360 Langmuir 1992,8, 360-365

Cross-Sectional Areas of Adsorbed N2, Ar, Kr, and 0 2 on

Carbons and Fumed Silicas at Liquid Nitrogen Temperature
Ismail M. K. Ismail
University of Dayton Research Institute, c/o Phillips LabIRKFC, Edwards Air Force Base,
California 93523- 5000
Received November 1,1990. In Final Form: November 26, 1991

Adsorption of Nz, Ar, Kr, and 0 2 at 76.6 K was investigated on Aerosil samples and three different
categories of carbons: graphitized, ungraphitized, and activated graphitized substrates. With the BET
equation, the appropriate ranges of relative pressure are defined and the corresponding experimental
cross-sectional areas, u (nm2/adsorbedspecies), are reported, based on u(N2) = 0.162. Values of u on the
carbons were fitted to physical models illustrating possible arrangements of adsorbed species on the
surface. With graphitized carbons, ~ ( 0 2=) 0.147 and u(Ar) = 0.138. With ungraphitized or activated
carbons, values of u are 0.157 for Ar, 0.214 for Kr (at P / P = 0.02-0.25), and 0.148 for 0 2 . With Aerosils,
however,the value of u(Ar) = 0.186, obtained by the calibration, is high and could not be fitted to a model.
An alternative approach has been considered. When comparing the present data with those reported in
the literature, the recommended new values on Aerosils (and other silicas) are u(Nz) = 0.143,~(02)= 0.147
and u(Ar) = 0.165 nm2/mol. The present results indicate that oxygen could be considered as the standard
gas for adsorption on carbons and silicas, provided that it does not change the carbon structure or its
surface characteristics when adsorbed at 77 K.

Introduction When the "apparent" BET value of Vm is computed at

Gas adsorption has been used for several decades to
estimate surface area of materials, to define type of
porosity, to compute pore volumes, and to calculate pore
size distributions. The most common adsorbates are
simple gases such as N2, Kr, Ar, 02, and C02. With the
adsorption isotherm of an adsorbate, X, the Brunauer-
Emmett-Teller (BET) equation' is normally used at
relative pressure, PIP", of 0.05-0.35 to estimate the mono-
layer volume, Vm. Multiplying Vm by the appropriate
cross-sectionalarea, u(X),along with other proportionality
constants, yields the BET surface area.
With graphitized carbons, this procedure has limitations.
For example, the adsorbed Kr atoms or N2 molecules near
the completion of the first monolayer undergo several
phase changes2including a change from commensurate to
incommensurate structure, and the isotherms exhibit a
kink2+ which may affect the linearity of the BET plot.
Since on graphitized carbons, the true area of an adsorbed
N2 molecule or Kr atom at 77 K near the monolayer
coverage, as determined by other experimental
techniques,'-13 is equivalent to the area of three graphitic
hexagons, 0.157 nm2, the true surface area is the product
of 0.157 and the volume of gas taken by the sample at the
beginning of the isotherm kink which represents the true
monolayer coverage. Nevertheless, the BET equation is
still useful to yield the same answer in a different way.
(1) Brunauer, S.; Emmett, P. H.; Teller, E. J.Am. Chem. SOC.1938,
60, 309.
(2) Thomy, A.; Duval, X.; Regnier, J. Surf. Sci. Rep. 1981, 1 , 1.
(3) Thomy, A.; Duval, X. J. Chim.Phys. 1970, 67, 286.
(4) Rouquerol, J.; Partyka, S.; Rouquerol, F. J. Chem. SOC.Faraday
Trans. 1 1977, 73, 306.
(5) Grillet, Y.; Rouquerol, F.; Rouquerol, J. J. Colloid Interface Sci.
1979, 70, 239.
( 6 ) Larher, Y. J. Chem. Phys. 1978, 68, 2257.
(7) Kjems, J. K.; Passel, L.; Taub, H.; Dash, J. G. Phys. Rev. Lett.
1974, 32, 724.
(8) Chinn, M. D.; Fain, S. C., Jr. Phys. Rev. Lett. 1977, 39, 146.
(9) Frenkel, D.; McTague, J. P. Annu. Rev. Phys. Chem. 1980,31,491.
(10) Diehl, R. D.; Fain, S. C., Jr. Phys. Rev. B: Condens. Matter 1982,
26, 4785.
(11) Stephens, P. W.; Heiney, P. A.; Birgeneau, R. J.; Horn, P. M.;
Moncton, D. E.; Brown, G. S. Phys. Rev. E Condens. Matter 1984,29,
3512.
(12) Fain, S. C., Jr. Carbon 1987, 25, 19.
(13) Bojan, M. J.; Steele, W. A. Langmuir 1987, 3, 116.
