Feb.

, 1938 Adsorption of Gases in Multimolecular Layers 309

fContribution from the Bureau of Chemistry and Soils and George Washington University]

Adsorption of Gases in Multimolecular Layers

By Stephen Brunauer, P. H. Emmett and Edward Teller

Introduction The Polarization Theory of DeBoer

I.
The adsorption isotherms of gases at tempera- and Zwicker
tures not far removed from their condensation According to DeBoer and Zwicker, the induced
points show two regions for most adsorbents: dipole in the fth layer polarizes the i + 1st
at low pressures the isotherms are concave, at layer, thus giving rise to induced dipole moments
higher pressures convex toward the pressure axis. and binding energies that decrease exponentially
The higher pressure convex portion has been with the number of layers. If we call the dipole
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

variously interpreted. By some it has been moment of a molecule in the i-th layer µ,·, it
attributed to condensation in the capillaries of follows that
the adsorbent on the assumption that in capil- w =
c,C‘ (1)
Downloaded via DONG-A UNIV on December 12, 2023 at 09:32:50 (UTC).

laries of molecular dimensions condensation can where and C are appropriate constants, C actu-
C\
occur at pressures far below the vapor pressure of ally being equal3 to µ,/µ, v The corresponding _

the liquid. By others such isotherms are believed binding energy is proportional to the square of the
to indicate the formation of multimolecular ad- dipole moment
sorbed layers. DeBoer and Zwicker1 explained =
CíC2' (2)
the adsorption of non-polar molecules on ionic where is another constant.
c% The equilibrium
adsorbents by assuming that the uppermost layer pressure of the nth. layer (top layer), pn, according
of the adsorbent induces dipoles in the first layer to Boltzmann’s law varies exponentially with the
of adsorbed molecules, which in turn induce binding energy of that layer and, if the only bind-
dipoles in the next layer and so on until several ing energy were that due to polarization, would
layers are built up. The isotherm equation which be given by the equation
they, and later Bradley,2 derived on the basis of pn =
cie~**,RT (3)
this polarization theory is practically the only It follows, therefore, that
quantitative expression that has been so far pro- = “

(4)
posed to account for multimolecular adsorption. lnoT wC2n
However, as we shall show in the first part of this which is identical with DeBoer and Zwicker’s
paper, the polarization of the second layer of equation
adsorbed gas by the first layer is already much too
small to constitute the major portion of the bind-
In
^ =
A2A," (4a)

ing energy between the two adsorbed layers, at if c5 be replaced by K¡po, —

cz/RT by Kt, and C2 bv

least in those instances in which the gas molecules
do not possess considerable permanent dipole One can substitute n = v/vm, where v is the vol-
moments. ume of gas adsorbed at pressure pn, and &m is
It seems to us that the same forces that produce the volume adsorbed in one complete unimolecular
condensation are chiefly responsible for the bind- layer. DeBoer and Zwicker,1 and subsequently
ing energy of multimolecular adsorption. On (3) According to DeBoer and Zwicker the µ values for all layers,
this assumption, in the second part of this paper except the first and the top layer, satisfy the equations
« = ¿(/Ui-i + µ.+ ) (la)
we shall carry out a derivation of the isotherm
Equation (1) is a solution of (la) if
equation for multimolecular adsorption by a ^ 1 -
Vl -
4&2
method that is a generalization of Langmuir’s treat-
_

2k
ment of the unimolecular layer. In the third DeBoer and Zwicker give the approximate expression

part of the paper we shall then apply the isotherm C = k/(l k2) -
(lc)
which for k 1 is a good approximation for (lb). It should be noted,
equation to a variety of experimental isotherms however, that when k 1 only a few layers are usually adsorbed
obtained by others and by us on a number of and, since equation (la) is not valid for the top layer, in this case
all of the equations are to be considered approximate. Bradley,
catalysts, catalyst supports and other adsorbents. using equation (lc), derived the values of 0.6075 and 0.615 for k
(1) DeBoer and Zwicker, Z. physik. Chem., B3, 407 (1929). for copper sulfate and aluminum sulfate, respectively, even though
(2) Bradley, J. Chem. Soc., 1467 (1936). the upper limit of k according to (lb) is 0.5.
310 Stephen Brunauer, P. H. Emmett and Edward Teller Vol. 60

DeBoer4,6 in several papers showed that various infinite wave length, and vs is the gram molecular
experimental adsorption isotherms could be fitted volume of the gas divided by Avogadro’s number.
by equation (4a). However, they could not We shall assume, to start with, that the separa-
evaluate Ki because in all cases, except one, the tion of the argon atoms in the adsorbed state is
surface, and therefore vm, was not known. In about the same as in the solid state. For a face-
the only case where the surface was known6 the centered cubic lattice, such as that of solid argon,
adsorption proved to be unimolecular in thickness, the distance r between the nearest neighbors is
and thus the application of the equation would given by the equation
not have been justified. r3 V2v, (6)
=

Bradley2 arrived at an equation identical to where vs is the molecular volume of the solid
(4a) except that Ki is taken as unity. By apply- divided by Avogadro’s number. It follows from
ing it to his data for argon adsorption on copper (5) and (6) that
sulfate and aluminum sulfate, and by estimating „
w ~
1
Ü» (7)
K /
-

vm from microscopic measurements of the diame- r3 23'3v v.

ters of the adsorbent particles, he evaluated For argon gas at 0° and 760 mm. pressure, 1 —

