heat of condensation (typically between –2 and –20 kJ/mol).

Chemisorption requires the formation of a chemical bond between an unsaturated surface and the
adsorbing molecule. The characteristics of chemisorption are thus strongly linked to those of a
chemical reaction. The chemisorption process can be an activated process (i.e. a process with a
substantial activation energy) and can be endothermic or exothermic. The heat of reaction is
typically in excess of 50 kJ/mol. Chemisorption is a chemical reaction between an adsorbing
molecule and the unsaturated surface. Thus, only monolayer formation is allowed.

Figure 2.7: Schematic representation of physisorption and chemisorption

Table 2.2: Characteristic features of physisorption and chemisorption


Physisorption Chemisorption
Similar to condensation process Formation of chemical bond between
unsaturated surface and adsorbing molecules
All surfaces Reactive surfaces
crit
All adsorbing fluids below T Reactive gases
Low temperature process High temperature process
Multi-layer formation Mono-layer formation
Reversible process Can be irreversible
Non-activated process Can be activated process
Exothermic process Can be exothermic or endothermic

2.3.1 Measuring adsorption on surfaces


The uptake of a gaseous species through adsorption on a surface can be measured in various
ways. The ultimate goal of the measurement is to estimate the amount of material adsorbed on
the solid in equilibrium with a gas phase at a certain partial pressure of the adsorbing component.
Thus, the critical features in each measurement technique is the accurate determination of the
amount taken-up by the solid through adsorption, the achievement of the equilibrium state and the
determination of the partial pressure of the adsorbing component.

2.3.1.1 Static volumetric method


The most commonly used technique is the volumetric measurement of the adsorption of a
gaseous species on the surface (see Figure 2.10).

26
PI
vacuum
Adsorbing gas

Sample
Figure 2.10: Principle of set-up for measuring adsorption isotherms using the static volumetric
method

The set-up consists of a vacuum line, a gas-dosing line and a line connected to the sample
holder. The gas-dosing section and the sample holder are initially evacuated. The valve to the
vacuum pump and to the sample holder is closed, and gas is added to the gas dosing system.
The pressure of in the gas dosing system is controlled, and knowing the volume of the gas dosing
system, the number of moles of gas in the gas dosing system is known (via e.g. ideal gas law).
Opening the valve to the sample holder will reduce the pressure, since the volume occupied by
the gas is now larger, and because gas is taken up by the solid (adsorption). Knowing the volume
of the gas-dosing system and the volume of the free space in the sample holder, the number of
moles of gas adsorbed on the solid can be determined. The volume of the free-space can be
determined by performing the adsorption experiment with a non-adsorbing gas (helium). The gas
dosing is repeated till the change in pressure is within certain limits. The achievement of the
equilibrium state is determined by looking at the pressure change over a given time interval.

In this method the determination of the pressure of the adsorbing species is directly measured.
The determination of the amount adsorbed depends on the accuracy of the determination of the
pressure, temperature, the volume of the gas dosing system, and the volume of the free space in
the sample holder. The determination of the equilibrium state proceeds through a pressure
measurement.

2.3.1.2 Dynamic flow method


In the dynamic flow method, the adsorbing gas at a given partial pressure is passed over the
solid. The concentration of the adsorbing gas in the effluent is measured (typically using a TCD;
see Figure 2.11).

In this set-up, a non-adsorbing diluent must be co-fed with the adsorbing gas in order to measure
the concentration of the adsorbing gas. The concentration of the diluent must be relatively large,
so that the volumetric flow rate through the detector does not change significantly when gas is
adsorbed. The concentration of the adsorbing gas in the effluent will approach the inlet
concentration of the inlet gas asymptotically. This can make the accurate determination of the
equilibrium up-take rather difficult.

27
Adsorbing gas

diluent
TCD

Sample holder

Figure 2.11: Principle set-up for measuring adsorption using the dynamic flow method

2.3.1.3 Gravimetric method


The amount of material adsorbed can in principal also be determined gravimetrically. The mass of
the solid exposed to an adsorbing gas at a given pressure will increase upon adsorption. The
gravimetric can be done in both flow systems and in static systems. In a flow system buoyancy
has to be taken into account. A static system, with an accurate measurement of pressure, would
allow independent measurement of pressure of the adsorbing species and the amount taken up.
The achievement of the equilibrium state can be determined by measuring the mass change as a
function of time.

2.3.2 Heat of adsorption


The adsorption of (gaseous) species on the surface of a solid is associated with a heat effect. In
the case of physisorption, the heat effects can be quite small, but chemisorption is typically
associated with large heat of adsorption.

The heat of adsorption can be determined experimentally by performing the adsorption


experiments in a calorimeter [23] or by performing the adsorption experiments at different
temperatures. In the latter case we can apply the Clausius-Clapeyron equations. In equilibrium
the Gibbs’ free energy of the species in the vapour phase equals the Gibbs’ free energy of the
species in the vapour phase:
GG  Gadsorbed
HG  T  SG  Hadsorbed  T  Sadsorbed
HG  Hadsorbed Hadsorption
SG  Sadsorbed  
T T

Furthermore, the change in the molar Gibbs’ free energy in both phases with respect to
temperature and pressure must be equal:
dGG  dGadsorbed
 SG  dT  V G  dp  Sadsorbed  dT  V adsorbed  dp
dp SG  Sadsorbed

dT V G  V adsorbed

28
Taking into account the relationship between the change in the entropy and the change in the
enthalpy, and further realising that the molar volume of the adsorbed phase is negligible
compared to the molar volume of the gas phase:
dp Hadsorption

dT T  VG

The molar volume of the gas phase is given by the ideal gas law:
dp p  Hadsorption d ln p  Hadsorption
 or 
dT R  T2 d1 R
T

The heat of adsorption can be determined by measuring the uptake as a function of pressure at a
given temperature. This experiment should then be repeated at various temperatures (see Figure
2.12). Reading off the pressures at a constant amount of material adsorbed at a given adsorption
temperature, and plotting in a semi-logarithmic plot the pressure versus the reciprocal of the
adsorption temperature, should yield a straight line with a slope equal to the heat of adsorption
relative to the gas constant. The heat of adsorption can thus be determined as a function of the
amount of material adsorbed.

