10

Adsorption

Afldr~ De.<;clIll11lp.<;

Suphie Julliufl

When a solid is placed in contact with a gaseous or liquid compound, forces

are exerted between the surface of the solid and the compound causing bonds
to be formed. This is the phenomenon called adsorption. The bonds can vary
in strength depending on the nature of the compounds and of the solid
involved. They can cause selective adsorption of certain components in mix-
tures. As we will see later, this selectivity can also be the result of steric hin-
drance or kinetic phenomena. A number of microporous solids with a large
specific area, and consequently with a large adsorption capacity, can be used
to take advantage of this highly specific adsorption selectivity in order to frac-
tionate mixtures.
In the last 30 years, adsorption has become a key separation technique in
industry and particularly in the oil and gas industry. There are two prime areas
of application for adsorption. The first is ultrapurification of feeds (obtaining
the non-adsorbed fraction with very high purity), and the second is in frac-
tionating mixtures that are difficult to separate by distillation: mixtures of iso-
mers with close boiling points, of compounds with different structures and of
gases (an alternative to cryogenic distillation).
The penetration of the adsorption process in industry is due first of all to
the development of adsorbents exhibiting very specific selectivity, in particu-
lar synthetic zeolites and molecular sieves. It is also because of the effort
made to develop efficient ways of implementing adsorbents that are suited to
the different applications contemplated and use highly optimized regeneration
procedures. These improvements have benefitted from the progress in adsorp-
tion theories made in the seventies allowing adsorption equilibria for mixtures
to be predicted on the basis of adsorption isotherms for pure substances.
More recently, the contributions made by modelling and dynamic simulation
have enabled hydrodynamic, kinetic and thermal aspects to be better taken
into account in order to describe the way adsorbent beds work.
526 Chapter 10. ADSORPTION

Table 10.1 which presents an attempt to classify the most important

adsorption processes used in the oil industry shows that there is a wide diver-
sity. They can be classified according to a number of criteria:
type of adsorbent used: zeolite, activated carbon, carbon molecular
sieve, alumina, silica gel;
type of selectivity involved: equilibrium, steric or kinetic selectivity;
state of the feed during processing: gaseous or liquid phase;
composition of processed feeds: purification processes (by adsorption of
impurities present in low concentrations), mass separation processes
(feeds including several per cent of adsorbable products);
regenerative or non-regenerative process;
regeneration method used: by temperature increase (TSA), by pressure
decrease (PSA), by stripping, by displacement with an adsorbable com-
pound or by a combination of several techniques;
continuous process (moving bed, simulated moving bed) or discontinu-
ous process (rapid or slow cyclic).
Table 10.1 shows that in practice only some solutions are compatible with
each other. Other criteria such as stability of the adsorbent, presence of trou-
blesome impurities, incorporation in a sequence of processes can also play an
important part in the choice of the technique to be implemented for a given
separation.
The objective of the chapter will be to point out the rules that allow a
choice to be made among these techniques, by presenting the different types
of adsorbents, the theoretical bases of adsorption, the main implementation
techniques and the industrial applications in the oil industry. For further
details readers can refer to works giving an overview of adsorption (Ruthven,
1984; Yang, 1987).
In the presentation of adsorption separation processes later on, "extract"
means the fraction that includes the most adsorbed product(s), and "raffinate"
means the fraction containing the least adsorbed product(s).

10.1 Adsorbents
Adsorbents used industrially are generally synthetic microporous solids: acti-
vated carbon, molecular sieve carbon, activated alumina, silica gel, zeolites
and bleaching clay. They are usually agglomerated with binders in the form of
beads, extrudates and pellets of a size consistent with the application that is
considered.
The most important physical characteristics of adsorbents are the pore
volume, pore distribution and specific area. For implementation it is also
important to know the packing density DP, the compressive strength and the
abrasion resistance.
 Chapter 10. ADSORPTION 527

I I

nC4=/iC4= nC4= Gas Zeolite 5A nC4= Displacement Cyclic

(shape)
Olefins/paraffins Olefins Liquid Zeolite X Olefins Displacement SCC
(equilibrium)
C8 Aromatics Paraxylene Liquid Zeolites X,Y Paraxylene Displacement SCC
(equilibrium)
Glucose/fructose Fructose Liquid Zeolite X Fructose Displacement SCC
- (equilibrium)
- Table
10.1 Main separations by adsorption (SCC means Simulated Counter-Current,see
Section 10.5.3.2).

The pore volume and distribution in pore diameter are determined by

helium and mercury porosimeters and by nitrogen adsorption (or desorption).
Because of its small size, helium can access all the voids and is used to deter-
mine the structural density DST of the solid by considering that its adsorption
is negligible.
A mercury porosimeter is used to measure the volume of mercury pene-
trating in the solid versus the hydrostatic pressure, and thereby to calculate
the grain density DG and the skeleton density DSK.
At atmospheric pressure, mercury fills only the interparticle volume (mea-
surement of grain density). When pressure is increased, it penetrates into
smaller and smaller pores according to the equation below:

(10.1)
528 Chapter 10. ADSORPTION

with:
r radius of the pore
P pressure
G interfacial tension (0.48 N/m)
8 angle of contact (180O)
This equation can be reduced to r = 9 600/P with r in nm and P in bar. For
example, for an intrusion pressure of 4 200 bar, the highest normally used, the
equivalent pore radius is 2.4 nm. In this way the macro- and meso-pore volume
and the skeleton density of the solid are measured.
Adsorption (or desorption) of nitrogen at 77 K is used to characterize the
size distribution of the pores between 10 and 150 8, via the capillary conden-
sation phenomenon, on the basis of the Kelvin equation. The adsorption
isotherm is also used to determine the Dubinin volume (volume of micropores
for zeolites) and the specific area (BET method)(Gregg, 1982).
Sometimes a more detailed characterization of the micropore size and vol-
ume is necessary, particularly for applications based on molecular sieving
(zeolites, molecular sieve carbon). Here adsorption measurement by probe
molecules of different sizes is used.
Pores of adsorbent are often classified according to their diameter as
macropores (> 500 A), mesopores (20 to 500 .k) and micropores (< 20 A).
Table 10.2 and Figures 10.1, 10.2 and 10.3 (Yang, 1987) present typical char-
acteristics of the major industrial adsorbents.

10.1.1 Activated Carbon

Activated carbon is manufactured from carbon-bearing materials such as coal,
oil coke, coconut shells, peat, etc. After an initial treatment and shaping, there
is a carbonization step at 400-500°C to eliminate most of the volatile matter,
then an activation step by partial gasification at 800-1 000°C to develop the
porosity and the surface area. A number of different gasification agents are
used such as CO,, steam and stack gas. The activated carbon manufactured in
this way is mainly used in adsorbing gases and vapors.
Another activation process which generally yields powdered activated car-
bon, used especially in water treatments, utilizes inorganic additives such as
phosphoric acid, zinc chloride and potassium sulfide.
The specific properties of activated carbon are due to its surface which,
contrary to that of the other main adsorbents, is non-polar or only slightly
polar, thereby giving little affinity for water (Fig. 10.2). In contrast, its large spe-
cific area (300 to 3 000 m2/g) allows the adsorption of large amounts of non-
polar products even in the presence of water. This explains its priority use for
adsorbing organic products in water and wet gas treatment processes
(Cookson, 1978).
The surface polarity can, however, be increased by treatments such as oxi-
dation, ion exchange, impregnation, etc.
 Chapter 10 ADSORPTION 529

BET surface Volume of Diameter of

Main
Adsorbent micropores micropores
applications
:cm3/100 g) (A)
50-60 10-30 Organics/air,
organics/water, CH,, CO/H,
Molecular 1 200-2 000 15-25 3.5-6.5 WN,
sieve carbon
Activated alumina 250-350 20-35 10-80 H,O/gases and liquids
Silica gel 700-850 40-50 22-26 H,O/gases
Zeolites 3A (K) 500-600 28 3 H,O/olefins, alcohol
Zeolites 4A (Na) 500-600 28 4 H,O, CO,/natural gas, air,
solvents
28 5 H,O, CO,, H,S/natural gas,
N,/O,, CH,, CO/H,,
nlisoparaffins
30 10 H,O, H,S, R.SH/natural gas,
LPC,H20, COJair
30 10 Paraxylene/C, aromatics

Table

PY Characteristicsand applications of main adsorbents.

LIVE GRAPH
Click here to view

0.10

.
6 0.08
m
Y

'0

8
0
m
Ea 0.04
0.06 / Activated carbon

E -
a 0.02 Silica gel

- Figure
10.1 Methane adsorption isotherms at 25°C on silica gel, zeolite 5A and activated
carbon (Ah'er Yang, 1987).
 LIVE GRAPH
Click here to view
530 Chapter 10. ADSORPTION

40
c
h

c
a,
e
8
U
ca 30
Y
P)
0
0
7-

ci
20
s
U
ea,
8
-
u
m
c
10

a
aE
0
0 20 40 60 80 100
Relative humidity (“A)

Figure
10.2 1 Water adsorption isotherms at 25°C on granular alumina (A), spherical
alumina (B), silica gel (C), zeolite 5A (DJ and activated carbon (E). 100%
relative humidity corresponds to a water partial pressure of 23.6 Torr (AFter
Yang, 1987).

