You are on page 1of 11

Chemical Engineering Science, Vol. 43, No. 12, pp.

3185-3195,
Printed in Great Britain.

KINETICS

Department

1988.
0

OF LOW-PRESSURE

METHANOL

SYNTHESIS

G. H. GRAAF,+
E. J. STAMHUIS
and A. A. C. M. BEENACKERSZ
of Chemical Engineering, State University of Groningen, Nijenborgh
l&9747
Netherlands
(Received

11 June

1987; acceptedfor

publication

1 June

ooo9-X09/88
$3.00 + 0.00
1988 Pergamon Press plc

AC Groningen,

1988)

Abstract-ThB kinetics of low-pressure methanol synthesis, starting from CO, CO, and hydrogen over a
Cu-Zn-Al catalyst,
were studied in a spinning
basket reactor
at p= 15-50
bar and
T= 21&245C. The results show that methanol can be formed from both CO and CO,.
Besides these two

commercial

reactions the water-gas-shift


reaction takes place. Based on these three reactions and a dual-site adsorption
mechanism,
48 kinetic rate models are derived. Hydrogen is believed to adsorb dissociatively. The
experimental results support this assumption. Based on X2-statistics and consistency tests a final kinetic rate
model

is selected.

This kinetic model

gives a significantly

better agreement

with the experimental

results than

kinetic models taken from recent literature.

INTRODUCTION

Kinetic data often play an important role in designing


a chemical reactor and methanol synthesis is no
exception
to this. Unfortunately,
there is still no
agreement in the literature on the kinetics of methanol
synthesis, not even for the same types of catalyst. The
objectives of this paper are to clarify which reactions
are involved in methanol synthesis and to derive
kinetic rate equations for these reactions.
LITERATURE

Although low-pressure
methanol synthesis is an
important industrial process, the kinetic studies on
this subject as published in the open literature are very
often conflicting. The role of CO, especially is insufficiently understood. This can be seen in Table 1.
Most models published up till now describe methanol
formation from CO only. The role of CO, in these
models, if present, is restricted to competitive adsorption on the active sites of the catalyst. Contrary to this
some authors (Dybkjaer, 1985; Chinchen et al., 1984)
claimed that methanol is formed from CO, only.
According to Dybkjaer this is because strong adsorption of CO,
prevents the co-adsorption
of CO.
Chinchen et al. based their conclusions on experiments with labelled carbon in CO,. A third group of
authors concluded that methanol is formed from both
CO and CO,. Liu et al. (1985) came to this conclusion
based on experiments with labelled oxygen in CO*.
Denise and Sneeden (1982) and Klier et al. (1982)
reached the same conclusion based on kinetic experiments.
Most of the authors mentioned in Table 1 presented
kinetic rate expressions. The rate expressions published more recently are listed in Table 2.

TPresent
address:
N.V.
Nederlandse
Gasunie,
Laan
Corpus den Hoom
102, 9728 JR Groningen,
Netherlands.
*To whom correspondence
should be addressed.

Seyfert and Luft (1985) (see Table 2) assumed a


Langmuir-Hinshelwood
mechanism in which CO and
H, are believed to be non-dissociatively
adsorbed on
the same kind of active sites. Methanol is made in a
two-step reaction: in the first step formaldehyde
is
formed in an equilibrium reaction; the second step, in
which adsorbed H, and adsorbed formaldehyde react
to form methanol, is believed to be rate-controlling.
Villa et al. (1985) (see Table 2) also assumed a
Langmuir-Hinshelwood
mechanism in combination
with non-dissociative adsorption of CO and H,. The
rate-controlling
step is believed to be a trimolecular
surface reaction between adsorbed CO and two adsorbed H, molecules as proposed originally by Natta
(1955).
Klier et al. (1982) (see Table 2) presented a kinetic
rate expression based on two synthesis routes. The
first term in their kinetic rate expression describes
methanol formation from CO and H,. Furthermore
they assumed that the active sites can be reduced to
inactive sites by a redox equilibrium involving CO and
CO,. The second term in their kinetic rate expression
describes methanol formation from C02.
The kinetic rate expression proposed by Dybkjaer
(1985) (see Table
2) is based
on a dual-site
Langmuir-Hinshelwood
mechanism in which H, is
believed to adsorb dissociatively and reacts with adsorbed CO,. Dybkjaer also reported the results of
studies on the chemisorption
of H,, H,O, CO and
CO2 on Cu-Zn<r
and Cu-Zn-Al
catalysts. H, and
H,O are believed to adsorb competitively on one kind
of active site, while CO and CO2 are adsorbed on a
second kind of active site.
In addition, Herman et al. (1979) noted the probability of a dual-site mechanism, because ZnO is a
good hydrogenation
catalyst that activates H, by
dissociative adsorption, while CO is poorly adsorbed
on ZnO, and because univalent copper is known to
adsorb CO but not H,. Matulewicz (1984) concluded
from initial-rate experiments that a dual-site mech-

3185

3186

G. H. GRAAF

Table I. Role of CO,

in methanol

as reported

rate

Zn-Cr
Zn-Cr
Cu-Zn-Al
cu-?
Cu-Zn-Al
Cu-Zn
Cu-Cr
Cu-E--AI
Cu-Zn-Al
Cu-Zn
Cu-Zn
Cu-&-Al,

Yes
Yes

Yes

co*

co
co + co,
co
co2

authors

Catalyst

Yes
No
Yes

Yes
Yes
Yes
Yes

Cu-Zn-Cr

expressions
for the methanol
formation
catalysts as found in recent literature
Kinetic