8 40
adsorption
slope = -H /R

6 30
, kJ/mol
ln(p/mm Hg)

adsorbed
V =
0.7 cm3
4 1.7 cm3 20
ads

2.7 cm3
H

2 10

0 0
2.6 2.8 3 3.2 0 0.5 1 1.5 2 2.5 3 3.5
-1 3
1000/T, K Vads, cm (STP)/g

Figure 2.12: CO-chemisorption on 10 wt-% Co/8 wt-% V2O5/-Al2O3 [24]


Top: Adsorption isotherms at 323K, 348K, 373 K
Bottom left: Adsorption pressure at a constant amount adsorbed as a function of
reciprocal temperature
Bottom right: Heat of adsorption as a function of amount of CO adsorbed

29
A coverage dependent heat of adsorption can be associated with the presence of different
surfaces in the catalyst samples, i.e. when the material contains poly-crystalline material [25],
However, a coverage dependent heat of adsorption has also been observed on single crystals
[26] (see Figure 2.13).

250

Integral heat of adsorption of


CO, -Hads,CO, kJ/mol
200 CO on Pt{100}-(1x1)

150

100
CO on Ni(100)

50
0 0.2 0.4 0.6 0.8 1
Surface coverage with CO
Figure 2.13: Integral heat of CO-chemisorption on single crystals (data: CO on Pt{100}-(1x1)
redrawn from [27], CO on Ni(111) redrawn from [28]

2.3.3 Modelling chemisorption


Chemisorption can be modelled as the reaction of an incoming molecule reacting with the
unsaturated, uncovered surface. Some assumptions have to be made in order to evaluate the
adsorption isotherm. The basic assumption underlying all types of isotherms is the collision of
molecules in the gas phase with the surface, and will with a given probability stick to the surface.
The adsorbed species will desorb from the surface due to vibrational motion.

2.3.3.1 Langmuir isotherm


The Langmuir-isotherm is the simplest isotherm. Following assumptions are being made in the
derivation of the Langmuir-isotherm:
1. Monolayer formation
2. Adsorption and desorption follow a single mechanism
3. Surface is energetically homogeneous
4. No interaction between adsorbed molecules

The reaction between a gas phase molecule A and a vacant site on the surface can be written as:
A+* A*
Thus, the concentration of sites covered with A (A*) can be obtained over the equilibrium
relationship:
CA *
K or CA *  K  p A  C*
p A  C*

The total number of sites is fixed and thus a site balance yields: CA *  C*  Ctotal,*

30
Substituting the equilibrium relationship into the site balance yields:
and CA *  Ctotal,*  K  p A
Ctotal,*
C* 
1  K  pA 1  K  pA

It is convenient to talk about the fraction of the sites covered with e.g. A:
CA * K  pA
A  
Ctotal,* 1  K  p A

At low partial pressure of A, the coverage of the surface with A increases linearly with the partial
pressure of A (the proportionality constant is the adsorption constant). At high partial pressure of
A, the coverage of the surface tends to 1.

By measuring the amount of material adsorbed as a function of the partial pressure of the
adsorbing gas A, the amount of material, which needs to be adsorbed to form a monolayer can be
determined.
VA,adsorbed K  pA
A  
VA,monolay er 1  K  pA

1 1 1 1
  
VA,adsorbed VA,monolay er K  VA,monolay er p A

Figure 2.14 shows schematically the Langmuir-isotherm. At low partial pressures the volume
adsorbed (or the number of moles or mass adsorbed) increases linearly with increasing pressure.
At high partial pressure the curve flattens out and reaches a plateau. That means that at this
plateau very little more can be adsorbed, despite a further increase in the pressure. This is taken
as an indication that the surface is fully covered and a monolayer of adsorbed material has been
formed. The plateau represents thus the volume adsorbed needed to form a monolayer.

Figure 2.14: Langmuir-isotherm

31
The shape of the Langmuir isotherm is strongly dependent on the value for the adsorption
constant K. With increasing value of K, the plateau is reached at lower pressures due to the
strong adsorption. The influence of temperature on the adsorption curve is opposite to the
influence of the adsorption constant K, since the adsorption constant typically decreases with
increasing temperature.

Example: Estimation of number of molecules adsorbed to form a monolayer


Supported nickel catalysts are typically used for the liquid phase hydrogenation of vegetable oil. The
catalyst supplier offers a novel 10 wt.-% Ni on activated carbon for this reaction. It is claimed that this
2
catalyst has a metal surface area of 1.5 m per gram of catalyst. In order to test the claim by the catalyst
manufacturer, you decide to perform H2-chemisorption. Following data were generated by H 2-chemisorption
o
at 100 C.

p, mbar 10 20 40 100 200 400


3
Vadsorbed cm (STP)/g 0.20 0.25 0.31 0.38 0.44 0.49

Determine the amount of hydrogen adsorbed to form a monolayer.

