ASSUMPTIONS OF THE B.E.T.

THEORY 17

T H E ASSUMPTIONS OF T H E B.E.T. THEORY’-*

W. G. McMILLAN
Department of Chemistry, University of California at Los Angeles, Los Angeles, Califomio
AND

E. TELLER
Los Alamos Scientific Laboratory, Los Alamos, New M e z i c o

Received August 10, 1960

INTRODUCTION

The theory of multimolecular adsorption proposed by Braunauer, Emmett,

and Teller (B.E.T.) (1) has been found to explain qualitatively many of the
common features of gas adsorption isotherms and to give a quantitative measure
of the surface area of the adsorbent. There remain, however, important dis-
crepancies between the theoretical and experimental isotherms. For the sigmoid-
shaped isotherms characteristic of multilayer adsorption on nonporous solids,
to which the present discussion will be restricted, the B.E.T. theory generally
predicts an adsorption which is too small a t low pressures and too large in the
multilayer region at pressures approaching saturation. The deviation a t low pres-
sures, which presumably results from an inhomogeneity of the experimental
surface, has already been the subject of a number of papers (4) and will not be
further considered here. We wish rather to direct attention to the other region
of discrepancy, and particularly to the two assumptions of the B.E.T. theory
which are to blame. The first of these assumptions states that although the energy
of adsorption (E1) into the first monomolecular layer is unique, that for all suc-
cessive layers is just the energy of liquefaction ( E L ) . This has recently been
critically considered by Halsey (2). He correctly argues that the extension of
the attractive forces emanating from the adsorbent into layers higher than the
first cannot be ignored, but must generally lead to energies of adsorption which
are greater than E L .I t is clear, however, that if this were the only change made
in the B.E.T. assumptions the resulting theory would predict in the multilayer
region an adsorption exceeding that of the B.E.T. theory, which, as has been
mentioned earlier, itself already exceeds the experimental values. Thus Halsey’s
modification, which is nevertheless essentially correct, gives, with the retention
of the other B.E.T. assumptions, a result still farther from the truth.
The explanation of this difficulty lies in the fact that so far there has been
corrected only one of a pair of partially compensating errors. The second as-
sumption of the B.E.T. theory requiring modification is that the effects of the
1 Presented a t the Twenty-Fourth National Colloid Symposium, which was held under

the auspices of the Division of Colloid Chemistry of the American Chemical Society a t
S t . Louis, Missouri, June 15-17, 1950.
2 This is a qualitative interpretation of a detailed mathematical treatment of the role of
surface tension in gas adsorption which will appear elsewhere.
 18 W. 0. hf-N AND E. TELLER

surface tension of the adsorbate can be neglected. Although this condition is not
expressly stated, it is tacitly demanded by the manner in which the model of
the adsorbed phase enters: the number of adsorption sites covered by exactly n
molecular layers is employed, but no distinction is made aa to whether or not
contiguous sites hold similar numbers of molecules; that they should hold similar
numbers is, of course, preferable if a minimal surface energy is a governing
factor.’
THE SURFACE ENERGY EFFECT
To formulate this objection i s a manner amenable to the necessary ‘modifica-
tion, we shall describe a state (“conformation”) of the adsorbed phase by specify-
ing the number of molecules in the “pile” over each adsorption site. The effect
of ignoring the surface energy is then to assign equal statistical weights to all
those conformations generated from any initial conformation by the shuffling
of the piles among the sites. Thus we find conformations, permitted in the B.E.T.
theory, in which the thickness of the adsorbed phase varies radically as we change
from one site to a neighboring one. W i l e some fluctuation in thickness is clearly
demanded by the thermal agitation, a nonzero surface energy imposes an upper
limit and thus restricts the volume in phase space accessible to the system. The
calculated chemical potential in the adsorbed phase is thereby increased and
the amount of adsorption is correspondingly decreased. Qualitatively, this is in
the right direction to improve the agreement with experiment. However, if we
make this modification without taking account of Halsey’s proposal, we find
that the theoretical isotherm is seriously overcorrected. It is thus apparent that
the B.E.T. theory owes its approximate agreement with experiment in this region
to a fortuitous cancellation of these two neglected effects.
The purpose of the present investigation waa to evaluate quantitatively the
effect of the inclusion of a nonzero surface energy in the theory. By the nature
of the approximations we make, the validity of the following discussion is re-
stricted to the region of large adsorption. For simplicity we shall first retain all
the B.E.T. assumptions (ignoring for the moment Halsey’s modification) with
the exception that the adsorbate shall be assigned a nonzero surface energy.
This has the consequence of associating with very irregular conformations (which
of course have an excessive area of the interface between vapor and adsorbed
phase) a small Boltzmann factor involving the (large) energy of distortion of
the surface. The more irregular the surface, the larger is its area, the larger is
the distortion energy, and the smaller is the resulting weighting factor for that
conformation.
In order to calculate this area for a given conformation, a Fourier analysis
of the surface is made. On the assumption that the inclination of the adsorbate
surface to the underlying adsorbent surface (assumed planar) is everywhere small,
the area of the adsorbate surface in excess of its projected area (Le., the area of
the adsorbent surface) is found to be just the sum of the surface area excesses
contributed by each Fourier component. For the present description we need
I Halsey implicitly asaumed an infinite surface energy. In almost all cases this &odd
lead to satisfactory resulta.
 ABMJMPTIONS OF "BE B.E.T. THEORY 19

