You are on page 1of 9

Advances in Colloid and Interface Science

76]77 Ž1998. 3]11

Adsorption methods for the characterization of


porous materials

Kenneth S.W. Sing


School of Chemistry, Uni¨ ersity of Bristol, Cantock’s Close, Bristol BS8 1TS, UK

Abstract

No other type of adsorption method can offer the same scope for the characterization of
porous solids as gas adsorption. Adsorption from solution measurements are easy to carry
out, but are often difficult to interpret. Although immersion calorimetry is experimentally
demanding, the technique can yield useful information provided that the corresponding
adsorption isotherm data are also available. Nitrogen Žat 77 K. is the most widely used
adsorptive for the characterization of porous materials. Although the Brunauer]
Emmett]Teller ŽBET. theory is based on an over-simplified model of multilayer adsorption,
the BET method continues to be used as a standard procedure for the determination of
surface area. Generally, the derived values of BET-area can be regarded as effective areas
unless the material is ultramicroporous Ži.e. containing pores of molecular dimensions.. It is
advisable to check the validity of the BET-area by using an empirical method of isotherm
analysis. In favourable cases, this approach can be used to evaluate the internal and external
areas. The computation of mesopore size distribution should be undertaken only if the
nitrogen isotherm is of Type IV. Because of network]percolation effects, analysis of the
desorption branch of the hysteresis loop may give a misleading picture of the pore size
distribution; also, a considerable range of delayed condensation is to be expected if the
pores are slit-shaped. Recent work on MCM-41, a model mesoporous adsorbent has
improved our understanding of the mechanisms of mesopore filling. Adsorptive molecules of
different size are required to provide a realistic evaluation of the micropore size distribu-
tion. Q 1998 Elsevier Science B.V. All rights reserved.

Keywords: Adsorbents; Mesopores; Micropores; Physisorption; Surface area

0001-8686r98r$19.00 Q 1998 Elsever Science B.V. All rights reserved.


PII S0001-8686Ž98.00038-4
4 K.S.W. Sing r Ad¨ . Colloid Interface Sci. 76]77 (1998) 3]11

Contents
1. Historical introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2. Physisorption by mesoporous adsorbents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3. Physisorption by micorporous adsorbents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
4. Adsorption at the solidrliquid interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
5. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

1. Historical introduction

At the beginning of this century many investigators continued to hold the view
that when a solution was brought into contact with a porous material, the uptake of
solute involved its penetration into the solid structure. Freundlich w1x was one of
the few who regarded the process as a form of ‘surface condensation’. His
empirical equation, which expressed the amount adsorbed as a function of the
equilibrium concentration Ži.e. the Freundlich adsorption isotherm., provided a
useful basis for comparing the adsorption of dyes and other solute molecules by
such adsorbents as charcoals, clays and silica gels. We may regard these investiga-
tions as some of the first quantitative attempts to use adsorption for the characteri-
zation of porous materials.
The work of Langmuir w2x led to a radical change in the interpretation of
adsorption phenomena. Langmuir was not the first to discuss the thickness of very
thin films. Rayleigh and others had much earlier come to the conclusion that
certain oil films on water were one molecule thick, but Langmuir’s great contribu-
tion was to bring together all the available evidence to support the unifying concept
of the monomolecular layer Žthe monolayer.. In the light of the Langmuir theory, it
seemed reasonable to suppose that the plateau of an adsorption isotherm Žof a
solute or gas. represented completion of the monolayer and thus provided a
measure of the monolayer capacity. It followed that, if the average area occupied
by each adsorbed molecule was known Žor assumed., it would be possible to
estimate the surface area of the adsorbent.
Over the past 75 years many investigators have adopted this approach in their
attempts to use adsorption measurements for the evaluation of the specific surface
area of porous materials. However, in 1916 Langmuir w2x had already pointed out
that the surface area of a highly porous adsorbent cannot be easily defined and, in
particular, that equations derived for plane surfaces should not be applied to the
adsorption by such adsorbents as charcoal. These reservations were reinforced in
1932 by Rideal w3x who drew attention to the importance of the accessibility of the
surface and to various problems which may arise in the application of adsorption
measurements Že.g. dependence on the surface structure ..
It was over 60 years ago that Brunauer and Emmett w4x made their first attempts
to determine the surface area of an iron synthetic ammonia catalyst by means of
low-temperature gas adsorption. Their work culminated in the publication in 1938
K.S.W. Sing r Ad¨ . Colloid Interface Sci. 76]77 (1998) 3]11 5