PIP" of 0.05-0.21, the product of Vm and the "apparent"
value of a(Nz), 0.162 nm2/mol,or a(Kr), 0.214 nm2/atom,
gives the same "true" surface area.14 This procedure is
especiallyconvenient with graphitic carbons whose Nz and
Kr isotherms a t 77 K do not have a distinct kink near the
completion of mon01ayer.l~
1, or a(Kr) on other forms of
Values of ~(0.2u(Ar),
carbons, e.g., ungraphitized and activated carbons as well
as coals and chars, and the limits of the BET equation,
have not been clearly defined. The literature indicates
that u(X) may vary from one substrate to another. The
limits of the BET equation could cover a P/P" range as
narrow as 0.0001-0.03815or as wide as 0.05-0.35,' and there
is a considerable scatter in a(X) values reported for
different adsorbates on carbons. For example, the liter-
ature cross-sectional areas (nm2/adsorbed species) are
0.1432and 0.23516 for Kr, 0.11117J8to 0.15119 for Ar, and
0.135 and 0.15918*20 for 0 2 . The most popular ones are
u(Ar) on graphitized carbons, 0.138, which has been
recommended by several investigator^,'^*^^-^^ and a(Kr) =
0.195.25 The 0.138 is commonly used with ungraphitized
and microporous carbons,15but as it will be shown here,
this value is not applicable to these classes of carbon.
The objectivesof this article are to determine the reliable
or most reasonable values of 4021, u(Ar), and a(Kr) on
different carbon and silica substrates at 77 K, and to report
the corresponding BET limits. The substrates examined
(14) Ismail, I. M. K. Carbon 1990,28, 423.
(15) Fernandez-Colinas, J.; Denoyel, R.; Grillet, Y.; Rouquerol, F.; Rou-
querol, J. Langmuir 1989,5, 1205.
(16) Deitz, V. R.; Berlin, E. J. Colloid Interface Sci. 1973, 44, 57.
(17) Singleton, J. H.; Halsey, G. D., Jr. J.Phys. Chem. 1954,58, 330.
(18) McClellan, A. L.; Harnsberger, H. F. J. Colloid Interface Sci.
1967, 23, 577.
(19) Smith, W. R.; Ford, D. G. J. Phys. Chem. 1965,69, 3587.
(20) Razouk, R. I.; EIGobeily, M. A. J. Phys. Chem. 1950, 54, 1087.
(21) Gregg, S. J.; Sing, K. S. W. Adsorption, Surface Area and Porosity,
2nd ed.; Academic Press: New York, 1982; p 76.
(22) Walker, P. L., Jr.; Foresti, R. J.; Wright, C. C. Ind. Eng. Chem.
1953,45, 1703.
(23) Aristov, B. G.; Kiselev, A. V. Russ. J.Phys. Chem. (Engl. Traml.)
.- - -, .. , - - -.
1963.37.1359. .
(24) Berezkina, Yu. F.; Dubinin, M. M.; Sarakhov, A. I. Akad. Nauk
SSSR, Ser. Khim. 1969, 2653; Bull. Akad. Sei. USSR, Diu. Chem. Sci.
(Engl. Transl.) 1969, 2495.
(25) Beebe, R. A.; Beckwith, J. B.; Honig, J. M. J. Am. Chem. SOC.
1945, 67, 1554.

0743-7463/92/2408-0360$03.00/0 0 1992 American Chemical Society

Letters Langmuir, Vol. 8,No. 2,1992 361
Table I. Cross-Sectional Areas of Oxygen and Argon on Different Carbons and Silicas at 76.6 K
nitrogen BET U(O2), O(Ar),
sample area: m2/g nm2/mol BET limit nm2/atom BET limit*
Ungraphitized Carbons (UGC)
N990 7.32 0.145 0.02-0.28 0.161 0.02-0.28
ST MT 8.82 0.149 0.02-0.28 0.158 0.02-0.25
N762 25.55 0.148 0.02-0.28 0.159 0.02-0.25
N550 37.02 0.145 0.03-0.27 0.160 0.02-0.28
v3 73.2 0.143 0.02-0.28 0.158 0.02-0.25
N330 80.4 0.145 0.02-0.28 0.159 0.02-0.28
average ~ ( 0 2=)0.146 average u(Ar) = 0.159
Graphitized Carbons (GC)
SP-1 2.01c 0.145 0.01-0.10 0.136 0.01-0.1
V3G 59.3c 0.148 0.01-0.10 0.137 0.01-0.1
GCC 74.1c 0.148 0.01-0.10 0.138 0.01-0.1
Graphon 83.3' 0.148 0.01-0.10 0.139 0.01-0.1
average ~ ( 0 20.147
) average u(Ar) = 0.138
Activated Graphitized Carbons (AGC)
GCC, 9.7% BO 114.2 0.148 0.001-0.10 0.153 0-0.1
GCC, 19.8% BO 120.2 0.147 0.01-0.10 0.154 0-0.1
GCC, 30.0% BO 119.2 0.149 0.01-0.10 0.156 0-0.1
GCC, 50.0% BO 113.8 0.149 0.01-0.10 0.158 04.1
average ~ ( 0 2 =) 0.148 average u(Ar) = 0.155
Fumed Silica
Aerosil-50 73.3 0.163 0.05-0.30 0.183 0.04-0.3
Aerosil-150 131.6 0.168 0.05-0.30 0.189 0.04-0.3
Aerosil-200 184.3 0.166 0.05-0.30 0.187 0.04-0.3
average 4 0 2 ) = 0.166 average a(Ar) = 0.186
Based on u(N2) = 0.162 nm2/mol.* Using PO of supercooled Ar. Based on a(N2) = 0.157 nmZ/mol.