Pore distribution is generally of the polymodal type, with the largest pro-
portion of the pore size being micropores from 10 to 25 8,for activated carbons
used in the gaseous phase (Fig. 10.3). For liquid phase applications, slightly
larger micropores (around 30 A) are used because diffusion is slower than in
the gaseous phase. These micropores lead into open macropores which allow
the adsorbate access to the entire particle.

10.1.2 Carbon Molecular Sieve, MSC

Carbon molecular sieves are activated carbon whose micropore size has been
adjusted by specific treatments to allow separations on the basis of molecular
sieving (pore diameter from 3.5 to 6.5 8,)(Juntgen, 1977). It can be prepared by
a large number of methods such as carbonization of polymers, cellulose,
anthracite, etc.
Commercial products are the result of modifying the coal-based activated
carbon manufacturing procedure, particularly by depositing coke at the
entrance of the micropores so as to narrow down the access diameter
(Fig. 10.4). These carbonaceous deposits are obtained by cracking hydrocar-
bons such as methane and isobutane under carefully controlled conditions.
 Chapter 10. ADSORPTION 531

Activated carbon

Figure
10.3 Pore size distribution of main adsorbents.

A B

-
Figure
10.4 Molecular sieve carbon (MSC) manufactured by Bergbau-Forschung (Afier
Jiintgen, 1977). A. N2 type MSC with a pore opening o f around 5 A, manufac-

The main application of MSC today is in the production of nitrogen from air.
The MSC pore diameter of about 5 8, that is used allows oxygen to be more
rapidly adsorbed than nitrogen.
532 Chapter 10. ADSORPTION

10.1.3 Activated Alumina

Activated alumina is produced by thermal dehydration of aluminum oxide
Al(OH),. Depending on the activation temperature (400 to 800°C), more or less
amorphous products with a surface area of 250 to 350 m2/g can be prepared.
An impure variety is also manufactured by thermal activation of bauxite.
Activated alumina has a network of micropores from 10 to 50 A in size
(Fig. 10.3), but a very large proportion of the pore volume corresponds to
larger pores. It has great affinity for water and is therefore used mainly in dry-
ing gases (Fig. 10.2). It is regenerated thermally at 200-250°C. For very severe
drying, zeolites are preferred, however regeneration is more difficult.

10.1.4 Silica Gel

Silica gel is prepared by adding a dilute inorganic acid (HCl or H2S03 to a
sodium silicate solution (Her, 1979). A gel is formed which is then washed,
dried and roasted. The product obtained is in the form of translucent grains
containing about 4 to 6% residual water, and is basically made up of a single
layer of hydroxide groups bonded to the surface silicon atoms in the form of
silanol groups Si-OH.
There are two commercial varieties: common silica gel (surface area
700-850 m2/g,average pore diameter 21-26 A) and what is called a lowdensity
product (300-350 m2/g, average pore diameter 100-150 A).
Like alumina, silica gel is a good adsorbent for drying. It is characterized by
great adsorption capacity at medium water partial pressure (Fig. 10.2) and a
low heat of adsorption (46kJ/mole of water) which allows it to be easily regen-
erated by heating to 150-2OO0C, instead of 250-300°C required by zeolites
(heat of adsorption of water: 73.4 kJ/mole of water).

10.1.5 Zeolites (Molecular Sieves)

Zeolites are microporous crystals of alkali or alkali earth aluminosilicates with
the following stoichiometry:
[(AlOd, (SiOdJ, wH,O
4 , r I

with R being the valence of the cation M,x and y integers with y/x equal to or
greater than 1, and w the number of molecules of water per cell.
The crystal lattice is built of a three dimensional arrangement of tetrahe-
drons [SO4] and [AlOJ connected by their apexes (Fig. 10.5a). These
sequences form polyhedrons that are arranged according to a given symmetry
to make the crystal. The void delimited by the aluminosilicate skeleton is orga-
nized into a network of cavities connected by pores of uniform size. The effec-
tive pore diameter, variable depending on the type of zeolite, is of the same
order of magnitude (3 to 10 A) as that of most organic molecules. This explains
the name "molecular sieve" given to these adsorbents. The M cations and the
water molecules make up the extra-skeletal species. The cations offset the
 Chapter 10. ADSORPTION533

Structure of A, X and Y type zeolites. a: Assembly of tetrahedrons in a zeolite.

b: Sodalite cage. c: A type zeolite. d: X and Y type zeolite.

negative charges generated by the aluminum atoms of the skeleton and can be
involved in exchanges to adjust adsorption properties. Their number depends
on the Si/Al ratio of the skeleton. The water occupying the pore network can
be desorbed by vacuum heating or dry gas stripping, thereby freeing a micro-
pore volume which accounts for 20 to 50% of the volume of the crystal.
There are a large number of natural zeolites (chabazite, mordenite, fauja-
site, etc.) and more than 150 have been synthesized (Barrer, 1978). We will dis-
cuss only the zeolites used in separations which are of the A (3A, 4A, 5A), X
and Y types.
534 Chapter 10. ADSORPTION

For these three types of zeolites, the base structural unit is a truncated
octahedron called a sodalite cage or p cage (Fig. 10.5b). In type A zeolites,
sodalite cages are interconnected tetrahedrally by four-sided prisms, whereas
in X and Y types they are connected via hexagonal prisms (Figs. 1 0 . 5 ~ and d).
In type A, the diameter of the central cavity (acage) is 11.4 A with eight-oxy-
gen-atom windows 4.4 A in diameter. In types X and Y, the cavity (12.5 A in
diameter) and the 12-atom openings (7.5 A in diameter) are larger, hence a
greater adsorption capacity and the possibility of adsorbing larger molecules.
The nature of the compensating cations is an optimization parameter of
adsorbent selectivity. For instance, the A zeolite is synthesized in the Na form,
which exhibits an effective pore diameter of 3.8 A (4A type). By exchanging K
for Na the 3A type can be prepared, which exhibits a smaller pore opening
because of the larger size of the K+ ion. This yields a zeolite that allows water
molecules to enter but excludes olefins and is used for drying olefinic cuts.
In contrast, by partially exchanging Ca for Na the 5A type is prepared which
exhibits a larger pore opening. It can adsorb linear paraffins and exclude
branched paraffins.
However, numerous applications are not based on this molecular sieving
effect but on the fact that certain compounds are adsorbed more strongly than
others. For example, nitrogen is adsorbed in preference to oxygen by the 5A
zeolite (by a factor of three) due to its quadripolar moment which creates a
very strong bond with the polar surface.
X and Y zeolites, which differ only in the Si/Al ratio (1 to 1.5 for X, 1.5 to 3
for Y), are also prepared in the Na form. An exchange of alkaline earths for Na
is also used to adjust adsorption selectivity, which in this case is based on
adsorption energy (separation of aromatics for example). The Y zeolite can be
modified by aluminum removal to Si/Al ratios of 50 to 100 to give it a
hydrophobic property. The zeolite can then be used instead of activated car-
bon in recovering organic vapors from wet air with the advantage of easier
regeneration by hot air.
These zeolites are produced by hydrothermal crystallization between 20
and 175°C from freshly prepared aluminosilicate gels. The crystals, generally
from 1 to 5 micrometers, are washed then compacted and shaped with an inor-
ganic bonding agent (clay) and finally calcinated. The final product is in the
form of beads or extrudates.
Zeolites are used in a great many purification and separation processes for
gaseous and liquid mixtures which will be described later on: drying, desulfu-
rization, decarbonation, hydrogen purification, 0,/N2 separation, nlisoparaffin
separation, aromatics separation.

10.1.6 Natural and Activated Clays

These adsorbents are designed for use in finishing treatments for lube oils to
improve their quality in different ways: bleaching, removing readily oxidizable
products, neutralization (for acid-refined lube oils), etc. A distinction is made
between natural and activated clays.
 Chapter 10. ADSORPTION535

The first category, also called fuller’s earth, consists of natural clays (atta-
pulgites, some montmorillonites or bauxites) with bleaching capacity once
they have been dried to a suitable degree of humidity which depends on their
nature.
The second category is made up of clays of the montmorillonite family that
have undergone an acid (sulfuric or hydrochloric) treatment. The treatment
eliminates alkali and alkaliearth cations and gives the clays their adsorbent
power as well as a catalytic activity that can be used to remove olefins.
Natural clays are mainly used in percolation and are regenerated by roast-
ing the adsorbed matter in the presence of large amounts of excess air.
Meanwhile, activated clays are employed in redistillation clay contact pro-
cesses and are disposed of after use.