Hydrogen adsorbs dissociatively on metal surfaces, i.e.

H2 + 2* 2H*
Thus, the concentration of sites covered with H (H*) can be obtained over the equilibrium relationship:
2
CH
K * or CH*  K  pH2  C*
pH2  C2*

The total number of sites is fixed and thus a site balance yields: CH*  C*  Ctotal,*
Substituting the equilibrium relationship into the site balance yields:
Ctotal,* K  pH2
C*  and CH*  Ctotal,* 
1  K  pH2 1  K  pH2

Thus, the volume adsorbed is given by:


CH* n 2  nH2 ,adsorbed VH2 ,adsorbed
 H*  
Ctotal,* ntotal,* 2  nH2 ,adsorbed,max VH2 ,adsorbed,monolay er

VH2 ,adsorbed K  pH2



VH2 ,adsorbed,monolay er 1  K  pH2

Plot the reciprocal of the volume adsorbed versus the reciprocal of square root of the hydrogen partial
pressure:
1 1 1 1
  
VH2 ,adsorbed VH2 ,adsorbed,monolay er VH ,adsorbed,monolay er  K pH
2 2

3
From linear regression, we obtain in the linearised plot an intercept of 1.493 g/cm and a slope of 11.115
3 . 0.5 3
g/(cm (STP) mbar ). Thus, the volume of hydrogen adsorbed to form a monolayer is 0.67 cm (STP)/g.
o
This corresponds to 0.03 mmol H2/g (using ideal gas law, 1 mmol at 0 C and 1 bar has a volume of 22.414
3
cm ) or 0.06 mmol H/g. The adsorption constant can be determined from the slope of the linear fit and is
-1
0.0138 mbar .

32
0.6 6

1/Vadsorbed, g/cm (STP)


Vadsorbed, cm (STP)/g

3
0.4 4
3

0.2 Adsorption isotherm 2 y = 1.4933+11.115/p0.5


R2 = 0.9997

0 0
0 50 100 150 200 250 300 350 400 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
0.5 -0.5
p, mbar 1/p , mbar

2.3.3.2 Temkin isotherm


In the derivation of the Langmuir-isotherm, it is assumed that the surface is energetically
homogeneous, i.e. the heat of adsorption is independent of the coverage. This is not always the
case (see Figure 2.11) [23,25]. The decrease in the heat of adsorption can be taken into account.
Consider the surface to consist of elements of size dS with different heats of adsorption. The total
coverage, , is then given by summing up the coverage i over the different size elements of the
surface with different energetics [29]:
1
   i  dS
0
Each size element has a particular heat of adsorption, Hads,i, and the adsorption in each element
follows the Langmuir-isotherm
Hads,i
K ads,i T   p K 0  e RT p
i  
1  K ads,i T   p Hads,i
1  K 0  e RT p

In the Temkin-isotherm, it is assumed that the heat of adsorption decreases linearly with the
coverage, i.e.
Hads,i  Hads,0  A  S

Development of Temkin isotherm


Thus, the coverage in an element is given by:
Hads,0  A S
K0  e RT p
i 
Hads,0  A S
1 K0  e RT p
and the total coverage can be evaluated by:

Hads,0  AS
1 1
K0  e RT p
   i  dS    dS
Hads,0  AS
0 0
1  K0  e RT p

The most convenient way to evaluate this integral is by changing the variables:

33
Hads,0  AS
  K0  e RT p
Hads,0  AS
A RT
d   K0  e  p  dS
RT
 A
d   dS
RT

Substituting this back into the integral:


Hads,0  A
 ce R T p
RT 1
  A

1 
 d
Hads,0
 ce R T p

Thus, the Temkin isotherm has the form

 Hads,0   
 
  
R  T  1 K0  e R T p  RT  1  K  0  p 
  ln    ln
A Hads,0  A A  A 
 RT  1 K RT  p 
 1  K 0  e  p    0  e 

Figure 2.15 shows the Temkin-isotherms. An increasing value of A/RT means a stronger
decrease of the heat of adsorption with coverage. Hence, the isotherms become flatter, since the
average heat of adsorption decreases and thus the average adsorption constant decreases.

Figure 2.15: Temkin-isotherm

The data obeying the Temkin-isotherm can roughly be fitted with a Langmuir-isotherm. This
means that the isotherm is being modelled with an average adsorption constant, i.e. an average
heat of adsorption. However, a number of noticeable deviations will be observed, since the heat of
adsorption is not constant. In the intermediate pressure range, the Langmuir-isotherm typically
overestimates the volume adsorbed, whereas at high pressure, the Langmuir-isotherms
underestimate the volume adsorbed.
2.3.3.3 Freundlich isotherm

34
The Freundlich isotherm is a purely empirical equation, which has been found to be obeyed by
many experimental data of adsorption.
1
  K pn

It can be easily seen that the Freundlich isotherm cannot strictly describe chemisorption, since the
coverage of the surface can become larger than one. The Freundlich isotherm predicts an
increase in the coverage with increasing pressure. This would correspond to a continuous growth
of a liquid film, which is thermodynamically inconsistent (vide supra).

The Feundlich isotherm has erroneously been rationalised based on an exponential distribution of
adsorption energies [30]. Yang [31] showed that for intermediate coverages the Freundlich
isotherm can be derived taking into account the interaction between adsorbed species and a
decrease in the desorption vibration.