here have been used extensively in this laboratory as
candidates for deposition of pyrolytic carbon26and silica27
on their surfaces. At some point during the investigation,
there was a need to know precisely the a(X) of these
adsorbates, especially when the surface characteristics
change, for example,from non-graphitic to graphitic after
the deposition of pyrolytic carbon,26or from graphitic or
non-graphitic to a silica-coated surface after the deposition
of Si02.27
Experimental Section
The adsorption isotherms of N2, 0 2 , Ar, and Kr at 76.6 K were
obtained by a Digisorb2600 volumetric apparatus manufactured
by the Micromeritics Instrument Co., Atlanta, GA. The samples
were initially outgassed at 573 K for 24 h and then flushed with
He. This process was repeated twice at 3 h evacuation with
intermediate adsorption runs of N2 at 76.6 K. By the end of this
treatment, the surface of substrates, especially the ungraphi-
tized and activated carbons, became stable and clean; they gave
reproducible results. For each carbon, the same sample was used
for the adsorption of Nz,02,or Ar at random. However, to obtain
the Kr isotherms, separate smaller samples (50-80 mg) were
examined. The number of experimental data points on each
isotherm varied between 40 and 85; they covered a relative
pressure range of 0.0014.95, depending on the adsorbate/
adsorbent system. Since the time taken for a single isotherm
was 20 h to 8 days, the saturation pressure of adsorbate was
updated every 6 h. Data analyses were performed using, when
appropriate, the saturation pressure of liquid Nz, liquid 02,solid
Kr, or supercooled Ar. It is admitted that since the adsorption
temperature (76.6 K)is lower than the triple point of Ar (83.8
K), the saturation pressure of solid rather than supercooled Ar
should have been used in the analysis. However, there are three
main reasons for adopting the supercooled state of Ar. First,
since the development of the BET theory, most investigators
have used the supercooled state;1J5~22-24~2*33a direct comparison
(26) Ismail, I. M. K.; Rose, M. M.; Mahowald, M. A. Carbon 1991,29,
575.
(27) Hoffman, W. P.; Phan, H. T. In Fundamental Issues in Control
of Carbon Gasification Reactiuity; Lahaye and Ehrburger,Eds.; Kluwer
Academic Publishing: The Netherlands, 1990.
(28) Pace, E. L.; Siebert, A. R. J. Phys. Chem. 1960,64, 961.
(29) Brown, C. E.; Hall, P. G. Trans. Faraday SOC.1971,67, 3558.
between their work and the present study would only be possible
with the same assumption. Second, most of the currently
available automated adsorption instruments are using PO of
supercooled Ar in their operating software; it is convenient for
the users to simply change the value of a(Ar) at the end of
calculations (whenever needed) instead of changing the entire
isotherm and computations from supercooledto solid Ar. Third,
whether PO of solid or supercooledAr is chosen for the calibration,
the correspondingproduct of u(Ar) and V,yields the same surface
area regardless of the chosen value of a(Ar) and the actual state
of Ar in the adsorbed layer.
The samples examined and their nitrogen BET surface areas
are listed in the first and second columns of Table I. The ranges
of PIP" were 0.05-0.21 for graphitized carbons and 0.024.28 for
the other materials. The table has four groups of materiale: the
first includes six ungraphitized carbon blacks, UGC, the second
includes four graphitized carbons, GC (three carbon blacks and
natural SP-1graphite), the third includes four activated graph-
itized Columbian carbon (GCC),AGC, and the fourth has three
Aerosils (fumed silica supplied by Degussa Co.). The AGC
samples were prepared by oxidation of the original GCC in air
at 773 K to different levels of burn-off (BO). This process has
developedmicroporosity.For example,a GCC samplewith 19.8%
BO has ita micropore surface area accounting for 39.1% of the
total BET area.%
Results
The third column lists values of a(O2) and the corre-
sponding P/P"ranges for the BET best fit. The range is
0.014.1 for GC and AGC samples and is higher for the
other materials, 0.02-0.28 for UGC and 0.05-0.30for Aero-
sils. The limits of best fit were based on the highest
correlation coefficient computed with each r u n at different
P / P ranges. Changing these limits has little effect on
monolayer volumes. For example, when the upper PIP"
limit was extended from 0.1 to 0.3,V , of GC increased by
0.5-1.0% in 02 but decreased by 1-4% in Ar. This means
(30) Payne, D. A.; Sing,K. S. W.; Turk, D. H. J. Colloid Interface Sci.
