Applied Catalysis A: General 179 (1999) 31±49

Methanol±steam reforming on Cu/ZnO/Al2O3 catalysts.

Part 2. A comprehensive kinetic model
Brant A. Peppley*, John C. Amphlett, Lyn M. Kearns, Ronald F. Mann
Department of Chemistry and Chemical Engineering, Royal Military College of Canada, Kingston, Ont., Canada K7K 7B4

Received 21 January 1998; received in revised form 18 August 1998; accepted 18 August 1998

Abstract

Surface mechanisms for methanol±steam reforming on Cu/ZnO/Al2O3 catalysts are developed which account for all three
of the possible overall reactions: methanol and steam reacting directly to form H2 and CO2, methanol decomposition to H2 and
CO and the water-gas shift reaction. The elementary surface reactions used in developing the mechanisms were chosen based
on a review of the extensive literature concerning methanol synthesis on Cu/ZnO/Al2O3 catalysts and the more limited
literature speci®cally dealing with methanol±steam reforming. The key features of the mechanism are: (i) that hydrogen
adsorption does not compete for the active sites which the oxygen-containing species adsorb on, (ii) there are separate active
sites for the decomposition reaction distinct from the active sites for the methanol±steam reaction and the water-gas shift
reaction, (iii) the rate-determining step (RDS) for both the methanol±steam reaction and the methanol decomposition reaction
is the dehydrogenation of adsorbed methoxy groups and (iv) the RDS for the water-gas shift reaction is the formation of an
intermediate formate species. A kinetic model was developed based on an analysis of the surface mechanism. Rate data were
collected for a large range of conditions using a ®xed-bed differential reactor. Parameter estimates for the kinetic model were
obtained using multi-response least squares non-linear regression. The resultant model was able to accurately predict both the
rates of production of hydrogen, carbon dioxide and of carbon monoxide for a wide range of operating conditions including
pressures as high as 33 bar. # 1999 Elsevier Science B.V. All rights reserved.

Keywords: Kinetics; Surface mechanisms; Methanol steam reforming; Fuel cells; Copper catalysis

1. Introduction cell. However, at the operating temperature of PEM

fuel cells, the electrocatalyst at the anode is extremely
Recently, the possibility of using an on-board sensitive to CO poisoning. For this reason it is highly
methanol±steam reformer to generate hydrogen for desirable to have a kinetic model of methanol±steam
a fuel cell engine in various transportation applications reforming which is able to not only predict the rate of
has resulted in an increased interest in the study of the production of hydrogen but also the rate of production
methanol±steam reforming process. The most promis- of undesirable by-product CO.
ing type of fuel cell for this application is the low An enormous amount of research has been pub-
temperature proton exchange membrane (PEM) fuel lished on the surface processes which occur on Cu/
ZnO/Al2O3 catalysts during methanol synthesis. Var-
*Corresponding author. E-mail: [email protected] ious reviews and extensive kinetic studies have been

0926-860X/99/$ ± see front matter # 1999 Elsevier Science B.V. All rights reserved.
PII: S0926-860X(98)00299-3
32 B.A. Peppley et al. / Applied Catalysis A: General 179 (1999) 31±49

