Journal of Colloid and Interface Science 263 (2003) 350356

www.elsevier.com/locate/jcis

New theoretical expressions for the five adsorption type isotherms

classified by BET based on statistical physics treatment
M. Khalfaoui, S. Knani, M.A. Hachicha, and A. Ben Lamine
Laboratoire de Physique Quantique, Dpartement de Physique, Facult des Sciences de Monastir 5019, Tunisia
Received 6 July 2002; accepted 28 January 2003

Abstract
New theoretical expressions to model the five adsorption isotherm types have been established. Using the grand canonical ensemble in
statistical physics, we give an analytical expression to each of five physical adsorption isotherm types classified by Brunauer, Emett, and
Teller, often called BET isotherms. The establishment of these expressions is based on statistical physics and theoretical considerations.
This method allowed estimation of all the mathematical parameters in the models. The physicochemical parameters intervening in the
adsorption process that the models present could be deduced directly from the experimental adsorption isotherms by numerical simulation.
We determine the adequate model for each type of isotherm, which fixes by direct numerical simulation the monolayer, multilayer, or
condensation character. New equations are discussed and results obtained are verified for experimental data from the literature. The new
theoretical expressions that we have proposed, based on statistical physics treatment, are rather powerful to better understand and interpret
the various five physical adsorption type isotherms at a microscopic level.
2003 Elsevier Inc. All rights reserved.
Keywords: Adsorption; BET type isotherms; Statistical physics; Grand canonical ensemble

1. Introduction
The aim of this work is to contribute to the investigation of molecular adsorption at solidgas and/or solidliquid
interfaces. This is in order to contribute to the understanding of this physical mechanism at the molecular level since
the study of adsorption is important for various physicochemical processes such as textile dyeing and depollution
of industrial liquid effluents [1]. The adsorption isotherm is
used to compare the adsorption capacity of adsorbents for
given substances. It is noted that there are two adsorption
categories: physical and chemical adsorption. In the case of
physical adsorption, the interactions between the adsorbate
and the adsorbent are electrostatic [2,3]. Chemical adsorption involves specific forces, such as those that are operative in the formation of chemical bonds. Three phenomena may be involved in physical adsorption: monomolecular
and multimolecular adsorption and condensation in pores or
capillaries. So the interpretation of adsorption studies can
be complicated. Brunauer and his co-workers [2] have clas* Corresponding author.

E-mail address: aben_lamin@yahoo.fr (A.B. Lamine).

0021-9797/03/$ see front matter 2003 Elsevier Inc. All rights reserved.
doi:10.1016/S0021-9797(03)00139-5

sified physical adsorption isotherms into five characteristic

types shown in (Fig. 1). In the IUPAC recommendations
there are six physical adsorption isotherm types [4]. We are
interested in our present work in modeling only the five
BET types of isotherms. The sixth type will be studied in
a later work since this isotherm presents several partial saturations, so the statistical physics treatment of this particular type is rather complicated. Several studies [2,5,6] were
made to look for model expressions that can describe these
various BET types of isotherms. The authors proposed empirical models that describe the isotherms met in various experiences. However, most of these models contain constants
that do not give any supplementary information describing
the adsorption process. Furthermore, some of these constants
that each expression found in the literature contains have no
physical significance or relationship with the physicochemical parameters intervening in the adsorption process in major cases. We focus in this work on finding similar expressions for each adsorption isotherm type, but we relate the
constants to the physicochemical parameters. For this purpose, we thought of reestablishing and/or of improving these
same expressions of suggested models, but based on theoretical treatments of statistical physics. The aim of the theoreti-

M. Khalfaoui et al. / Journal of Colloid and Interface Science 263 (2003) 350356

351

effect of various physicochemical parameters on adsorption

has also been investigated.