Ki, and obtained k values of about 0.6. He con- =

278 X 10 ~6,8 and by using the density of solid
cluded that the very strong polarization of ad- argon at 40 °K.9 one obtains
sorbed argon explained the relatively thick ad- a/r3 =
0.029 (8)
sorption layers (at half of the saturation pressure If we that the separation of the argon
assume
more than 30 layers) that he believed he was ob- atoms in the adsorbed state is the same as in the
taining. Emmett and Brunauer7 have already liquid state, a/r3 becomes even smaller.
pointed out reasons for believing that the surface The ratio of the induced dipoles in two succes-
areas of the samples used by Bradley were prob- sive layers is given by the equation
ably many times greater than he estimated, and (w + O/w C d(a/r3) (9) = —

that therefore his estimate of the thickness of his where d is a constant dependent upon the geo-
films was high by possibly a factor of 20 or so. It metrical structure of the adsorbed layers and the
will now be shown that the ratio of the strength relative orientation of the induced dipoles. If
of the dipoles induced in the ith layer to that the adsorbed argon atoms build up in a close-
induced in the dipoles of the t-lst layer is of the packing of spheres and the dipoles in one layer
order of 0.01 for argon rather than 0.99, a value are all oriented in the same direction and per-
that one obtains from Bradley’s k 0.615 for
=
pendicular to the surface, then the value calcu-
aluminum sulfate. Accordingly the polarization lated for d becomes —0.35.10 The minus sign
of one layer of argon by the next lower layer is far (8) “International Critical Tables,” Vol. VII, p. 11.
too small to constitute the major portion of the (9) Ibid., Vol. I, p. 103.
distance
(10) The x component of an electric field produced at a

binding energy between successive adsorbed layers. \s(xi -

The estimation of the magnitude of C, the ratio p by a dipole µ pointing m the x direction is µ
jJ
of the strength of the induced dipoles in two suc- where x\ and x% are x coordinates of the dipole and the point at which
the field is calculated. We take the x direction perpendicular to the
cessive layers, can be carried out in the following surface, and we denote by r, as before, the distance between the
manner. The field of a dipole µ at a distance r nearest neighbors. Then x\ xt for atoms in the same layer is zero, and
—

for atoms in successive layers is \/2/3r for close-packing of spheres.

is proportional to µ /r3, and the dipole, µ,, induced Then the field produced in the center of an atom in the ¿th layer by
in a molecule of polarizability a at that distance all the other atoms in the same layer is µ/ (1/ 9), where the sum- —

mation has to be extended over all the other atoms of the layer.
is proportional to µ a/r3. The ratio µ /µ is The field produced in the center of this same atom by all the atoms
therefore proportional to the dimensionless num- /2r2 1 \
in the ¿-1st and ¿ + 1st layer is (ju< -1 -j- jue' +1)
—

ber a/r3, and the ratio of the binding energies in PV*

The numerical evaluation of the infinite sums yields the results
two successive layers to a2/r6. —
«(11.1/r8) and (µ<_ + µ<+ )(8.357-8,823/r8). The dipole µ;
The polarizability of a gas molecule, a is given induced in the atom by the sum of these forces is

by the equation Pi =
pi I — . µ»
—
0.466 (µ,·_ + w+l))
2 =
(n —

l)t'„ (5) giving for a/r9 0.029 the result

where n is the index of refraction extrapolated to
µ< = —0.35
^ (µ<- + Mi+i)
(4) DeBoer, Z. physik. Chem., BIS, 134 (1931).
(5) DeBoer, ibid., B14, 149 (1931). Comparison with equation (la) in the footnote to page 2 shows that
(6) DeBoer, ibid., BIT, 161 (1932). k - 0.35 a/r9, and since k 1, therefore k is approximately equal to
(7) Emmett and Brunauer, This Journal, 99, 1553 (1937). C.
Feb., 1938 Adsorption of Gases in Multimolecular Layers 311

signifies that the dipoles in successive layers point for multimolecular layers that is similar to
alternately toward the surface and away from the Langmuir’s derivation for unimolecular layers.
surface. From (8) and (9) we obtain C —0.01
= The equations obtained appear not only to repre-
and therefore K\ (of equation 4a), or C2 (of sent the general shape of the actual isotherms,
equation 4) equals about 1 X 10~4. Hence the but to yield reasonable values for the average
binding energy that can be attributed to polariza- heat of adsorption in the first layer and for the
tion is negligibly small already in the second volume of gas required to form a unimolecular
layer. layer on the adsorbents.
In the calculation of d it was assumed that In carrying out this derivation we shall let so,
the dipoles can be localized in the centers of the j'i, Si, .
s¡,
. .
represent the surface area that is
. . .

atoms, and are induced by forces acting at the covered by only 0, 1, 2, layers of ad- .

