Professional Documents
Culture Documents
_ J _ {Znm)*{kT)*bg{T)
(1)
l-d h3 va(T)
350 W. J. C. Orr
and Strachan (1935) and Lennard-Jones and Devonshire (1936), who con
sider a one-dimensional case in which an isolated atom vibrates normal to
the surface, the potential energy of the system being represented by a
Morse curve,
V = D[e-2*(r-re) _ 2e-*<r- re)]. (2)
With this assumption the discrete vibrational levels are known provided the
parameters D and k can be otherwise determined. Considering th at the
effect of the thermal motion of the solid on V is small and may be treated by
first order perturbation theory, they calculate further the probabilities of
atoms making transitions from one discrete vibrational level to another or
evaporating into or condensing from the gas phase by the emission or
adsorption of quanta from the solid. In order to provide data to which these
various theoretical developments would be directly applicable the present
experiments on the adsorption of non-polar gases such as argon, oxygen and
nitrogen on the alkali halide crystals, KC1 and Csl were undertaken. The
choice of non-polar spherically symmetrical gases avoided the complications
which the presence of permanent dipoles would introduce, while the choice
of alkali halide crystals was influenced by the following factors: (1) they were
easily obtained analytically pure, (2) the lattice structures and such pro
perties of the individual ions as their ionic radii, optical polarisabilities and
magnetic susceptibilities were well known, (3) KC1 was an example of a
simple cubic and Csl of a body centred cubic crystal so th at the influence of
the crystal type on adsorption behaviour could be examined. Further the
previous experimental work on such crystals was very scarce. Durau (1928)
had made some measurements on the adsorption of argon, oxygen and
nitrogen at room temperatures on NaCl crystals but since the adsorption at
these temperatures is extremely small, accurate isotherms could not be
obtained and no heats of adsorption were calculated. Later, Tompkins
(1936) measured the adsorption of oxygen, carbon monoxide, nitrogen and
argon among other gases on NaCl at low pressures and liquid air tem pera
tures but obtained improbably small values for the heats of adsorption.
Some accurate experiments by Lenel (1933), however, gave heats of ad
sorption close to the order of magnitude expected for argon on KC1 and
other salts. The variation of the heat of adsorption with the amount adsorbed
was however not examined. Single isotherms for argon on KC1 which show a
qualitative similarity with the present results have been described by
Bradley (1936) and by Brunauer and Em m ett (1937).
Downloaded from http://rspa.royalsocietypublishing.org/ on October 29, 2018
E x pe r im e n t a l r esu l ts
352 W. J. C. On-
temperature then gave the volume S to approximately the same accuracy.
In order to determine the change in free volume of the salt in S due to con
traction, between room and liquid air temperatures, calibrations were
carried out at liquid air temperatures using helium, for which the adsorption
at 80° K is quite negligible.
reservoirs
pumps
capillary
—oil
manometer
„__ oil
manometer
""manometer
suction
pump
V= p Z P - A „ a,
VR
X= ^FpJ<Ds Z p -Vsl>)>
Vr = vr + V % ^ ,
where the subscripts!? and Srefer to the reference and salt limbs r
£ being the volumes of the bulbs at room temperature (TA) and vR S the
volumes of the capillaries ^M;
T VR sare
adsorption temperature ( TL).D RS are the volumes of the dos
E P is the sum of the successive pressures admitted via TXT%to the bulbs.
A Ug = 8 is the pressure difference on M in cm. of Hg a t room tempera-
PlLg
Downloaded from http://rspa.royalsocietypublishing.org/ on October 29, 2018
354 W. J. C. Orr
be set to regulate at any desired temperature by adjusting the mercury
level at D, the temperature being determined from the vapour pressure
(measured on the millimetre scale E),by referenc
these gases. For the present experiments the vapour pressure data of Cath
(1918) were used for liquid oxygen and those of Giauque and Clayton (1933)
for liquid nitrogen. In operation the temperature was controlled to within
0-02° K and perfectly steady readings of the oil level at M (figure 1) were
obtained.
^ 10
logio P (cm.)
F ig ure 3. Log x-\og p isotherms for argon, oxygen and nitrogen on KC
p (cm.)
F igure 4. Behaviour of the KC1 isotherms in the low-pressure range.
Downloaded from http://rspa.royalsocietypublishing.org/ on October 29, 2018
356 W. J. C. Orr
intervals. A few results showing the trend and the relative shape of the
log x-log p curves derived from these experiments are given in figure 6.