Authors
Seyfert

by several

Adsorption
of CO*?

co
co
co
co
co +co,
co + co,
co

Natta (1955)
Bakemeier et al. (1970)
Leonov et al. (1973)
Schermuly and Luft (1977)
Denise and Sneeden (1982)
Klier et al. (1982)
Monnier et al. (1984)
Chinchen et al. (1984)
Villa et al. (1985)
Liu eC al. (1985)
Seyfert and Luft (1985)
Dybkjaer (1985)

2. Kinetic

synthesis

Carbon source
for methanol

Authors

Table

et al.

on

Cu-containing

rate expression
rCH,oH =

&of

and Luft (1985)

i% -SCH,OHI~;,

(~,+~,f,,+~3~2+~4fCHJOH+~5fCOfH~+~6fC0~~2
(p=80-140
Villa et al. (1985)

Dybkiaer

T=235-265C)

.&of ;, --fcn,o~IK;,

(Al+Azfco+A,f,oz+A,~~fH,)3
(p=3&95

Klier et al. (1982)

bar,

bar,

r=215-245C)

~,~:(p,mJPccd3

A,A:(PcoPH~

-PPCHK,HIK~,)

(1985)

anism is more likely than a single-site mechanism. He


also concluded that no distinction can be made between molecular H, adsorption and dissociative H,
adsorption.
Liu et al. (1984) reported an inhibiting
effect of water on the methanol production.
The
results of Dybkjaer (1985) are in agreement with this
observation.
From the authors mentioned in Table 2 only Villa et
al. and Dybkjaer presented kinetic rate expressions for
the water-gas-shift
reaction. These rate expressions
are listed in Table 3.
From the literature survey presented here it follows
that still no agreement exists in the open literature on
the kinetics of methanol synthesis. We are of the
opinion that this lack of agreement is mainly caused
by the complicating
effects of the simultaneously
proceeding
reactions. Due to the presence of the
water-gas-shift reaction it is in no way a simple matter
to conclude
unambiguously
whether methanol
is
formed from CO, CO,
or both. This paper will
quantify the relative importance of CO and CO, in the

synthesis of methanol. Additionally,


new kinetic rate
expressions are presented for the reactions involved in
methanol syntheis. Finally, these new rate expressions
will be compared with those listed in Tables 2 and 3.

THEORY

Reaction

schemes

and kinetic

rate expressions

Without knowing whether methanol is formed from


CO, CO, or both, the safest way of writing down a
reaction scheme is to include both routes. Because the
Cu-Zn-Al
catalysts are known to catalyse the watergas-shift reaction as well, this reaction should be
modelled too.
Therefore the following three reactions are the basis
for the derivation of the kinetic rate expressions:
(A)

CON-+ 2H, = CH,OH

(1)

(B)

CO,

+ H2 = CO +H,O

(2)

(C)

CO,

+ 3H, = CH,OH

+ H,O.

(3)

Kinetics
Table

of low-pressure

methanol

3187

synthesis

3. Kinetic rate expressions for the water-gas-shift


reaction
Cu-Zn-Al
catalysts as found in recent literature
Kinetic

Authors
fco&,

Villa et al. (1985)

on

rate expression
r&410=

-fHzofcoIG
A5

(p=3&95

bar, T=215-245C)

(C4)
(C5)
(C6)

From

the results of Dybkjaer


(1985),
Herman
et al.
and Matulewicz
(1984) all reactions are assumed to be based on a dual-site
LangmuirHinshelwood
mechanism. On site 1 CO and CO,
adsorb competitively,
while on site 2 H, and H,O
adsorb competitively. The adsorption of methanol is
assumed to be negligible. H, is believed to adsorb
dissociatively. However, it is quite straightforward to
derive alternative kinetic rate expressions that are
based on molecular adsorption
of H,. It is now
possible to write down the elementary
reactions
necessary for the overall reactions (AHC).
Adsorption equilibria:

(1979)

co

+ sl = COsl

(4)

co,

+ sl = co+1

(5)

H, + 2~2 = 2Hs2
H,O

c,1,,0t = c,, + ccor1 + cco2s1


cs2.,01

vA2
CHoH.A2

&2

cHs2

(=

Reaction
(Al)
(A2)
(A3)
(A4)

&OK,,

(1 +&of,,

CfCOfH,

f Kc,,_&o,)(l

-.&+3H/(fHz

+ J$WJ2

K;,

(Bl)
(B2)

COsl

+Hs2=HCOsl
+s2
+ Hs2 = H,COsl
+ s2
H,COsl
+ Hs2 = H,COsl
+s2
H,COsl
+Hs2=CH,OH
+sl +s2.

(Cl)
(C2)
(C3)

+s2
+ H,Os2.

(8)
(9)
(10)
(11)

U2)
(13)

(C):

CO,sl +Hs2=HCO,sl+s2
HCO,sl + Hs2 = H,CO,sl
H2C02sl +Hs2 =H,CO,sl

+ s2
+s2

[fco,fH,

-fCHIOHf20/tfH*KDpl)]

(22)
(23)

(24)

c2 1

(B):

CO&
+Hs2 =HCO,sl
HCO,sl
+ Hs2 = COsl

Reaction

k&JkozKHz

11

+ KHIoS*O)

(l+K,ofco+Kco,~~~)(l+K~~ff,j~+K,,ofH,o)

(A)

HCOsl

Reaction

4h0,

(21)

CHzOa2.