2. Theoretical development

Fig. 1. The five types of adsorption isotherm, I to V, in the classification of

Brunauer, Emett, and Teller [2].

cal development is to give physical meaning to the constants

that these models contain and therefore to facilitate and lead
toward understanding the adsorption process at the molecular level. Furthermore, the new theoretical expressions have
a more simple form than those established such by Fowler
and Guggenheim [7] and BeBoer Hill [8] derived forty or
fifty years ago based also on statistical physics. The proposed simple forms will allow the experimental isotherms
to be easily fit since the experimental adsorption isotherms
are generally represented in the form of adsorbed quantity versus concentration or pressure. So the new theoretical expressions reinforce more the contribution of statistical
physics to the modeling of the adsorption isotherms. Each
new theoretical expression can predict a suitable experimental isotherm very well. This theoretical base will allow us
to better interpret and to deduce information about physical adsorption at the molecular level, which remains unapproachable by means of empirical methods. The equations
established in an ad hoc manner are oversimplified and suffer from the drawback that they cannot bring out the physical
realities. Another disadvantage associated with these models
is that they cannot fit some cases of experimental adsorption
isotherms well. In order to overcome these drawbacks, some
researchers [3,9] attempted either to find other expressions
for some of these models or to improve those existing based
on statistical physics. The theoretical modeling of adsorption isotherms is crucial to the design and scale-up of practical adsorption processes. In our theoretical development, we
rely on a statistical physics treatment and some hypotheses
are assumed to simplify the theoretical calculations. In this
paper, we present the methodology used to find new and simple theoretical expressions that are able to describe and fit
well the five types of isotherms of physical adsorption. The

In this section an analytical expression for each type of

isotherm will be established using a statistical physics approach. It will be seen that such a treatment provides expressions for all the different model parameters, thereby enabling
physical interpretations of the values derived for the same.
In order to treat adsorption problems using statistical
physics methods, it is necessary to make some assumptions
as the bases of an initial postulate. As a first approximation,
the adsorbed molecules are treated as an ideal gas [35,9],
since the mutual interaction between the adsorbed molecules
will be neglected and so their concentration will be relatively
low. This hypothesis is made for both gas and liquid systems.
Adsorption involves an exchange of particles from the
free state to the adsorbed state. Its investigation cannot be
performed without employing the grand canonical ensemble
in order to take account of the particle number variation
through the introduction of a variable chemical potential
in the adsorption process. Although the overall system is
canonical, the canonical ensemble and the grand canonical
are equivalent since it is macroscopic. This means that
it is at the thermodynamic limit and fluctuations may be
neglected. This allows the system to be studied using the
grand canonical ensemble with the results being interpreted
canonically [5,9].
2.1. General statistical physics treatment
The following statistical treatment that will be presented
is applicable to both liquid and gas systems. In liquid systems we determine the solubility of the adsorbed molecules
in solution and in gaseous systems we determine the saturated vapor pressure from the experimental curve fitting as
we will describe.
In order to employ statistical physics methods to treat
the adsorption model, we consider that a variable number
of molecules, c per unit of volume, is adsorbed onto NM
receptor sites located on a unit mass of the adsorbent. In
our treatment, our departure point is the grand canonical
partition function describing the microscopic states of a
system following the physical situation in which this system
is placed [5,9],


e()Ni ,
Zgc =
(1)
Ni

where is the receptor site adsorption energy, is the chemical potential, Ni is the receptor site occupation state, and
is defined as 1/kB T , where k B is the Boltzmann constant
and T is the room temperature. This last grand canonical
partition function led us to determine the average site occupation number and therefore the average number of the

352

M. Khalfaoui et al. / Journal of Colloid and Interface Science 263 (2003) 350356

adsorbed molecules on N M receptor sites per surface unit

based on the method used by Diu et al. in 1989 [5]. In our
investigation, we take account only of the translation partition function,
3/2

Zgtr = 2mkB T / h2
(2)
,
where m is the adsorbed molecule mass and h is the Plancks
constant. The translation partition function, for an ideal gas,
can be expressed by the saturated vapor pressure as
3/2

v
v
Zgtr = 2mkB T / h2
(kB T )eE /RT eE /RT
= Pvs eE

v /RT

(3)

where E v is the vaporization energy of one adsorbed mole

and Pvs is the saturated vapor pressure. The vaporization energy of on ideal mole is confused with the dissolution energy [9]. For liquid system we replace the saturated vapor
pressure by the solubility cs of the adsorbate [2,18] and the
vaporization energy E v by the dissolution energy E d
and we can write
Zgtr = cs eE

d /RT

(3 bis)

Then the new expressions that will be presented in this paper may be used for both liquid and gaseous systems. For
the gaseous system it is a question of saturated vapor pressure and for the liquid system it is a question of solubility.
The total grand canonical partition function related to N M
receptor sites per surface unit, which we assume identical,
is equal to ZgNM . The occupation number is given by the
following formula [39]:
No = kB T

ln ZgNM

(4)

