Colloid & Polymer Science Colloid & Polymer Sci 264:362-367 (1986)

Methanol adsorption by amorphous silica alumina

in the critical temperature range
M. Kuczynski, A. van Ooteghem, and K. R. Westerterp
Chemical Reaction Engineering Laboratories, Department of Chemical Engineering, Twente University of Technology, Enschede, The
Netherlands

Abstract: The methanol adsorption capacity of an amorphous silica-alumina was meas-

ured using an equilibrium technique. The experimental temperature range was of 140 to
260 ~ and the pure methanol pressure range was 0.1 to 1.2 MPa. A multilayer adsorption
was found, also for temperatures above the critical temperature of the adsorbate. Based
on the Jovanovic adsorption model, the mean residence times of the adsorbed molecules
were calculated. Surprisingly, the heat of adsorption was found to be independent of the
temperature in the multilayer adsorption range.

Key words: Methanol [67-56-I], high temperature adsorption, silica-alumina, multilay-

ers, adsorption heat.

Notation larly at high absorbate pressures and temperatures.

Symbols Units Generally, the reliability of the existent theoretical
a = parameter defined by Equation (7) expressions for adsorption equilibria decreases rapidly
a' = parameter defined by Equation (10) Pa with increasing pressure. In this paper, we present the
a(T) = parameter in the Peng-Robinson equation Pa/(mol/m3) 2 results of our adsorption measurements for methanol
b = parameter defined by Equation (8)
b' = parameter defined by Equation (11) pa < on the low alumina, LAC-25 cracking catalyst of
fi(T) = parameter in the Peng-Robinson equation m 3 tool-1 AKZO Amsterdam. We consider the multilayer
c = constant in the BET equation adsorption theory of Jovanovic to be valid also above
k = Boltzmann constant, being 1.3806 910-23 j K -1 the critical temperature of the adsorbate. Therefore,
K = characteristic constant in the Peng- we used his adsorption isotherm expressions to fit our
Robinson equation B

m = mass of one adsorbate molecule kg

experimental data.
p = pressure Pa
q = adsorption capacity g/g
Q = heat of adsorption J mol -i Experimental
R = gas constant, being 8.314 J mo1-1 K-'
T = absolute temperature K The experimental system is schematically shown in Figure 1.
V = molar volume m 3 mo1-1 Prior to each experiment, the solid sample was dried for several
x = relative pressure (= P/Po) hours at 280 ~ under vacuum, weighted and placed in a 570.10 -6
o = active molecule area m2 m 3 autoclave. Subsequently, the air was evacuated from the auto-
r = residence time clave at 280 ~ and the apparatus was cooled to room temperature.
o) = acentric factor A precisely measured volume of liquid methanol was then sucked
into the autoclave. The valves V1 and V2 are used to evacuate the air
and to inject the methanol samples, respectively. After evacuation
Introduction and injection, plugs were installed on the valves in order to prevent
leaking. The pressure in the autoclave was measured at several tem-
A theoretical interpretation of the experimental peratures between 140 and 260 ~ each time after thermal stabiliza-
equilibrium data for adsorption is still difficult, particu- tion of the system. An electronic pressure transducer was used for

W18
Kuczynski et aL, Methanol adsorption by amorphous siSca alumina in the critical temperature range 363

in which

a(T) =0.45724 R_Tc22 (1 + K(1-T,))2,Nm4mo1-2

Pc
vacuum (2)
I Tc
K = 0 . 3 7 4 6 4 + 1.54226 co--0.26992 co2 (3)
H2 ~ Vl 1__ir /
and

b(T) = b(Tc) = 0.07780 -Rp~, m 3tool -1 (4)