centers of the atoms. Furthermore, the calcula- sorbed molecules. Since at equilibrium 50 must
tion was made only for the case in which the remain constant the rate of condensation on the
dipoles in a given layer all point in the same direc- bare surface is equal to the rate of evaporation
tion and are perpendicular to the surface. Though from the first layer
it is realized that for slightly different assump- aipso =
biSie~Bl/BT (10)
tions and for different arrangements and orienta- where p is the pressure, Ei is the heat of adsorp-
tions of the dipoles the value of d will differ from tion of the first layer, and ai and bi are constants.
the above, it seems certain that for gases for This is essentially Langmuir’s equation for uni-
which a/r* is as small as for argon, the portion molecular adsorption, and involves the assump-
of the binding energy due to polarization forces tion that ai, bi, and E, are independent of the
operating between the various layers of adsorbed number of adsorbed molecules already present in
gas must be very small. For molecules with the first layer.
larger polarizability a/rz will, of course, be larger At equilibrium Si must also remain constant,
than for argon. DeBoer chose iodine for his
Ji can change in four different ways: by con-
experiments because of its larger polarizability. densation on the bare surface, by evaporation
Even in this case, however, since /r3 is about 0.1 from the first layer, by condensation on the first
it is unlikely that the energy of binding due to
layer, and by evaporation from the second layer.
polarization will constitute an appreciable portion We thus obtain
of the binding energy of the second layer of ad-
a¡ps¡ + bísle~E'/nr =
biSie~E*/RT + a,pso (11a)
sorbed iodine. On the other hand, if the ad-
sorbed gas has a large permanent dipole it is where the constants a2, b2, and E2 are similarly
possible that many layers may be successively defined to au bx and E¡. From (10) and (11a)
polarized by the mechanism of DeBoer and follows
Zwicker. This case has been treated by Bradley.11 02 psi =
baSie~E*/RT (11)
It may be well at this point to call attention the rate of condensation on top of the first layer is
to the fact that equation (4a) in its logarithmic equal to the rate of evaporation from the second
form and with K2 equal to unity has been used layer.12 Extending the same argument to the
frequently to present adsorption data. Thus a second and consecutive layers we obtain
plot of log log p/po against the amount adsorbed a¡pSí =
bss3e"E‘/Rr (12)
has often been found to be linear over a consider-
able range of p/p<¡ values. In view of our con- üips¡-1 =
biSie~E¡/RT

clusions relative to the polarization theory it The total surface area of the catalyst is given by
seems that such representations are to be re-
garded, for the present, as empirical though _
00i.o* <13>

interesting relationships. (12) Equation (11) could have been arrived at immediately by
applying the principle of microscopic reversibility, because the
. number of evaporation processes by which sz decreases and si in-
Generalization of Langmuir’s Theory creases must be equal to the number of condensation processes by
to Multimolecular Adsorption which ss increases and si decreases. It follows also from the same
principle that every other process that can occur (such as, for
With the help of a few simplifying assumptions example, the sliding over of a molecule from one layer into another),
it is possible to carry out an isotherm derivation will be balanced by a reverse process of equal frequency. Thus the
equilibrium, expressed in (11), will not be influenced by such further
(11) Bradley, J. Chem. Soc., 1799 (1936). processes.
312 Stephen Brunauer, P. H. Emmett and Edward Teller Vol. 60

and the total volume adsorbed is ent. To make v = ®, when p po, —

x must be
equal unity. Thus from (19)
to
® i = o
isi (14)
(Po/g)eBh/RT l, and
where v0 is the volume of gas adsorbed on one x =
p/po (27)
square centimeter of the adsorbent surface when Substituting into (26) we obtain the isotherm
it is covered with a complete unimolecular layer equation
of adsorbed gas. It follows that
^
+ (e l){P/Po)\
- V":
P) 11
- -

(Po
v V The following considerations show that equa-
(15)
Avq Vrr
tion (28) gives, indeed, an S-shaped isotherm.
/ J·* I
The constant c, as a rule, will be large compared
=

where vm is the volume of gas adsorbed when the to unity, and therefore the isotherm will consist
entire adsorbent surface is covered with a com- of two regions. The low pressure region will be
plete unimolecular layer. concave to the pressure axis, and for p «C po
The summation indicated in equation (15) can equation (28) reduces to
be carried out if we make the simplifying assump-
tions that
Eo Ez ...£, £l (16)
--(^)/(1+¿») (29)

which is a special form of the Langmuir equation.

- = =

where EL is the heat of liquefaction, and At higher pressures, as p approaches po, v becomes
(17) large, and the curve becomes convex to the pres-
Ü2 O3 &i sure axis.
g being an appropriate constant. This is equiva- For the purpose of testing, equation (28) can
lent to saying that the evaporation-condensation be put in the form
properties of the molecules in the second and
_±__|_ ^_tC.
r _

(A)
higher adsorbed layers are the same as those of v{po p) f»c vmc po
~

the liquid state. We can express now Ji, s2, s3, s¡ .

Equation (A) is in a convenient form, since a

in terms of s0. plot of p/v{Po p) against p/pa should give a
—

V) II where y
<-0
=
{ai/h)pe*i/RT (18) straight line, whose intercept is 1 /vmc and whose
H =
m where x =
(p/g)e1:i-',tT (19) is 1 Thus from the slope and
,

slope (c )/vmc.
—

ss = =
X2Si (20) the two
intercept constants vm and c can be
=
XSi-1 = = 1
S{ =
yx'~1 = ex's 0 (21)
evaluated, the former being the volume of gas
where y íil eE,~ Eh/RT (22) ad-
X !h required to form a complete unimolecular
obtain sorbed layer, the latter, as we shall see later, being
Substituting into equation (15) we
approximately equal to eEl EL/RT. The ap-
~

V
C5o
'
/ Ji- =
1
ÍX¿
plication of this equation to experimental data
(23) will be given in the next section.
ío
l1+ß - * If the thickness of the adsorbed layers cannot
The summation represented in the denominator is exceed some finite number n, then the summation
merely the sum of an infinite geometric progres- of the two series in equation (23) is to be carried
sion to n terms only, and not to infinity. Instead of
X equation (A) one obtains then
00 ¿ =
=

r (24)
t’¡pCx Í1 —

(« + l)x’* -f- wx,l

+ 1
(B)
Concerning the summation in the numerator we (1 —

x) ) 1 + (c l)x
—
cxn+1 —

may note that where x p/po, as before, and the constants

=

vm and c have also the same meaning as before.

" i =
1

It follows therefore that

.