G, Kcoz KHz(fco+fHz -fH,ofcolKk


J
H0B2=(1+K~Of~o+.~~2fC0~)(~+K~~fj;l:/2+~~~ofH~~)
&,O, cz =

(20)

Kinetic rate expressions can be obtained by choosing


rate-controlling
steps for each overall
reaction
[(AHC)]
and assuming that all the other elementary
reactions are at equilibrium. For instance, if reactions
(A2), (B2) and (C2) are chosen to be the rate-controlling steps, the following kinetic rate expressions
are obtained.

(7)

(17)
(18)
(19)

Although these schemes contain some reactions with


equal stoichiometry
[e.g. eqs (10) and (18)], these
reactions are regarded as being different. Assuming
the total number of sites 1 and 2 is constant per weight
of catalyst and neglecting terms originating
from
intermediate products, the following equations are
obtained:

(6)

+s2 = H,Os2.

H,C02sl
+ Hs2 = H,COsl + H,Os2
H,COs1+Hs2=H,COsl
+s2
H,COsl +Hs2 =CH,OH
+ sl +s2.

(14)
(15)
(16)

Since all the elementary reactions involve sites 1 and 2,


the denominators of all the resulting rate expressions
are identical. The kinetic constants kiZr k&, and k&
are in fact compounded. For example kX2 is calculated
as follows:
k
k:,,AzKA,%.tots.
A2 -

(25)

In eq. (25) k;,. A2 is the surface reaction rate constant


based on the elementary reaction (A2), K,,
is the
equilibrium constant of the elementary reaction (Al),
c,r, (,,r is the total number of sites 1 per weight of
catalyst, and s is the number of neighbouring sites 1
and 2, which is of relevance because reaction can only
occur between species adsorbed at adjacent sites 1 and
2 (Froment and Bischoff, 1979).

G. H. GRAAF et al.

3188

Based on the reaction schemes given above and


assuming that adsorption or desorption steps are not
rate-controlling there are 48 possible combinations of
kinetic rate expressions. Such a combination is called
a kinetic model. The only differences between the
kinetic rate expressions are the driving-force groups.
These driving-force groups are given in Table 4.

Parameter
estimation
and model discrimination
Each kinetic model given in the previous section
contains seven kinetic constants, which have to be
estimated from experimental
results. These experimental results are sets of the following data:
I
rcHaOn

I
9 rHrO)

P? Yco 1 Yco2 > YH2. YCH+3H,

YHzO.

From the temperature, the total pressure and the mole


fraction fugacities of each component are calculated
by the Soave-Redlich-Kwong
equation
of state
(Soave, 1972). For a chosen kinetic model the reaction
rates for methanol and water can be calculated using
these fugacities in combination
with the estimated
values of the parameters
(kinetic constants). The
equilibrium constants Kg, and K& in the kinetic rate
expressions are taken from Graaf et al. (1986). Because
reaction (C) is the stoichiometric sum of reactions (A)
and (B), K& can be written as follows:
K;,

= K;, ICOp=.

OF

(Go - Go 1;1
FCH,OH

SARR

(27)

-&OH

=
r&H,OH

PHz0

+WF

-kO

Go

Rate-controlling
step

Corresponding driving-force
group

(Al)
(A21
643)
6441
031)
(B2)
(Cl)
(C2)
(C3)
(C4)
(C5)
tC6)

Once a set of optimal constants was found for a


given model, the variances of this model for the
methanol production rate and the water production
rate were calculated:

p=j=1

&H~OH

r&OH

13

(29)

N-m

(25)

For the parameter


estimation
a direct-search
algorithm was developed in which the parameters are
adjusted towards optimum values. The adjustment
steps were taken as fractions of the parameter values
ranging from 0.5 to 0.01. The objective functions were
chosen as follows:

+ WF

Table 4. Driving-force groups of kinetic rate expressions


for reactions (A), (B) and (C)

I 1.

(28)

lj/

In these equations WF is a weighting factor; for


WF = 0 only the methanol production rate is considered in fitting the parameters, for WF = 1 both
methanol and water production
rates with equal
weights are considered, and for WF = cc only the
water production rate is considered.
was used because the statistical methods
OF,,,
applied in this paper are based on variances and thus
on sums of squares of residuals. However, a wellknown disadvantage of the sum-of-squares regression
is that large reaction rates and large residuals have the
greatest contribution
in the fitting procedure. This
problem vanishes using OF,,,,
in which the sums of
absolute values of the relative residuals are minimized.
For the final results of the chosen kinetic model
OF SARRwas used.

(30)
The values of these variances are due to experimental
inaccuracies and to a lack of fit of the kinetic model
used.
The variances of all models were tested for their
equality with Bartletts X2-test (Bartlett, 1937). As has
been pointed out by Dumez et al. (1977) this test is not
a true adequacy test: models that are retained may not
be adequate but simply the best of a series of inadequate models. For this reason a model that passed
the X2-test was subsequently
tested by two other
methods:
(a) Physico-chemical
constraints.
(b) Residual analysis.
(a) The estimates of the kinetic parameters must
have physico-chemical
meanings. This led to certain
rules for the estimates of kinetic parameters which are
summarized
below (Boudart,
1972; Vannice et al.,
1979; Kapteyn,
1980).
Reaction rate constants:
k = Aexp[-EE,/(RT)J

(31)

rule 1: k>O

(32)

ruie 2: E, > 0.
Adsorption

equilibrium

K = exp (A%:,,IR)
rule 3: K > 0

(33)
constants:
exp C- AHLJ(R

(34)
(35)

Kinetics of low-pressure methanol synthesis


rule 4: - AH&

> 0

rule 5: 0 < - AS&,, -z S&,.