published in the last decade [1±4]. The continuation of

various debates concerning the reaction mechanism
for over three decades is an indication of the complex-
ity of the process. Less has been published concerning
the water-gas shift reaction on Cu/ZnO/Al2O3 but the
debate over whether the mechanism is associative via
an intermediate formate or regenerative via a redox Fig. 1. Rate-determining step for mechanism of methanol synth-
reaction involving a special form of copper has con- esis. From Waugh (1992) based on the work of Millar et al. [14].
tinued for a similar period of time [4±7]. By contrast,
the literature to date on methanol±steam reforming is
relatively limited and only a few studies have CO2 via a bidentate surface formate. The rate-deter-
attempted to develop kinetic models based on surface mining step (RDS) in Waugh's model is the hydro-
reaction mechanisms [8±13]. genation of this surface formate species to
This paper presents the development of a compre- formaldehyde or the adsorbed species, hydroxycar-
hensive model for the kinetics of methanol±steam bene ion. Millar et al. [14] used infrared spectroscopy
reforming on Cu/ZnO/Al2O3 catalysts. Surface to prove the existence of the intermediates involved in
mechanisms for the main reactions were developed Waugh's mechanism. Fig. 1 is a schematic drawing of
using information from the literature. Progressive the RDS for the mechanism of methanol synthesis
improvements were made to the mechanisms based from CO2 proposed by Waugh [1]. The formation of
on observations from an extensive experimental pro- methanol from CO and H2, which historically was
gram. A set of Langmuir±Hinshelwood rate expres- considered to be the reaction path for methanol synth-
sions were then derived based on a steady-state esis, was reported to be 100 times slower than metha-
analysis of the ®nal surface mechanisms. The kinetic nol synthesis from CO2 and H2.
and thermodynamic parameters for these rate expres- According to the concept of micro-reversibility [15]
sions were then obtained by non-linear least squares it should, theoretically, be possible to use the reverse
regression. The model is referred to as comprehensive of the methanol-synthesis mechanism for the metha-
because it accounts for both the rates of production of nol±steam reforming process. Attempts to do this,
hydrogen and of the by-product carbon monoxide. however, were unsuccessful, probably due to the large
differences between the operating conditions of the
two processes. For example, in methanol±steam
2. Previous studies of surface mechanisms on reforming the average partial pressure of methanol
Cu/ZnO/Al2O3 catalyst is much greater and the partial pressure of CO is much
less than during methanol synthesis; hence, it would
2.1. Methanol synthesis be expected that competitive adsorption of methanol
with CO and CO2 would have a less signi®cant effect
Based on extensive temperature-programmed reac- on the kinetics of methanol synthesis than on the
tion and desorption studies, as well as in situ FTIR kinetics of methanol±steam reforming. Indeed, in
surface studies, Waugh with various co-workers devel- the review by Skrzypek et al. [2] of the various kinetic
oped a surface mechanism which proposes that the models which have been proposed for methanol synth-
active site for the predominant reaction in methanol esis, only three of the 15 models included terms for the
formation from synthesis gas is solely metallic copper adsorption of methanol.
[1]. According to Waugh, the other components of the Reversible methanol-synthesis mechanisms could
catalyst do not play an active role in the reaction be applied to methanol±steam reforming by simply
mechanism but are crucial to maintaining a crystal adding terms for the adsorption of other intermediates
form which provides a large stable surface area with if the difference in adsorption effects was the only
small copper crystallites. issue. However, recent works by Kalchev et al. [5] and
The model of Waugh also proposed that methanol is Clausen et al. [16] have shown that the chemical state
synthesised almost entirely by the hydrogenation of of Cu/ZnO/Al2O3 catalysts is strongly affected by the
 B.A. Peppley et al. / Applied Catalysis A: General 179 (1999) 31±49 33

reducing potential of the reactant mixture. The much this table. Two key pieces of information which will be
higher partial pressure of CO and the lower partial used in developing the mechanism for methanol±
pressure of steam during methanol synthesis would steam reforming on Cu/ZnO/Al2O3 catalysts are:
result likely in the catalyst surface being in a different 1. CH3OH, H2O, CO2 and CO all competitively
chemical state than during methanol±steam reform- adsorb. This indicates that there is at least one type
ing. Considering these factors, it is unlikely that the of active site common to the adsorption of all
reverse of methanol-synthesis rate expressions would these species.
be able to explain the kinetics of methanol±steam 2. Hydrogen does not adsorb competitively with the
reforming. other products and reactants but appears to have a
unique mode of adsorption.
2.1.1. Adsorption of reactants and products in
methanol synthesis
Despite the fact that reversible rate expressions for 2.2. Water-gas shift reaction on Cu/ZnO/Al2O3
methanol synthesis failed to provide an adequate
model of the kinetics of methanol±steam reforming, The exact nature of the surface mechanism for the
the literature on the surface processes which occur water-gas shift reaction on Cu/ZnO/Al2O3 catalysts
during methanol synthesis provided a signi®cant has been a matter of debate for many decades. The
amount of information for developing surface central issue is whether the reaction proceeds via an
mechanisms for the methanol±steam reforming pro- associative mechanism or a regenerative mechanism.
cess. Skrzypek et al. [2] have reviewed a large number In the associative mechanism, H2O and CO react to
of publications concerning the adsorption of various form an adsorbed intermediate surface formate which
reactants, products and possible intermediates in then decomposes to form H2 and CO2. In the regen-
methanol synthesis on Cu/ZnO/Al2O3 catalysts. erative mechanism, also known as the redox mechan-
Table 1 provides a brief summary of their review. A ism, copper oxide reacts with CO to form CO2 and
great deal of information about the surface processes copper metal. Water then dissociates to produce H2
occurring on Cu/ZnO/Al2O3 catalysts is contained in and a surface oxygen which re-oxidises the copper.