Adsorption isotherm based on an exponential decline of the heat of adsorption


The linear decrease in the heat of adsorption used in the derivation of the Temkin-isotherm can
be replaced by an exponential decay-function:
Hads,i  Hads,0  eAS
Thus, the coverage in an element is given by:
Hads,0 e  A S
K0  e RT p
i   A S
Hads,0 e
1  K0  e RT p
and the total coverage can be evaluated by:

Hads,0 e  A S
1 1
K0  e RT p
   i  dS    dS
Hads,0 e  A SS
0 0
1  K0  e RT p

At low coverage the integral can be evaluated analytically:


Hads,0 e  A S
1
   K0  e RT  p  dS
0

The most convenient way to evaluate this integral is by changing the variables:
  e AS
d  A  eAS  dS  A    dS

Substituting this back into the integral:


Hads,0 
 e  A
1 RT
   K0  e  p  d
 1
A

35
 e  A
 K 0  p  
2 3
Hads,0   1  Hads,0    1  Hads,0   
   ln             
A  RT 4  RT  18  R  T  
  1
 
     
2 3
 Hads,0 A 1  Hads,0  2A 1  Hads,0  3A 
  K 0  p  1   e 1  
 
  e  1   
 
  e  1  
       

R T A 4 A  R T  18 A  R T  
Thus, at low coverage an exponential decline in the heat of adsorption predicts a linear increase in the
coverage with increasing pressure (similar to Langmuir and Temkin-isotherm).

A full analysis of an exponential decrease in the heat of adsorption can be obtained from numerical
evaluation of the integral describing the coverage as a function of the surface sites. This description will
however not lead to the Freundlich isotherm.

Isotherms predicted using an exponential decline of the heat of adsorption

2.3.3.4 DFT-derived adsorption isotherms


Chemisorption of molecules on surfaces results in the formation of a chemical bond between the
surface and the adsorbed species resulting in a change in the electron distribution in the surface
and possible within the adsorbed species. A polarisation in the adsorbed species will result in long
range electro-static interactions in addition to the short-range re-distribution of the electrons of the
adsorbed molecules. Adsorbate-adsorbate interaction and surface reconstruction may alter the
uptake of the adsorbing gas as a function of pressure and temperature. The Temkin and
Freundlich-isotherm do account for the change in the heat of adsorption with increasing surface
coverage, but the continuous decrease in the heat of adsorption with increasing coverage is only
valid for long-range interactions. It should however be noted that the short-range interaction
between adsorbates may be much stronger. The interaction can be taken into account using a
simple lattice model (see Figure 2.16).

The strength of adsorption is reduced with increasing coverage due to adsorbate-adsorbate


interactions has been confirmed using DFT-calculations [32] and can be attributed in some cases
to the repulsion between the adsorbed species [33]. The interaction of atomic oxygen on Pt(111)
also shows strong repulsion (see Figure 2.16). The Pt(111) surface can be seen to consists of
sites in the shape of equilateral parallelograms. These sites are each surrounded by 4 edge
sharing sites and 4 corner sharing sites. Atomic oxygen present on the corner sharing sites hardly
influences the strength of adsorption of oxygen on the central site. The adsorption becomes
noticeable weaker when atomic oxygen is also present on the edge sharing site. The close
proximity of adsorbed, atomic oxygen on Pt(111) leads to surface reconstruction.

36
O 0.111 0.222 0.333 0.444
Eads -1.073 -0.986 -0.880 -0.734
(eV/O)

Figure 2.16: Influence of adjacent atomic oxygen on the heat of adsorption of atomic oxygen on
Pt(111) relative to gas phase O2 as estimated using DFT (Functional: PAW-PBE;
Ecut-off: 400 eV; k-points: 6x6x1; smearing: 1st order Methfessel-Paxton with  = 0.2
eV)

2.3.3.5 Multi-component adsorption


Chemical reaction may involve the adsorption of two different chemical species. The co-
adsorption of various species is typically modelled assuming ideal Langmuir behaviour for both
components yielding a random mixture on the surface. For example the adsorption of species A
and B on a surface can be viewed as a competition of these species for the same type of active
site:
A+* A*
B+* B*
Thus, the concentration of sites covered with A (A*) and B (B*) can be obtained from the two
equilibrium relationships:
CA * CB *
KA  KB 
p A  C* pB  C *

or CA*  K A  pA  C* CB*  KB  pB  C*

The total number of sites is constant and thus a site balance yields: CA*  CB*  C*  Ctotal,*

Substituting the equilibrium relationships into the site balance yields:


Ctotal,*
and K A  pA
C*  CA *  Ctotal,* 
1  K A  pA  KB  pB 1  K A  pA  KB  pB
KB  pB
CB*  Ctotal,* 
1  K A  pA  KB  pB
The assumptions associated with the Langmuir isotherm are typically problematic at high
coverage (vide supra) corresponding to industrial relevant conditions.

Multi-component adsorption does not necessarily lead to the formation of a mixed co-adsorbed
state [34], and the formation of islands each containing one of the adsorbing molecules [35]. A
mixed co-adsorbed state will only occur, when the repulsion between the co-adsorbing molecules
is less than the repulsion experienced by these molecules in a compressed state.