Generally, the adsorption reaction should include a stoichiometric coefficient n, such that
nA + S
An S,

2.2. Adsorption isotherm of Type I

Type I isotherms show a fairly rapid rise in adsorbed
quantity with increasing concentration (or pressure) up to
saturation; i.e., they are characterized by a plateau that is
nearly or quite horizontal. To choose an adequate model,
we notice that the representative curve of Type I (Fig. 1)
is for saturation and that the cavity is negative [39]. For
this type of isotherms, Khalfaoui et al. [3] supposed that the
site is occupied by one molecule or empty. It was assumed
that any given receptor site can be empty or occupied by one
part molecule in accordance with the monolayer model. The
grand canonical partition function in this case, for only one
site, has the form
zgc = 1 + e(+) ,
where is the magnitude of the adsorption energy level of
one receptor site, is the chemical potential, and is the
Boltzman factor.
Using the previous definition of the adsorbed quantity
expression against the concentration of the adsorbed molecules, we obtain the formula
Na =

nNM
,
1 + (c1/2 /c)n

where n is the number of adsorbed molecule(s) per site, NM

is the receptor site density, and c1/2 is the concentration
at half saturation. The parameter c1/2 is related to the
adsorption energy as [39]
c1/2 = Pvs eE

N a = nN o .

(5)
n

We can define the anchorage number as = 1/n, i.e., the

number of sites on which the molecule can be adsorbed.
In the following sections, we present for every type of
isotherm the adequate grand canonical partition function
and afterward the new expression of adsorbed quantity
as a function of concentration or pressure, since the two
variables are related. After this, we present the results and
discussions that the treatment of statistical physical gives,
and in consequence the new expressions so obtained.

a /RT

(7)

where Pvs is the saturated vapor pressure, R is the ideal

gaseous constant, and T is the room temperature. E a
is the molar adsorption energy from the dissolved state.
Equation (7) takes the following form in the case of a liquid
system,
c1/2 = cs eE

where n is the number of the adsorbed molecules A per

site S.
If there are n molecule(s) anchored per site, where the
value of n can be inferior or superior to 1, the total number
of adsorbed molecules is given as

(6)

a /RT

(7 bis)

where cs is the solubility of the adsorbate.

We can note that the adsorption is governed by a law
similar to that of Fermi and Dirac [5,6] since the site is empty
or occupied by only one molecule. If we take n equal to
unity, we obtain the expression of Langmuir which is usually
used to model this type of isotherm.
This model was established by Ben Lamine and Bouazra
[9] and was enough to model this type of isotherms. It was
demonstrated [3] that the last model is more general than that
of Langmuir. The effect of each parameter on the adsorption
isotherm was studied by Khalfaoui and his collaborators [3].
Afterward, we give the effect of each parameter on the
adsorption isotherm.
The stoichiometric coefficient n describes and interprets
the adsorption equilibrium. Khalfaoui and his collaborators [3] showed that for high concentrations (c > c1/2 ), the

M. Khalfaoui et al. / Journal of Colloid and Interface Science 263 (2003) 350356

353

higher n is, the less important the adsorbed molecule quantity is. For low concentrations, the inverse phenomenon occurs; i.e., when n decreases, the anchorage number, which
is defined as the inverse of n, increases, so it is more difficult to find a great number of empty sites near saturation
at the same time. This fact is inverted at low concentration
since the majority of sites are empty and a molecule with a
great number of anchorages is easily attracted by energetic
considerations.
The steric coefficient N M is related to the adsorbed quantity at saturation N asat by the expression
N asat = N M nNi max ,
where Ni max is the layer number. For fixed n and Ni max
parameters, the increase of N M can improve the adsorption
capacity at saturation.
The energetical parameter c1/2 is closely related to the
adsorption energy level. It is shown [3] that at a fixed concentration value, the lower c1/2 is, the higher the adsorbed
quantity is. This fact is valuable for all concentration values
far from the saturation. So we can conclude that the receptor sites which have the stronger energy will be rapidly and
easily occupied.
Such as experimental isotherms that possesses the shape
of Type I, we can quote the case of adsorption of iodine
from aqueous potassium iodine solution on zinc chlorideactivated carbons [10] and the adsorption of ethane on Ajax
activated carbons [11]. By fitting the experimental isotherms
(Figs. 2a and 2b) we have found that the adjustment
coefficient in the first case is close to 0.98 and in the second
case is close to 0.99. We notice that our new theoretical
expression is in good agreement with experimental data and
is also valid for both liquid and gaseous systems.
2.3. Adsorption isotherm of Type II
For the adsorption isotherm of Type II, to choose a suitable model, we can note that the curve indicates a multianchorage adsorption and the concavity is negative; i.e., the
value of n is less than unity at low concentrations. At high
concentrations, the adsorbed quantity is divergent; i.e., there
is a condensation phenomenon. It is therefore a multilayer
model (the occupation number goes from zero to infinity). If
we suppose that the first molecule is adsorbed with the energy 1 , and the others are adsorbed with the same energy
2 , with the condition that 1 > 2 > 0, the grand canonical
partition function related to one receptor site is then written
zgc =