Table 1
Summary of review by Skrzypek et al. [2] on adsorption of reactants during methanol synthesis on Cu/ZnO/Al2O3 catalysts

Species Type
H (kJ molÿ1) Competitive adsorbates Remarks

CO Non-dissociative 50±70 All other species except H2 Adsorbs perpendicular to surface

with C towards surface
CO2 Dissociative 80±128 All other species except H2 Surface oxygen enhances adsorption
by formation of carbonates
60 (as COÿ
2)
H2 Dissociative 40±50 ± Forms bonds directly with
metal atoms
H2 O Non-dissociative and 20 (ˆ0.7)±67 (ˆ0) All other species except H2 Heat of adsorption is a linear function
dissociative of surface coverage, 
CH3OH Dissociative 50±80 All other species except H2 Hydrogen adsorption is stabilised by
adsorption on ZnO to form OH(a)
HCOOÿ ± ± All other species except H2 Forms on all components of the
catalyst either by the reaction of
H2 and CO2 or by OHÿ and CO
CH3Oÿ ± ± All other species except H2 Formed by dissociative adsorption
of CH3OH
HCOÿ ± ± All other species except H2 Formed by reaction of CO with
the hydrided metal
34 B.A. Peppley et al. / Applied Catalysis A: General 179 (1999) 31±49

A number of investigators, such as DuÈmpelmann proposed an associative mechanism for the water-
[12], Waugh [1], and Vanden Bussche and Froment gas shift reaction which proceeds through an inter-
[4], claim that the matter has been settled. Rhodes et mediate formate. DuÈmpelmann found that this
al. [6], however, in an extensive review of the litera- mechanism agreed with his observations of the
ture, have shown that there are a number of issues kinetics of the water-gas shift mechanism under typi-
which still have not been resolved. One of the most cal operating conditions for methanol±steam reform-
confounding factors is that the crystal structure of the ing. Because of the issues concerning the structure of
catalyst varies for the range of experimental condi- the catalyst under the operating conditions, discussed
tions which have been used for the kinetic studies above, it was decided to use this mechanism as the
reported in the literature. Rhodes et al. concluded that basis for the mechanism of the water-gas shift reaction
both regenerative and associative mechanisms may be mechanism developed in this work with only minor
occurring at comparable rates for some catalyst con- variations which will be described below.
ditions, while certain catalyst-conditioning treatments
may cause one or the other of the mechanisms to 2.3. Methanol±steam reforming
dominate.
Campbell and Ernst [7] studied the steady-state Until recently, the literature on the kinetics and
kinetics of the forward and reverse water-gas shift mechanism of methanol±steam reforming on Cu/
reactions on Cu(1 1 1) and the Cu(1 1 0) single-crystal ZnO/Al2O3 catalysts was quite limited. Santacesaria
surfaces at medium pressure (1±2000 Torr). They and CarraÁ [8] published a paper which used an empirical
reported evidence of a hydrogen-induced surface approach to develop an expression for the rate of
reconstruction or phase transition which strongly disappearance of methanol. Their rate expression
affected the rate of reaction. Campbell and Ernst made was of the form of a Langmuir±Hinshelwood expres-
the interesting suggestion that the mechanism may be sion but it was not derived from an explicit mechan-
regenerative but that there is a hydrogen-assisted ism. Amphlett et al. [9±11] reported studies of both the
mode of adsorption for CO2. Hence, CO2 adsorption thermodynamics and the kinetics of methanol±steam
may involve a H±CO2 bond that would be indistin- reforming on Cu/ZnO/Al2O3 catalysts. Again, how-
guishable from the formate intermediate which is the ever, an explicit surface mechanism was not reported,
basis of the associative mechanism. although site blocking by CO adsorption was dis-
The idea of a hydrogen-induced surface reconstruc- cussed.
tion of the surface assisting with the adsorption of CO2 DuÈmpelmann [12] developed a mechanism for
introduces the question of the crystal structure of the methanol±steam reforming which included the direct
catalyst and its importance to the mechanism of the formation of CO2 by the reaction of methanol and
water-gas shift reaction. Kalchev et al. [5] studied the steam as well as the water-gas shift reaction. He
relationship between the structure of Cu/ZnO/Al2O3 assumed that the rate of the decomposition reaction
catalysts and their activity for the water-gas shift was negligible. Using the method of analysis proposed
reaction. They used a wide array of physical char- by Boudart and DjeÂga-Mariadassou [17], DuÈmpel-
acterisation methods to show that, for Raney copper mann concluded that the RDS for methanol±steam
catalysts, the activity of the catalyst for the water-gas reforming is the reaction of adsorbed formaldehyde
shift reaction appears to depend on the formation of a via either of the following surface reactions:
hydroxy-carbonate structure on the copper crystal kHCOOH…1†
surface. This further reinforces the idea that the dis- CH2 O…1† ‡ O…1† @ HCOOH…1† ‡ S1 (1)
kÿHCOOH…1†
tinction between an associative versus a regenerative
mechanism for the water-gas shift reaction may be an kCHO…1†
arbitrary choice depending on whether the hydroxy- CH2 O…1† ‡ S1 @ CHO…1† ‡ H…1† (2)
kÿCHO…1†
carbonate crystal structure of the catalyst surface is
considered as part of the mechanism. For the WGS reaction the RDS was reported to be
DuÈmpelmann [12], in his investigation of metha- the surface reaction between adsorbed CO and surface
nol±steam reforming on a Cu/ZnO/Al2O3 catalyst, hydroxyls to produce a formate species:
 B.A. Peppley et al. / Applied Catalysis A: General 179 (1999) 31±49 35