of the Brunauer]Emmett]Teller ŽBET. theory w5x, which has subsequently at-


tracted an enormous amount of attention } both support and criticism. Indeed,
the BET method is still used as a standard procedure for surface area determina-
tion, although the BET theory is now known to be based on an over-simplified
model of multilayer adsorption.
A novel heat of immersion method for surface area determination was intro-
duced in 1944 by Harkins and Jura w6x, which at the time was claimed to be an
‘absolute method’ since it did not depend on the evaluation of monolayer coverage.
The idea was to coat a solid surface with a thick multilayer so that its surface
enthalpy would become identical to that of the pure liquid. Immersion in the same
liquid should then result in the release of an amount of energy which would simply
depend on the removal of an area of liquid surface. Although the method attracted
a great deal of immediate interest, in the 1940s experimental techniques were not
sufficiently advanced to provide reliable calorimetric data. The method was dif-
ficult to apply and its potential value therefore remained unexploited for many
years.
The use of gas adsorption for pore size analysis was pioneered by Foster w7x and
further developed by Shull w8x and Barrett et al. w9x. The latter computational
procedure Žthe BJH method. probably remains the most popular method for pore
size analysis } or more strictly, for the evaluation of the mesopore size distribu-
tion. This may be regarded as the ‘classical’ approach since it is dependent on the
application of the Kelvin equation. The range of validity of the Kelvin equation
remains a much discussed and unresolved problem.
Langmuir’s early comments concerning the nature of the adsorption by highly
porous materials were not fully appreciated until the 1950s when Dubinin w10x
provided unambiguous evidence that the mechanism of physisorption in very
narrow pores is not the same as in wider pores or on the open surface. By studying
the properties of many activated carbons, Dubinin was able to identify three groups
of pores Žoriginally termed micropores, transitional pores and macropores.. The
IUPAC w11x classification of pore width, which is based on the principles developed
by Dubinin w12x and by Gregg and Sing w13x is as follows: micropores - 2 nm;
mesopores, 2]50 nm; macropores ) 50 nm.
In the 1960s the growing awareness of the specificity of physisorption was mainly
due to the systematic work of Barrer w14x and Kiselev w15x. By studying the
adsorption of a wide range of polar and non-polar molecules by zeolites and other
well-defined adsorbents they were able independently to demonstrate the impor-
tance of the adsorbent]adsorbate and adsorbate]adsorbate interactions in addi-
tion to that of the pore structure.
The five hypothetical types of physisorption isotherms originally proposed by
Brunauer et al. w16x in 1940 were incorporated into a more practical classification
by the IUPAC w11x in 1985. The reversible Type I isotherm was originally known as
the Langmuir isotherm because it appeared to represent the Langmuir mechanism
of monolayer coverage. In fact, most Type I isotherms can be attributed to
physisorption by microporous adsorbents with relatively small external surfaces. In
contrast, Type II isotherms ŽS-shaped or sigmoid. are monolayer]multilayer
6 K.S.W. Sing r Ad¨ . Colloid Interface Sci. 76]77 (1998) 3]11

isotherms, which are normally obtained with non-porous or macroporous adsor-


bents. The more common Type IV isotherms are given by adsorbents which possess
mesopore structures, such as many silica gels and other porous oxides. The
remaining isotherms in the IUPAC classification exhibit particular features, which
indicate weak gas]solid interactions in the case of Types III and V isotherms and a
high degree of surface uniformity in the case of the stepwise Type VI isotherm.