where cois the acentric factor for methanol. In Figure 2

our experimental data are compared with the predic-
tions of the Peng-Robinson and the Beattie-Bridgeman
[2] equations of state. On the vertical axis the procen-
tual differences in the predictions of pressure are plott-
Fig. 1. The experimental set-up. Significance of symbols: .4 = 570- ed, with respect to the measured ones. Other equa-
10.6 m 3 stainless steel autoclave; V1 = valve for evacuation; V2 = tions of state, e. g. the virial equation, gave differences
methanolinjection valve; O = autoclave oven; Hi, H2 = heating ele- of more than 2 ~ with respect to our results. Equation
ments; TC = thermocouples; P electronic pressure transducer; S = (1) gives also a good fit of the p - V - T data for methanol
sample
found in Literature [2, 3], and therefore we used it in
the calculations.
Based on the most recent vapour-liquid equilibrium
data for methanol by Ambrose et al. [4] we recalculat-
the pressure measurements. The temperature of the transducer was ed a value of 0.564 for the acentric factor of methanol,
kept constant at 200 _+ 0.1 ~ in order to prevent thermal drift of its with the relation:
calibration line. This procedure was repeated for various amounts
of methanol injected, from 1 to 7.10-6 m 3. The weight of the solid co = - log10 Cpsaq
~ r j%=o7 - 1.
sample was 8.75 g. After the experiments, the composition of the
gaseous autoclave contents was determined by means of gas chro-
matographic and NMR measurements. No decomposition pro-
ducts of methanol could be detected although dimethyl ether, for
example, might be expected at the higher temperature levels investi-
gated. There was no adsorption hysteresis observed, either. The
same experimental procedure, but without the adsorbent, was used 1.0--
to obtain the most appropriate equation of state for methanol in the
Q
pressure and temperature range of our adsorption measurements.
0

Results o.~

0
The amount of methanol adsorbed was calculated
from the difference between the methanol pressure -~ ~ a a
measured without and with the solid sample in the o 1.0

autoclave. Our experimental data obtained without ~ o_ 'a''' 0:9 .... I ; Tr

4 A
the adsorbent were compared to the predictions of O
various equations of state for methanol. The experi- * A O
O
mental data agree well with the two-constant Peng- 8 -o5
Robinson equation of state [1]:
Fig. 2. Comparison of the experimental p-V-T data with the pre-
dicted values with the Peng-Rohinson (A) and Beattie-Bridgeman
RT a(T) (O) equations of state. 5.5.10 - 6 m 3 methanol injected into the 570.
P - V -. b V ( V + b) + b ( V - b) (1) 10 - 6 m 3 autoclave
364 Colloid and Polymer Science, VoL 264 9No. 4 (1986)

Table 1. Surface area distribution of the low alumina cracking cata- 0.192 + 0.006 g/g. This indicates a multilayer adsorp-
lyst a)
tion. Another feature of our adsorption isotherms sup-
Pore diameter range Surface area porting the multilayer adsorption hypothesis is the
nm mVgram convexity of the isotherms toward the pressure axis in
the range between the lines AA' and B B ' in Figure 3
2.0- 3.0 92.6 and the apparent existence of an inflexion point in the
3.0- 4.0 101
4.0- 5.0 138.1
left part of this area. Such S-shaped isotherms are com-
5.0- 6.0 89.2 monly considered to correspond with multilayer
6.0- 7.0 38.4 adsorption.
7.0- 9.0 31 This inference is in disagreement with the essential
9.0-13.0 10 assumption of the multilayer adsorption theories as
~) These results were obtained in a BET apparatus. It was
given by Hiittig [6], Brunauer, Emmett and Teller [7]
assumed that all pores have a cylindrical shape. and modifications of the BET theory [8-12]. Accord-
ing to these authors multilayer formation should be
c impossible in our pressure and temperature range,
which is above the boiling point of methanol.