e(Ni +Ni ) .

(8)

Ni =0

Using the occupation number as defined in general statistical

physics, we obtain the adsorbed quantity against concentration as
nNM
,
Na =
(9)
[(c1 /c)n (c1 /c2 )n + 1][1 (c/c2 )n ]

Fig. 2. Adsorption isotherm of iodine (a) on zinc chloride-activated

carbons [10] and adsorption isotherm (b) of ethane on Ajax activated
carbons [11]. The open circles are the experimental data and the line is
the simulation data.

with c1 = zgtr e1 and c2 = zgtr e2 . We can also write

a
a
c1 = Pvs eE1 /RT and c2 = Pvs eE2 /RT .
We notice that when concentration value is equal or near
to the one of c2 , the adsorbed quantity is infinite and so
the capillary condensation is reached. For the values of the
concentration lower than the one of c2 , the expression for
Na is only dependent on the parameter c1 in addition to
the parameters n and N M . We obtain therefore the same
expression as related to Type I.
This category of adsorption isotherm depends on two
adsorption energy levels, so we can conclude that the first
proportion that adsorbed at low concentrations needs an
energy different from that related to the second proportion.
We can predict with this model that the energy related to
the first adsorbed proportion is higher than that related to
the second proportion. Indeed, the first adsorbed proportion
requires a high energy value 1 and to increase this number
of molecules a weak energy value is sufficient.
In the investigation of the effect of hydrocarbon chain
unsaturation on the water adsorption characteristics of the

354

M. Khalfaoui et al. / Journal of Colloid and Interface Science 263 (2003) 350356

Fig. 3. Adsorption isotherm a of water on lipid molecules. The open circles

are the experimental data [12] and the line is the simulation data.

phospholipids [12], some adsorption isotherms (adsorption

of water on lipid molecules) that were established have the
same shape as Type II. We notice that the new theoretical
expression given by Eq. (9) is in agreement with one of these
adsorption isotherms for water on lipid molecules since the
adjustment coefficient is close to 0.94 (Fig. 3).
2.4. Adsorption isotherm of Type III
For this adsorption type isotherm, we suppose that every
receptor site is occupied by an infinity of molecules with the
same energy . The grand canonical partition function of
one receptor site is written as
zgc =

e()Ni .

(10)

Ni =0

The adsorbed quantity expression is then

Na =

nNM
,
(co /c)n 1

(11)

with co = zgtr e = Pvs eE /RT . We have used the

notation co in order to differentiate between this particular
concentration and the one related to Type I isotherms since
the two parameters c1/2 and co have the same expression but
not the same role in the plot of N a .
We note that toward the value of co , the adsorbed quantity
becomes infinite and the isotherm diverges for this value
of concentration. When the number of molecules is big,
a condensation phenomenon may result. This is similar to
BoseEinstein condensation [5,6]. The adsorbed quantity
has the same behavior as that of the distribution for bosons
in statistical physics; i.e., the receptor site can be occupied
by a infinity of molecules and a condensation phenomenon
appears.
Typical experimental adsorption isotherms of Type III
have been obtained in the case of adsorption of carbon
a

Fig. 4. Adsorption isotherm of carbon tetrachloride on mesoporous gel. The

open circles are the experimental data [13] and the line is the simulation
data.