2. Physisorption by mesoporous adsorbents

The characteristic shape of a Type IV isotherm is the result of surface coverage


of the mesopore walls followed by pore filling. The onset of capillary condensation
Žthe pore filling process. is indicated by an upward departure of the isotherm from
the multilayer Type II isotherm for the same gasrsolid system. The plateau at
higher prp 0 is attained when the mesopore filling is complete.
If the Kelvin equation is used to evaluate the mesopore size, it is necessarily
assumed that a simple relationship relates the meniscus curvature to the pore
shape and size. Thus, if the pores are cylindrical, the meniscus shape is hemispheri-
cal; but if the pores are slit-shaped, the meniscus becomes hemicylindrical. Other
assumptions involved in the computation of the mesopore size distribution are that
the pores are rigid and that a standard multilayer correction curve can be applied
w17x. To obtain the mesopore volume from the amount adsorbed at the plateau, the
condensate is assumed to have the same density as the liquid adsorptive at the
operational temperature.
A long-standing problem in mesopore analysis is the interpretation of the
various hysteresis loops associated with most Type IV isotherms. For many years it
was thought that the desorption branch of the loop represented thermodynamic
equilibrium and therefore should be adopted for the pore size analysis. This
practice is now questionable since the path of the desorption branch is often
dependent on network-percolation effects w18x. On the other hand, it is known that
on the adsorption branch delayed condensation is the result of the persistence of a
metastable multilayer w19x, this effect being especially pronounced in slit-shaped
pores.
It is of great advantage to have independent information on the pore shape and
connectivity, but the size and shape of the loop itself can also give a useful
indication of the predominant pore fillingremptying mechanisms. Thus, a narrow
Type H1 loop, with almost vertical and parallel branches, is generally associated
with delayed condensation and very little percolation hold-up, whereas a much
broader Type H2 loop Žwith a very steep desorption branch. has the typical
features associated with network-percolation.
Recent work w20]23x on MCM-41, a model mesoporous adsorbent, has revealed
that it is possible to obtain well-defined reversible Type IV isotherms. The pore
structure of MCM-41 is in the form of hexagonal arrays of uniform tubular
channels of controlled width. Most attention has been given so far to a form of
MCM-41 with 4-nm pores. With samples of this grade of MCM-41 the reversible
K.S.W. Sing r Ad¨ . Colloid Interface Sci. 76]77 (1998) 3]11 7

capillary condensationrevaporation of nitrogen at 77 K occurs over the narrow


range of prp 0 s 0.41]0.46. By applying the Kelvin equation and correcting for
multilayer thickness, we obtain pore diameters of 3.3]4.3 nm, which agrees well
with a mean value of approximately 4 nm derived from the volumersurface ratio.
A number of other adsorptives Žargon, oxygen, carbon dioxide, sulfur dioxide and
various alcohols. have been found to give Type H1 hysteresis loops w20,21x with the
4-nm MCM-41. In another investigation by Branton w23x, it was found that the
carbon tetrachloride isotherms determined at temperatures between 273 and 303 K
on a 3.4-nm pure silica form of MCM-41 exhibited steep and very narrow hysteresis
loops, whereas the corresponding isotherm at 323 K was completely reversible.
When they are confined to very narrow ranges of prp 0 , the reversible pore-fill-
ing steps appear to be equivalent to first order phase transformations, the prp 0
being dependent on the adsorptive and the temperature. It seems significant that
for a given adsorptive the characteristic prp 0 is remarkably close to the lower limit
of closure of the hysteresis loop Ži.e. the limiting chemical potential controlling the
stability of the capillary condensate.. We may conclude that the reversible stepwise
filling of the mesopores in MCM-41 is due to: Ž1. the absence of pore blocking
effects; Ž2. the tubular pore shape; and Ž3. the particular range of its narrow pore
size distribution in relation to the temperature and properties of the adsorptive.

3. Physisorption by microporous adsorbents

As already noted, in the IUPAC classification of pore size the upper limit of the
micropore width was placed at approximately 2 nm. It turns out that this limiting
width is somewhat arbitrary since the mechanism of pore filling is not determined
by pore width alone.
The high prp 0 plateau of a well-defined Type I isotherm always extends over a
wide multilayer range. The great majority of Type I isotherms can be attributed to
micropore filling, but there are a few systems Že.g. butanol on alumina. with which
a form of ‘gas phase autophobicity’ inhibits the development of the multilayer. We
can be sure that this behaviour is not possible with simple adsorptive molecules
ŽAr, Kr, N2 , O 2 , etc.. or even with alkanes and other larger molecules of low
polarity. With these adsorptives, the low multilayer slope is a direct consequence of
a small external area.
It is now apparent that Type I isotherms can be broadly divided into two groups.
Ultramicroporous carbons and molecular sieve zeolites exhibit high adsorption
affinity, their isotherms generally having a high degree of rectangularity with the
plateau approached at very low prp 0 . The second group of Type I isotherms are
given by many ‘supermicroporous’ activated carbons and oxide xerogels which have
somewhat wider pores. In this case, the initial part of the isotherm is less steep and
the approach to the plateau more gradual. However, this difference in isotherm
shape is not really controlled by the absolute pore width: in fact, it is dependent on
the pore width and geometry in relation to the size, shape and electronic character
of the adsorptive molecules.
8 K.S.W. Sing r Ad¨ . Colloid Interface Sci. 76]77 (1998) 3]11