10.2 Adsorption Thermodynamics

10.2.1 Adsorption of Pure Substances
In order to characterize adsorption phenomena, the adsorbed quantity is
expressed in the following way:
(10.2)

where q can be expressed in mol/g or in g/g (i.e. mol/kg or kg/kg in the SI sys-
tem of units). At a fixed temperature this is called an adsorption isotherm.
Adsorption phenomena are schematically divided into two main categories
depending on the type of bonds between the adsorbate molecules and the
solid: physisorption and chemisorption.
Chemisorption involves electron transfers and can be considered as a
chemical reaction in a two-dimensional space.
Physical adsorption occurs when the forces generated by the solid’s sur-
face are physical in nature (van der Waals forces, electrostatic forces).
Generally speaking, physical adsorption involves an attraction between the
adsorbate molecules and those making up the surface of the adsorbent solid.
Interactions between adsorbed molecules can occur, especially when the
molecular layer is dense enough. The layer of molecules can also interact with
non-adsorbed molecules. A second adsorbed layer can be formed, then a
third, and so on. Physical adsorption is therefore potentially a multilayer phe-
nomenon. An increase in temperature causes more intense molecular agitation
and consequently promotes molecule desorption. Moreover, a decrease in the
partial pressure of the substance that is to be adsorbed in the external phase
(or a decrease in concentration) also promotes desorption. Physical adsorp-
tion is therefore a phenomenon which takes place at moderate temperatures
and is totally reversible. Additionally, the heat involved is relatively small and
because of this the processes are more readily reversible. In separation pro-
cesses by adsorption, only physical adsorption is considered.
536 Chapter 10. ADSORPTION

Most isotherms observed to date can be described according to one of the

six IUPAC classification types (Sing, 1988)Fig. 10.6).
Reversible type I isotherms are concave with respect to axis P/Pa and the
adsorbed quantity approaches a limit when P/Pa approaches 1. These
isotherms are found for microporous solids that have a relatively small out-
side surface area (e.g. activated carbons, zeolites and certain porous oxides),
and the limiting adsorption value is related to the micropore volume rather
than to the outer surface area.
Reversible type 11 isotherms are conventionally obtained with non-porous
or macroporous solids. They are representative of multilayer adsorption.
Point B, the departure point of the linear part of the middle of the isotherm,
corresponds to single layer packing and to the beginning of multilayer adsorp
tion.
Reversible type Ill isotherms are convex with respect to axis P/Pa over all
the domain and therefore do not exhibit any point B. These isotherms are not
frequent and are found for systems where adsorbent-adsorbate interactions
are strong.
The main characteristic of type IV isotherms is the hysteresis correspond-
ing to capillary condensation in the mesopores associated with adsorption on

I A1
L
1
Figure
10.6 IUPAC classification o f
isotherms (Sing, 1988).

Relative pressure (PlP")

 Chapter 10. ADSORPTION537

the surface, as for type I1 isotherms. Type IV is frequent for numerous meso-
porous industrial adsorbents.
Type V isotherms are rare. They are found for certain porous solids in sys-
tems where adsorbent-adsorbate interactions are weak.
Step-wise type VI isotherms are characteristic of non-porous surfaces
where adsorption occurs in stages for each layer. The height of the steps rep-
resents the adsorption of a layer in the simplest case and is constant for two
or three layers. The slope of the steps depends on temperature conditions.
A lot of theories and models have been developed to interpret these differ-
ent types of isotherms, with resulting equations that can be used to predict the
quantities adsorbed based on a minimum number of experiments.
A detailed discussion of these models is found in the work by Gregg and
Sing (1982).
a. The Langmuir Approach
The Langmuir approach stems from kinetic considerations and was developed
in 1918 by assuming that the adsorption system is a dynamic equilibrium
where the evaporation rate (desorption) is equal to the condensation rate
(adsorption). The Langmuir equation remains the most commonly used cor-
relation in adsorption separation processes, since it allows type I and I1

.
isotherms to be represented. The hypotheses it is based on are as follows:
Molecules are adsorbed on a finite number of well defined sites.

..Each site can adsorb only one single molecule (single layer).
All the sites are equivalent from an energy standpoint.
There is no interaction between close neighbors.
The Langmuir equation expresses the fractional surface coverage (versus
the pressure and a constant) in the following form:

(10.3)

At low pressure it is reduced to the Henry equation, 0 = BP and at high

pressure, 0 approaches 1.
Taking more realistic considerations into account, numerous modifications
of the equation have been proposed. An example to be mentioned is the
Langmuir-Freundlich equation, which uses the Langmuir model and includes a
correlation known under the name of the Freundlich isotherm: q = kP1/".The
Langmuir-Freundlich equation is written:

(10.4)

The Langmuir and Freundlich isotherms contain two adjustable parame-

ters, while the Langmuir-Freundlichisotherm has three. This makes it superior
to a certain extent for correlating data in a wide range of pressures and tem-
peratures.
538 Chapter 10. ADSORPTION

At temperatures lower than the critical gas temperature, type I1 isotherms

predominate over type I. This indicates that multilayer construction begins
before the single layer is completely formed. To take this phenomenon into
account, a two parameter isotherm model was developed by Brunauer, Emmet
and Teller (1938), which is basically an extension of the Langmuir equation. In
this model, each site can adsorb 0 to i molecules which are not mobile on the
surface, and the main assumption lies in the fact that the heat of adsorption is
constant after the first layer and equal to the heat of liquefaction. The BET
equation is expressed as indicated below:

(10.5)

Even if the BET equation is seldom used to fit experimental data, it still
remains the most widely used implement for measuring the surface areas of
microporous solids when the size of the adsorbed molecule is known (for
example 16.2 A* for the nitrogen molecule at 77 K). When the vapor pressure
and the amount adsorbed for different pressure values are known, the amount
adsorbed q1 corresponding to an adsorbed single layer is found by plotting
P / q ( P - P“) versus PIP“ and by assuming the surface is covered with spherical
molecules. In this way the value of the solid’s surface area can be found out.
b. The Gibbs Approach
In the Gibbs approach, the description of adsorption equilibrium (adsorbed
phase-gaseous phase) is patterned after the approach used for vapor-liquid
equilibrium, by using the concepts of spreading pressure and of adsorbed
phase. The adsorbed phase hypothesis consists in considering all the
adsorbed molecules as a homogeneous phase with the adsorbent being com-
pletely inert. As a result, it is thermodynamically logical to apply the equality
of chemical potentials at equilibrium.
The spreading pressure K is defined simply in the case of a two dimensional
film (Adamson, 1976) as being the difference in surface tension between the
adsorbent alone and the adsorbent covered with adsorbate, in equilibrium
with a gaseous phase at pressure P. For a volume adsorption, the spreading
pressure is defined by means of the following thermodynamic relations:

where U the internal energy; H,the enthalpy; F, the Helmholtz free energy and
G, the Gibbs free energy are all expressed in J. A is the surface area of the
adsorbent in m2. K is expressed in N/m.
The Gibbs-Duhem equation is therefore written as follows at constant tem-
perature:
- A dn + n dp = 0
p being the chemical potential (Jlmol) and n the number of moles adsorbed
(mol).
 Chapter 10. ADSORPTION 539

At equilibrium, with a gaseous phase assumed to be ideal, the equality of

chemical potentials gives:

This relation, known as the Gibbs adsorption isotherm, allows transition

from an equation of state of the adsorbed phase to an adsorption isotherm.
There are two possibilities. First, we have an n(P) isotherm that can be inte-
grated and we can calculate the equation of state of the adsorbed phase relat-
ing n to n. Second, by stating an equation of state model of the adsorbed phase
by analogy with the equations used in vapor-liquid equilibrium, we calculate
the adsorption isotherm with the Gibbs isotherm in the form:
RT n (P)
(10.8)
P A
If for example we apply the ideal gas equation to the adsorbed phase,
nA = nRT, we get the following from the previous relation:

( = hence 7c = KP

(10.9)

and we get q proportional to P at constant temperature, i.e. Henry’s law.