37
2.3.4 Physisorption
Physisorption describes the process involved in the process without the formation of a chemical
bond between the adsorbed species and the surface. Figure 2.17 depicts the difference between
physisorption and chemisorption of hydrogen on a metal surface. At infinite distance from the
surface H2 is in its molecular form and the energy is the sum of the energy of the metal and metal
surface and the energy of H2. Upon the approach of the hydrogen molecule to the surface, the
attractive van der Waals force (due to the polarization of the surface and the hydrogen molecule)
becomes noticeable. A further approach to the surface leads to repulsion and thus an increase in
the energy. However, hydrogen can dissociate, after passing the energy barrier, and form H on
the surface. A closed approximation to the surface leads to a repulsion. Increasing the distance
from the surface will result in an increase in the energy. The difference between the approach
energy of the H2-molecule and the energy at large distance of the H-species, is the dissociation
energy for hydrogen.

Figure 2.17: Potential energy diagram for the physisorption and chemisorption of hydrogen on a
metal surface (redrawn from [36])

2.3.4.1 Classification of physisorption isotherms


Brunauer et al. [37] proposed to classify the physical adsorption isotherms in five different types.
The classification is based on the observed differences in the shape of the isotherms and relates
the shape of the isotherms to a mean pore size of the adsorbent and the intensity of the
adsorbate-adsorbent interaction. The five different types are shown in Figure 2.18.

 Isotherm type I represents microporous adsorbents and are characterised by a Langmuir-


type behaviour and an increase in the amount adsorbed at high relative pressure
 Isotherm types II and III are observed for macroporous adsorbents. The adsorbate-
adsorbent interaction is in the case of isotherm type III very weak (and is rather unusual)
 Isotherm types IV and V are observed for mesoporous adsorbents. The adsorbate-
adsorbent interaction is in the case of isotherm type V very weak (and is rather unusual)

38
Type I Type II Type III Type IV Type V
Vadsorbed

0 1 0 10 1 0 1 0 1
p/p0

Figure 2.18: Classification of physical adsorption isotherms according to [37]

2.3.4.2 Hysteresis in physisorption isotherms


The filling of pores upon increasing the relative pressure and the emptying of pores upon
decreasing the relative pressure leads for meso-porous materials to an observable hysteresis in
the physisorption isotherm, which is due to capillary condensation in the meso-pores (capillary
condensation will not take place in very small pores, since the thickness of the adsorbed layer
already corresponds to the diameter of the pore; capillary condensation in very large pores will not
be observed, since it takes place at relative pressures close to the saturation pressure –see
Figure 2.5).

The shape of the hysteresis yields information on the pore structure in the porous material. The
types of hysteresis have been classified as depicted in Figure 2.19. Each group is related to the
general shape of the pores.
 Type A hysteresis is associated with pores with an approximately constant pore radius.
The pore does not have to be cylindrical, but might be cubical or hexagonal in shape.
 Type B hysteresis is associated with so-called ink-bottle shaped pores with a very small
neck and with slit-type pores
 Type C hysteresis is derived from type A hysteresis, where the pore diameter is not
constant but becomes continuously smaller. Therefore, the desorption-branch of the
hysteresis is no longer sharp, but a steady change is observed, since at each pressure a
part of the pore is emptied.
 Type D hysteresis is derived from type B hysteresis, where the slit opening is changing,
i.e. the plates are not parallel.
 Type E hysteresis occurs frequently. It corresponds to ink-bottle pores, cavities between
the close packed spheres.

39
Figure 2.19: Classification of hysteresis observed with physical adsorption isotherms according
to [38]

2.3.4.3 BET-isotherm
Physisorption is classically been modelled using the Brunauer-Emmet-Teller (BET)-isotherm [39].

p
c
Vadsorbed p0
 
Vmonolay er  p   p 
1    1  c  1  
 p   p 
 0   0 

Derivation of BET-isotherm
In the derivation of the BET-isotherm it is assumed that the surface is energetically homogeneous and that
the surface is covered with a number of layers of adsorbate. i is defined as the fraction of the surface that
st
is covered with i layers of adsorbate. The heat of adsorption for the formation of the 1 layer is assumed to
be larger than the heat of adsorption for the consecutive layers.
Fraction of surface covered with:

4 3 2 layers 1 layer 0 layers

40
The total surface coverage is given by sum of all coverages of the surface with i layers:
i 
  i
i0
and the total amount of material adsorbed is given by:
i 

Vadsorbed,total
 i  i
 i 0
Vadsorbed,monolay er i 
 i
i0

At equilibrium, the net rate of formation of the fraction of the surface which is covered with 0 layers is zero,
i.e. the rate of formation of this fraction of the surface (by desorption of an adsorbed species from a fraction
of the surface covered with 1 layer of adsorbate) equals the rate of consumption of this surface (by
adsorption):

k1,desorption  1  k1,adsorption  p  0 or 1  K1,adsorption  p  0


k
( K1,adsorption  1,adsorption )
k1,desorption

A similar approach can be taken for the fraction of the surface, which is covered with 1 layer of adsorbate.
This layer is formed by the adsorption onto the surface, which is covered with 0 layers of adsorbate and by
desorption of a molecule from the surface covered with 2 layers of adsorbate. This layer is consumed by
adsorption of a molecule from the gas phase onto this layer or by desorption of a molecule from this layer:

k1,adsorption  p  0  k2,desorption  2  k2,adsorption  p  1  k1,desorption  1

Substituting the obtained relationship between the fraction of the surface covered with 1 layer and the
fraction of the surface covered with 0 layers:

k2,desorption  2  k2,adsorption  p  1
2  K 2,adsorption  K1,adsorption  p2  0

th
This argument can now be extended to the net rate of formation of the i layer:

k 2,desorption  i  k2,adsorption  p  i1 (since the heat of adsorption is constant beyond layer 2)
i  Ki2,1adsorption  K1,adsorption  pi  0