In all experiments adequate time was given for thermal equilibrium to be
established between the salt and the low temperature bath. The salt was left
in contact with the initial dose of gas for 2-3 hr. before readings were com
menced. For subsequent doses the pressure became steady within about
*<—N.
1
3
O
p (cm.)
F igure 5. Oxygen on KC1. Behaviour of the isotherms near the
condensation pressure.
« 05
No67-00°
0,83-00'
Ar8500°
logic P
F igure 6. Log a;-log p isotherms for argon, oxygen and nitrogen on Csl.
Heats of adsorption
The heats of adsorption as a function of were calculated from the above
experiments using the Clausius-Clapeyron equation in the form
2-303AH r 1 in
[logio P lM x R L*i
358 W. J. C. Orr
From figures 7, 8 and 9 it is evident th at the maximum experimental
error in AH is ±150 cal. which is of the order expected from a possible error
of ±0-5% in x and ± 0-01° K in T, the heats being calculated from iso
therms differing only by from 2 to 5°K.
The star in figure 7 which indicates the heat of adsorption found experi
mentally by Lenel (1933) will be seen to be in good agreement with the
present results.
Table 1
xm
x 106 g.mol.
A 02 n2
KC1 8-3 8-4 7-2
Csl 4-5 4-0 4-0
•s 2500
F ig ure 7. H eats of adsorption as a function of 6 for argon on KC1 <£) and on Csl 0 .
D isc u ssio n
In two recent papers (Roberts and Orr 1938; Orr 1939) detailed theoretical
calculations have been carried out of the adsorption behaviour of argon on
KC1 and Csl surfaces and the following results have been obtained:
(1) The energies of adsorption and the equilibrium distances of an isolated
Downloaded from http://rspa.royalsocietypublishing.org/ on October 29, 2018
ctbe 8. Heats of adsorption as a function of 6 for oxygen on KC1 (•) and on Csl © .
g
A H cal./g.mol.
360 W. J. C. Orr
atom at various positions on these surfaces were calculated directly and are
collected in table 2.
T a bl e 2
~E 0 ~E
»equil. cal./ Surface f'equil. cal./
KCl A g.mol. Csl layer A g.mol.
Above the centre 3-215 1593 Above the centre Cs+ ions (1-75) 3168
of a lattice cell of a lattice cell
Above the m id 3-44 1314 I - ions 2-60 2682
point of a lattice
edge
Above a K+ ion 3-48 1423 Above a lattice Cs+ ions 3-65 1100
point
Above a Cl- ion 3-74 1233 I - ions 4-08 1361
(2) In the case of KC1 it was shown that, in the formation of a primary
monolayer, atoms would be adsorbed from 0 to about 0-5 above the
centres of lattice cells but th a t as 0 tended to unity the arrangement would
pass over to a close-packed hexagonal pattern. In this configuration, the
adsorbed atoms being distributed a t random with respect to the lattice
periodicities, the energy of adsorption normal to the surface at = 1 would
av
be a weighted average of values in column 3, table 2, namely —E — 1412 cal.
KCl
The calculated behaviour is shown in figure 7, curve 1. I t should be noted, in
this case, th at the electrostatic field is zero from = 0 to 0-5 and th at from
0 = 0-5 to 1-0 the electrostatic contribution to the mutual energy of the
molecules in the adsorbed layer is less than 3 % of the mutual energy due to
van der Waals forces.
(3) In the case of Csl both Cs+ and I - faces are present in equal amounts.
I t was shown th at in both cases the most stable arrangement over the range
6 = 0 to 1*0 would be that with atoms adsorbed above the centres of lattice
cells. The calculation of the variation in heat in the actual physical case
where the two surfaces are competing for atoms was carried out making use
of equation (1). The result is plotted as curve 2 in figures 7 and 8. The
distinct difference between this curve and curve 1 is due to the important
part played by the electrostatic potential of the crystal in this case. Dipoles
orientated in parallel are induced in the adsorbed molecules and these give
rise to a repulsive potential which, although amounting only to 16 % of the
van der Waals contribution to the mutual energy at 1 in the case of an
I - face, becomes 33 % greater than the van der Waals field in the case of the
Cs+ face at the same value of 6.