(36)
(37)

(b) The residuals &oH


- r,&,H
and i;120-rnao
should be normally distributed with zero mean. Also,
the residuals should have no trend effects as a function
of any of the independent variablesf,,
f& ,fH2, fCHsOH

and hilo.
EXPERIMENTAL
Equipment

The kinetic study was carried out with a spinning


basket reactor as described by Tjabl et al. (1966). A
commercial
catalyst was used (Haldor Topsoe Mk
101). Properties
of this catalyst were reported by
Dybkjaer
(1981). A simplified flow scheme of the
equipment
is given in Fig. 1. The reactant feeds
(prefabricated
mixtures of CO, CO, and Hz) were
drawn from gas cylinders (1). The pressure in the
reactor (3) was adjusted with a pressure reducer (2).
The spinning basket reactor was heated electrically
and thermostatted
by a proportional
thermal controller (4). A small part of the product stream was led
to an on-line GLC (1 I). The flow rate of this part was
measured with a soap bubble meter (10) and regulated
with two needle valves placed in series (9). The remaining part of the product stream was passed through a
condenser of -40C (7) and a gas-liquid separator (6).
The methanol and the water formed in the reactor
were condensed almost completely and stored in a
cooled vessel (8). The gas flow through the reactor was
adjusted with a needle valve (13) which was placed
between a pressure reducer (12) and a back pressure
regulator (14). The gas flow was measured with a wet
gas meter (15). The product lines were heated electrically where necessary in order to avoid unwanted
condensation of methanol and water (see Fig. 1). The

3189

reactant feed could be sampled for analysis through a


reactor bypass (16) using a needle valve (17).
Analysis

A schematic drawing of the GLC apparatus is given


in Fig. 2. Gas samples of 1 ml were injected. The
column temperature as well as the sampling valve
temperature were maintained at 100C. Helium was
used as a carrier gas. The column (2 mm i.d., 6 m long,
packed with Porapak Q) was connected to a thermal
conductivity detector and a flame ionization detector
placed in series. Calibrations
of the detectors were
carried out each day in order to assure accurate
analysis. Hydrogen was not determined directly in the
analysis, but from the material balance:
Y Hz --

1-_Cy,

GZH,).

(38)

Measurements

The kinetic experiments were always carried out


under steady-state conditions.
External mass- and
heat-transfer limitations were negligible during the
experimental conditions chosen. This was both calculated and experimentally
verified. At temperatures
above 245C intra-particle diffusion limitations were
observed. Therefore, these experimental
results will
not be dealt with in this paper. A subsequent paper on
the subject of the intra-particle diffusion limitations in
methanol synthesis will be presented in the near
future.
For each experiment the material balances over the
reactor for hydrogen, carbon and oxygen were calculated. The deviations in these material balances were
always very small, usually less than 5%.
A broad range of experimental
conditions
was
examined in order to gain a good insight into the

Fig. 1. Flow scheme of the equipment used for the kinetic experiments:
1 =gas-cylinder,
2=pressure
reducer, 3 = spinning basket reactor, 4 = thermostat,
5 = manometer,
6 = gas-liquid
separator,
7 = cooler,
8 = storage vessel, 9 = needle valves, 10 = soap bubble meter, 11= GLC, 12 = pressure reducer, 13 = needle
valve, 14 = back pressure regulator, 15 = wet gas meter, 16 = bypass, 17 = needle valve.

3190

G. H. GRAAF et al.

fC0

fHzO

K P; fco,fcl,

apparatus
used:
I= sampling
valve,
Pig.
2. GLC
2 = Porapak Q column, 3 = TCD, 4 = FID, 5 = integrator,
6 = recorder.

kinetics. These conditions are briefly summarized in


Table 5. It was assumed that the spinning basket
reactor behaved as a perfect mixer. Justification of this
assumption is given by Tjabl et al. (1966).
Reaction rates for water and methanol were calculated from simple mixed-flow material balances over
the reactor:

4 P
,
PCH,OH= YCH,OHWR

RESULTS

in methanol

20

10

3.0
1030/W
md

L.0

50

synthesis

The amount of water formed in the methanol


synthesis as a function of the gas flow rate shows some
peculiar features (see Fig. 3). Here, the quantity put at
the vertical axis is a dimensionless measure for the
amount of water related to the water-gas-shift equilibrium. If the water-gas-shift reaction is at equilibrium,
its numerical value will be one. Under certain conditions more water is formed than is predicted thermodynamically. We see only one possible explanation for
this phenomenon:
in addition to the water-gas-shift
reaction a second water-yielding
reaction proceeds.
Since no by-products
in detectable amounts were
formed in these experiments, the surplus of water must

kg

Fig. 3. Water
formation
in methanol
synthesis: (0)
p = 50 bar, (0) p = 30 bar; (a) p = 15 bar. Symbols = results
of feed 7 (see Table 5). Lines = calculated with model A3B2C3
after correcting for the difference in activity of the catalyst
used in feed 7 (with respect to methanol).

result from the direct synthesis of methanol from CO,,


which indeed yields water.
Still another interesting feature can be detected
from Fig. 3. In some experiments (marked with an
arrow in Fig. 3) the water content is about the same as
predicted from the chemical equilibrium of the watergas-shift reaction. In this situation no driving force is
left for this reaction. Furthermore, it should be emphasized that the water-gas-shift reaction is not a fast
reaction compared with the methanol formation reactions. Otherwise, the water content would be close to
equilibrium under all conditions. For these reasons,
the contribution of the water-gas-shift reaction to the
amount of water formed will be negligible for the
experiments marked with an arrow in Fig. 3: all the
water formed will be the result of the methanol
formation
from CO,.
Since hydrogenation
of CO
yields only methanol, we can now calculate the
amounts of methanol formed from CO and CO,,
respectively. The results of these calculations are listed
in Table 6. They prove unambiguously that methanol
is produced from both CO and CO,. It also follows
that none of the two independent synthesis routes is
relatively negligible.