Thus,
i  i 

Vadsorbed,total
 i  i  i  Ki2,1adsorption  K1,adsorption  pi  0
 i0  i 0
Vadsorbed,monolay er i  i 
 i 0   Ki2,1adsorption  K1,adsorption  pi  0
i0 i1
K1,adsorption i 
  i  Ki2,adsorption  pi
Vadsorbed,total K 2,adsorption i0

Vadsorbed,monolay er K1,adsorption i i
1   K 2,adsorption  pi
K 2,adsorption i1

41
i  x i  x
Realising that  xi  1  x and  i  xi  (for 0<x<1)
i1 i 0 1  x 2
Therefore,

K1,adsorption
 K 2,adsorption  p
Vadsorbed,total K 2,adsorption

Vadsorbed,monolay er  K1,adsorption 
  K 2,adsorption  p 
 
1 
K 2,adsorption
1  K  p

  1  K 2,adsorption  p2

 2,adsorption 
 
 
K1,adsorption
 K 2,adsorption  p
Vadsorbed,total K 2,adsorption

Vadsorbed,monolay er   K1,adsorption  
1  
  K 2,adsorption

 1  K 2,adsorption  p   1  K 2,adsorption  p
 

   

.
In this formulation, it can be seen that the volume adsorbed will go to infinity, when K 2,adsorption p approaches
one (it should be noted that the other term in the denominator is always larger than zero, since the heat of
adsorption on a surface covered with zero layers is assumed to be larger than the heat of adsorption on a
surface covered with one or more layers of adsorbate). This will occur if p approaches the saturation
pressure at a given temperature. Thus, the reciprocal value of K2,adsorption corresponds to the saturation
pressure, and the equation can be rewritten as:
K1,adsorption p

Vadsorbed,total K 2,adsorption p0

Vadsorbed,monolay er   K1,adsorption  
1    1  p   1  p 
  K 2,adsorption  p0   p0 
   

The constant ‘c’ in the BET-isotherm therefore represents the ratio of the adsorption constant for adsorption
on the surface, which is not covered with adsorbed material, relative to the adsorption constant for the
adsorption on the surface covered with one or more layers of adsorbed material.

Figure 2.20 shows the principal shape of the BET-isotherm. At low pressure a behaviour
approximating the Langmuir isotherm is observed, if the adsorption constant on the surface
covered with 0 layers is much larger than the adsorption constant for the adsorption on the
surface covered with one or more layers. At high pressures a strong increase in the volume
adsorbed is observed, due to the predicted large extent condensation taking place on the surface.

42
Figure 2.20: Principal shape of the BET-isotherm

It should however be noted that in principal the BET-isotherm predicts physically unrealistic
situations, since it was assumed in the derivation that the physisorbed layer can grow up to
infinity. This means that a continuous growth of a liquid-like film may occur, which is
thermodynamically inconsistent. Furthermore, the assumption that the reciprocal value of the
adsorption constant for adsorption on the surface covered with one or more layers corresponds to
the saturation pressure leads to deviations of the measured isotherm from the ideal BET-isotherm
at high values for p/p0.

2.3.4.4 Alternative models describing physisorption


The inconsistency of the BET-isotherm has led to the development of a number of alternative
models based on statistical thermodynamics taking into account the energetic adsorbate-
adsorbate interaction, the energetic solid-adsorbate interaction, the pore size and geometry [40-
43].

2.3.5 Interpretation of N2-adsorption isotherms


Figure 2.21 shows a typical N2-adsorption isotherm. At low pressures (relative to the saturation
pressure), the true physisorption is observed, i.e. the built-up of the multi-layers of physisorbed
materials. With increasing relative pressure a strong increase in the amount of nitrogen adsorbed
is observed, due to capillary condensation. Upon decreasing the relative pressure, the emptying
of the pores is observed at a different pressure (see 2.1.2).

43
800

Vadsorbed , cm3(STP)/g
600
decreasing
pressure
400

200
increasing
pressure
0
0 0.2 0.4 0.6 0.8 1
p/p0
Figure 2.21: N2-adsorption isotherm at 77K on Davisil grade 636 [37]

Porous materials are often characterised using the physisorption of nitrogen at its normal boiling
temperature (77 K). Other gases, such as argon and krypton, can be used as well. These gases
have a lower vapour pressure at 77 K and can thus be used to characterise micro-porous and low
surface area materials. The use of gases with a lower vapour pressure would give a better
separation between the relative pressures at which physisorption and at which capillary
condensation occurs.

2.3.5.1 Determining the BET-surface area


The N2-adsorption isotherm comprises of the formation of a physisorbed layer on the surface and
capillary condensation. The BET-isotherm does not describe capillary condensation and therefore
only data points at relative pressures below the occurrence of capillary condensation can be taken
into account.