"~ o.154
t , ~---" ~\r_--.18o
As is well known the BET equation given by Equa-
tion (5)
I '\ F - - - - ' ' " q6 \ _ - - e ' ~ X ) - q _ C~

OaO ~, , 1'~'- .a ~ , . _ - - , ~ ' 4 o

qm (1 - x) (1 - x + cx) (5)
~. , / W , / ~,- . . . . . .o~-~ is quite inaccurate for relative pressures higher than 0.6
i/ ~ 7/~" ~.r--- k,.,

C~05
-- /d
~ /''4G / . .
/' , J~ /
~" .9
--/ '~
x ,
or 0.7. Usually r -> 1 [7], further the term (1 - x) in the
denominator of the BET expression becomes signifi-
iz z
'zz<..
/
cant at high pressures making the isotherm convex
towards the pressure axis. Although (1 - x) gives a
o correct "boundary" behaviour of the isotherms at
0 0!5 I i!0 I
PNIeOH/ MPa pressures tending to the saturation pressure of the
adsorbate, the quantitative description of the multi-
Fig. 3. The adsorption isotherms for methanol on the silica-alumina
cracking catalyst. Parameter is the temperature in ~C
1
layer adsorption is still inaccurate if ~ becomes
significant. Generally, the modifications of the BET
theory all introduce correction terms to the heat of
The pore diameter distribution of the adsorbent was liquefaction for the subsequent adsorbed layers, still
determined using the usual BET technique; the results using the saturation pressure as a temperature depend-
are given in Table 1. The measured adsorption iso- ent parameter. In this manner, the saturated vapor
therms are shown in Figure 3. pressure is being ascribed to a "liquid-like" adsorbed
phase consisting of a number of adsorbed layers having
different binding energies. This approach is rather
Discussion and interpretation
dubious in our opinion.
It is obvious that the complicated profile of our iso- In 1969 Jovanovic [13] proposed a new adsorption
therms cannot be described by any of the usual adsorp- theory for a homogeneous surface without lateral in-
tion isotherm expressions, such as the BET isotherm teractions between the adsorbed molecules. TheJova-
or its modifications, for example. novic isotherm is expressed by the following equation:
A surface area determination by the BET method
gave the value of 505 _ 9 m2/g solid. Taking the value q = qm {1 - exp ( - ax)} exp (bx) (6)
of 0.212 nm 2 for the effective area of one adsorbed
methanol molecule [5], the maximum possible mono- where q is the adsorbed amount, qm is the monolayer
layer capacity is 0.127 g MeOH/g solid, while our ex- adsorption capacity, and x is the relative pressure,
perimental resuks for the adsorption capacity go up to defined as the ratio of the adsorbate pressure to its satu-
Kuczynski et al., Methanol adsorption by amorphous silica alumina in the critical temperature range 365

ration pressure. The term 1- exp ( - ax) represents the multilayer adsorption parameter b can be obtained
monolayer adsorption, and the term exp (bx) is the from the convex segment of the adsorption isotherm,
correction for the multilayer adsorption. The parame- because in a semi-logarithmic plot of q vs. x a straight
ters a and b describe the monolayer and muhilayer line will be obtained with qm as the intercept of the y-
adsorption, respectively: axis and with b as the slope. In our case the experimen-
tal points in the region between the lines AA' and B B '
a = o r Po/1/2n m k T (7) in Figure 3 form straight lines in the semi-logarithmic q
vs. p plot as shown in Figure 4. The values of qm and b'
b = o rL po/1/2~ m k T (8) were calculated by linear regression; the results are
presented in Table 2. It can be seen that the monolayer
where o is the active area of one adsorbed molecule, Po capacity qm decreases with increasing temperature.
is the saturation pressure, and r and rL are the mean Our experiments did not cover the low pressure range
residence times in the adsorbed state in the first and in where the adsorption isotherms are expected to be
the higher layers, respectively. concave toward the pressure axis. Therefore, we are
Jovanovic assumed for the case of a multilayer not able to determine experimentally the monolayer
adsorption that the molecules adsorbed in the higher adsorption parameter a'.
layers do not deprive the molecules in the underlying
layers of their capability for desorption. The adsorbed
phase is thus considered as a zone of compressed gas
rather than a liquid-like phase. Therfore, the pheno- 0.2-
menon of muhilayer adsorption has no relation at all to 140
/ 160
the saturation pressure and Equations (6)-(8) have to .180
be rewritten for multilayer adsorption as follows: "~ 0.1
f ,'~ ~200
,.,f220
/ J /~ 240
q = qm {1 -- exp(1 -- a'p)} exp (b'p) (9) ~ //"( ~ ~ 260
8
,~ o o~
with