1973, 43, 287.
(31) Rouquerol, J.; Rouquerol, F.; Peres, C.; Grillet, Y.;Boudellal, M.
Characterization of Porous Solids: Societyof Chemical Industrv: Letch-
work, U.K., 1979; p 107.
(32)Ali, S.; McEnaney, B. J. Colloid. Interface Sci. 1984, 107, 355.
(33) McEnaney, B. Carbon 1987,25, 69.
362 Langmuir, Vol. 8, No. 2, 1992 Letters
Table 11. Apparent Cross-Sectional Area of Kra on

- 1
Carbons at 76.6 K Using Po of Solid Kr
,23 b- Average value
relative pressure range
sample 0.02-0.1 0.02-0.2 0.024.3
ungraphitized carbons
N990 0.200 0.209 0.226
ST MT 0.190 0.199 0.211
N762 0.191 0.203 0.222
N550 0.194 0.203 0.223
v3 0.200 0.210 0.224
N330 0.201 0.210 0.222
average u(Kr)
graphitized carbonsb
0.196
NIA
0.206
0.2146
0.221
NIA
'
U

.I9 -
x
X
I
I
I
I
I
-

0.2706 NIA ,
Aerosil-50, -150,-2W6 N/A
a Values of u(Kr)are expressed in nmzlatom. 6 From ref 14between
relative pressures of 0.05 and 0.21.

that by extending the BET limits, the computed value of

) lowered by 1% from 0.147 and 0.145 nm2/molecule,
~ ( 0 2is
but a(Ar) is raised by as much as 4% from 0.138 to 0.143
for GC or from 0.155 to 0.162 nm2/atom for AGC. The
important point in the table is that the average value of
4 0 2 ) on the three categories of carbons is the same; 0.147
f 0,001 nm2/mol. On Aerosils, 4 0 2 ) is higher, 0.166 nm2/
mol.
With argon (fourth column), the average value of a(Ar)
on silica, 0.186 nm2/atom, is also higher than on all other
carbons. However, a(Ar) is influenced by the type of
carbon surface under consideration, 0.138 f 0.002 nm2/
atom on GC, and averages to 0.157 f 0.002 nm2/atom on
UGC and AGC. Unlike 02,the packing density of Ar is
dependent on surface characteristics. The value of 0.138
nm2/atom is in excellent agreement with values that have
been reported by previous i n v e s t i g a t o r ~ ~ Jand
~ - ~ ~ - ~of
observed experimentally using neutron ~ c a t t e r i n g .It~ is
~ nm2/atom,14the same value may arbitrarily
interesting to note that the value of 0.157 nm2/atom on
UGC and AGC happened to be exactly equivalent to the
area of three graphitic hexagons. This fortuity does not
necessarily mean the adsorbed Ar atoms are in registry
with the carbon surface.
With Kr adsorption on UGC, the BET plots were never
linear at any PlPO range; instead, they showed a curvature
through the entire range, 0.01-0.35. Thus, for the Kr/
UGC system, the BET theory deviates. The plots were
slightly convexto the pressure axis; as the relative pressure
increased, the slope increased but the intercept decreased.
Since the increase in slope exceeded the decrease in
intercept, the net effect was a lowering in the apparent
value of V, with increasing Kr relative pressure (Table
11). The calculated "apparent" value of a(Kr) on each
carbon increased by extending the range from 0.02-0.1 to
0.02-0.2 and finally to 0.02-0.3. Dependence of the
"apparent" average a(Kr) on PIP" is displayed in Figure
1,which can be utilized to select the proper value of a(Kr)
according to the upper relative pressure range chosen for
the BET analysis.
It is noted in the table that with each column, a(Kr) is
scattered around an average value. The scatter is attrib-
uted to two main reasons: weight of the samples and
inconsistency of the carbon batch. First, to perform the
Kr analysis in a reasonable length of time, the weight of
the samples whose surface areas are high had to be reduced
to 50-80 mg. A fluctuation of f l mg in sample weight
yields as high as 2 7% error in V,. Second, each lot of the
UGC has some range of particle size distribution. The
small samples used for Kr analysis could have a slightly
(34)Taub, H.; Carneiro, K.; Kjems, J. K.; Passell, L.; McTague, J. P.
Phys. Reu. E S o l d State 1977,16, 4551.
different size distribution and, hence, yield a slightly
different V , than the corresponding larger samples used
for N2, 02,and Ar analysis. Earlier tests revealed that
changing the sample caused V , to fluctuate by fl-4%.
Based on these two reasons, the maximum error in
estimating u(Kr), from one sample to another, could be as
high as 8 % ;this explains the scatter in a(Kr)values shown
in Figure 1and listed in Table 11. In spite of this scatter,
the average values of a(Kr) can be used for other Kr/
ungraphitized carbon systems at 77 K to estimate the
surface area. First, the BET analysis is performed at any
selected relative pressure range to yield an "apparentn
V,. Then, with the corresponding "apparent" value of
a(Kr) taken from the figure, the true surface area can be
computed.