tetrachloride on a mesoporous gel [13]. We have fitted

the experimental isotherm and we have found that the
adjustment coefficient is equal to 0.99 (Fig. 4).
2.5. Adsorption isotherm of Type IV
We note that the adsorption isotherm of Type IV is the
same as that of Type I. The only difference between the
two types of isotherms is that the concavity in the case of
Type IV is positive at low concentrations and the curvature
is against the power n with the condition that the value of n
is greeter than unity. For this type, we can also suppose that
the adsorption takes place with only one energy level or a set
of energies but very closed since the isotherm is for only one
landing.
It was demonstrated [3] that this category of isotherms
could not explain that the receptor site did not contain an infinity of molecules and on the other hand it is an intermediate
case between Types I and III. The adsorbed molecule number per site is finite. For example, if we suppose that the site
can be empty or occupied by one molecule or by two molecules, the grand canonical partition function is written as
zgc = 1 + e(+) + e2(+),

(12)

and therefore, the adsorbed quantity is

Na = nNM

(c/c1/2)n + 2(c/c1/2)2n
.
1 + (c/c1/2)n + (c/c1/2)2n

(13)

The same behavior of the theoretical isotherm could be obtained if we increased the occupation number to more than 2.
Adsorption isotherm of benzene on a mesoporous sieve
[14] is a good illustration for Type IV. We can note
that Eq. (13) is in good agreement with the experimental isotherm with an adjustment coefficient equal to 0.95
(Fig. 5).

M. Khalfaoui et al. / Journal of Colloid and Interface Science 263 (2003) 350356

355

Fig. 5. The open circles are the experimental data [14] and the line is the
simulation data.

Fig. 6. The open circles are the experimental data [13] and the line is the
simulation data.

2.6. Adsorption isotherm of Type V

procedures and deduced for such considered experimental

isotherms. We are restricted in our present study to giving
just some examples of adsorption isotherms and not discussing the results of the fitting since our first principal goal
is to establish the new theoretical expressions. We can determine, thanks to the new expressions the number of adsorbed
layers. This result could also be calculated previously using
other methods [15]. Thus, we are able to deduce the steric
factors n and N M that describe the anchorage of the adsorbate, i.e., whether the adsorbed molecule is parallel (n < 1)
or perpendicular (n > 1) to the adsorbent surface. We note
also that thanks to the proposed expressions for each type
of adsorption isotherm we can determine the mean adsorption energy, which may include the interaction and mobility
energies together. In spite of this fact the Volmer equation
includes the mobility, the Fowler and Guggenheim equation
includes the interaction energy, and the two-dimensional van
der Waals relationship contains both structure of energies
separately [16]. This step of modeling that we have reached
could be improved later by modifying the expression for the
grand canonical partition function, taking into account other
degrees of freedom in addition to the translation one. We emphasize that our work is an attempt to create a new statistical
treatment of physical adsorption, which has a starting-point
character. This new statistical treatment will be improved by
modifying the initial assumptions that we considered as the
basis of an initial postulate.
We consider examples to compare the proposed equations
by experiment. In each case, the fitting of the experimental
data was carried out by a nonlinear least-squares optimization procedure. The application of the expression describing Type IV gave new interpretations of dye adsorption onto
modified cotton [17] by studying the evolution of the physicochemical parameters n, N M , and c1/2 as a function of the
grafted nitrogen rate. It was shown that thanks to the statistical treatment, the investigation of the adsorption process at
the molecular level becomes possible and easy. The behav-

The adsorption isotherm of Type V is similar to that of

Type II except that the behavior at low concentrations is at
the power of n with a value inferior to unity. The molecules
which are adsorbed at low concentrations present a sort of
dimerization and not a multianchorage as could be foreseen
for the Type II. The difference between Types II and V is that
a multianchorage occurs in the first case and a dimerization
appears in the case of the second type.
Type V was obtained in the case of adsorption of water
vapor on oxygenated carbon heated in vacuo at 200 C [13].
The adjustment coefficient with the new theoretical expression that proposed for this type is equal to 0.99 (Fig. 6).

3. General discussion
We have found the adequate expression model for each
of the five adsorption type isotherms over the entire range
of vapor adsorption, from zero to the saturation vapor pressure. The new expressions that we have established can be
also used to fit the adsorption isotherms in the case of a
liquid system. Indeed, the state of the adsorbate in the adsorbed phase varies with temperature [1]. Above the critical temperature, the adsorbate is vapor-like and we can use
one of the above theoretical expressions with the notation
of the solubility. Below the critical temperature, the adsorbate behaves like a compressed gas and we introduce the
notation of the saturated vapor pressure. Thanks to the theoretical modeling, we can well understand the behavior of
each type of isotherm and thus the corresponding adsorption phenomenon at microscopic level. Form the experimental isotherms we can, using the adequate model, determine
the physicochemical parameters related to the adsorption
process. All the values of these parameters, in the various
equations already determined, may be obtained by fitting