a'=or/r (10)

b' = o zL/1/2~ m k T . (11) 0.~

Equations (9)-(11) are essentially the original expres-

sions obtained byJovanovic. He used the relative pres- 001 i
sure x in Equations (6)-(8), because, as he writes, in
o o'.2 ' 0'.4 ' d6 '
PMeOH/ MPa
some cases it is more convenient to use relative instead
Fig. 4. The experimental adsorption capacity as a function of the
of absolute pressures. In our opinion, the only advan- methanol pressure. Parameter is the temperature of the system in
tage of Equations (6)-(8) is that they can be easily used ~
for the elaboration of the literature data, as was done
by Jovanovic in his subsequent paper [14]. A number
of examples of the use of the Jovanovic equations in a Table 2. The values of qm and b'
dimensionless form can also be found in the Literature
[15-17]. We believe that the equations ofJovanovic are T qm b'. 10 6
also valid above the critical temperature of the adsor- K gMeoH/gsolia Pa-1
bate and therefore that the introduction of the satura-
413 0.0256 4.92
tion pressure is inappropriate. 433 0.0226 4.39
If we only consider the lines to the left of the line B - 453 0.0199 3.95
B' in Figure 3, our experimental isotherms can be treat- 473 0.0175 3.59
ed as type II isotherms according toJovanovic's classi- 493 0.0152 3.30
fication. Jovanovic [13,14] and Yaronets and Yaronets 513 0.0131 3.07
533 0.0112 2.83
[15] showed that the monolayer capacity qm and the
366 Colloid and Polymer Science, VoL 264. No. 4 (1986)

Table 3. The mean residence times of the adsorbed molecules of QL/R is obtained from the slope of the straight line,
provided O~ is independent of the temperature. The
T rL" 101~
K s
result is shown in Figure 5. It can be seen that O~ is
indeed independent of temperature. After linear
413 5.06 regression we found a value o f - 6430 J/mol for QL.
433 4.63 This is equal to the heat of condensation of methanol at
453 4.26 237.7 ~ just below the critical temperature [19]. Fur-
473 3.96
493 3.72 ther, from the linear regression we derive a value of
513 3.52 7.76.10 -11 S for r(o)L. This is a very large value when
533 3.32 compared to the value of 10 -13 s as suggested by de
Boer [18] for this parameter.
A quantitative interpretation of our adsorption iso-
therms in the high methanol pressure range (on the
fight of the line BB' in Fig. 3) is much more difficult.
As the active area of one adsorbed molecule o is in- Their stepwise form is analogous to the type W and
dependent of the temperature [13] the mean residence VIII isotherms in the dassification of Jovanovic.
times of the adsorbed molecules can be calculated However, he explains the occurrence of the irregular
using Equation (11). In compliance with the theory of steps by capillary condensation phenomena. As
Jovanovic we took a value of 0.424 nm 2for o, being the already mentioned above, the pressure and tempera-
effective area of one admolecule multiplied by two. ture range of our experiments is far removed from the
The calculated values of ZLas a function of the temper- region of methanol condensation and capillary con-
ature are shown in Table 3. densation is out of the question. In our opinion, the
Based on Equation (12), the heat of adsorption can be existence of the two "retortion points" on every iso-
calculated for the multilayer range [13,14,18]: therm - which agree with the points of intersection of
the isotherms with the lines BB' and CC' in Figure 3 -
(12) is possibly related to the pore diameter distribution
ZL=r(o)Lexp(~--T)"
and surface area distribution of the adsorbent as given
in Table 1.
Substituting Equation (12) into Equation (11)and plot- A peculiarity can be observed in the shape of the iso-
ting b'l/T- vs. 1/T in a semi-logarithmicplot, the value therms in Figure 3: the ratio of the adsorption capacity
at the "retortion point" to the calculated monolayer
capacity is nearly equal for every isotherm. For the first
"retortion points" - corresponding to the line BB' -
lOft
this ratio approximately amounts to 5.5 to 5.9 while
for the higher "retortion points" - line CC' - it lies be-
tween 7.3 and 7.6.
9- The decrease in the slope of the isotherms in the
higher region is possibly caused by the fact that at suffi-
n ciently high adsorbate pressures the pores with diame-
v
ters smaller than a certain value are completely filled
with the adsorbate. For cylindrical pores, the number
L: of adsorbed molecules in each succeeding layer will
decrease, while for crevice pores it will remain con-
stant. We do not know which pore shapes actually
occur in our solids. Another difficulty is to ascribe a
value to the thickness of one adsorbed layer. Moreov-
6x10- er, we do not know whether the layer thickness will
119 ' 2!1 ' 2!3 '
remain the same for each subsequent layer. Due to
1000 /K_I
T these unknown factors a quantitative description of the
Fig. 5. The temperature dependence of o'ci./l/'~--mk in Equation isotherms for the higher adsorbate pressures is impos-
(11) sible.
Kuczynski et al., Methanol adsorption by amorphous silica alumina in the criticaltemperature range 367