~ ~cr(Kr)
~ ~ is0.214However, since with GC the apparent value
be recommended (for the sake of simplicity) for UGC and
AGC, provided that the BET analyses are executed at a
P/PO range of 0.02-0.25, as illustrated with the dotted line
in Figure 1. As mentioned earlier, this analysis is based
on PO of solid Kr.
Discussion
u(X) on Graphitized Carbons. It is now established,
from neutron diffraction,"^^^ X-ray ~ c a t t e r i n gand , ~ low-
~~~
energy electron diffraction experiments38-39 that on graph-
itized carbons, a(N2) = a(Kr) = area of three graphitic
hexagons = 0.157 nm2/adsorbed species. In oxygen, the
monolayer density varies with coverage and adsorption
t e m p e r a t ~ r e . For
~ ~ oxygen the 2-D phase diagram on
graphitized carbons is complicated, it has 9 monolayer
phases and 6 triple points. At 10-50 K, the density is
0.065-0.085 molecule/A2,which corresponds to a(Od values
of 0.154-0.118 nm2/molecule.40 From the computer sim-
ulation at 30-70 K,4l a condensed layer is formed first on
carbons with a density of 0.074molecule/A2which can be
compressed to 0.112 molecule/A2. A t liquid nitrogen tem-
perature, the density may be taken as 0.074 molecule/A2,
(35) Eckert, J.; Ellenson, W . D.; Hastings, J. B.; Passel, L. Phys. Rev.
Lett. 1979, 43, 1329.
(36)Horn, P. M.; Birgeneau, R. J.; Heiney, P.; Hammonds, E. M. Phys.
Rev. Lett. 1978, 41,96i.
(37)Specht, E. D.; Sutton, M.;Birgeneau, R. J.; Moncton, D. E.; Hom,
P. M.Phys. Reo. €3: Condens. Matter 1984,30, 1589.
(38) Diehl, R. D.; Samuel, M. F.; Fain, S. C., Jr. Phys. Reu. Lett. 1982,
48, 177.
(39)Diehl, R. D.; Fain, S. C., Jr. J. Phys. Chem. 1982, 77, 5065.
(40) Toney, M. F.; Fain, S. C., Jr. Phys. Reu. E Condens.Matter 1987,
36, 1248.
(41) Bhethanabotla, V. R.;Steele, W.A.Phys.Reu.B: Condens.Matter
1990,41, 9480.
Letters
Table 111. ComDarison between Theoretical and Exmrimental Cross-sectional Areas
theoretical
adsorbate dtda o w GCC
NZ(liquid) 0.162 0.153 0.162
Nz (solid) 0.138
Ar (liquid) 0.138 0.135 0.138
Ar (solid) 0.128
Kr (liquid) 0.150 0.155 0.18w
Kr (solid) 0.143 0.21d
02 (liquid) 0.136 0.136 0.148
02 (solid) 0.121
a d t d = 1.091 ( M / O N ) ~nm2/adsorbed
/~: mecies. d t 2 ) = 0.06354(TJPc)2/3;
carbons. e Activated graphitized carbons. f From ref 14.-
which yields ~ ( 0 2 =
) 0.135 nm2/mo1ecule.4l This value is
slightly lower than the "apparent" value of 0.147nm2/mol
listed in Table I. The discrepancy between the two values
(-8%) indicates that the BET equation is not perfectly
adequate to define the true values of Vm for 02 adsorption
on GC.
In the case of Ar, there is some controversy in the
literature. The reported values of a(Ar) in nm2/atom are
0.132 from volumetric adsorption measurements: 0.125
from adsorption calorimetric measurements: and 0.134
from neutron ~ c a t t e r i n g . ~
From
~ the recent work of
DAmico et al.42at 44-50 K and 0.81 monolayer coverage,
the computed value is 0.19. It is possible that with
increasing adsorption temperature to 77 K and at Vm =
1, a(Ar) decreases to the neighborhood of -0.14 nm2/
atom. Apart from D'Amico's work, there is an overall
good agreement between the experimental BET value
obtained here (0.138nm2/atom) and most of the other
values either reported in the literature on GC or computed
theoretically using the bulk properties of supercooled Ar.
Therefore, the BET model is presumably adequate enough
to define a true value of V, for Ar on GC.
u(X) on Ungraphitized a n d Activated Graphitized
Carbons. This class of carbons has a heterogeneous
surface composed of basal planes and active sites whose
population is 5-15 % of the totalsurface depending
on the sample and its level of BO. From Table I, each
average value of ~ ( 0 2 or
) a(Ar) on the two classes of carbon
is practically the same regardless of the extent of surface
heterogeneity. The average values are 0.147 nm2/02
molecule and 0.157f 0.002nm2/Ar atom. This finding is
somewhat surprising because the active surface area of
the UGC and, hence, the degree of heterogeneity were
found to increase with increasing the starting surface area
of the carbon.26 It is possible that the contribution of
active surface area to the overall isotherm takes place at
the early stages of adsorption, i.e., at very low P/P"values
(<0.01).The effect probably diminishes at PlP" > 0.01.