IX*
.

=
X —
d
dx /
>
jí =
1
X* =
72-
(1 —
x
x)2
(25)
A rather plausible interpretation of the constant
n is that the width of the pores, cracks and capil-
laries of the adsorbent sets a limit to the maxi-
(26)
frn (1 -

X)(l -
X + CX) mum number of layers that can be adsorbed even
If we deal with adsorption on a free surface, then at saturation pressure. It is conceivable that
at the saturation pressure of the gas, po, an in- other causes than the diameters of the capillaries
finite number of layers can build up on the adsorb- can put such a limitation to the maximum
Feb., 1938 Adsorption of Gases in Multimolecular Layers 313

number of layers that can build up. However, neighborhood of the critical point), whereas on a
even if this be the case, the result, equation (B), surface ttmltimolecular layers will form far below
may still remain valid. the saturation pressure. This difference is to be
Equation (B) has two important limiting cases, accounted for on the basis of surface tension con-
both of which we shall make use of later. When siderations. In the gas phase the large free
1, it reduces to the Langmuir type equation
= surface energy of small clusters makes their
(29) and when =°
(free surface), it reduces to = formation improbable. On the other hand, in
equation (A). Furthermore, it should be noted adsorption the “liquid surface” is practically com-
that when x has a small value, and n is as large plete after the first layer has been adsorbed, so
as 4 or 5, equation (A) becomes a very good ap- that during the formation of successive layers
proximation to (B). To use equation (B), there- hardly any surface tension has to be overcome.
fore, one should plot the experimental isotherm
in the low pressure region according to the linear
III. Application of the Theory of Multimolecular
form of (A), evaluate c and vm from the slope Adsorption to Experimental Data
and intercept of the straight line, then using these A. S-Shaped Isotherms.—In several recent
values in equation (B) solve for the best average papers7,14·15 we have published low temperature
value of n. We shall apply this equation also to van der Waals adsorption isotherms for some or all

experimental data in the next section.18 of the gases nitrogen, oxygen, argon, carbon
The formation of multimolecular layers on an monoxide, carbon dioxide, sulfur dioxide, and
adsorbent below the saturation pressure, accord- butane on some 30 samples of catalysts, catalyst
ing to the present view, is analogous in many supports, and miscellaneous materials, including
respects to the formation of multimolecular various promoted and unpromoted iron synthetic
clusters in a non-ideal gas. There is, however, ammonia catalysts, metallic copper catalysts,
an important difference. In the gas the number pumice, nickel oxide supported on pumice, nickel
of multimolecular clusters is negligible compared on pumice, copper sulfate pentahydrate, anhydrous

to the number of single molecules (except in the copper sulfate, potassium chloride, crystalline
chromium oxide, chromium oxide gel, glaucosil,
(13) For the sake of completeness we include here two other
equations. silica gel, soils and soil colloids, dried powdered
1. If we assume that £2, the heat of adsorption in the second layer,
is still somewhat larger than the heat of liquefaction, (£3 £4 = =
bacteria, Darco (decolorizing carbon), and char-
. .
£l); and further assume that the packing in the first layer is
. coal. On all of these substances, except charcoal,
different from the packing in the higher layers, then the adsorption
isotherm equation for a free surface becomes typical S-shaped isotherms have been obtained,
vmcx ( + (b 5)(2x X2) )
- -
.
the low pressure portion of the isotherm being
( J to the pressure axis, the higher pressure
(1 x) \ 1 + (c l)x + (b 1 )cx° \
- - -
concave
where 6 eEi~ El>/RT, and
«=
Vm^Am. where vm(0 is the volume
=
region convex to the pressure axis, and the inter-
of gas adsorbed in the first layer when it is completely covered, and
vm is the volume adsorbed in any higher complete layer. All the
mediate region approximately linear with respect
other terms in the equation have the same meaning as in (A) and to pressure. To these S-shaped isotherms equa-
(B). It is to be expected that is not very different from unity.
Setting 1, and b > 1, equation (C) gives a faster rise in 9 with
=
tions (A) and (B) have been applied. Typical
increasing pressure than (A) does. Example of this has been found plots of equation (A) are shown in Fig. 1 for the
in the adsorption of butane on silica gel (see third part of this paper),
2. In the derivation of equation (B) we assumed that n was the adsorption of nitrogen at 90.1°K. on a variety of
maximum number of layers that can build up on any part of the ad- adsorbents, and in Fig. 2 for the adsorption of
sorbing surface. As long as the capillaries of an adsorbent are fairly
uniform in diameter, equation (B) can describe the entire course of
various gases on silica gel. Between relative
the isotherm with a fair degree of accuracy. If, on the other hand, pressures (p/po) of 0.05 and 0.35 the plots are
the size of the pores in an absorbent varies considerably from the
average, the experimental points will fail below the theoretical closely linear. From them the values of vm and
curve calculated with the help of equation (B) at lower pressures, c can be evaluated as explained in the previous
and above the curve at higher pressures. The equation which takes
care of the varying size of the capillaries is
section. From c one can obtain an approximate
i>mcx \ \ 1
—

(n 4- I)*" -f ««"->-1} value for Ex EL.16

—

"
(1 X) Z—Jn =
l ( 1 + (c —

l)x cxn+1 (
—
(14) Emmett, Brunauer and Love, Soil Science, to be published.
(15) Brunauer and Emmett, Tais Journal, 59, 2682 (1937).
(D)
(16) According to equations (17) and (22)
where vm is the volume of gas adsorbed in the complete first layer
and 0i, 02,
maximum of 1, 2,
0i are the fractions of the total surface on which a
...