The BET-isotherm can be linearised to obtain the amount of gas adsorbed to form a monolayer:

p
p0

1

c  1 
p
 p  c  Vmonolay er c  Vmonolay er p0
Vadsorbed  1  
 p0 

Thus, a plot of the BET-isotherm in this form should yield a linear line. From the slope of the line
and the intercept, the volume of gas adsorbed to form a monolayer can be estimated.
int ercept 
1
slope 
c  1
c  Vmonolay er c  Vmonolay er
slope 1 1
c 1 Vmonolay er  
int ercept c  int ercept slope  int ercept

44
0.01

p/p0/(Vadsorbed *(1-p/p0)),
0.008

g/cm3 (STP)
0.006

0.004

. -5
0.002 intercept = 9.72 10
. -2
slope = 1.40 10
0
0 0.2 0.4 0.6
p/p0

Figure 2.22: Plot of the linearised form of the BET-isotherm of Davisil grade 636 [37] in the
range of relative pressures between p/p0=0.05-0.6 (data from Figure 2.21)

The BET-isotherm is typically applied in the relative pressure range of 0.05-0.3 to determine the
volume adsorbed to form a monolayer. The maximum value of the relative pressure is however
dependent on the type of material and lower values might be taken.

Thus, for the data from [44] the volume of nitrogen adsorbed to form a monolayer equals 70.8 cm3
(STP)/g and c equals 145. The value of the constant ‘c’ should be typically between 50 and 300
when using nitrogen at 77 K [45]. A high or even a negative ‘c’-value is indicative of micro-pores
and the BET-equation cannot be applied without appropriate modifications.

Seeing that the value of ‘c’ is relatively large, the linearised form of the BET-equation can be
approximated by:
p
p0 1 p
 
 p  Vmonolay er p0
Vadsorbed  1  
 p0 
This allows for the so-called single point BET-surface area. This introduces a slight overestimation
for the obtained volume for the formation of the monolayer, but the advantage is the speed of the
analysis.

Knowing the volume of the gas adsorbed to form a monolayer, the number of moles of N 2 forming
a monolayer can be estimated using the ideal gas law (1 mmol  22.414 cm3 (STP)). Thus, in this
example the number of moles of nitrogen forming a monolayer equals 3.16 mmol.

The BET-surface area can now be estimated if the area occupied by a single adsorbed molecule
is known. The mean area per molecule, , is given by
2
 M 3
  4  0.866   
 4  2  N  0.5 
 Av 

45
with M, the molecular weight, NAv, Avogadro’s number and  the density of the liquid adsorbate.
The mean area of an adsorbed N2-molecule is 16.2 Å2 (Kr (77K): 21.0 Å2; Ar (87K): 14.2 Å2; CO2
(273.15K): 17.0 Å2) [45]. Thus, Davisil grade 636 has a BET-surface area of 308 m2/g.

2.3.5.2 Pore size distribution – BJH method


The occurrence of capillary condensation allows the computation of a pore size distribution from a
physisorption isotherm. This is of course only possible, if capillary condensation takes place at a
measurable relative pressure. Thus, the method is limited to the determination of the pore size
distribution of meso-pores (the pore diameter of micro-pores is too small and capillary
condensation does not occur, whereas the diameter of macro-pores is too large, and capillary
condensation only takes place at pressures very close to the saturation pressure). The pore size
distribution can be calculated from both the adsorption-branch and the desorption-branch of the
hysteresis using the (modified) Kevin-equation (see 2.1.2).

In the model developed by Barrett, Joyner and Halenda [46] (BJH-method), it is assumed that the
walls in the pores are covered with a physisorbed layer with a thickness t, which upon reaching
the capillary vapour pressure is suddenly filled with liquid. The thickness of the liquid layer on the
surface increases thus suddenly from a thickness t to a thickness equal to the pore diameter.
Upon desorption the thickness of the liquid layer suddenly drops from a thickness equal to the
pore diameter to a thickness t.

The thickness of the liquid layer on the surface is thus crucial for the determination of the pore-
size distribution. Harkins and Jura [47] showed that in the region where a liquid film is formed on a
surface:
 p  A
log   B 
p 2
 0 Vadsorbed
The statistical thickness of the film is related to the volume adsorbed over the surface area. For a
range of solids and adsorbed materials it was found that:
0.5
 
 
 13.99  [45].
t 
 p 
 0.034  log  
 p 
  0 
Alternative models to calculate the thickness of the adsorbed film have been developed as well
[45].

The pore size distribution estimation according to the BJH-method requires a condition at which
the pores are completely filled. This is typically arbitrarily set at a relative pressure of 0.995. Upon
reduction of the pressure in the desorption process the volume adsorbed decreases. The amount
lost in each step represents the volume of the liquid in the pores removed, i.e. the volume of liquid
corresponding to the pore volume minus the volume remaining behind as a liquid layer on the
surface. The thickness of the layer is estimated e.g. with the correlation given by Harkins and Jura
[47]. The pore radius is estimated using the Kelvin equation
Vliquid L V
r  t 2 
RT  p 
ln 
 p0 

46
The same procedure can be applied to the adsorption branch. This method allows the estimation
of another surface area (BJH-surface area) by assuming cylindrical pores from which the average
pore length can be computed. Using the cylindrical geometry, a surface area can be obtained.