In the same way, a number of equations of state have been used (van der
Waals, Volmer, Viriel) and have resulted in new isotherm formulations.

c. The Potential Theory

A third approach was developed by Dubinin et al. (1960, 1965) from a premise
initially proposed by Polanyi (Adamson, 1976). The isotherms resulting from
this theory are used in measuring micropore volumes and obtaining the char-
acteristic curve, which is assumed to be independent of temperature.
Polanyi was the first to represent the surface force field by equipotential
contour lines above the surface, with the space between two equipotential sur-
faces corresponding to a constant adsorbed volume. The molecules adsorbed
in the micropores are considered as being in a phase liquified by the adsor-
bent’s force field and so the properties of this liquid phase differ from those of
a saturated liquid phase. The free energy difference between the adsorbed
phase and the liquid phase at the same temperature is called adsorption
potential E and is calculated in the following way:

E = RTIn (7) (7)

x RTIn (10.10)

where f is the saturation fugacity and f the fugacity in the adsorbed phase.
E is expressed in J/mol.
540 Chapter 10. ADSORPTION

For a given adsorbate-adsorbent system, there is only one relation, inde-

pendent of temperature, between the adsorbed volume W (m3/kg) and the
adsorption potential E:
w = f (&)
This relation, whose function is not specified, is called the characteristic
curve and can be used to represent equilibrium data in a wide range of tem-
peratures and concentrations by one single curve. The characteristic curve
can be established from a reference temperature and allows the adsorption
isotherms to be predicted at other temperatures for the same adsorbate-
adsorbent system. Based on the following relation:
W=nV,
where V, is the molar volume of the adsorbate (m3/mol), to get the charac-
teristic curve, nV, has to be plotted versus RT In (P/P).An example of this
type of representation is given in Figure 10.7.
The method undeniably has advantages related to its predictive feature,
but it does also have three important drawbacks:
The expression of the characteristic curve can not be reduced to Henry's
law for low concentrations and this makes it inconsistent from a therme
dynamic standpoint.
The methods of calculating the molar volumes of the adsorbed phase
should be used with care.
The assumption that the characteristic curve is independent of tempera-
ture is not always verified, particularly for polar adsorbates.
Dubinin has stated a semiempirical function to represent isotherms
according to this theory, which is formulated as seen below:

w= woexp (- $) (1 0.1 1)

where Wo is the micropore volume of the adsorbent (m3/kg) and p is an affin-

ity coefficient that characterizes the adsorbate's polarizability, or:

(10.12)

where:
D = kR2T2/p2 (D is dimensionless)
This is known as the Dubinin-Radushkevich equation and the coefficient p
is taken as a translation coefficient that allows all the characteristic curves of
different adsorbates to be reduced to one single curve per adsorbent. The
Dubinin-Radushkevich equation has been used successfully on zeolites and is
especially useful in determining the micropore volume W, that all experimen-
tal gas isotherms converge toward. The problem of extending this theory to
the supercritical domain (impossible to calculate the saturation pressure and
the molar volume in the adsorbed phase) has been discussed by numerous
authors and Table 10.3 indicates some of the solutions proposed.
 Chapter 10. ADSORPTION 541
p
I
L

Y
h t -

Y
0

0
0
0
?
0
0
-0 0
O1 2
0
0
L 0
?
0
0
0
z
.oc 0
0
F- 0
co
0
0
0
(D
0
.O
7
I
0
0
0
I
d
0
0
0
cu
0
542 Chapter 10. ADSORPTION

Calculation of Calculation of
Authors Curve plotted molar volume in saturation
adsorbed phase pressure/fugacity
T < T,: T < T,:
same as for saturated liquid vapor pressure
T > T,: T > T,:
take van der Wads b con. = 0.4 T/b
stant @: van der Waals constant)

Dubinin (1960, nV, versus Same as Polanyi T < T,:

1965) RT vapor pressure
-In PIPa T > T,:
Vm = TtP"

Toth (1962) nV, versus Same as Polanyi T < T,:

RT vapor pressure
-In P I P a T > T,:
Vm extrapolation of Clapeyron
I I equation

(1950)
I Same as for saturated liquid T < T,:
vapor pressure
T > T,:

Maslan et al.
I ",nV, versus Calculated by equation 01
extrapolation of Antoine
equation
T < T,:
(1953) RT state vapor pressure
-In f / P T > T,:
Vm extrapolation
~

Grant et Manes nV, versus Same as for saturated liquid T < T,:
(1964) RT at its normal boiling point vapor pressure
-In f / P T > T,:
Vm extrapolation
Cooket I nV,,,versus T < T,: T < T,:
Basmadjan same as for saturated liauid vapor pressure
(1964) T > T,: T > T,:
tangent to V , versus In T extrapolation by equation of
state using Vvalue
T < T,: T < T,:
(1980) = Vat normal boiling point Kirchhoff equation
-In f/P T,<T<T,: T > T,:
= RTJSP, extrapolation of this equa-
tion
Table
Calculating the molar volume and the saturation pressure to plot the charac-
teristic curve.

10.2.2 Mixture Adsorption

The models and correlations describing competitive adsorption are very
important in designing separation units. They need to be capable of predicting
the amounts adsorbed for each of the mixture components on the basis of pure
substance isotherms in the pressure and temperature domain of the process.
 Chapter 10. ADSORPTION 543

Equations of the Langmuir type can be used; they can easily be extended
to mixtures. To make for an easier presentation, the method will be given for a
binary mixture in the following discussion.
The fractional coverage for component 1 is given by the formula below:
BlPl (1 0.13)
0, =
1 + BlPl + BzP2
with pirepresenting the partial pressure of component i.
The fractional coverage of the second component is obtained in the same
way.
To calculate the amount of each component adsorbed, all that has to be
done is to use the relation 0; = 9i/9i,max. However, Broughton (1948) and
Kemball et al. (1948) have shown that to preserve thermodynamic consistency
for the Langmuir equation, it is necessary for 9i,max for all components.
= 9j,max
The Langmuir-Freundlich model can be applied to mixtures in the same
way:
B;pf.Int
0;= (10.14)
1 +cBjp,!lni
j=1
n

The BET theory has also been applied to mixtures, but the complexity of
the resulting equations has restricted its use.
Adsorbed Solution Models are also used. The Gibbs formalism was applied
to mixtures by Myers and Prausnitz in 1965. This led to relation (10.15) below
by writing the equality of chemical potentials in the two phases for each of the
components, using the activity coefficients y;, similar to those for the liquid
phase:
Py;= P?(n)y;x; (1 0.15)
In this relation, Pf(7c) represents the equilibrium pressure of component i,
pure and adsorbed at temperature T and spreading pressure 7c of the mixture
(i.e. the pressure that would give the same adsorbed quantity of i if it were the
only component in contact with the adsorbent).
The most well known model is the Ideal Adsorption Solution (IAS) which
represents the adsorbed phase as an ideal phase (y; = l), thereby allowing a
pseudeRaoult law to be obtained. By using this hypothesis it is possible to cal-
culate mixture adsorption from pure substance isotherms. For a binary mix-
ture, the following equations can be obtained:
0,=PT X1 (1)
Py2=pzx2 (2)
nsi = ~1 ( P f ) RT (3)
nsi = ~2 (Pz) RT (4)
x 1+ x , = 1 (5)
544 Chapter 10. ADSORPTION

Equations 3 and 4 are given by the pure substance isotherm model n = f(P*).
The system obtained has 5 unknowns ( x ~x2, , PT, P;,n) and 5 equations that
can be solved.
In some cases, wide deviations from ideality have been encountered when
the IAS theory has been used for some mixtures. This is why a number of
authors have applied the conventional solutions usually employed in mod-
elling liquid phases to calculate activity coefficients. The Margules, van Laar,
Wilson and UNIQUAC equations have been used successfully by Costa et al.
(1981) to calculate binary and ternary mixtures exhibiting wide deviations
from ideality. It should, however, be remembered that these methods are not
predictive and so it is necessary to have experimental data on mixtures.
The Dubinin-Radushkevich equation has been extended to mixtures by
Bering (1963, 1965) in the following form:

n, = Cn;= -
WO (10.16)
c x ; v,;

where:
Vmi is the molar volume of component i in the adsorbed phase (m3/mol)
n, is the total number of adsorbed moles (mol/kg)
This relation is not sufficient, however, and further equations must be used
to describe the system (Bering, 1963). This semi-empirical correlation can be
useful for non-polar adsorbents.
All these methods come up against the limits of a representation of adsorp-
tion equilibrium as an equilibrium between two homogeneous phases, which
does not take the heterogeneity of the solid surface into account. Other rep-
resentations are now being studied which take the energy of each adsorption
site into account and use molecular modelling, but their implementation in
process simulation is far from practicable for the time being. Semi-empirical
correlations of the Langmuir or Langmuir-Freundlich type remain the most
commonly used.

10.2.3 Heat of Adsorption

Adsorption is always accompanied by a release of heat. In physical adsorption
it is equal to the adsorbate’s latent heat of condensation plus the adsorbent’s
heat of wetting, The heat of wetting comes fully into play only for the first
molecules adsorbed. The heat of adsorption is seen to decrease as adsorption
proceeds and finally approach the heat of condensation. It is important to
know this value in order to simulate processes since it is involved in enthalpy
or heat balances.
A distinction should be made between three different values:
the total heat of adsorption for a specified adsorbed amount;
 Chapter 10. ADSORPTION545

the differential heat of adsorption (dh,,> related to adsorption of a

gas molecule at a given temperature and fractional coverage:
d On P)/dT = dhads/RT2;
the integral heat of adsorption.