Downloaded from http://rspa.royalsocietypublishing.org/ on October 29, 2018
Further, the diameters of an argon and an oxygen atom derived from the
densities of the liquids at 87° K are 4-05 and 4-00 A respectively. Thus it can
be anticipated th at the behaviour of these two gases in weak van der Waals
adsorption fields will be very similar indeed. A comparison of figures 7 and 8
will show th at this is true of adsorption on KC1 and Csl. In the case of
nitrogen it would be expected from the lower heat of interaction of a pair of
nitrogen molecules and also from the lower heat of evaporation of the
liquid compared with argon or oxygen, that the adsorption heat of nitrogen
would be lower than th at of the other two. A similar conclusion is indicated
from theory, for taking account of the increased diameter of nitrogen
(namely, 4-38 A at 85° K) compared with argon and oxygen, the van der
Waals energy at the equilibrium distance calculated, alike by the London
formula (1930) and the Kirkwood-Muller formula (1936), gives a smaller
negative energy than in the other two cases. I t is thus surprising to find th at
the measured heats of adsorption AH, for nitrogen are on the average about
400 cal. greater than those of oxygen and argon. The result is, however, not
peculiar to the present experiments but has been observed by Barrer (1937)
with adsorption on graphite, by Wilkins (1938) on platinum, by Zeise (1928)
Downloaded from http://rspa.royalsocietypublishing.org/ on October 29, 2018
362 W. J. C. On-
on glass. An observation of possible significance in this connexion is th a t of
Vegard (1934), who finds evidence, from the crystal structure of solid nitro
gen, of a small polarity comparing very exactly with th at found for the iso-
steric molecule carbon monoxide. Although the explanation of such a
polarity is not clear, if the effect is real, then it might be expected to be
operative also for nitrogen in the adsorbed state and would explain the
observed high heats of adsorption. In some unpublished, adsorption ex
periments on LiCl, nitrogen and carbon monoxide isotherms behaved very
similarly and differed quite markedly from the shape of argon isotherms.
A direct comparison of the theoretical calculations with the experimental
results may no w be made, taking first the case of argon on KC1. From figure
7 it is evident th at the calculations reproduce quite exactly both the shape
and the absolute magnitude of the experimental curve from d = 0-5 to 1-0.
The divergence of the curves a t lower values of 0 is most probably to be
ascribed to adsorption taking place in cracks and other irregularities in the
crystals. Initially molecules will be adsorbed on the most energetic positions,
th a t is, in the bottoms of deep fissures where the adsorption heat may rise to
av av
a maximum of ^ —2 E —2<x —3060cal. ( —E = 1412 has been previo
KCl K C1
defined and — cl —
240 cal. is the energy of two argon atoms at contact
From the extent of divergence of the experimental curve it may be estimated
th at about 15 % of the surface of the sample of crystals used consists of
cracks of from one to two molecular diameters width. Where the fissures are
wider than this the adsorption energy will not differ significantly from ad
sorption on the plane surface. The number of molecules which can be ad
sorbed in a second layer will thus be approximately 85 % of I t is further
evident, comparing figures 7 and 8, th at the same calculation reproduces
equally well the shape of the (AH, 6) curve in the case of oxygen. In absolute
magnitude the values of AFIfor oxygen are only about 100 c
those for argon. A further fact consistent with the interpretation we are
here giving to the results is th at the values of xm for argon and oxygen are
closely equal, being 8-3 and 8-4 x 10~6g.mol. respectively. This would be
expected from the similarity in their radii. In the case of nitrogen on KCl,
besides the fact th at the absolute magnitude of is greater than th at
expected theoretically, the variation of with is somewhat greater than
would be calculated. However, it is not permissible to draw definite conclu
sions from the latter fact since a consideration of the shape of the isotherms
(see figure 3) in the region near xm will show th at a very slight experimental
error would give rise to large variations in the heat of adsorption. The value
of xm = 7-2 x 10~6g.mol., however, is significantly smaller than that found
Downloaded from http://rspa.royalsocietypublishing.org/ on October 29, 2018
Connected with the distinct U-shaped appearance of the curves for these
gases on KC1 is the fact th at the (p,x) isotherms all hav
flexion in the region between 6= 0 to 1-00 (see figure 4). It h
by Fowler (19366) and by Wang (1937) that an adsorption isotherm will not
show an inflexion in the range where AH (or y, eqn. (1)) is steadily decreasing.