Conclusions 3. Bhattacharyya D, Thodos G (1964) J Chem Eng Data 9:530

4. Ambrose D, Sprake CHS, Townsend R (1975) J Chem Ther-
The Peng-Robinson equation of state was found to rood 7:185
give an accurate representation of the experimental 5. Susic MV, Vucelic DP, Pausak SV, Karaulic DB, Milakovic-
data for methanol, at reduced pressures of up to 0.15 Vucelic V (1969) J Phys Chem 73:1975
6. Hiirtig GF (1948) Monatsh Chemie 78:177
and in the reduced temperature range of 0.8 to 1.05. 7. Brunauer S, Emmett P, Teller E (1938)J Am Chem Soc 60:309
Based on the recent literature data a value of 0.564 was 8. Anderson RB (1946) J Am Chem Soc 68:686
found for the acentric factor of methanol, as distinct 9. Smith RN, Pierce C (1948) J Phys Chem 52:1115
from the value of 0.556 which is commonly reported 10. Steele WA (1956) J Phys Chem 25:819
in the literature. 11. Pickett G (1945) J Am Chem Soc 67:1958
12. Lunde PJ, Kester FL (1975) Chem Eng Sci 30:1497
From the experimental results obtained it can be 13. Jovanovic DS (1969) Kolloid ZZ Polym 235:1203
concluded that mukilayer adsorption is possible at 14. Jovanovic DS (1969) Kolloid ZZ Polym 235:1214
temperature above the boiling point, even beyond the 15. Yaronets YA, Yaronets M (1980) Bull Acad Sci USSR Div of
critical temperatures of the adsorbate. The Jovanovic Chem Sci 29:870
16. Jaroniec M, Sokolowski S, Waksmudzki A (1976) Roczn Che-
multi]ayer adsorption model was found to be valid in mii 50:779
this temperature range. The heat of multilayer adsorp- 17. Sokolowski S, Jaroniec M, Waksmudzki A (1976) Roczn Che-
tion calculated from his model was found to be tem- mii 50:1149
perature independent. We have not found any experi- 18. (ie Boer JH (1968) The Dynamical Character of Adsorption,
mental adsorption data obtained at conditions com- Oxford University Press, London
19. Radosz M, Lydersen A (1980) Chem Ing Tech 52:756
parable to those of our experiments in the literature, so
we have no support from other experiments for this Received August 27, 1985;
rather surprising result. accepted September 23, 1985

Acknowledgement Authors' address:

The authors thank the Twente University of Technology and the M. Kuczynski
Royal Dutch Shell group for the financial support. Chemical Reaction Engineering Laboratories
Department of Chemical Engineering
References Twente University of Technology
P.O.Box 217
1. Peng DYu, Robinson DB (1976) Ind Eng Chem Fundam 7500 AE Enschede, The Netherlands
15(I):59
2. HirschfelderJO, Curtiss CF, Bird RB (1966) Molecular Theory Correspondence should be directed to K. R. Westerterp, same
of Gases and Liquids, 3rd ed, J Wiley & Sons, New York address.