References:
[1] U. Wilkenhöner, D.W. Gammon, E. van Steen, Proc. 13th Int. Zeolite Conf., Montpellier
2001, (A. Galarneau, F. Di Renzo, F. Fajula, J. Vedrine, eds.), Stud. Surf. Sci. and Catal.
135 (2001), 368.
[2] E.L. Viljoen, M. Claeys, E. van Steen, unpublished results (2004)
[3] E.W. Washburn, Proc. Natl. Acad. Sci. USA 7 (1921), 115
[4] R.G. Jenkins, M.B. Rao, Powder Technology 38 (1984), 177
[5] S.P. Rigby, R.S. Fletcher, S.N. Riley, Appl. Catal. A: General 247 (2003), 27
[6] S.P. Rigby, R.S. Fletcher, S.N. Riley, Chem. Eng. Sci. 59 (2004), 41
[7] R. Pirard, C. Alié, J.-P. Pirard, Powder Technology 128 (2002), 242
[8] R.C. Weast, M.J. Astle, W.H. Beyer (Eds.), Handbook of Physics and Chemistry, 67th Ed.,
CRC Press, Boca Raton, F20-F35.
[9] J.C.P. Broekhoff, J.H. de Boer, J. Catal. 9 (1967), 8.
[10] S.Z. Qiao, S.K. Bhatia, X.S. Zhao, Microp, Mesopor. Mater. 65 (2003), 287
[11] A. Huertam O. Pizio, S. Sokołowski, J. Chem. Physics 112 (2000), 4286.
[12] IUPAC, J. Colloid Interface Chem.: Pure Appl. Chem. 31 (1972), 578
[13] W. Romanowski, Surf. Sci. 18 (1969), 373.
[14] J.C.W. Swart, P. van Helden, E. van Steen, J. Phys. Chem. C 111 (2007), 4998.
[15] P. van Helden, R.A. van Santen, E. van Steen, J. Phys. Chem. C (in press).
[16] H. Wise, J. Oudar, “Material concepts in surface reactivity and catalysis”, Dover
Publications Inc. Mineola, New York (2001)
[17] A. Titmuss, A. Wander, D.A. King, Chem. Rev. 96 (1996), 1291
[18] E. van Steen, M. Claeys, M.E. Dry, J. van de Loosdrecht, E.L. Viljoen, J.L. Visagie, J.
Phys. Chem. B 109 (2005), 3575-3577
[19] R. van Hardeveld, F. Hartog, Surf. Sci. 15 (1969), 189
[20] B. Ducourty, M.L. Occelli, A. Auroux, Thermochimica Acta 312 (1998), 27-32
[21] K. Foger, ‘Dispersed metal catalysts’, in “Catalysis, Science and Technology”, (J.R.
Anderson, M. Boudart, Eds.), Vol. 6, p.227, Springer Verlag, Berlin (1984)
[22] C.H. Bartholomew, R.B. Panell, J. Catal. 65 (1980), 390
[23] B.E. Spiewak, J.A. Dumesic, Thermochimica Acta 290 (1996), 43-53
[24] S.T. Zwane, MSc-thesis, University of Cape Town (2004).
[25] O. Beeck, Disc. Faraday Soc. 8 (1950), 118
[26] W.A. Brown, R. Kose, D.A. King, Chem. Rev. 98 (1998), 797
[27] Y.Y. Yeo, C.E. Wartnaby, D.A. King, Science 268 (1995), 1731
[28] J.T. Stuckless, N. Al-Sarraf, C.E. Wartnaby, D.A.J. King, Chem. Phys. 99 (1993), 2202
[29] I. Quiňones, G. Guiochon, J. Colloid Interf. Sci. 183 (1996), 57
[30] D.O. Cooney, Chem. Eng. Comm. 94 (1990), 27
[31] C.H. Yang, J. Colloid Interf. Sci. 208 (1998), 379
[32] T.C. Bromfield, D. Curulla Ferré, J.W. Niemantsverdriet, ChemPhysChem 6 (2005), 254
[33] P. van Helden, PhD thesis University of Cape Town (2008)

[34] P. van Helden, E. van Steen, J. Phys. Chem. C 112 (2008), 16505
[35] H. Conrad, G. Ertl, J. Kuppers, Surf. Sci. 76 (1978), 323
[36] M.E. Davis, R.J. Davis, ‘Fundamentals of chemical reaction engineering’, McGraw-Hill,
Boston (2003)
[37] S. Brunauer, L.S. Deming, W.S. Deming, E. Teller, J. Amer. Chem. Soc. 62 (1940), 1723

47
[38] J.H. de Boer, ‘The shapes of capillaries’ in ‘The structure and properties of porous
materials’ (D. Everett, F.S. Stone, Eds.), Butterworths, London (1958), p. 68, cited in A.J.
Lecloux, ‘Texture of catalysts’ in “Catalysis, Science and Technology”, (J.R. Anderson, M.
Boudart, Eds.), Vol. 2, p.171, Springer Verlag, Berlin (1981)
[39] S. Brunauer, P.H. Emmett, E. Teller, J. Amer. Chem. Soc. 60 (1938), 309
[40] J.P. Olivier, Carbon 36 (1998), 1469
[41] M. Kruk, M. Jaroniec, K.P. Gadkaree, Langmuir 15 (1999), 1442
[42] M.L. Ocelli, J.P. Olivier, J.A. Perdigon-Melon, A. Aroux, Langmuir 18 (2002), 9816
[43] P. Kowalczyk, A.P. Terzyk, P.A. Gauden, R. Leboda, E. Szmechtig-Gauden, G.
Rychlincki, Z. Ryu, H. Rong, Carbon 41 (2003), 1113
[44] A. Saib, MSc-thesis, University of Cape Town (2001)
[45] P.A. Webb, C. Orr, ‘Analysis methods in fine particle technology’, Micromeritics Instrument
Corporation, Norcross (1997)
[46] E.P. Barrett, L.G. Joyner, P.P. Halenda, J. Amer. Chem. Soc. 73 (1951), 373
[47] W.D. Harkins, G. Jura, J. Amer. Chem. Soc. 66 (1944), 1366

48

S-ar putea să vă placă și