1 layers can build up. Because of the large

...
t =
^ha. eB,-iL/Rr
nuriiber of constants in equation (!D), its practical use is difficult. From the nature of the constants 01, <z?, bi and bs it is evident that
However, fortunately, equation (B) is usually a good enough ap- aifn/bxat will not differ much from unity, and therefore £i —£l
proximation to describe the actual experimental isotherms. will be 2.303 log c.
314 Stephen Brunauer, P. H. Emmett and Edward Teller Vol. 60

Table I
Values of Constants for Adsorption of Nitrogen
at 90.1°K. on Various Adsorbents
Curve Vm Point Ei-El
in incc. B in cal. per B.

Substance Fig. 1 per g. cc. per g. mole,

Unpromoted Fe catalyst 9731 0.13 0.12 794 Sole

Unpromoted Fe catalyst 973II 1 0.29 0.27 774

Fe-Al208 catalyst 954 2.86 2.78 894
Fe-Al2Os catalyst 424 2 2.23 2.09 774
Fe-AlíOs-KüO catalyst 931 0.81 0.76 911
Fe-Al208-K20 catalyst 958 3 .56 .55 854
Fe-K20 catalyst 930 .14 .12 772 cc.

Fused Cu catalyst 4 .05 .05 776

Commercial Cu catalyst
10*

.09 .10 834

X

Cr2Oa gel 5 53.3 50.5 738

6.09 6.14
P)

Cr208 glowed 835 -

Silica gel 6 116.2 127.0 794

(P.

Table I gives the constants vm and E\ Eh —

P/V
for the six adsorption isotherms represented in
Fig. 1 as well as for six other isotherms not shown
in the figure. The constants vm in the table are
given for one gram of the adsorbent rather than
for the entire sample. Curves l-6> 0.05 0.10 0.15 0.25 0.30
0.20 0.35
Curve 7> 0.10 0.20 0.30 0.40 0.50 0.60 0.70
P/Po.
Fig. 2.

a complete unimolecular layer of adsorbed gas.

Since the constant vm, evaluated from the straight
line plots of equation (A), is also the volume of
gas corresponding to a unimolecular layer, values

/ ».
Fig. l.

In several previous publications Emmett and

Brunauer concluded that the beginning of the
approximately linear portion of the experimentally
determined low temperature adsorption isotherms
(point B in Fig. 3 for example), corresponds to
the covering of the surface of the adsorbent by Fig. 3.
Feb., 1938 Adsorption of Gases in Multimolecular Layers 315

of vm and of the volume of gas adsorbed at point straight line connecting the 760-mm. adsorption
B on the various isotherms should in general agree point for nitrogen at —183° with the origin on a
closely with each other. Columns 3 and 4 of plot of equation (A) such as is showm in Figs. 1
Table I give a comparison between the vm values and 2, is equal to l/vm with an error of no more
and the adsorption volumes corresponding to than 5% on all solid adsorbents with which we
point B for twelve isotherms. It is evident that have worked that gave S-shaped isotherms (enu-
the agreement is very satisfactory, the two seldom merated at the beginning of this section). This
differing by as much as 10%. is the same as saying that ym z>(l p/po), so = —

The last column of Table I reveals an interest- that, for example, is 2/s of the volume of gas
ing constancy in the value of the heat of adsorp- adsorbed when p/pa =
l/z-
tion in the first layer for different adsorbents. Table II
For nitrogen —

£L is uniformly 840 =±= 70 cal.

Values Different
Ex—Ex, for Gases
Since £l is about 1330 cal., is therefore 2170 * Na at A at CO2 at C4H10 at
Substance -183° -183° -78° 0°
70 cal. for nitrogen on all twelve adsorbents.
AI2O3 promoted Fe
At first thought this uniformity may seem rather
catalyst 954 894 704 1580 1870
surprising. It must be remembered, however, Silica gel 794 594 1335 1930
that is by definition an average heat of ad-
After the value of vm is obtained from the
sorption for the first layer. Furthermore, as is
evident from Figs. 1 and 2, for p/p0 values smaller isotherm, multiplication of the number of mole-
than about 0.03 the experimental points do not cules required to form a unimolecular layer by the
fall on the linear plot, in other words, equation average area occupied by each molecule on the
surface gives the absolute value of the surface
(A) breaks down for the most active points on the
area of the adsorbent. In Table III the surface
surface. (The reason for this is probably that
the assumption that is independent from the area in square meters per gram has been calcu-
amount of gas already adsorbed in the first layer, lated from the isotherms of different gases ad-
becomes untenable for the most active part of the sorbed on silica gel, using the vm values listed in
surface.) Hence , as obtained from the linear Column 4, and the cross sectional areas of the
adsorbed molecules calculated from both the
plot, must be regarded as the average heat of
adsorption for the less active part of the adsorb- density of the solidified gas and of the liquefied
ing surface. gas.17 It is evident from Columns 5 and 6 that
It is interesting to note that, as shown in Table the areas calculated from one gas agree approxi-
II, the value of —

EL is approximately con- mately "with those calculated from another gas.

stant for a given gas regardless of the chemical On the basis of solid-like packing the average
nature of the adsorbent. Thus on such different surface area obtained from the seven isotherms is
substances as a singly promoted iron synthetic 440 square meters per gram of silica gel, and the
ammonia catalyst and silica gel, the values of maximum deviation from the average is 15%;
on the basis of liquid-like packing the average
—