10.3 Transfer Phenomena in Adsorbers

Virtually all adsorption separation processes take place in fixed beds (adsor-
bers) containing the shaped adsorbent (beads or pellets). Dynamic modelling
of the phenomena involved in adsorbers must take into account: the variation
in interstitial velocity (u), concentration (9and temperature (T), versus time
(t) and the position in the bed (z).
The models used must account for mass and heat balances and equilibrium
isotherms as well as they can for obvious economic reasons. The resistance to
mass and heat transfer both inside and outside the adsorbent grains is of great
importance for process feasibility. Any model describing an adsorber needs to
include:
an adsorption isotherm model;
a mass and a heat balance in the fluid phase;
a mass and a heat balance in the adsorbent pellets.
All the equations are coupled together of course and the boundary condi-
tions are written for each special case (TSA, PSA, etc.). The equations below
are written for one single adsorbable component.

10.3.1 Balance Equations

For a spherical pellet, the equilibrium model gives:

where:
T(R,t) represents the temperature at a time t and a position R in the pel-
let (K)
C(R,t) represents the concentration at a time t and a position R in the pel-
let (mol/m3 of fluid)
q’(R,t) represents the amount adsorbed at a time t and at a position R in
the pellet (mol/m3 of pellet)
The notation q’(mol/m3 of pellet) is to differentiate it from
q(mol/kg of pellet) (Eq. 10.2). The two values are related by q’ = qps
with pg= density of the pellet (kg/m3of pellet).
546 Chapter 10. ADSORPTION

The mass balance for a crown shape sliced out of the spherical pellet is
written:
a2C(R,t) 2 aC(R,t) aq'(R,t)
(10.17)
D e ( T + R 7)=7
where De is the effective diffusivity (m2/s).
Equation 10.17 can also be written as:

(1 0.18)

In writing this equation the accumulation of mass in the gaseous phase is

disregarded.
This equation applies to solids exhibiting a uniform pore structure such as
molecular sieve carbons, aluminas and silicas but not zeolites. Zeolites have
dual porosity of macropores and micropores as indicated in Figure 10.8. Here
two mass balances should be written, for the inside of the pellet and the inside
of the crystallite, linked by continuity at the crystallite surface. These equa-
tions are written:

(10.19)

where r represents the radius inside the crystallite, with rc as the radius of the
crystallite, and:
i a
(10.20)
at
where:
f'(R,t) is the average value of 9'(R,t) in the pellet
r is the radius inside the crystallite, with rc as the radius of the crys-
tallite
R is the radius inside the pellet with R, as the radius of the pellet
D', is the diffusion coefficient in the crystallite micropores
D,8 is the diffusion coefficient in the pellet macropores
a is the pellet macroporosity (fraction of the pellet volume not occu-
pied by crystallites)
The three possibilities of transfer resistance are schematically shown in
Figure 10.9c, but in many cases only one or two of these limitations can be con-
sidered (Figs. 10.9a and b).
The mass balance of the fluid phase is expressed in the following way:

-+ a~ c(Z,t) + -ac(z,t)
aZc(z,t)
-0,
at2 ataz
l-&
+- k ~ [ C ( z , t-) C(z,R,,t)] = 0 (10.21)
&
 ChaDter 10. ADSORPTION547

Crystallite

Grain

---
Figure
10.8 Representation of an adsorbent pellet in a bed.

with:
4 coefficient of axial dispersion, which will be introduced later on
in Section 10.3.2.1 (m2/s)
V superficial velocity of the fluid (m/s)
C(z,t) concentration of fluid circulating between the pellets (mol/m3 of
fluid)
E fraction of void in the bed
Q outside surface area of pellets per bed volume (m2/m3of bed)
k average film transfer coefficient (m/s)
C(z,R,,t) concentration at the surface of the pellet ( R i radius of the pel-
let)(mol/m3 of fluid)
t time (s)
Z position in the bed (m)
548 Chapter 10. ADSORPTION

Fluid phase

Boundary layer

I I I I l l
I I I I I I Concentration profile
---y Cpml I
--I I I I I
I I I I l l
I I I I l l
I I I I l l
I I I I l l
I g’(z,RgJ) 19’(Zm)I I
.-Ld I L I I Deposit profile
I I I I l l
I I I I l l
I I I I l l
I I I I l l
I I I I l l
I I I I l l
I I I I l l

Figure
10.9a Diffusion barrier around the pellet (excluding any other limitations).
 Chapter 10. ADSORPTION549

Grain
(at position z
in the bed)

Concentration profile

.-- ~I-1-
I I I
I I I I 1 I
I I I I I I I
1 I I I I 1 I
I’(Z,RgJ)
I
I
I I
I
/--
I
I - I I
1 -
I I
I Deposit profile

I
I I I
I
I
Rg
I I
I I
t R
I
I
0
I
I
R t Rg
I
I
I
I .
R+dR R+dR

Figure
10.9b Diffusion barrier in the macropores (excluding any other limitations).

As for the mass balance, the heat accumulation in the gas is disregarded.
For an adiabatic bed, the heat balance of the fluid phase yields:
aqz,t)
+-
aT(t,t) + - (
1; E ) ,%a
- [ T(Z,t) - T(z,Rg,t)]= 0 (10.24)
az at PF cP?

with:
T(z,t) temperature of the gas at a position 2; and a time t (K)
T(z,R,,t) temperature at the pellet boundary (K)
,% heat transfer film coefficient (W/(m2.K))
PF density of the fluid (kg/m3>
CP? heat capacity of the fluid (J/kg.K))
It is possible to introduce a second-order term which is practically negligible.
 /
0
/ -
- \

/ / Grain \ Crystallite
(at position R in the grain)

II II
Rg
I 1I
R
I
0
I 1I
R
Il
Rg
Il +-++I- O r

-
Figure
1 0 . 9 ~ Diffusionbarrier on three levels (outside, rnacropores, rnicropores).
 Chapter 10. ADSORPTION 551

The heat balance equations are coupled together by the relation:

he ( Rg
= %?[T(z,t) - T(z,Rg,t)] (10.25)

The mass and heat balance equations are coupled together by dependence
on concentration in the vapor phase and on the amount adsorbed with tem-
perature.

10.3.2 Transfer Resistance Models

If the adsorbent has a dual pore distribution (macro and micro) there can be
three types of resistance to mass transfer in an adsorber: intragranular resis-
tance, intracrystalline resistance and the film effect. In addition there is the
axial dispersion effect which also affects separation efficiency.

10.3.2.1 Axial Dispersion

Generally speaking the different mechanisms that contribute to axial disper-
sion are merged together in one single coefficient 0,. A detailed study of the
topic has been carried out by Langer et al. (1978). The main idea is that if the
effects related to packing heterogeneity are disregarded, two major mecha-
nisms cause the phenomenon: molecular diffusion and turbulence around
adsorbent pellets. The axial dispersion coefficient is often expressed as a func-
tion of the Peclet number Pe:
UL
Pe = -
4
with:
u superficial velocity of the fluid (m/s)
L adsorber length (m)
0, axial diffusion coefficient (m2/s)

10.3.2.2 Film Coefficients

The mass and heat transfer rate on the surface of a solid is limited by a film
whose thickness is determined by hydrodynamic conditions (see Chapter 3).
Transfer phenomena in a film are represented by a Fick law of the following
form:
N = k ( C - C*)
with N: flux (mol/(m2.s))
34’
- = ka [C(z,t)- C(z,Rg,t)] (10.26)
at
552 Chapter 10. ADSORPTION

The transfer coefficient is calculated using the dimensionless Sherwood

number Sh for mass transfer and the Nusselt number Nu for heat transfer. They
are expressed as follows:
2Rgk
S h = - with D,: molecular diffusivity
Dln
2RgX
N U = - with hF:heat conductivity of the fluid
hF
The value of Nu and of Sh is 2 in the first order for low Reynolds number
values, but there are correlations in the literature to calculate them according
to the usual hydrodynamic parameters (Re, Pr,Sc) (Ranz and Marshall, 1952;
Wakao and Funazkri, 1978).