kHCOO…1† 3. Development of a comprehensive mechanism

HO…1† ‡ CO…1† @ HCOO…1† ‡ S1 (3)
kÿHCOO…1† for the reactions of methanol and steam on Cu/
ZnO/Al2O3 catalyst
DuÈmpelmann [12] derived a rate expression for
methanol±steam reforming of the form The objective of developing a comprehensive sur-
  face mechanism for methanol±steam reforming is to
KCH3 OH pCH3 OH
rR ˆ kR 2 ; (4) be able to account for the variation in the CO content
KH2 pH 2
of the product gas as well as the rate of production of
where , the adsorption term, is the fraction of active hydrogen. In a previous paper we have shown that it is
sites available. The order of importance of adsorption, necessary to include reaction paths which can account
which presumably is associated with the degree of for the following three overall reactions [18]:
surface coverage during reaction, was reported as
k‡R
follows: CO2>H2OCH3OHCO>0. CH3 OH ‡ H2 O @ CO2 ‡ 3H2 (I)
kÿR
Jiang et al. [13] explicitly de®ned a set of elemen-
tary surface reactions which could be analysed to k‡W
CO ‡ H2 O @ CO2 ‡ H2 (II)
obtain a true mechanistic Langmuir±Hinshelwood rate kÿW
expression. Jiang et al. carried out a kinetic analysis of k‡D
the reaction mechanism followed by a regression CH3 OH @ CO ‡ 2H2 (III)
kÿD
analysis of kinetic measurements of the rate of steam
reforming on a Cu/ZnO/Al2O3 methanol-synthesis Only two of these reactions are linearly independent
catalyst. They obtained the following rate expression: and any one of the reactions can be expressed as the
 .
1=2
 .
1=2
 sum or difference of the other two. Because of this it is
kCH3 O KCH3 OH KH2 pCH3 OH pH2 ÿ T
2 not possible to measure the rate of each reaction
rR ˆ   .  .  2 CS ;
1=2
1 ‡ KCH3 OH KH2
1=2 1=2 1=2
pCH3 OH pH2 ‡ KH2 pH2 separately, except in the special instance where one
reaction is at equilibrium. The starting point for the
(5)
mechanisms which were ®nally developed, therefore,
where only the adsorption of methanol and the adsorp- relied heavily on fundamental information from sur-
tion of hydrogen were found to have statistically face studies and adsorption studies found in the lit-
signi®cant effects on the rate of reaction. The kinetic erature. The reaction rate expressions derived from the
expression developed by Jiang et al. [13] predicts the surface mechanisms could only be evaluated against
rate of disappearance of methanol and the rate of the experimental results as sums or differences of
formation of CO2. They claim that the process is rates. Nevertheless, experimental data provided a
100% selective for CO2 and that the rate of the number of signi®cant insights into the form of the
WGS reaction is negligible. Although for industrial mechanism and the importance of speci®c surface
processes the rate of CO production could be con- processes. Details of the experimental data and the
sidered negligible, for low temperature fuel-cell appli- analysis have been presented in a previous publication
cations, where very low levels of CO contamination [18].
can severely poison the anode electrocatalyst, the
decomposition reaction and the WGS reaction must 3.1. Nature of the active sites
be taken into account [18].
Despite this simpli®cation used by Jiang et al. [13], In developing surface mechanisms for Reactions
the surface mechanism which they propose provides (I)±(III), the basic assumption was made that there are
the best explanation for the observed kinetic behaviour two distinct types of active sites on the surface of the
and also accounts for the high rate of methyl formate catalyst [18]. One type of site is assumed to be active
production which occurs when the S/M ratio is low. for methanol±steam reforming and the water-gas shift
Their surface mechanism provided the starting point reaction. It is believed that this type of site is created
for the further development of the more comprehen- by the partial decomposition of a hydroxy-carbonate
sive kinetic model presented in this work. phase that supports the hydroxylation of the inter-
36 B.A. Peppley et al. / Applied Catalysis A: General 179 (1999) 31±49