On these classes of carbon, there is no direct experi-
mental technique to determine ~ ( 0 2or) a(Ar) as has been
the case with GC. Therefore, the values listed in Table
I, which were obtained through the BET calibration with
a large number of experimental data points, are probably
accurate and reasonable enough to estimate the surface
areas of UGC and AGC as well as coal chars.
In order to fit ~ ( 0 2and
) a(Ar) to physical models, let
us first compare the experimental cross-sectional areas
against the theoreticalvalues, u(t1) and a&), listed inTable
111. There are two methods to calculate the theoretical
values. In the first one, the liquid (or solid) density, p, of
(42)DAmico, K.L.;Bohr, J.; Moncton, D.E.;Gibbs, D.Phys. Reu.
B: Condens. Matter 1990,4I,4368.
Langmuir,Vol. 8, No. 2, 1992 363
~~ ~ ~ ~ ~~
experimental a(X);nm2/species
UGCd AGCe silicas silicas/GC
0.162 0.162 0.162 1.OOO
0.159 0.155 0.186 1.348
N/A 0.234 1.300
0.214 NIA 0.270 1.262
0.146 0.148 0.166 1.121
nm2/adsorbedspecies. Graphitized carbons. Ungraphitized
adsorbate and its molecular (or atomic) weight, M,are
used
a(t,) = 1.091(M/pN)2/3
where N is Avogadro's number and 1.091 is the "tradi-
tional" hexagonal packing factor. If the adsorbed species
exhibit a different packing configuraton, the value of u(t1)
will change accordingly. In the second method, the critical
temperature, Tc (in K), and critical pressure, P, (in bar),
are used to estimate the van der Waals cross-sectional
a(t2), using
u(t,) = 6.354 ( T c / P c ) 2 / 3
As shown in Table 111,the agreement between a(t1) and
a&) is satisfactory which means that regardless of the
type of physical parameters used, there is practically one
theoretical value for each gas which may or may not agree
with the value obtained by the calibration. However, by
adjusting the packing factor in eq 1 from the "traditional"
hexagonal configuration to others, u(t1) could be altered
until it agrees with, or becomes closer to, the value
computed in this study. For example, the average value
of u(O2) on UGC and AGC is 0.147nm2/mol;a few percent
higher than u(t1). If the adsorbed oxygen molecules were
packed with only four nearest molecules on the surface
and eight in the liquid, the packing factor raises from 1,091
to 1.192,44and a(t1) increases from 0.136 to 0.148 nm2/
mol. The agreement between the latter value and the
0.147 listed in Table I suggests that the adsorbed oxygen
molecules on UGC and AGC are arranged in a square (not
hexagonal) two-dimensional packing. That is, the increase
in 402)may be due to a decrease in coordination number
on the surface. The experimental values of u(Ar) on these
carbons average to 0.157 f 0.002 nm2/atom, which is
smaller than the 0.138found on GC. Once again, the ad-
sorbed Ar atoms are packed looser on the heterogeneous
than the homogeneous (graphitic) carbon surfaces. In
conclusion, the surface heterogeneity of carbons, or the
presence of a high population of active sites, tends, on the
average, to spread out the Ar atoms or 02 molecules ad-
sorbed on the entire surface (active sites and basal planes).
As mentioned earlier, the BET plots for Kr on UGC
were convex to the relative pressure axis. The deviation
from linearity is a result of the assumptions used in deriving
the BET equation. That is, the heat of adsorption during
monolayer formation is constant, and the lateral inter-
action between the adsorbed species is negligible. UGC
have a high degree of surface heterogeneity, the heat of
adsorption should decrease with coverage yielding BET
(43)Hill, T.L.J. Chem. Phys. 1948,16, 181.
(44)Livingston, H. K.J. Colloid. Interface Sci. 1949,4 , 447,
364 Langmuir, Vol. 8,No. 2, 1992 Letters
un

70
60
t\ N762
--b
N550
--c
4-v 3

-p 50
2
0 40

30

20

0 ' 1 1 1 1 / 1 1 1 1 " " 1 1 " I " 1 1 1 1 ~ I I ( " ~ " " " " in
..
018 023 028 033 038 043 048 053 058 .I6 .I8 .2 .22 .24 .26 .28 .3
P P PIP
Figure 2. Variation of C with relative pressure below 0.06. Figure 3. Variation of C with relative pressure at 0.15-0.30.