Table 5. Experimental conditions in the present study (catalyst Cu-Zn-Al)


Feed composition
Feed
1
2
3
4
5
6
7

sc

(39)

In eqs (39) and (40), p and T correspond


to the
conditions at which 4, is measured, being atmospheric
pressure and room temperature.

Water formation

00
0

Yco

YCO,

YH,

0.065
0.053
0.220
0.120
0.179
0
0.092

0.26 1
0.047
0.155
0.02 1
0.067
0.115
0.105

0.674
0.900
0.625
0.859
0.754
0.885
0.803

P
(bar)
15,
15,
15,
15,
15,
15,
15,

30,
30,
30,
30,
30,
30,
30,

1039%lW
(m3s-lkg-)

(&
50
50
50
50
50
50
50

483.5,
483.5,
483.5,
483.5,
483.5,

499.3,
499.3,
499.3,
499.3,
499.3,
483.5
499.3

516.7
516.7
516.7
516.7
516.7

1-6
16
16
l-6
l-6
0.3-7
0.14

Kinetics
Table 6. Relative

of low-pressure

amounts of methanol

6.9
8.2
7.5

YCHlOH

7.54
7.54
7.54

of Table

iments

5 were

showed

temperature
collected.

used,

a constant

the data
The

parameter

because

replicated

catalyst

activity.

of about

experAt

each

30 experiments

estimation

was carried

were
out at

temperature
for all 48 kinetic
models
given in
Table 4. However,
the results of these calculations
were very dependent on the initial guess values of the
parameters. A careful1 analysis of this phenomenon
showed that ill-guessed initial parameter values led to
solutions in which one of reactions (A) and (C) [eqs (1)
and (3)] was completely
neglected. As was shown
above, this is essentially wrong. For this reason the
data were screened for experiments
in which the
water-gas-shift reaction was approximately at equilibrium (within 10%). As explained above the amounts of
methanol produced from CO and from CO, were
calculated from these experiments. The ratio of the
kinetic factors could be calculated from these results
for all kinetic models. For instance, the kinetic models
A2BlC2
and A2B2C2 yield the following equation:
each

t(=

k&z Kc,,

GL,

kA2 &OK,,

y,,oDF,,
= (YCH,OH - y,,o)DFc,

YH.0

0.0109
0.0122
0.0110

Parameter
estimation
and model discrimination
In a first series of computations the results of feeds
l-5

synthesis

formed from CO and from CO,


(see Table 5)

lOSKi

l~3LXJL,ol(&0*fH,)

methanol

(41)

The parameter estimation was carried out again, with


while a was not involved
k& Kcoz K,, = akaz K,,K,,
in the fitting procedure but calculated from the experiments for which the water-gas-shift
reaction was
approximately
at equilibrium. WF was chosen to be
0.5. It should be noted that the fitting results were
almost independent of values of the WF ranging from
0.1 to 2. This revised approach gave considerably
better results: based on the X2-test at a 95% confidence
level six models were retained from the original 48
models. For these six models the parameter estimation
was repeated
for all three temperatures
simultaneously. Here it was assumed that all parameters
follow an Arrhenius temperature dependency. Initial
guess values of the parameters were based on the
results of the parameter estimation at each temperature. The data consisted of the results of 89 experiments. Now, a was no longer excluded from the fitting
procedure. Based on the X2-test three models were
retained at a 95% confidence level. These models are
given in Table 7.
In order to discriminate between these three rival
models, the results of the experiments with feed 6 (see
Table 5) were used. These results were not used for the
parameter estimation, because the catalyst activity
was different during these experiments
(compared

3191
for experiments

% CH,OH
from CO

0.0061
0.0074
0.0067

Table

Kinetic

7. Kinetic

model

A3BlC2
A3BlC3
A3B2C3

with feed 7

% CH,OH
from CO,

44
39
39

56
61
61

models

that passed the X*-test

P&,0,)
(~1

(r&J)+
(%)

7.9
6.4
6.4

28.7
26.8
24.2

These deviations are defined


function, OFsARR [eq. (28)].

by the objective

Table 8. Relative catalyst activities with respect to methanol


and water

Kinetic model
A3BlC2
A3BlC3
A3B2C3

Activity for
methanol
1.45 + 0.27
1.34+0.05
1.36 +0.04

Activity for
water
1.75 +0.30
1.38 kO.07
1.35kO.05

with the catalyst activity during experiments with


feeds l-5). For each experiment of feed 6 (18 experiments) and each kinetic model of Table 7 the relative
methanol activity and water activity were calculated.
The activity of methanol or water is defined as being
the ratio of the observed rate of formation and the
calculated rate of formation using one of the kinetic
models in combination
with the estimated values of
the parameters. The results of these calculations are
listed in Table 8.
For the correct kinetic model equal catalyst activities might be expected for both methanol and
water. Based on the results listed in Table 8 in
combination with those listed in Table 7 we conclude
that the best kinetic model is A3B2C3.
In Fig. 4 rates of methanol and water production as
predicted by model A3B2C3 are compared with the
experimental results of feed 6. As can be seen there is a
good agreement between the model calculations and
the experimental data. The same agreement can be
seen in Fig. 3: the solid lines were calculated
with
model A3BZC3 in combination
with a correction for
catalyst activity (with regard to methanol).
A thorough residual analysis on model A3B2C3,
which is not given here, showed that trending effects of
the residuals as a function of any of the independent
variables were absent. The residuals were also normally distributed with zero mean.
It turns out that the kinetic model can be simplified,
because the number of free sites 2 is negligible, which