10.3.2.3 Intraparticle Transport Phenomena

We will deal only with mass transfer phenomena in this part, for we generally
consider that the temperature gradient in an adsorbent pellet (and even more
so in a crystal when a zeolite is involved) is negligible, given the heat conduc-
tivity of the adsorbent, with external transfer (film) being limiting.
Four diffusion mechanisms can be identified distinctly on the macropore
scale, and the diffusion coefficient to be taken into account is often a combi-
nation of the four. They are: Knudsen diffusion, molecular diffusion, surface
diffusion and the Poiseuille effect. Diffusion in a gas is partly the result of col-
lisions of molecules with each other (molecular diffusion) and with pore walls
(Knudsen diffusion).
Molecular diffusion can be readily calculated for gases with the Chapman-
Enskog correlation:

(10.27)

with:
Dm molecular diffusivity (cm2/s)
T temperature (K)
MA and MB molecular weight (g/mol)
P total pressure (atm)
om collision diameter (A)
Qm collision integral (dimensionless)
The values of these different parameters can be found by referring to the
work by Bird, Stewart and Lightfoot (1960).
For liquids, more caution is required in using the Stokes-Einstein and
Nernst-Einstein equations, but empirical or semiempirical correlations can be
utilized (Reid, Prausnitz and Poling, 1987).
Knudsen diffusion comes into play when the pore effect is significant. This
effect is generally considered as predominant when the average free path of a
 Chapter 10. ADSORPTION553

gas molecule is greater than the adsorbent’s pore diameter. Diffusivity is cal-
culated in this case by the formula below:

(10.28)
with:
DK Knudsen diffusivity (m2/s)
r, pore radius (m)
M molecular weight (kg/mol)
T temperature (K)
It is generally assumed that diffusivity is ce-alated according to formula
(10.29) when neither of the two effects mentioned above predominates.
1 1 1
-=- +- (1 0.29)
Dm DK
Furthermore, the preceding equations were written for cylindrical pores in
a real adsorbent. Pore configuration is heterogeneous, so an empirical tortu-
osity coefficient is defined that serves to calculate the effective diffusivity
value to be used in models in the following way:
D0
D,= -
0
with:
0 pellet porosity (a, for macroporosity)
z tortuosity coefficient, whose value often ranges between 2 and 6
Molecular and Knudsen diffusions imply a flow of gas inside a pore.
Added to this there is diffusion in the adsorbed layer, which is known as
surface diffusion. If the adsorbed layer is thick (Costa et al., 1985), surface dif-
fusion can account for 80% of the diffusion phenomena. The stronger the
adsorption, the more significant the surface diffusion, in particular when pore
diameters are small, i.e. when Knudsen diffusion is involved. Surface diffusiv-
ity depends on the amount adsorbed of course and corresponds to adsorbed
molecules moving from site to site. It is activated by temperature and follows
a classical Arrhenius law. Sladek et al. (1974) have developed correlations that
can be used to calculate surface diffusivity values.
The Poiseuille effect is often negligible in adsorbers, since it comes into
play only when there is a significant pressure difference through a pellet. In
fact, if there is a total pressure gradient through the particle, there can be a
direct contribution by a forced convective flow through the macropores. This
effect is generally negligible in filled columns because the pressure drop on an
individual particle is very low. Equivalent diffusivity is written as follows in
this case:
P?
D=-
~ I J
554 ChaDter 10. ADSORPTION

where:
F represents the average pore radius (m)
p is the viscosity of the fluid (Pas)
Diffusion in zeolite micropores has been studied by numerous authors
because it is often the limiting step in adsorption processes that use zeolites
(Breck, 1984; Barrer, 1978; Ruthven, 1984). The factors that influence intracrys-
talline diffusion are listed below:
concentration in the crystal;
temperature:
geometry and dimensions of zeolite canals;
size, shape and polarizability of diffusing molecules:
distribution, size, charge and number of compensation cations;
the crystal’s structural defects.
For a specified gas-zeolite system, intracrystalline diffusivity follows the
Fick law and therefore depends on temperature and concentration, which are
the two main parameters.

10.4 Dynamic Adsorber Behavior: Bed Profiles

and Breakthrough Curves
The non-stationary nature of adsorption processes is their particular feature.
This is why it is necessary during process engineering to carry out dynamic
adsorption studies under conditions hydrodynamically similar to those in real
applications. These frontal chromatography studies usually consist in mea-
suring the exiting flow rates in response to a step injection (breakthrough
curves) or pulse injection (pulse test) (Fig. 10.10).
The adsorbate concentration at any point in the bed at each time depends
on the progression of the adsorption front in the bed. The shape and abscissa
of breakthrough curves are crucial in designing adsorption processes where
adsorption and desorption cycles alternate, without having the saturation
front breakthrough at the end of the bed. In fact, the sharpest possible adsorp
tion front is sought so as to achieve better bed efficiency. A breakthrough
curve for a gas containing one single adsorbate is modelled by solving mass
and heat balance equations in the two phases (fixed and mobile, or adsorbed
and gaseous). Numerous publications in the literature deal with this particular
problem (Ruthven, 1984; Rodrigues and Tondeur, 1981). In contrast, reverse
breakthrough curves related to desorption of adsorbed species have not often
been covered in the literature Pasmadjian, 1975). The shape of the break-
through curve stems from the nature and type of isotherm of course, but resis-
tance to mass and heat transfer tend to make the front more dispersive.
 Chapter 10 ADSORPTION 555

Inlet profile Inlet profile

Outlet profile Outlet profile

t t

Figure
10.10 Dynamic experiments. A. Pulse test response. B. Step test response.
Breakthrough curve.

10.4.1 Equilibrium Theory

The factors controlling the behavior of a breakthrough curve are basically
thermodynamic and kinetic in nature. Initially, equilibrium can be considered
to be instantaneous, to such an extent that column performance depends on
equilibrium data alone. In this case, the mass balance equation is:

which finally gives:

(10.30)

This equation gives the propagation velocity of concentration C versus the

slope of the equilibrium curve (&f/aC), and the interstitial velocity u.
To take into account the influence of the isotherm's shape on the propaga-
tion of the concentration front in the bed, isotherms are classified into three
groups as indicated in Figure 10.11:
type 1 isotherm, linear, with d2q'/dC2 = 0;
556 Chapter 10. ADSORPTION

-
Figure
10.1 1 Shape o f isotherms yielding differentconcentration waueh-onts.
1. Linear. 2. Favorable. 3. Unfauorable.

type 2 isotherm, favorable, with d2q'/dC2 < 0;

type 3 isotherm, unfavorable, with d2g'/dC2 > 0.
Type 1 is generally only a special case of type 2 for low concentrations. The
adsorbents used in separation exhibit type 2 isotherms. If equilibrium follows
a type 3 isotherm, the mass transfer zone broadens as the front propagates
through the bed. Meanwhile, if the isotherm is favorable, the front's progres-
sion is of the self-sharpening type, referred to as a compressive wave.
Figure 10.12 shows these differences in behavior.
One of the most useful parameters in sizing adsorption units is the time it
takes the discontinuity to get through the bed, called the stoichiometric time
fst or retention time. If the time to get through the column is T (T = EV/F, with V
the volume of the column and Fthe volume flow rate), this stoichiometric time
can be related to the column's adsorption capacity by the following formula:
1-E Q
with 5= - - (1 0.31)
E C

with Q corresponding to the adsorption capacity at concentration C, which

allows adsorption capacities to be calculated under actual operating condi-
tions.
The limits of this equilibrium theory method can be distinguished very
clearly in the following example of sizing for an adsorber to purify a feed with
a flow rate F. The pollutant with a concentration Co is to be adsorbed so as to
have an exiting gas containing less than 0.5 Co during a time f, given that the
 Chapter 10. ADSORPTION 557

-
Figure
10.12 Influence o f isotherm type on concentration profiles in an adsorber.
A. Compressive wave, favorable isotherm. B. Dispersive wave, unfavorable

solid's adsorption capacity is Q. To find out how much molecular sieve needs
to be implemented, the equilibrium theory is used to give the adsorbent vol-
ume with the formula below:
co -
V= - 5 Ft (10.32)
Q 1+5
This formula is true if the equilibrium theory can be applied and if the
isotherm is of the favorable type. Otherwise (Fig. 10.13), the amount of adsor-
bent implemented will have to be increased. We will have:
1 +K5
tEp= -tst (10.33)
1 +5
hence the larger adsorbent volume:

(10.34)

Two conclusions can be drawn from this simple example:

the contribution of breakthrough type experiments is very important in
sizing adsorption units;
not only the adsorption capacity Q for concentration Co, but also the
whole adsorption isotherm must be taken into account in sizing.
558 ChaDter 10. ADSORPTION

-
L

Figure
10.13 Equilibrium model. Influence of isotherm type on the breakthrough curve.

This is why more sophisticated models representing breakthrough curves

have been developed. The values that will best represent the adsorber can be
calibrated from these models and from experimental results. This representa-
tion is usually based on the following parameters (see Fig. 10.14): break-
through time tBp,width of the mass transfer zone 62 and leakage 1. The values
62 and I are calculated from exit profiles and from the value of the interstitial
velocity in the bed. The shape of the breakthrough curve is the result of the
influence of both thermodynamic and diffusional parameters.

10.4.2 Modelling Breakthrough Curves

Adsorber sizing is based on the breakthrough curves. The aspect of the curves
depends on factors such as: flow rate, pellet diameter, temperature, pressure,
concentration in the feed, etc. The effect of these variables can be taken into
account by models and the objectives of modelling are as follows:
predict adsorbers’ dynamic behavior;
optimize operating conditions;
go from the laboratory unit to a larger scale (pilot plant, industrial unit).
There are a large number of models in the literature and the diversity of
notations, assumptions made and mathematical solution techniques can be
 Chapter 10. ADSORPTION 559

C
-6Z-
I
CO --- -I- ------ '

0.95 Co -I- ------

0.05 Co

tBP t o r z = vt

Figure
10.14 Descriptive parameters of a breakthrough curve.

puzzling. The work by Yang (1987) presents the Rosen and Thomas models
which provide analytic solutions for the different balance equations by making
simplifying hypotheses.
Multicomponen t Breakthrough Curue
It is quite a complex job to calculate adsorptiondesorption systems for sev-
eral components because there is no analytic solution to the problem. The
equations can be solved numerically by computer and there are currently two
approaches:
solve all equations in time and space (there is even commercial software
that can build relatively sophisticated bed models (ADSIM));
use models with theoretical stages (Fig. 10.15).
In the approach by theoretical stage models, the adsorber is divided into a
number of perfectly agitated cells. In each of these stages the temperature and
concentration are constant and the number of theoretical stages used to
describe the column is an approximate representation of axial dispersion and
transfer resistance phenomena.
560 Chapter 10 ADSORPTION

-
-+-8-l-
I -.-
>n+l
-I -1
z + dz
Z cfl 8 t l 1 Differentrepresenta-
tions of an adsorber
for modelling.
A. Continuous repre-
sentation.
B. Theoretical stages.