In figure 10, curves a and 6 represent two simple Langmuir curves, a with
and 6 with y2, where Xi > X2 - Curve c represents the behaviour of an isotherm
where y is falling from yxtowards y2. Further, Fowler (1936a) has shown that
Downloaded from http://rspa.royalsocietypublishing.org/ on October 29, 2018
364 W. J. C. Orr
when the molecules in their equilibrium positions in the first layer attract
one another, the isotherms will always have one inflexion, as shown in curve d.
In the present experiments AH falls initially and subsequently rises due to
lateral interactions. Thus we expect the isotherm first to follow a curve such
as c and later to bend over to one similar to d. The resultant effect will be as
shown in curve e, which reproduces all the essential features of the isotherms
such as those in figure 4, found experimentally in the case of KC1. In the
case of Csl, the very slight maximum in the curves is apparently insufficient
to produce a noticeable inflexion in the initial portion of the isotherms (see
figure 6). I t may be concluded th at for the range 0 to 1*0 all the distinc
tive features of the results can be quite adequately explained on theoretical
grounds.
I t now remains to discuss the behaviour of the systems in the range from
the completion of a monolayer to the bulk condensation of the gas on the
solid. In table 3 are given values of the ratio of the equilibrium vapour
pressure (p) to the saturation vapour pressure of the bulk gas phase (ps ) for
representative isotherms a t amounts adsorbed corresponding to 1, 2, 3 and 4
apparent layers, the area available in the second and subsequent layers
being assumed equal to 85 % of th at in the first.
T able 3
P/Ps Bulk
"S condensed
No. of “ layers” 1 2 3 4 phase
Argon Solid
(79-35° K) 0-25 0-69 0-90 0-96 m.p. = 83-9° K
Argon
(86-89° K) 0-22 0-63 0-80 0-90 Liquid
Oxygen
(79-32° K) 0-20 0-65 0-80 0-85 Liquid
Nitrogen
(73-57° K) 0-012 0-44 0-70 0-85 Liquid
The similarity between oxygen and argon and the contrast in their be
haviour to th at of nitrogen is here very clearly brought out, the relatively
very low value of PiPsat one monolayer, in the case of nitroge
indicative of the much stronger interaction of these molecules with the
surface. I t is also significant to note th at p/ps for nitrogen reaches values
similar to those of oxygen and argon when about three or four apparent
layers are adsorbed, thus illustrating the very rapid falling off in the in
fluence of the surface field. The fact th at the vapour pressure does not reach
Downloaded from http://rspa.royalsocietypublishing.org/ on October 29, 2018
Using the value PdPs = 0*69 from table 3, we obtain (Sads —Serys) =
entropy units. This figure is comparable with the entropy changes occurring
at transitions between two different crystalline species, and it is a measure of
the increase of disorder in a bimolecular layer over th at of crystalline argon.
This disorder results from the adjustment which must be made in passing
from the hexagonal pattern which we have attributed to the first layer of
argon on KC1, or the cubic pattern in the case of argon on Csl, to the face
centred structure of the bulk crystal. If the second layer were condensing in
the form of localized crystallites no such distinct entropy difference could be
expected. We thus conclude that, subsequent to the formation of a first
layer, which we have already shown has a well defined range of existence, we
must suppose that, in the case of argon below its melting point, a fairly
distinct second layer is also built up. In the cases of oxygen and argon
(above 83-9° K), the second layer is possibly not quite so distinct as in the
previous case, while in the case of nitrogen, where the relative vapour pres
sure is so much smaller than in the case of argon, it seems likely that the
second layer may be more distinct. By the third or fourth apparent layer the
effect of the surface field must be quite unimportant and the geometric
relations characterizing the bulk phase, whether solid or liquid, should be
almost completely satisfied. From this stage onwards the growth of the
adsorbed phase may take place at random. These results are in harmony
with most recent experiments on multimolecular adsorbed gas layers and
may be considered as lending further weight to the viewpoint that relatively
few layers, no more than four or five at the most, are adsorbed before bulk
condensation of the adsorbed phase sets in and, in particular, as giving
Vol. 173. A. 24
Downloaded from http://rspa.royalsocietypublishing.org/ on October 29, 2018
366 W. J. C. Orr
support to Bangham’s (1938) hypothesis of the laminated character of
these primary surface layers.
S um m ary
R eferences
1. I ntroduction