plots concave to the relative pressure axis. On the other Zoo00

hand, the lateral interactions between Kr atoms adsorbed
on graphitic surfaces are well documented in the literature;
--
*
+
StMT
N762
N550
there is no reason to believe that these interactions do not -+ v3
4 N330
exist on UGC. If the first few Kr atoms are merely ad-
sorbed on the active sites a t low relative pressures below
0.01, as with the GC,2,3then the first point on the isotherm
is reached where Kr on UGC is adsorbed essentially on 2 10000
i)
the remaining (uncovered) graphitic surface. In this case,
the role of lateral interactions becomes important, the
heat of adsorption tends to increase to a maximum near
monolayer completion, and the BET plots becomes convex
to the relative pressure axis. Evidently, during mono-
layer formation of Kr on UGC, the offset caused by the
lateral interactions far exceeds that caused by surface .05 .06 .07 .OS .09 .I .I 1 .I2 .I3 .I4 . I5
heterogeneity; the net effect through the entire P I P range P P
would be an increase in heat of adsorption. To confirm Figure 4. Variation of C with relative pressure at 0.05-0.15.
this point, the heats of adsorption have to be measured
experimentally using, for example, an adsorption calo- P I P , C is low in the latter case (<loo)and keeps decreasing
rimeter. They can also be computed if the experimental with increasing P I P ; the drop in C occurs gradually over
isotherms were measured at two temperatures. Unfor- a considerably wide relative pressure range. At moderate
tunately, both capabilities are not available to the author P I P , 0.05-0.15, the monolayer builds up apparently to a
at the present time. However, the following attempt was completion, the value of C abruptly attains a maximum
made to support the previous discussion. (whose value could be as high as 20 OOO), then it quickly
The values of C in the BET equation were examined at falls down as shown in Figure 4. The relative position of
different P I P values up to 0.30. The term C is expressed the maximum with respect to P I P axis depends on how
as exp(qi- qL)IRT,where qi is the isoteric heat of adsorption fast the measurements are taken and how far the exper-
and q L is the latent heat of condensation. When the curved imental points are spread apart. The value of C at
BET plots are convex to the relative pressure axes, then maximum,,, C correlates neither to the nature of sample
as PIP" increases, the slope of the tangent progressively nor to its surface area. However, it is evident from the
increases, and the value of C is elevated. That is, (qi- QL) figures that there are continuous changes in C or heats of
increases with P I P . This could be satisfied by one or adsorption (qi-qL) as the measurements were progressing.
more of the following combinations: (1)qi increases but This is probably the main reason for obtaining a curvature
QL is constant; (2) qi increases but QL decreases; (3) qi in the BET plots.
increases at a faster rate than q L ; (4) qi decreases at a o(X) on Aerosil Samples. With Aerosils, the value of
smaller rate than QL. The assumption that QL remains a(Ar) = 0.186 nm2/atom is believed to be high, especially
constant cannot be easilyjustified with Kr on UGC because because it could not be justified with a reasonable physical
it is not valid for GC where several phase transitions occur, model. The Aerosil samples have a high population of
and different amounts of heat are supplied to or given by functional groups, mainly silanols (Si-OH).45p46 These
the system during the buildup of the mon0layer.~83 groups did not decompose appreciably by heating to 573
Further, it is not a simple experimental task to separate K, i.e., before performing the gas adsorption experiments.
the effect of qi from the net heat of adsorption if q L is Therefore, the adsorbed N2 molecules, because of their
changing. However, for the current purpose, an exami- high quadrupole moment, could interact with the silanol
nation of overall trend given by the C values can shed groups and become localized on the surface with the mo-
more light on the reasons for observing the convex BET lecular axis of N2 perpendicular to the surface. In this
plots. The assumption is made here that the C value in case, the rotation of adsorbed molecules is probably
the BET equation does, indeed, have some merit. restricted. The N2 molecular width (0.33 nm) is smaller
Figures 2 and 3 illustrate the variations of C with P I P ~ ~~ ~~~

during monolayer buildup ( P I P < 0.06) and after its (45) Technical Bulletin Pigments. Report #11 (Basic Characteristics),
Degussa Co., 1981.
completion (PIP = 0.15-0.30). While in the former case (46) Valencia, E.; Maldonado, A. J. Chem. SOC.,
Faraday Trans. 1990,
the value of C is high (100-550) and slightly increases with 86, 539.
Letters
than the length (0.44 nm); when the adsorbed molecules
are localized on the surface in a stand-up position, the
value of a on Aerosils is expected to be smaller than 0.162
nm2/mol, and there is enough evidence in the literature
to support this p 0 i n t . 4 ~ ~ ~
To propose a physical model, it was also anticipated
that the adsorbed Ar atoms could be in registry with the
underlying surface forming, for example, a square lattice
configuration, as it has been noted for the adsorption of
methane on MgO s u r f a ~ e s Yet, . ~ ~ this
~ ~ ~concept is not
applicable to the Aerosil samples because they are amor-
p h o ~with
s ~ no
~ specificcrystallographicstructure or lattice
parameter. This situation leaves us with one remaining
option; justifying a smaller value for a(Ar) on Aerosils.