10.5 Different Techniques for

Implementing Adsorption
As shown in Table 10.1, adsorption processes can be classified according to a
number of criteria. The adsorbent regeneration method and the mechanical
configuration of the unit are the most important characteristics in describing
the basic implementation techniques.

10.5.1 Adsorbent Regeneration Methods

Four methods are used to desorb adsorbed products. Figure 10.16 shows that
a decrease in the amount adsorbed at equilibrium can be achieved either by
increasing temperature or by lowering the partial pressure (or the concentra-
tion) of the adsorbate, or by a combination of these.
Increasing temperature: this is generally done by flushing the bed with
a preheated gas (sometimes a liquid). In most cases, the effect of higher
temperature is amplified by the effect of lower partial pressure. This
method is at the root of TSA processes (temperature swing adsorption),
which are basically of the slow cyclic type because of the time required
to heat and cool the beds.
Lowering total pressure: this method is used in PSA processes (pressure
swing adsorption), which generally work in rapid adiabatic cycles. It is
suitable when the adsorbate is not too strongly adsorbed.
 Chapter 10. ADSORPTION561

Figure
Adsorbent regeneration
methods.
A. Increasing tempera-
ture (TSA).
B. Lowering pressure
(PSA) or purging by
inert gas.
C. Combination TSA-
Partial pressure PSA or purging by a hot
inert gas.

Stripping with an inert gas: the driving force is the lower adsorbate par-
tial pressure, but the total pressure or temperature are not necessarily
modified. To keep inert gas in the adsorber voids from polluting the raf-
finate produced in the following adsorption step, the raffinate itself is
often used as purge gas. This type of purge is often combined with TSA
and PSA techniques.
Displacement by an adsorbable compound: regeneration is made easier
not only by the reduction in the adsorbate’s partial pressure (or concen-
tration), but also by competitive adsorption of the displacing agent. The
displacing agent is in turn displaced by adsorbable species during the
adsorption step. This regeneration method is used when the adsorbate is
very strongly bonded to the adsorbent or exhibits limited thermal stabil-
ity. It can be implemented in the vapor or in the liquid phase in cyclic,
chromatographic or simulated moving bed processes.

10.5.2 Cyclic Processes

10.5.2.1 TSA and Inert Gas Purge Processes
The TSA cycle is the oldest one used in adsorption. It requires at least two
adsorbers working alternatively in adsorption and desorption. In some cases
a third adsorber is used as a guard bed so that a bed can be completely satu-
rated before it is regenerated. This type of set up can prove to be advanta-
geous, especially when the mass transfer zone takes up a large proportion of
the bed. However, the additional costs entailed in having a third bed are not
warranted for most purification applications.
562 Chapter 10. ADSORPTION

The bed that is being regenerated is heated by circulating preheated gas in

an open or closed loop, in the reverse direction to that of the adsorption step.
A large number of variations in flow schemes are used depending on local con-
straints and mainly on the nature of the heating gas: nitrogen, air, steam, feed
gas, or product of adsorption step. Figure 10.17 illustrates a few examples of
flow diagrams.

--_
Figure
10.17
I
E A process. Examples of operation in gas dehydration (Atier Carter, 1979).
a. Regeneration by the Feed at the same pressure. b. Regeneration by the prod-
 Chapter 10. ADSORPTION 563

10.5.2.2 PSA Processes

The basic PSA process cycle has two versions known by the name of their
inventors: Skarstrom and Cuerin-Domine. The simpler Skarstrom cycle
(Skarstrom, 1972) is used in drying air. It has two adsorbent beds, with two
steps for each bed (Fig. 10.18): pressurization by the feed, then injection of the
feed at the bottom and production for bed I, depressurization from the bottom
and then purge from top to bottom by part of the product for bed 11. Each bed
therefore undergoes two half cycles of the same duration.
The major improvements in this basic cycle are the result of the introduc-
tion of cocurrent depressurizing and pressure equalization steps.
Depressurization cocurrently with the adsorption step allows increased
recovery of the least adsorbed species by partially evacuating the interstitial
volumes. At the same time it yields higher purity for the most adsorbed
species recovered during regeneration. Care must, however, be taken to inter-
rupt the adsorption phase soon enough so that the adsorption front is not too
close to the bed outlet.
A pressure equalization step between two beds mainly aims to minimize
mechanical energy consumption for the cycle. In preparation for the regener-
ation phase, a high pressure bed is placed in communication with a lower pres-
sure, regenerated bed in order to partially repressurize it and prepare it for the

B (high pressure)
4

Figure
PSA process. The 10.18 I
Skarsh-om cycle with

e
two beds.
A. The most adsorbed
product. B. The least
adsorbed product. -I- pressure)

Feed (A, 6)
564 Chapter 10. ADSORPTION

coming adsorption step. The more beds involved in pressure equalization, the
greater the compression energy savings. There is, of course, a trade-off
between energy savings and the additional cost of more beds.
Figure 10.19 shows an example of a unit with four beds (Union Carbide pro-
cess), used in purifying hydrogen, along with the different sequences of the
cycle (Cassidy and Holmes, 1984). Each bed alternatively undergoes the fol-
lowing steps:
Bed 1: High pressure adsorption with part of the product used to repres-
surize IV.
Bed 111: Co-current depressurization and pressure equalization with W, fol-
lowed by depressurization once again and use of the effluent to purge 11.
Bed 11: Countercurrent depressurization then purge.
Bed 1V: Countercurrent repressurization by pressure equalization with 111
then by some of the product from I.

product

EG: pressure equalization 1 and 2 A Co-current stream

DP: depressurization 1 and 2 7 Counter-current stream
RP: repressurization

Figure
10.19 PSA process. Sequences o f a cycle with four beds and two pressure equaliza-
tion steps (After Cassidy and Holmes, 1984).
 Chapter 10. ADSORPTION 565

Figure 10.20 indicates the pressure variations that each bed undergoes dur-
ing a cycle and the average duration of the different steps. Contrary to the TSA
process, it works in short cycles. The faster each operating cycle is, the
smaller the amount of adsorbent to be implemented for a given separation. A
number of constraints related to adsorption kinetics, pressure drop in the
beds and adsorbent mechanical strength do, however, impose limits. The
adsorption sequence does not generally last less than two minutes.
This operating principle has been extended to up to 10 beds by Union
Carbide in its Polybed process used in purifying hydrogen.
In most cases it is necessary to pretreat feeds so as to remove products
that would be retained too strongly by the adsorbent, especially when zeolites
are used (steam, heavy hydrocarbons, etc.), gradually poisoning it.
Pretreatment is generally carried out by adsorption with an adsorbent spe-
cially designed for the impurity to be eliminated. Increasingly an attempt is
made to incorporate pretreatment in the PSA process by using two different
adsorbents in the same adsorber. For example, a mixed silica gel5A zeolite
bed is used in producing oxygen from air. The silica gel retains steam and C 0 2
reversibly and thereby protects the zeolite. In the same way, the 5A zeolite bed
is protected from heavy hydrocarbon pollution with a bed of activated carbon
in hydrogen purification.
Numerous other modifications of the PSA process have been proposed to
find solutions in special cases. The following can be mentioned among others:
Purging by the most adsorbed compound, which improves the purity of
the product recovered during desorption, but entails an expenditure of
mechanical energy.
Equalizing temperature between the adsorption and desorption steps by
operating in very short cycles or by using a number of devices (heat con-
ducting bed, heating systems).
Most commercial PSA processes use equilibrium adsorption selectivity: air
drying, hydrogen purification, nlisoparaffins separation and oxygen produc-
tion from air. One process uses kinetic selectivity based on differences in dif-
fusivity in the adsorbent’s pores: nitrogen production from air with molecular
sieve carbon.