El are for nitrogen 840 =*= 50 cal., for argon

650 =*= 55 cal., for carton dioxide 1460 =*= 120 surface is 500 square meters per gram, and the
cal., and for butane 1900 =*= 30 cal. Furthermore, maximum deviation is 10%. In Table III,
as is evident from Figs. 1 and 2, the values of the Column 7, is listed the value of adsorption at
intercepts, 1 /vmc, are small in all cases. The Point B on the seven adsorption isotherms. It
wall be seen that for six isotherms the agreement
consequence of these facts is that a single adsorp-
tion point for any gas for which an approximate between vm and the volume at Point B is very
value of —

£L is known will enable one to fix vm good, but for butane the former is twice as large
within comparatively narrow limits for any finely as the latter. For some as yet unknown reason
divided solid known to be characterized by S- the empirical choosing of Point B as correspond-
shaped isotherms. For example if the highest ing to a unimolecular layer apparently gives
erroneous results for butane on silica gel. The
pressure point of Curve 4, Fig. 1, were connected
by a straight line to the origin, the slope of the vm value, on the other hand, as is evident from

resulting line (and therefore the value of ym) Table III, yields a surface area for the silica gel
would differ from that for the curve as now drawn (17) For a table of these cross sectional molecular areas and
details of calculation see Emmett and Brunauer, This Journal,
by no more than 3%. Indeed, the slope of a 69, 1553 (1937).
316 Stephen Brunauer, P. H. Emmett and Edward Teller Vol. 60

that is consistent with that calculated from the Fig. 3, a value of re (re =
5) fits the data well up
other gases. to a certain relative pressure (p/pa =
0.58), but
Table
III at higher relative pressures higher values of re
Value of Constants for Adsorption Isotherms on give the better fit (re= 7 at p/pa =
0.72). If one
Silica Gel takes an average value of re =
6, the theoretical
Surface in sq.
m./g. El-El curve will be in error to the extent of +5% at
Solid Liquid Point cal.
Temp. Curve Vm, pack- pack- B per
x =
0.58, and —7% at x =
0.72. On the basis
Gas °C. in Fig. 2 cc./g. ing ing cc./g. mole
of the capillary interpretation this means that
n2 -195.8 6 127.9 477 560 135.3 719
the iron catalyst in question has pores of con-
n2 -183 3 116.2 434 534 127.0 794
A -183 2 119.3 413 464 122.0 594 siderably varying diameters. On the other hand,
02 -183 4 125.1 410 477 132.0 586 an adsorption isotherm of oxygen at —183° on the
CO -183 5 121.2 449 550 132.0 973 adsorbent Granular Barco G (activated carbon),
co2 -
78 1 99.0 378 455 102.3 1335 could be fitted with a value of re = 2.2 from x =

C4H,o 0 7 58.2 504 504 28.1 1930

0.02 to 0.93, and none of the experimental points
At relative pressures in excess of about 0.35 to were off the curve as much as 4%, indicating a
0.50 the plot of experimental data according to rather uniform pore size in this adsorbent.
equation (A) deviates with increasing pressures For substances giving S-shaped isotherms re
more and more strongly from the straight line. values have been found to range from the low
The points deviate in the direction of there being value of 2 for Darco decolorizing carbons to values
too little adsorption at a given p/pa value to of 8 or 9 for some silica gel samples. It is in-
conform to equation (A). It becomes therefore teresting to note in this connection that for the
necessary to use instead equation (B). As already adsorption of butane on silica gel, as shown in
explained one can first evaluate c and rm by Curve 7, Fig. 2, the higher pressure deviations from
applying equation (A) up to p/pa 0.35, and =
the linear plot are in the direction of too much
can then proceed to use these values in equation adsorption rather than too little adsorption. One
(B) to obtain by the method of trial and error would expect just such deviation if the heat of
the value of re that gives best agreement with the adsorption in the second layer were still appreci-
experimental points. In certain instances, as in ably greater than the heat of liquefaction rather
than equal to it, as was assumed in the deriva-
tions of equations (A) and (B).16
In order to be able to calculate from one iso-
therm another at a different temperature, one
must examine how c, vm and re change with tem-
perature. The dependence of c on temperature
is exponential since c is approximately equal to
eEi-EL/RT' an(j ^ changes only slightly
with temperature. vm would be expected to vary
with temperature owing to the thermal expansion
of the adsorbed layer as the temperature increases.
As a first approximation one may assume that
vm changes with temperature as dL2/3> where
dL is the density of the liquefied gas. By the same
argument re should vary approximately as dLI/3,
which is, indeed, a very slight variation. It was
found empirically that the variation in re is ac-
tually negligible. With these assumptions the
adsorption isotherms for singly promoted iron
catalyst 954 at —195.8° have been calculated
from the —183° isotherms for nitrogen and argon.
The results are shown in Fig. 4. The agreement
between the calculated curves and the experimen-
Fig. 4. tal points appears to be quite satisfactory.
Feb., 1938 Adsorption of Gases in Multimolecular Layers 317