G. H. GRAAF et al.

3192

kr,. c3 = (4.36 + 0.25) x 10


- 65,200 + 200

x exp

RT

Kco=(7.99f

IO3r
mol

1.28) x lo-
58,100 f 600

x exp

Skg-'

RT

K,,=(1.02+0.16)x
x exp

OO
V

301

2.0

10
8

L-0
I

5.0

x exp

means that
(42)

can now be expressed kinetically

CH30H
= A3
(1 + Kcofco
k6,

B2 Km

&o,_G,

-_A,,o&o/K;z

=
(1 +KcoJzo

J
CHsoH*C3= (1
c3

&oJLod

k;S.ca &o,Cfco,f%2

CfX2

+(K.r,olk~~2)f.,01

- fCu,o&,ol(fH3j2K;~)l

Kc0 fco + K cozfcoz)

IX:

A3

(50)

>

lo--
104,500+

1100

RT

>

(51)

N-m

FIm,N-m.0.99,.

(43)

+W,,JKAI:)f,,ol

(44)
(45)

The reaction rate constants are marked with the


subscript ps (pseudo), because they now contain the
adsorption
equilibrium
constant of hydrogen. The
parameter estimation was carried out again for this
simplified form of model A3B2C3. It should be noted
that the model predictions as presented in Figs 3 and 4
did not change noticeably
after the simplification
mentioned above. The following results were obtained:
&,,

RT

(52)

as

kbs,,, Kc, C&offi~,-.Lx,odfA:


K;, )I
+ Kco&oz) Cfh;2+(K~,olk~~~K,,ol

?J

( = &,o.

SSR to.991= SSR,r + SSR,i,

A3B2C3

&zO.BZ

67,400 f 600

The Arrhenius diagrams are given in Figs 5 and 6. The


results of the parameter estimation per temperature
are also plotted in Figs 5 and 6. The differences
between these results and the results obtained from the
parameter estimation for all temperatures are justified
by the confidence intervals.
The confidence intervals in eqs (46) and (47) were
calculated from

Fig. 4. Reaction rates for methanol and water: (0) and (a),
p=50 bar, (Cl) and (m) p=30 bar, (0) and (A) p=15 bar.
Open symbols = reaction rates for methanol, results of feed 6
(see Table 5). Closed symbols= reaction rates for water,
results of feed 6 (see Table 5). Lines = calculated with model
A3B2C3 after correcting for the difference in activity of the
catalyst used in feed 6 (with respect to methanol).

Model

lo-

K,,o/K~~=(4.13f1.s1)x

Kg2fH:12 + KH2&*0.

(49)

>

6.0

103~,IW
n?< kQ

1Q

(48)

>

(2.69kO.14)
x exp

x exp

x IO

- 109,900 f 200
RT

(46)

>

- 123,400 + 1600
RT

>

(471

In this equation Ftm.N-,,,0.991 is Fishers F-value with


[m. N-m]
degrees of freedom at a 99% significance
level (Fisher, 1958). The confidence intervals were
obtained by varying one parameter at a time and
holding all the other parameters at their optimal
values.
The results of the parameter estimation were used to
check whether the kinetic model follows the physicochemical constraints. It can be seen from eqs (46H51)
that rules 14 [eqs (32)-(36)]
are obeyed.
From the pre-exponential factors of the adsorption
constants the entropies of adsorption for CO and CO,
were calculated from eq. (34). Together
with the
boundary
values from eq. (37) these adsorption
entropies are listed in Table 9.
Clearly, the adsorption entropies have reasonable
values. For hydrogen and water, only the ratio of
adsorption constants was determined: this gives no

Kinetics of low-pressure methanol synthesis

3193

2LO

230

oc

220
I

210
I

-2

i04 k
- mot
-0.8
- 0.7

0.8

0.7
0.6
0.5
-

E? k$bar

-112

- 0.6
- 0.5

Fig. 5. Reaction rate constants vs temperature: (0) k&,,,(n)


kb,.,,,
(0) VP,..,.
Symbols =regression per
temperature. Lines = regresslon with all temperatures.

From
the results of feed 6 the adsorption of
hydrogen, which was assumed to be dissociative, can
be studied to a greater extent. Because feed 6 did not
contain CO, it may be assumed that methanol
is
formed almost exclusively from CO*. This was confirmed by model calculations, which are not given
here. After rearrangement of eq. (45) the following
equation is obtained:

$ii
k;,, ca Km2
r&O(l

Cfc,,f,:/
+&of,,

-fc,ofH,ol(f~~2K;,)1
+ KC02fC02)fH:/2

Thus on plotting the left-hand side of eq. (53) against


a straight line should be obtained. As can be
seen from Fig. 7, the results are in complete agreement
with our expectations, thus supporting the assumption that hydrogen is adsorbed dissociatively.

f*O /fix2

Fig. 6. Adsorption constants vs temperature: (0) Kc,,, (A)


Kc,,, (0) K,,oIK,, I/*.Symbols = regression per temperature.
Lines = regression with all temperatures.
Table 9. Adsorption entropies of CO and CO,

-AS,,,
Compound

co
CO,

(J mole1 K- )
116.7
133.9

S&(500 K)t
(J mol- K-l)
213.2
243.9

tTaken from Stull et al. (1969).


useful information
about the adsorption
entropies,
however. Therefore, we may conclude that the kinetic
model A3B2C3 obeys all the physico-chemical
constraints.