356

M. Khalfaoui et al. / Journal of Colloid and Interface Science 263 (2003) 350356

iors of each parameter cannot be interpreted without an understanding of its physical significance thanks to statistical
physics.
We have, on the other hand, shown the difference between
the behavior of various types of isotherms that have almost
the same curvature except at weak concentrations. This difference is explained by the fact that the adsorption mechanism of the first molecules changes from one isotherm type
to another. For this reason, the choice of a suitable model
changes from one isotherm type to another. We noticed that
this difference is mainly governed by the effect of the anchorage number of the adsorbate, i.e., by the geometry and
the topography of the adsorption, and also by the value of the
adsorption energy. In the case of the isotherm types that did
not present saturation level, except Type II, a second energy
level is necessary to adsorb an infinity of molecules. In that
case, the adsorption process begins with a high energy and
ends with less energy. In all cases, we suppose that the anchorage number stays the same for the entire concentration
range.
These new theoretical expressions are not in contrast
with the others that were established ago based on statistical
physics, but they give new forms and reinforce the statistical
physical theory in the field of adsorption mechanisms.

4. Conclusion
New theoretical expressions are proposed using a statistical physics treatment to maintain a good fitting for physical
adsorption isotherms classified by BET. These models are
thermodynamically consistent and the physicochemical parameters intervening in the adsorption process are taken account in the theoretical treatment. They overcome the drawbacks in the ad hoc expressions that did not contain the
physical realities. The behavior of these proposed expressions is compared to experimental adsorption isotherms pub-

lished in the literature. It is found that, using a statistical

physics treatment, the new theoretical models are able to predict the topography and the geometry by which the gaseous
molecules are adsorbed on the surface. The difference between the isotherm types was presented and discussed. It was
found that the proposed statistical treatment is sufficiently
flexible to give a good representation of experimental data.
The statistical physics modeling is a powerful way to analytically understand these curve behaviors.

References
[1] J. Avom, J. Ketcha Mabdcam, M.R.L. Matip, P. Germain, AJST 2
(2001) 17.
[2] C. Tien, Adsorption Calculation and Modeling, in: Series in Chemical
Engineering, ButterworthHeinemann, London, 1994.
[3] M. Khalfaoui, M.H.V. Baouab, R. Gauthier, A.B. Lamine, Adsorpt.
Sci. Technol. 20 (2002) 1731.
[4] IUPAC Commission on Colloid and Surface Chemistry Including
Catalysis, Pure Appl. Chem. 57 (1985) 603619.
[5] B. Diu, C. Guthmann, D. Lederer, B. Roulet, Physique Statistique,
Hermann, Paris, 1989.
[6] L. Couture, R. Zitoun, Physique Statistique, Ellipses, 1992.
[7] R.H. Fowler, E.A. Guggenheim, Statistical Thermodynamics, Cambridge Univ. Press, Cambridge, 1952.
[8] H. BeBoer, Introduction to Statistical Thermodynamics, Addison
Wesley, Reading, MA, 1960.
[9] A. Ben Lamine, Y. Bouazra, Chem. Senses 22 (1997) 67.
[10] M.N. Alaya, M.A. Hourieh, A.M. Youssef, F. El-Sejariah, Adsorpt.
Sci. Technol. 18 (2000) 2742.
[11] Q. Shizhang, Xijumttu, AIChE J. 46 (2000) 1743.
[12] L. Gordon, Jendrasiak, R.L. Smith, Cell. Mol. Biol. Lett. 5 (2000) 35.
[13] S.J. Gregg, K.S.W. Sing, Adsorption, Surface Area and Porosity,
Academic Press, London, 1982.
[14] X.S. Zhao, Q. Ma, G.Q.M. Lu, Energy Fuels 12 (1998) 1051.
[15] J. Tth, Adv. Colloid Interface Sci. 55 (1995) 149151.
[16] S. Ross, J.P. Olivier, On Physical Adsorption, Interscience, New York,
1964.
[17] M. Khalfaoui, M.H.V. Baouab, R. Gauthier, A.B. Lamine, Adsorpt.
Sci. Technol. 20 (2002) 3347.
[18] S. Knani, DEA, Facult des Sciences de Monastir, 2002.