Unfortunately one cannot always safely assume responds to a vm value of 50.6 micromoles of
that vm changes with temperature as dL2/3. The acetone on a 14.67 g. sample of vitreous silica.
change in vm sometimes may be considerably The specific surface of this same sample was
greater, and sometimes less. An example of the 4690 sq. cm. according to Palmer and Clark’s
former is afforded by the adsorption isotherms measurements of the rate of solution in hydro-
of McGavack and Patrick18 for sulfur dioxide on fluoric acid. If one uses for the area occupied by
silica gel at temperatures between —80 and 40°. one acetone molecule on the surface the value
In Table IV, Column 2, the vm values are listed 26.9Á.2 (which is obtained from the density of
for the six isotherms. It will be noted that vm liquid acetone at 25°), one obtains from the above
for —80° is 1.43 times as great as for 40°, whereas vm value a specific surface of 5640 sq. cm., which
it should only be 1.15 times as great if vm would is about 20% larger than Palmer and Clark’s value.
have changed as dL2/3· It is also to be noted If on the other hand one uses for the area oc-
that in these isotherms the Ex £L value de- —

cupied by an acetone molecule the value 20.5A.2,

creases markedly with decreasing temperature. obtained by N. K. Adam for close-packed films
on water of long-chain compounds terminating
Table IV in the CO-CHg group, one obtains for the specific
Value of Constants of Adsorption of Sulfur Dioxide surface 4290 sq. cm., which is about 8% smaller
on Silica Gel (McGavack and Patrick) than the value of Palmer and Clark. These data
Temp. Vm, ¿L, £. -
El
EL, Ei
°
C. cc./g. g./cc. cal./mole cal./mole cal./mole suffice to show that the agreement is very good,
40 91.6 1.327 1789 4940 6730 and in our opinion is a weighty confirmation of
30 96.6 1.356 1750 5190 6940
the method that they used for measuring the
0 105.8 1.435 1705 5840 7540
-34.4 118.6 1.522 1587
surface of vitreous silica.
-54 129.2 1.573 1458
1.642
B. Adsorption Isotherms on Charcoal
-80 131.0 1364
3.5 for all six isotherms.
=
On charcoal, in contrast to all other adsorbents
thus far tried by us, no S-shaped isotherms were
The experimental data of Bradley2 for argon obtained. The slopes of the isotherms on char-
on anhydrous copper sulfate, as well as our own coal decrease continually as the pressure in-
for this system,18 also can be represented satisfac- creases, and in the neighborhood of saturation
torily by equation (B). Ex £L was found to
—
become practically zero. Such isotherms can be
be 745 cal. for Bradley’s isotherms, 777 for our fitted with equation (B) if 1. If we interpret
=

own, n was about 5 for both cases. The surface the limitation in the number of layers that can
of our copper sulfate sample was larger; vm for build up on a surface as due to capillaries in the
Bradley’s isotherms was 6.4 X 10-8 mole of adsorbent, then we must conclude that the pores
argon per mole of copper sulfate, our value of or capillaries in charcoal are exceedingly narrow,
vm was 10.8 X 10-8 mole per mole. It should not more than one or two molecular diameters in
be noted that the value of r>m for Bradley’s argon thickness.20
isotherm on copper sulfate which we consider to If equation (B) reduces to the form
=
1,
be a unimolecular layer was interpreted by Brad- p/v Pa/cvm + p/vm
=
(E)
ley as being about 18.5 layers. which is clearly a special form of the Langmuir
Recently Palmer and Clark19 published ad- equation. In Fig. 5 six isotherms on charcoal
sorption isotherms of acetone, benzene, and other are plotted according to equation (E), p/v being
organic vapors on samples of vitreous silica whose taken as ordinate and p as abscissa. It will be
total surface area had been measured by com- noted that the plots are accurately linear from
paring the initial rate of solution in hydrofluoric p/pa =
0.1 clear up to the saturation pressure.
acid of the powdered sample with the rate of Table V shows the constants vm evaluated from
solution of a sample whose area was known. If (20) Perhaps it is not immediately obvious that the evaporation-
one plots the adsorption isotherms of acetone on condensation equilibria expressed in equations (10), (11), and (12)
can be applied to cases where the width of the capillaries is not
vitreous silica according to equation (A) one ob- more than one, two or a few molecular diameters. The principle
of microscopic reversibility, however, shows that these equations
tains a very good straight line whose slope cor- remain valid even though the processes of direct condensation from
(18) McGavack and Patrick, This Journal, 42, 946 (1920). or evaporation into the gas-phase play only a small part in the total

(19) Palmer and Clark, Proc. Roy. Soc. (London), A149, 360 equilibrium. The effect of these narrow capillaries will be merely
(1935). to slow down the establishment of equilibrium.
318 Stephen Brunauer, P. H. Emmett and Edward Teller Vol. 60

the slopes of the plots of equation (E). If we tration, the curves in Fig. 6 for the data of Gold-
disregard the butane run for the moment, the man and Polanyi22 on the adsorption of ethyl
vm values calculated from the isotherms are con- chloride by charcoal are shown. From the 0°
sistent with each other, giving a mean surface of isotherm plotted according to equation (E) values
708 square meters per gram of charcoal if we use of vm and c were obtained. Assuming that vm is
the same at —15 and 20° as at 0°
(since the temperature range is rather
narrow), and that c, as before, can
be represented by the exponential
A -

el/rt , we have calculated the iso-

therms for —15.3 and 20°. The cal-
culated values are represented by the
curves, the points are experimental.
Here, as in previous instances, equa-
tion (E) appears applicable only to
that portion of the isotherm which
corresponds to pjpa values greater than
0.1. In the higher pressure region,
however {p/po being between 0.1 and
0.95), the calculated isotherms agree
quite well with the observed points.
400 600 For more extended calculation of
Pressure, mm.
the temperature dependence of these
Fig. 5.
Langmuir-type isotherms it would be
molecular cross-sectional areas calculated from necessary to make assumptions in regard to the
the solid state, and 829 square meters per gram if temperature dependence of vm or to determine z?m
the molecular areas are calculated from the over a sufficiently wide temperature range to be