Comparison

with literature

Parameter estimation was also carried out with the


models taken from the literature given in Tables 2 and
3 using the experimental data of feeds 1-5. Because
Seyfert and Luft (1985) and Klier et al. (1982) have not
presented kinetic rate expressions for the water-gasshift reaction, these literature models were completed
with the kinetic rate expression for this reaction as
given by Villa et al. (1985).
The optimal parameters were determined for these
models. Using these optimal parameters the deviations for the methanol and water production rates
were calculated. These values are summarized in
Table 10.

3194

G.

H.

GRAAF

The experiments further support the assumption of


dissociative hydrogen adsorption.
catalyst applied in this
At least for the commercial
study, the kinetic model proposed here explains the
experimental
results with a significantly
improved
accuracy as compared with the kinetic models proposed by Seyfert and Luft (1985), Villa et al. (f985),
Klier et al. (1982) and Dybkjaer (1985).

5.0
L.H.5 ~1531
1

et al.

LO
3.0

thank Haldor Topsoe A/S, Lyngby,


Copenhagen, Denmark for delivering their methanol synthesis catalyst Mk 101 and the N.V. Nederlandse Gasunie,
Groningen, Netherlands, for delivering gas mixtures for

Acknowledgements-We

1
OO

0.7

0.2

03

calibration purposes.

fCl,o/fCI:2
bar

NOTATION

I2

pre-exponential
factor
kinetic
constants
in literature
expressions
concentration,
mol kg-
driving force
energy of activation, J mol-
partial fugacity, bar
experiment index
reaction rate constant
adsorption
equilibrium
constant,
bar- ; e.g. for CO:

Fig. 7. Adsorption of hydrogen and water: (0) p = 50 bar,


( 0) p = 30 bar, (A) p = 15 bar. Symbols = results of feed 6 (see
Table 5). Line = best fit based on SSR.

A I...6
c

DF
Table 10. Accuracies of the kinetic models taken from recent
literature compared with the model proposed in this study

:
j

Kinetic model from

Ki

Seyfert and Luft (1985)


Villa et al. (1985)
Klier et al. (1982)

Dybkjaer (1985)
This study

10.8
12.3
10.0
14.7
6.4

100
100

57
167
24

These deviations are defined by the objective function


OFSARR [es- (WI.

Comparing
these results with the results of mode1
A3B2C3
it is obvious
that the latter describes the
kinetics in methanol synthesis much better. This was
confirmed by the X2-test: using this criterion the four
models
from
the literature
were rejected,
thus
favouring model A3B2C3.

K co=
K Al ...
K B1 . .
K cl-

K A4

. K 82
K C6

K,
m
N

OF
Pi
p
r

CONCLUSIONS
Experimental evidence shows that methanol can be
formed simultaneously
from both CO and CO2 in
low-pressure methanol synthesis.
The experimental
results on the methanol synthesis kinetics can be explained
by a dual-site
Langmuir-Hinshelwood
mechanism, based on dissociative hydrogen adsorption and three independent
reactions: methanol formation
from CO, methanol
formation from CO, and the water-gas-shift reaction.
Depending
on which elementary reaction step is
rate-controlling
in each of these three parallel reactions, 48 different kinetic models are possible. Based
on X2-statistics and consistency tests a final model was
selected.
The kinetic parameters could be determined
as
functions of temperature between 210 and 245C. The
values of these parameters are not in conflict with the
physico-chemical
constraints.

R
S

S
S
T
W

WF
Y
:
AH
AS
$

elementary
constant

If I
COG1

__

CCOSl

reaction

EP

equilibrium

e.g.

chemical equilibrium constant based


on partial pressures
number of parameters
number of experiments
objective function
partial pressure, bar
total pressure, bar
reaction rate per weight of catalyst,
mols-kg-
gas constant (8.314), JmolmKP
number of neighbouring sites
variance
entropy, J mol - 1 K -
temperature, K
weight of catalyst, kg
weighing factor
mole fraction
ratio of kinetic constants
relative error
enthalpy change, J mol-
entropy change, Jmol- K-
gas flowrate at standard temperature
and pressure (25C, 1.013 bar), m3 s-

Superscripts
0

indicates standard pressure (1 ,013 bar)