RP: repressurization

Adsorption

--
Figure
10.20 PSA process. Pressure diagram in each bed for a cycle with four beds and two
pressure equalization steps.
566 Chapter 10. ADSORPTION

10.5.2.3 Displacement Processes

Figure 10.21 shows a simplified flow diagram for a cyclic displacement pro-
cess. During regeneration the displacing agent D is partially adsorbed and
causes adsorbate A to be desorbed. Then during the adsorption step, D is
removed by the adsorbate A and ends up mixed with the least adsorbed prod-
uct B. Two distillation columns must be associated with the adsorbers to
regenerate the desorbent D and purify the products.
The choice of displacing agent is governed by two constraints: its affinity
toward the adsorbent must be similar to the adsorbate's and it must be read-
ily separated from A and B by distillation.
Even though this process is somewhat complex due to the distillation oper-
ations associated with it, it is used in separating hydrocarbons on molecular
sieves when adsorbate desorption is difficult or when it can cause coking of
the adsorbent. Some examples of this are: heavy nlisoparaffins and paraf-
fins/olefins separations. The displacement agents used are hydrocarbons and
ammonia. The processes are generally run in isothermal and isobaric cycles.

? Distillation

Adsorption Desorption

f! A
Feed
A+B

D Distillation

Cyclic displacement process (After Keller 17, 1983). A. The most adsorbed prod-
uct. B. The least adsorbed product. D. Desorbent.
 Chapter 10. ADSORPTION 567

10.5.2.4 Chromatographic Processes

Liquid or gas chromatography is a highly efficient separation technique due to
the implementation of a very large number of theoretical stages. It can there-
fore take advantage of relatively low adsorption selectivity and carry out diffi-
cult separations.
Numerous attempts have been made to extrapolate this technique
(Valentin, 1981; Canetsos and Barker, 1993). They have resulted in industrial
preparative chromatography processes used mainly in the area of fine chemi-
cals to produce small outputs. These processes have not developed in the oil
and gas industry, which is characterized by high throughputs, since they are
not competitive with continuous simulated moving bed processes.
As a reminder, mention can be made of some successful developments: the
N-iself process by Elf Aquitaine pernard et al., 1981) for separating nlisoparaf-
fins from light naphtha, and the Asahi process (Seko et al., 1982) for separat-
ing paraxylene and ethylbenzene from an aromatic C, cut. These processes
have not had any industrial application.

10.5.2.5 Parametric Pumping and Cycling Zone Adsorption

These two techniques are mentioned here only as a reminder. They have not
yet given rise to any industrial implementation, despite the large body of the-
oretical work that they have inspired (Yang, 1987).
Parametric pumping utilizes adsorption with a single adsorbent bed
undergoing very rapid inversions in flow direction along with variations in
temperature or pressure.
Cycling zone adsorption uses the same principles of varying temperature
or pressure between different zones, but without reversing the flow direction
of the streams.
The techniques come under the PSA and TSA heading. The degrees of free-
dom in their implementation are very limited in comparison with conventional
processes and this makes them difficult to put into practice.

10.5.3 Continuous Processes: Moving Bed

and Simulated Moving Bed
In fixed bed cyclic processes, the composition and flow rate of the different
streams vary during the cycle, and this sometimes makes them difficult to
incorporate in a sequence of processes. Moreover, adsorbent use is far from
optimized, because some parts of the bed are not used during part of the cycle.
A moving bed of adsorbent flowing countercurrently with the fluid allows
a steady state system to be established. The adsorbent is theoretically used
more efficiently because operating conditions are controlled in all the different
zones of the adsorber. In actual fact, real moving beds are seldom used indus-
trially, mainly due to problems of adsorbent attrition and transport. Their use
is restricted to treatment of gases. Simulated moving bed processes are not
subject to these drawbacks and still have the advantage of countercurrent
568 ChaDter 10. ADSORPTION

operation @roughton and Gembicki, 1984). They are used for separations in
the liquid phase.

10.5.3.1 Moving Bed Processes

The first industrial example of moving bed adsorption is the Hypersorption
process (Berg, 1951; Szirmay, 1975), implemented in the fifties for recovering
ethylene from FCC gas with activated carbon. Circulation of the solid e s t a b
lishes four zones in the adsorber: from top to bottom a cooling zone, an
adsorption zone, a rectifying zone, a heating and stripping zone. The process
has been phased out due to problems with the adsorbent’s mechanical
strength.
Based on the same principle, the Purasiv HR process (Fig. 10.22) consists
of a series of fluidized beds of activated carbon beads circulating countercur-
rently with the gas. It is used in recovering solvent vapors from polluted air.
The main advantage of the technology is that the temperature can be varied
between the adsorption and the regeneration zones.
In the field of air purification, a rotating adsorber technology is becoming
more and more developed. It is designed on the basis of the rotating heat

Non-adsorbed gas

A+
Feed

Heating
steam

Desorption
steam

Condensate
Carrier gas

Figure
10.22 Fluidized moving bed process. Purasiv HR process (AFter Keller 14 1983).
 Chapter 10. ADSORPTION 569

exchanger: the adsorbent is placed in a fixed bed in different sectors of a

wheel-shaped adsorber that rotates slowly on its axis. The adsorbent "circu-
lates" in this way from an adsorption zone to a regeneration zone where it is
contacted with hot air. The use of this technology is restricted to the treatment
of effluents that do not pose any leakage problems.

10.5.3.2 Simulated Moving Bed Processes

The simulated moving bed (SMB) or simulated countercurrent (SCC) is at the
basis of the Sorbex type of processes developed by UOP in the sixties
(Broughton, 1978; de Rosset et al., 1981). Although it can theoretically be
applied in the gaseous phase, its use is today restricted to liquid phase sepa-
rations. The adsorbent is regenerated with a displacement technique by a des-
orbent that has to be separable by distilling the extract and the raffinate.
To give a better understanding of how SMB works, we will analyze an actual
counter current between the adsorbent and the liquid phase (Fig. 10.23). The
aim is to separate a feed containing at least two components A and B into two
fractions. The means used is a column where an adsorbent that is selective
toward A circulates from top to bottom, while from bottom to top a fluid phase
circulates that is enriched in desorbent D at the bottom. The adsorbent stream
tends to carry product A, the most strongly adsorbed, toward the bottom of
the column while product B, less strongly adsorbed, is carried upward by the
liquid stream. Near the bottom of the column, product A is displaced by D,
thereby allowing an extract made up of A + D to be drawn off and a regener-
ated adsorbent to be sent back up to the top of the column. A raffinate made
up of B and D is drawn off above the feed injection point.
Lastly in the actual countercurrent, four zones can be distinguished, each
one playing a different part and having different liquid flow rates and compo-
sitions:
Zone 1: A is desorbed by D and extracted from the column.
Zone 2: B is desorbed and removed by the liquid flow.
Zone 3: Adsorption of A, B and D are in the liquid phase.
Zone 4: Adsorption of B, only D is in the liquid phase and can be recircu-
lated.
We have presented a set up with four zones, but in some cases a simplified
three-zone configuration is used, or five zones with injection of a reflux of
product A which enhances extract purity.
When the system has reached steady state running conditions, fixed con-
centration profiles are established and an extract and a raffinate with a con-
stant average composition are drawn off. The problem is to find the solid and
fluid flow rates that allow all of A to be recovered in the extract and all of B in
the raffinate.
In actual fact, circulating the adsorbent causes attrition problems and dis-
torts the concentration fronts. This is why a technique called simulated mov-
ing bed is used. Here, the solid remains stationary in the column where it is
570 Chapter 10. ADSORPTION

-
Figure
10.23 Simplified flow diagram of an actual counter current.
 7px
t= 0 t = permutation time

I Concentration

0
Concentration
a
0 .
0 w

2
0

,0

~-
Liquid
-
\ O x + Mx ’ recycle -

I I
0
0

0
0
0
0
0 ’, o x +
08

0
8
0 Mx
8
I
I
I

0 Paraxylene (Px) Ortho- and Metaxylene (Ox + Mx), Ethylbenzene (EB) 0 Desorbent

1 10.24 I Simplified flow diagram o f a simulated counter current.

572 ChaDter 10. ADSORPTION

Figure
10.25 Sorbex process: simulated moving bed adsorption (After de Rosset et al.,
1981).

established, the point where the extract is drawn off must always be located
in the zone where product B has just disappeared, whereas the point where
the raffinate is drawn off follows the zone where product A has disappeared.
As before, the quality of separation is governed by proper regulation of fluid
flow rates for a given permutation time.
Figure 10.25 shows how this concept is implemented in the Sorbex process
(Broughton, 1978). Depending on the applications, the adsorber can have from
8 to 24 beds. A rotating valve distributes the four streams (five in some ver-
sions with a reflux). In the new Eluxyl paraxylene separation process devel-
oped by the Institut Frangais du Petrole, distribution is handled by individual
computercontrolled valves (Hotier et al., 1994). The same is true for the
Aromax process developed by Toray Industries which additionally uses an
assembly of adsorbers where the liquid flows horizontally (Otani et al., 1971).
The simulated moving bed technique is widely used for difficult separa-
tions in the refining and petrochemical industries (nlisoparaffins, para-
xylene/C8 aromatics, olefins/paraffins), but also in the agro-food industry
&lucose/fructose) and on a smaller scale in manufacturing fine chemicals and
pharmaceuticals.