The micropores in activated carbons tend to be slit-shaped. In the ultramicro-


pores of this shape there is a significant overlap of adsorption forces provided that
the pore widths are not much larger than two molecular diameters w24x. This is
manifested in the form of enhanced adsorption energies, which are responsible for
the steepness of the isotherm at low prp 0 . This process has been termed ‘primary
micropore filling’ w25x. Supermicropores in the range of approximately two to five
molecular diameters are filled by a combination of surface coverage at low prp 0
and a cooperative process Žor quasi-multilayer adsorption. at higher prp 0 Žto
; 0.2.. Since it involves two overlapping stages, supermicropore filling is not a first
order transition, as in the case of capillary condensation.
Various procedures have been proposed w26x for the evaluation of the micropore
volume from experimental isotherm data Že.g. the Dubinin]Radushkevich plot.,
but it should be kept in mind that in practice these are empirical methods. Thus,
the observation that a particular equation fits part of an experimental isotherm at a
certain temperature is insufficient evidence to validate the underlying theory.
Another type of empirical approach is to plot the amount adsorbed against the
corresponding standard data determined on a non-porous reference adsorbent.
This form of comparison plot Žor a s-plot. has been found useful for the assessment
of micropore capacities and, with some systems, internal and external areas w27x.
To convert the micropore capacity into the micropore volume, it is usually
assumed that the pores are filled with the adsorptive in its normal liquid form }
as in mesopore filling. It is evident that in pores of molecular dimensions, this
practice does not allow for the dependency of molecular packing on both pore size
and shape. Therefore it is advisable to use the term apparent micropore volume
and refer to the adsorptive and operational temperature.
For the reasons already given, nitrogen adsorption alone cannot be expected to
give more than a qualitative indication of the micropore size distribution. To obtain
a more detailed picture of the range of micropore size, a number of probe
molecules of different size is required. This number can be minimised by taking
account of the stages of micropore filling and hence focus attention on the
mechanisms of pore filling rather than on the apparent pore size.

4. Adsorption at the solid r liquid interface

As we have seen, adsorption from solution measurements have been used for
many years for the characterization of activated carbons, clays and some other
materials. The interpretation of the experimental data is complicated by the fact
that both solute and solvent molecules are present at the solidrliquid interface.
Thus, the decrease in solute concentration, which normally occurs when the
adsorbent and solution are brought into contact, may not give a simple measure of
the amount of solute adsorbed w28x. However, with some systems the solute is
preferentially adsorbed to the extent that it covers virtually all the surface at low
concentration.
K.S.W. Sing r Ad¨ . Colloid Interface Sci. 76]77 (1998) 3]11 9

Different solutionrsolid adsorption systems give a variety of isotherm types,


which were classified by Giles et al. w29x. Some isotherms have the characteristic
Type I appearance Žwith the equilibrium concentration replacing the gas pressure..
If the plateau then extends over an appreciable range of concentration it is likely
to correspond to monolayer completion. This behaviour, which is exemplified by
the adsorption of a long chain fatty acid from an alkane solution on alumina w30x,
appears to be consistent with the classical Langmuir mechanism.
Unfortunately, there are a number of reasons why adsorption from solution does
not provide a reliable method for the determination of surface area:
1. Unless the plateau is attainable, any attempt at extrapolation Žor application of
the Langmuir equation. is unlikely to yield the true monolayer capacity of the
solute.
2. Most surfaces are heterogeneous and the surface coverage is not uniform.
3. The orientation of the adsorbate molecules may vary considerably from one
surface to another and adsorbate]adsorbate interaction often determines the
structure of the adsorbed phase.
4. Adsorption of the solvent is dependent on its interaction with both adsorbent
and adsorbate.
Few attempts have been made to investigate the effect of porosity on an
adsorption from solution isotherm. However, an exploratory investigation w31x has
indicated that it is possible to apply the a s-method to iodine and salicylic adsorp-
tion data for the evaluation of the micropore capacity and external surface area.
Immersion calorimetry is another useful technique for the characterization of
some porous materials. The heat of immersion Žor heat of wetting. is defined as the
amount of heat evolved when a known mass of an outgassed solid is completely
immersed } but not dissolved } in a given liquid. When determined under
conditions of constant pressure and temperature it can be regarded as the enthalpy
of immersion, which in turn is directly related to the integral enthalpy of adsorp-
tion for the equivalent gas]solid system.
The magnitude of the enthalpy of immersion is dependent on the nature of the
liquid]solid interactions and the extent of the available surface. In principle, if the
areal enthalpy of immersion is known for the given liquid]solid system, it should
be possible to determine the specific surface area by a single heat measurement.
Indeed, in the early work it was assumed that the areal enthalpy of immersion in a
particular liquid was almost entirely controlled by the properties of the liquid, but
it is now evident that the surface chemistry and pore structure can both play an
important role. For example, the enthalpy of immersion of an oxide in a polar
liquid is highly dependent on its pretreatment w32x, whereas it is the micropore
structure which is the more important when an activated carbon is immersed in a
polar or non-polar liquid w33x.
Improvements in the design and operation of immersion calorimeters have made
it possible to revisit the Harkins]Jura method w8x. In the work of Partyka et al. w34x
it was found that two molecular layers of water are sufficient to screen the
10 K.S.W. Sing r Ad¨ . Colloid Interface Sci. 76]77 (1998) 3]11