indicates calculated value

Kinetics

of low-pressure

Subscripts

ads
Al

adsorption
. ..A4

Bl...B2

indicates
methanol

rate-controlling
from CO reaction

indicates

rate-controlling

water-gas-shift
Cl...C6

indicates

CO,
CH,OH

step

indicates
indicates

component
comnonent

CO
CO,

indicates

component

CH;OH

indicates

component

H,

indicates

component

H,O

indicates

gaseous

of

maximum

value

min

minimum

value

Ps
SARR

pseudo
based

CO,

component

COz,

H,,

or H,O

max

on

relative

sum

of absolute

values

of

residuals

sr

surface

SSR

based

Sl

site

s2

site 2
total
indicates
in KS,
indicates

methanol

G,
indicates
in Kg,

methanol

step

component

CH,OH

of the

at,equilibrium

EQ
gas
HZ
Hz0

tot
1

of

reaction

rate-controlling
from CO, reaction

methanol
co

step

reaction
on sum of squares

of residuals

from CO reaction

water-gas-shift

reaction

from CO,

in

reaction

REFERENCES

Bakemeier, H., Laurer, P. R. and Schroder, W., 1970, Development and application of a mathematical model of the
methanol synthesis. Chem. Engng Prog. Symp. Ser. 66(98),
l-10.
Bartlett, M. S., 1937, Properties of sufficiency and statistical
tests. Proc. R. Sot. A 160, 268-282.
Boudart, M., 1972, Two-step catalytic reactions. A.I.Ch.E.
J.
18, 465-478.
Chinchen, G. C., Denny, P. J., Parker, D. G., Short, G. D.,
Spencer, M. S., Waugh, K. C. and Whan, D. A., 1984, The
activity of copper-zincoxide-aluminiumoxide
methanol
synthesis catalyst. Prep. Pap. Am. them. Sot. Div. Fuel
Chem. 2!J(5), 178-188.
Denise. B. and Sneeden, R. P. A., 1982, Hydrocondensation
of c&bondioxide IV. J. Molec. Cat&
15, 359-366.
Dumez, F. J.. Hosten, L. H. and Froment, G. F., 1977, The use
of sequential discrimination
in the kinetic study of l-

methanol

synthesis

3195

butene dehydrogenation.
Ind. Engng Chem. Fundam.
16,
298-301.
Dybkjaer,
I., 1981, Topsoe
methanol
technology.
Chem.
Econ. Engng Rev. 13(6), 17-25.
Dybkjaer, I., 1985, Design of ammonia and methanol synthesis reactors. Paper presented at the NATO
conference
on chemical reactor design and technology, Canada.
Fisher, R. A., 1958, Statistical
Methodsfor
Research
Workers,
13th edition. Hafner, New York.
Froment, G. F. and Bischoff, K. B., 1979, Chemical
Reactor
Analysis
and Design, p. 98. J. Wiley, New York.
Graaf, G. H., Siitsema, P. J. J. M., Stamhuis, E. J. and Joosten,
G. g. H., 1986, Chemical equilibria in methanol synthesis.
Chem. Engng Sci. 41, 2883-2890.
Herman, R. G., Klier, K., Simmons, G. W., Finn, B. P., Bulko,
J. B. and Kobylinski,
T. P., 1979, Catalytic synthesis of
methanol from CO/H.. J. Catal. 56, 407409.
F., 1980.
The
MetatheSis
of Alkenes
over
Kapteyn,
Rheniumoxide-Aluminia,
p. 77. Dissertation. Amsterdam.
Klier, K., Chatikavanij, V., Herman, R. G. and Simmons, G.
W., 1982, Catalytic synthesis of methanol from CO/H,. J.
Catnl. 74, 343Lj60.
Leonov, V. E., Karavaev, M. M., Tsybina, E. N. and
Petrishcheva, G. S., 1973, Kinetics of methanol synthesis
on a low-temperature catalyst. Kinet. Katal.
14, 970-975.
Liu, G., Willcox, D., Garland, M. and Kung, H. H., 1984, The
rate of methanol production on a copper-zincoxide
catalyst. The dependence on the feed composition. J. Catal. 90,
139-146.
Liu, G., Willcox, D., Garland, M. and Kung, H. H., 1985, The
role of CO, in methanol synthesis on Cu-Znoxide:
an
isotope labeling study. J. Catal. 96, 251-260.
Matulewicz,
E. R. A., 1984, Kinetics
and Spectroscopic
Investigations
of Propene Metathesis
and Methanol
Synthesis.
Dissertation, Amsterdam.
Monnier, J. R., Apai, G. and Hanrakan, H. J., 1984, Effect of
CO,
on the conversion
of HZ/CO to methanol
over
copper+hromia
catalysts. J. Catol. 88, 523-525.
HydroNatta, G., 1955, Synthesis of methanol, in Catalysis:
genation
and Dehvdrogenation
(Edited by P. H. Emmett),
pp. 349411. Rheinhoid, New York.
_
Schermulv,
0. and Luft, G., 1977, Untcrsuchung
der
Niederiruck-Methanolsynthese
im Treibstrahlreaktor.
Chemie-IngrTechn. 49, 907.
Seyfert,
W. and Luft, G., 1985, Untersuchungen
zur
Methanolsynthese
im Mitteldruckbereich.
Chemie-lngrTechn. 57, 482-483.
Soave. G.. 1972. Eauilibrium
constants from a modified
RedlichLKwong
etuation
of state. Chem. Engng Sci. 27,
1197-1203.
Stull, D. R., Westrum, E. F. and Simke, G. C., 1969, The
Chemical
Thermodynamics
of Organic
Compounds,
pp.
219-220. Wiley, New York.
Tjabl, D. G., Simons, 3. B. and Carberry, J. J., 1966, Heterogeneous catalysis in a continuous stirred tank reactor. Ind.
Engng Chem. Fundam. 5, 171-175.
Vannice, M. A., Hyun, S. H., Kalpakci, B. and Liauh, W. C.,
1979, Entropies of adsorption in heterogeneous
catalytic
reactions. J. Catal. 56, 358-362.
Villa, P., Forzatti, P., Buzzi-Ferraris,
G., Garone, G. and
Pasquon, I., 1985, Synthesis of alcohols from carbonoxides
and hydrogen, Ind. Engng Chem. Process
Des. Dev. 24,
12-19.

You might also like