liquefied gases. The maximum deviation of sur- able to estimate it for other temperatures. For
face areas from the mean for the seven
isotherms is about 9% on the basis of
solid packing, and 8% on the basis of
liquid packing. Butane adsorption on
charcoal is decidedly smaller (about
25%) than one would expect on the
basis of the estimated cross-sectional
area of the butane molecule. It seems
probable that some of the capillaries
are so small in diameter that molecules
as large as butane cannot enter them.
Similar phenomena were noted by other
investigators before; a discussion of
these, under the name “persorption,” is
given by McBain.21
Equation (E), as mentioned before, is 200 300
of the form of Langmuir’s equation, with Pressure, mm.
the usual constant k replaced by c/pa- Fig. 6.
Since the temperature dependence of pa,
c, and ym is known, or can be evaluated, equation
some substances, furthermore, EL will vary
Ei —

considerably with temperature and cause calcu-

(E) can be used to calculate an isotherm at one
lated isotherms to deviate more or less seriously
temperature from that at another. As an illus-
from the experimental points. With these limita-
(21) McBain, "The Sorption of Gases by Solids," George Rout-
ledge and Sons, Ltd., London, 1932, p. 169. (22) Goldman and Polanyi, Z. physik. Chem., 132, 321 (1928).
Feb., 1938 Synthesis of 1,4-Dimethylphenanthrene by Cyclodehydration Methods 319

lions it seems that equation (E) ought to be useful Summary

for calculating Langmuir isotherms of van der 1.A critical discussion of the polarization
Waals adsorption at one temperature from those theory of multimolecular adsorption is presented.
at another. It is shown that the adsorption energy due to
attraction of dipoles induced into a non-polar gas
Table V like argon is insufficient to constitute a major por-
Values of vm for Adsorption Isotherms on Charcoal tion of the binding energy between adsorbed layers.
2. Derivation of adsorption isotherm equations
Temp., Curve Solid Liquid
Gas °C. in Fig. 5
I'm,
cc,/g. packing packing for multimolecular adsorption are carried out on
n2 -195.8 4 181.5 677 795 the assumption that the same forces that produce
n2 -183 1 173.0 646 795 condensation are also responsible for multimolecu-
A -195.8 215.5 746 804 lar adsorption.
A -183 5 215.5 746 839 3. Numerous applications of the equations are
02 -183 6 234.6 767 894
CO -183 3 179.5 665 820 given to experimental adsorption isotherms ob-
co2 -
78 2 185.5 707 853 tained by other investigators as well as by us.
C4H10 0 63.0 545 546 Washington, D. C. Received November 19, 1937

[Contribution from the Chemical Laboratories of Columbia University and the College of the City of
New York]
The Synthesis of 1,4-Dimethylphenanthrene by Cyclodehydration Methods
By Domenick Papa, David Perlman and Marston T. Bogert

Ina recent article, Akin, Stamatoff, and Bo- as the final step, and in both cases obtained a
gert1 reported the synthesis of 1,4-dimethylphen- dimethylphenanthrene whose properties agreed
anthrene from ^-xylene by the familiar Pschorr with the 1,4- compound of Akin, Stamatoff, and
reaction, in which they obtained a product quite Bogert, and without any evidence of the migra-
different from that secured from the same initial tion of one of the p-xylene methyl groups under
material by Bardhan and Sengupta,2 by another the conditions of our experiments. Our suggested
series of reactions, and to which the latter investi- explanation of Bardhan and Sengupta’s results,
gators assigned the same constitution. therefore, is not supported by this evidence.
The experimental results published by Akin, These two syntheses were as shown in the chart.
Stamatoff, and Bogert indicated, however, that
the Bardhan and Sengupta hydrocarbon was Experimental
probably the 1,3-dimethyl isomer, previously pre- beta-(£-Xylyl)-ethanol, was prepared from bromo-p-
xylene and ethylene oxide by the Grignard reaction,
pared by Bogert and Stamatoff,3 and by Haworth,
following the method described by Dreger5 for the syn-
Mavin, and Sheldrick.4 thesis of n-hexyl alcohol, in a yield of 61% and a boiling
At the time, it was suggested by Akin, Stamatoff, point of 108-111° at 4 mm. On a redistilled sample,
and Bogert, although they could not rearrange the physical constants were: b. p. 105-106° at 3 mm.,
their own product to the 1,3-dimethyl isomer by d25, 0.9946, w25d 1.5286, caled. 46.30, > obsd.
selenium fusion, that perhaps something of this 46.48. Bardhan and Sengupta, who synthesized it in a
different way, gave the b. p. as 110-113° at 5 mm.
kind had happened in the final step of the Bard-
han and Sengupta synthesis. We have therefore
µ . Caled, for ,, : C, 79.94; H, 9.39. Found:
C, 79.64; , 9.40.
synthesized the 1,4-dimethylphenanthrene by two Phenylurethan.—Small white rosets of thin needles
slightly different methods, from ^-xylylethyl from petroleum ether, m. p. 79-79.5° (corr.).
cyclohexanols, dehydrogenating the resulting Anal. Caled, for C,7Hi902N: N, 5.20. Found: N,
octahydro compounds by either sulfur or selenium, 5.29.
(1) Akin, Stamatoff and Bogert, This Journal, 59, 1268 (1937). beta- f^-Xylyl)-ethyl bromide (I) was obtained from the
(2) Bardhan and Sengupta, J. Chem. Soc., 2520 (1932). above alcohol and hydrobromic acid, according to the
(3) Bogert and Stamatoff, Rec. trav. chim., 52, 583 (1933).
(4) Haworth, Mavin, and Sheldrick, J. Chem. Soc., 454 (1934). (5) Dreger, “Organic Syntheses." Coll. Vol. I, 1932, p. 299.