underlying surface of many adsorbents. It therefore becomes possible to apply the


technique to mesoporous solids by avoiding the complication of capillary condensa-
tion. In combination with gas adsorption, the Harkins]Jura method becomes a
powerful tool for the characterization of porous materials.
If the overall surface area is at least 10 m2 , the enthalpy of immersion is not
difficult to measure with the aid of a modern calorimeter. However, the recorded
values are of scientific value only if the measurements are made under carefully
controlled conditions and the changes of state clearly recorded w35x.

5. Conclusions

At present, no other type of adsorption method can offer the same scope for the
characterization of porous materials as gas adsorption. Adsorption from solution
measurements are relatively easy to carry out, but are often difficult to interpret.
However, they are esential if the adsorbent is to be used for the treatment of liquid
media. Although enthalpies of immersion are more difficult to determine, they can
yield useful information provided that the corresponding adsorption isotherm data
are also available.
The concept of ‘surface area’ is of great practical value, but as indicated by
fractal analysis, derived values of surface area should not be regarded as ‘absolute’
areas. It is remarkable that the BET-nitrogen method continues to be the favoured
procedure for determining the overall ‘effective’ surface area of adsorbents, cata-
lysts and various other porous materials. It is generally agreed that the method is
unreliable when applied to an ultramicroporous adsorbent, which contains pores of
molecular dimensions Že.g. a molecular sieve carbon or zeolite.. It appears that
nitrogen adsorption can be used to assess the internal surface area of a supermi-
croporous material provided that it can be established that monolayer adsorption
has preceded cooperative pore filling.
The computation of mesopore size distribution should be undertaken only if the
isotherm is of Type IV. Even in this case, little is to be gained by the use of an
elaborate computational procedure. The decision as to which branch of the
hysteresis loop should be used remains somewhat arbitrary unless the pore shape
and interconnectivity are known. If this information is not available, it is probably
safer to use the adsorption branch, although the possibility of delayed condensa-
tion should be considered.
No reliable procedure has been developed for the computation of the micropore
size distribution from a single isotherm. It is preferable to employ adsorptive
molecules of different size as molecular probes and take account of the different
stages of pore filling. Further progress in the elucidation of the mechanisms of
pore filling is likely to depend on the development of model pore structures,
refined theoretical analysis with the aid of density functional theory and computer
simulation and also the application of high resolution experimental techniques
including adsorption calorimetry.
K.S.W. Sing r Ad¨ . Colloid Interface Sci. 76]77 (1998) 3]11 11