There are three possible ways to achieve this goal. First,
assume a(Ar) on Aerosils = 0.138 nm2/atom as given by
eq 1 and as noted for the GC. In this case, a(N2)and a(O2)
are reduced to 0.120 and 0.123 nm2/mol, respectively.
Second, assume a(Ar) = 0.157 nm2/atom as noted for the
UGC and AGC, in this case, a(N2) and 4 0 2 ) become 0.137
and 0.140 nm2/mol,respectively. Third, assume ~(02) has
a universal value on all surfaces, 0.147 nm2/mol as noted
for all different classes of carbon. With this assumption,
the corresponding values of a(Ar) and a(Nz) on Aerosils
become 0.165 and 0.143 nm2/mol, respectively. With all
three assumptions, the proposed or computed values of
a(X) could be fitted to physical models describing the
arrangement of adsorbed species on the surface. There-
fore, physical modeling in itself cannot be used as the only
criteria for justifying one configuration over the other. I t
may also be added that regardless of the criteria used to
qualify one of the three possibilities, there is no direct
experimental technique for measuring a(X) on Aerosils;
thus all values are relative.
The third possibility can, however, be more substan-
tiated than the other two. First, for ~ ( N z=) 0.143 nm2/
mol, the packing factor in eq 1 has to be lowered from 1.09
to 0.97, suggesting that most of the NZmolecules in the
monolayers of Aerosils are adsorbed in the solid form whose
a value is 0.138 nm2/mol (Table 111). Alternatively, the
adsorbed N2 molecules may be present in the liquid form
~ ~ ~~~~
(47) Gregg, S. J.; Sing, K. S.W. AdsorptionSurface AreaandPorosity;
Academic Press: New York, 1982; pp 76-77.
(48) Kiselev, A. V. Discuss. Faraday SOC.1965, 40, 205.
(49) Furlong, D. N.; Rouquerol, F.; Rouquerol, J.; Sing, K. S. W. J.
Chem. SOC.,Faraday Trans. 1980, 76, 774.
(50) Coulobm, J. P.; Madih, K.; Croset, B.; Lauter, H. J. Phys. Rev.
Lett. 1985,54, 1536.
(51) Jung, D. R.; Cui, J.; Franke, D. R.; Ihm, G.; Kim, H. Y.; Cole, M.
D. Phys. Reu. B: Condem. Matter 1989, 40, 11893.
Langmuir, Vol. 8, No. 2, 1992 365
in a square two-dimensional packing with four nearest
neighbors on the surface and six in the liquid; that is, with
a packing factor = Second, with a(Ar) = 0.165
nm2/mol,the adsorbed atoms are present in a square two-
dimensional packing with 4 nearest neighbors on the
surface and 12 in the liquid (packing factor = 1.26). This
value of 0.165 is highly supported by other investigators
who performed N2 adsorption on Aerosils and other silica
samples.18*23.4649 Third, lowering a(N2) from 0.162 to0.143
nm2/mol requires a proportionality factor of 0.89. From
the adsorption of neopentane on GC and a factor
of 0.87 could easily be computed. Thus, by performing
the calibration using other hydrocarbon adsorbates,
a(N2)had to be lowered. Fourth, the proposed new values
of ~ ( N z )~, ( O Zand) , a(Ar) on Aerosils, and probably on
other forms of silica, conform to one pattern. With the
three gases, the adsorbed molecules in the monolayer are
all arranged in a square two-dimensional packing with
four adsorbed species on the surface and 6 (for Nz) 8 (for
0 2 1 , or 12 (for Ar) in the liquid. Thus, there is a consistency
in the packing pattern on the surface which is essentially
dictated by the silanol groups rather than the adsorbate.
Conclusions
1. It appears that oxygen is another universal adsorbate
at 77 K with 4 0 2 ) = 0.147 nm2/mol on GC, UGC, and
AGC. The present work suggests that this value could
also be applicable to Aerosil samples.
2. a(Ar) is 0.138 nm2/atom on nonporous graphitized
carbons and 0.157 nm2/atom on all other forms of carbons
(porous and nonporous); however, a(Ar) is different on
the Aerosil samples examined.
3. For adsorption of Kr on ungraphitized carbons, the
BET plot is not linear; however, if the data were analyzed
at P/PO of 0.02-0.25, using PO of solid Kr, then u(Kr) is
0.214 nm2/atom. This value is also valid for GC.
4. The followingvalues are recommended for adsorption
of gases on silica surfaces at 77 K: a(N2) = 0.143, ~ ( O Z ) ,
= 0.147, and a(Ar) = 0.165 nm2/adsorbed species.
Acknowledgment. The financial support of Air Force
Office of Scientific Research under Contract Number
F04611-83-C-0046with the Phillips Laboratory is highly
appreciated. I thank Professor William Steele for his
technical advise during the preparation of this manuscript
and Dr.Rodney Taylor (Columbian Chemicals Co.) for
providing most of the unactivated carbon samples.
(52) Carrott, P. J. M.; Roberta, R. A.; Sing, K. S.W. Langmuir 1988,
4, 740.