know only that this total surface area excess is a function of the amplitudes of the
Fourier waves. The surface deformation energy for a given conformation is
then given as a product of this function and the energy per unit surface.
The virtue of this analysis is that it yields for the deformation energy a simple
expression in place of the complicated one we would have obtained by ad-
hering to the original description of the surface in terms of molecular piles.
Once the Boltzmann factor appropriate to a given conformation is obtained,
one is in a position to evaluate the partition function for the adsorbed phase.
Our retention of the other B.E.T. assumptions permits the energy of the adsorbed
phase to be resolved into a sum of noninteracting terms and thus makes it neces-
sary to calculate only that factor in the total partition function which is de-
pendent upon the surface conformation. This phase of the treatment assumes
the only difference between the adsorbed phase and the reference state of the
semi-infinite liquid to lie in the fact that the amplitudes of the surface waves of
the former are restricted not only by the surface energy but also by the finite
thickness of the layer. On permitting the surface energy to take on the value
zero, as in the B.E.T. treatment, only the latter amplitude restriction remains;
this may be shown to result in an adsorption isotherm which becomes asymptotic
to the B.E.T. isotherm as the pressure approaches saturation. Using the symbol
Y for the average number of molecular layers adsorbed, and z for the relative
pressure (p/po)-Le., the ratio of the equilibrium adsorption pressure p to the
saturation pressure po-this asymptotic form is:
- In z = v-l (1)
The evaluation of the partition function for a nonzero surface energy involves
somewhat greater mathematical complexity, but again the result may be simply
stated. In this case the isotherm equation has the form
- In z = KY-* (2)
wherein the proportionality factor K is a function of the temperature T,the
energy c per unit surface, and the interlayer spacing T :

K kT/18mT2 (3)
THE ISOTHERM EQUATION

Before attempting a comparison of equation 2 with experiment we shall return

for a moment to Halsey's point concerning the energy of adsorption in layers
higher than the first. Hill (3) has quantitatively evaluated the influence of the
extension of the adsorbent attractive forces to these higher layers on the assump-
tion that the adsorbed phase may be represented as a liquid-like slab of uniform
thickness. On this basis he obtains an isotherm equation which, curiously enough,
is identical in form with equation 2, but which has a very different proportionality
factor, K ' . The isotherm equation embodying both this effect and the one due to
the surface energy may be shown to be simply
- lnz = + (K K')Y-* (4)
20 W. G . MCMILLAN AND E. TELLER

This follows from the fact that, for small amplitudes at least, the Fourier waves
do not appreciably alter the average distance of a given molecule from the ad-
sorbent surface; this permits the surface energy and the “altitude energy” to be
treated as independent and leads to the simple additivity in equation 4.Halsey
(21, using a more qualitative route, had obtained an equation of this general func-
tional form but with the power of Y left undetermined, and showed that the data
of Harkins and Jura for adsorption of nitrogen on titanium dioxide do indeed
give a straight line when log(- In 2) is plotted against log Y. The slope of this
line he found to be approximately -2.7, which is not in bad disagreement with
the theoretical value of -3.

tension term ( K -
A crude numerical evaluation of the quantities K and K‘ shows that the surface
5 X lo+) is far outweighed by the term of Hill (K’
Moreover, the latter term is also of the order of the experimental value (about 4,
1). -
from the data cited above). Plotted in the usual form of Y vs. x, equation 2, with
K as in equation 3, gives a curve which lies far below the B.E.T. isotherm. The

inclusion of the surface energy correction alone is thus too drastic, and must be
accompanied by Halsey’s modification to give reasonable agreement with ex-
periment.
From the form of equation 3 it is evident that K becomes less important rela-
tive to K‘ as e increases. This process is accompanied by a lessening of the surface
irregularities, and in the limit where e + m , only the term of Hill remains. This
term was, however, already completely dominant for any reasonable value of e,
and it is thus clear that Hill’s assumption of an adsorbed phase of uniform thick-
ness is completely justified. Inasmuch as the two treatments lead to equations of
identical form, such justification rests on a quantitative evaluation of the surface
energy term which was our aim in this work.
SUMMARY

A qualitative discussion is given of two simplifying assumptions of the B.E.T

theory and the consequences of their replacement by a more exact approach. An
isotherm equation, valid in the region of multilayer adsorption, is obtained which
takes account both of the effect of the surface energy of the adsorbate and of the
extension of the adsorbent attractive forces beyond the first layer. The differ-
ences between this equation and one of identical functional form obtained earlier
by Hill are usually negligible numerically, but the present treatment supplies the
quantitative evaluation necessary for a justification of Hill’s model.
REFERENCES
(1) BRUSAUER, S.,EMMETT, P. H . , A N D TELLER, E.: J . Am. Chem. SOC.60, 309 (1938).
(2) HALSEY, G.: J. Chem. Phys. 16, 931 (1948).
(3) HILL,T . L . : J. Chem. Phys. 17, 590 (1949); 17,668 (1049).
(4) M C ~ T I L L AW.S , G.: J. Chem. Phys. 16, 390 (1947).
W A L K E R , W . C. A N D ZETTLEAIOYER, A . C.: J. Phys. & Colloid Chem. 63, 47 (1948);
63, 58 (1948).