References
w1x H.M. Freundlich, Z. Phys. Chem. 57 Ž1907. 385.
w2x I. Langmuir, J. Am. Chem. Soc. 38 Ž1916. 2221.
w3x E.K. Rideal, The Adsorption of Gases by Solids, Disc. Faraday Soc. 1932, p. 139.
w4x S. Brunauer, P.H. Emmett, J. Am. Chem. Soc. 57 Ž1935. 1754.
w5x S. Brunauer, P.H. Emmett, E. Teller, J. Am. Chem. Soc. 60 Ž1938. 309.
w6x W.D. Harkins, J. Jura, J. Am. Chem. Soc. 66 Ž1944. 1362.
w7x A.G. Foster, Trans. Faraday Soc. 28 Ž1932. 645.
w8x C.G. Shull, J. Am. Chem. Soc. 70 Ž1948. 1410.
w9x E.P. Barrett, L.G. Joyner, P.H. Halenda, J. Am. Chem. Soc. 73 Ž1951. 373.
w10x M.M. Dubinin, Q. Rev. Chem. Soc. IX Ž1955. 101.
w11x K.S.W. Sing, D.H. Everett, R.A.W. Haul, et al., Pure Appl. Chem. 57 Ž1985. 604.
w12x M.M. Dubinin, in: P.L. Walker ŽEd.., Chemistry and Physics of Carbon, vol. 2, Marcel Dekker, New
York, 1966, p. 51.
w13x S.J. Gregg, K.S.W. Sing, Adsorption, Surface Area and Porosity, Academic, London, 1982, p. 25.
w14x R.M. Barrer, J. Colloid Interface Sci. 21 Ž1966. 415.
w15x A.V. Kiselev, Discuss. Faraday Soc. 40 Ž1965. 205.
w16x S. Brunauer, L.S. Deming, W.S. Deming, E. Teller, J. Am. Chem. Soc. 62 Ž1940. 1723.
w17x S.J. Gregg, K.S.W. Sing, Adsorption, Surface Area and Porosity, Academic, London, 1982, p. 132.
w18x A.V. Neimark, in: F. Rodriguez-Reinoso, J. Rouquerol, K.S.W. Sing, K.K. Unger ŽEds.., Characteri-
zation of Porous Solids II, Elsevier, Amsterdam, 1991, p. 67.
w19x G.H. Findenegg, S. Gross, Th. Michalski, in: J. Rouquerol, F. Rodriguez-Reinoso, K.S.W. Sing,
K.K. Unger ŽEds.., Characterization of Porous Solids III, Elsevier, Amsterdam, 1994, p. 71.
w20x P.J. Branton, P.G. Hall, K.S.W. Sing, H. Reichert, F. Schuth, K.K. Unger, J. Chem. Soc., Faraday
Trans. 90 Ž1994. 2965.
w21x P.J. Branton, P.G. Hall, K.S.W. Sing, Adsorption 1 Ž1995. 77.
w22x P.I. Ravikovitch, S.C. O’Domhnaill, A.V. Neimark, F. Schuth, K.K. Unger, Langmuir 11 Ž1995.
4765.
w23x P.J. Branton, K.S.W. Sing, J.W. White, J. Chem. Soc., Faraday Trans., 93 Ž1997. 2337.
w24x D.H. Everett, J.C. Powl, J. Chem. Soc. Faraday Trans. I 72 Ž1976. 619.
w25x P.J.M. Carrott, K.S.W. Sing, in: K.K. Unger, J. Rouquerol, K.S.W. Sing, H. Kral ŽEds.., Characteri-
zation of Porous Solids I, Elsevier, Amsterdam, 1988, p. 77.
w26x F. Rodriguez-Reinoso, Pure Appl. Chem. 61 Ž1989. 1859.
w27x K.S.W. Sing, Colloids Surf. 38 Ž1989. 113.
w28x D.H. Everett, Pure Appl. Chem. 58 Ž1986. 967.
w29x C.H. Giles, D. Smith, A. Huitson, J. Colloid Interface Sci. 47 Ž1974. 755.
w30x J.H. de Boer, G.M.M. Houben, B.C. Lippens, W.H. Meijs, W.K.A. Walgrave, J. Catal. 1 Ž1962. 1.
w31x J. Fernandez-Colinas, R. Denoyel, J. Rouquerol, Adsorpt. Sci. Technol. 6 Ž1989. 18.
w32x A.C. Zettlemoyer, K.S. Narayan, in: E.A. Flood ŽEd.., The Solid-Gas Interface I, Marcel Dekker,
New York, 1967, p. 145.
w33x J. Fernandez-Colinas, R. Denoyel, Y. Grillet, J. Vandermeersch, J.L. Reymonet, F. Rouquerol, J.
Rouquerol, in: A.B. Mersmann, S.E. Scholl, Fundamentals of Adsorption III, Engineering Founda-
tion, New York, 1991, p. 261.
w34x S. Partyka, F. Rouquerol, J. Rouquerol, J. Colloid Interface Sci. 68 Ž1979. 21.
w35x C. Letoquart, F. Rouquerol, J. Rouquerol, J. Chim. Phys. 70 Ž1973. 559.

You might also like