mediate surface formates. The second type of site data in this way [19]. On the other hand, information
(which may simply be the face-centred cubic form from adsorption studies and other methods of surface
of elemental copper) was assumed to primarily sup- analysis can be used to develop a mechanism and
port the decomposition reaction. Further work con- postulate a most likely RDS directly. The substantial
cerning the relationship between the crystal phase amount of information from methanol-synthesis
structure of the catalyst and its activity and selectivity research concerning surface processes occurring on
is on-going. supported copper catalysts provides an insight into the
Based on these assumptions concerning the active surface mechanism for methanol±steam reforming to
sites on the catalyst surface, a mechanism is proposed an extent that is rare in studies of heterogeneous
for methanol decomposition that is independent of the catalysis. In particular, a number of in situ FTIR
mechanism for the other two reactions. The elementary studies of the surface of copper catalysts in the pre-
surface processes involved in the mechanisms for the sence of various reactive atmospheres have shown
methanol±steamreactionandtheWGSreactionoccuron which intermediates are most readily formed and
Type 1 active sites while the decomposition reaction which are the most stable [1,14].
occurs on a separate and distinct Type 2 active site. The adsorption of methanol on oxidised and
reduced copper on silica catalyst has been studied
3.2. The RDS for methanol decomposition and using in situ FTIR spectroscopy. Millar et al. [20] were
methanol±steam reforming able to show that methanol readily adsorbs dissocia-
tively on copper to form adsorbed methoxy species at
The most frequently used method for determining temperatures as low as 295 K. At higher temperatures
the RDS for a given reaction mechanism is to test the methoxy groups dehydrogenate to form formalde-
various derived kinetic expressions for goodness-of-®t hyde, or its isomer oxymethylene, which subsequently
against kinetic data based on a non-linear regression of is converted to a formate [20,21]. As shown in Fig. 2,
a rate model. It has been shown, however, that there we have con®rmed these observations with diffuse
are numerous pitfalls and uncertainties associated re¯ectance in situ Fourier transform infrared spectro-
with discriminating between kinetic models using rate scopy (DRIFTS) which shows that methoxy and

Fig. 2. FTIR spectra of surface of Cu/ZnO/Al2O3 catalyst after methanol±steam adsorption and heating from 1208C to 2008C, obtained using
a Nicolet 510P with a Harrick environmental cell and a ``praying mantis'' diffuse reflectance attachment.
 B.A. Peppley et al. / Applied Catalysis A: General 179 (1999) 31±49 37

formate groups dominate the surface of the catalyst 3.4. Elementary reactions occurring on Type 1 and
when exposed to methanol and steam. Type 1a sites
Millar et al. [14] proposed a surface mechanism for
the formation of methyl formate from methoxy which The following surface mechanism for the metha-
involves the dehydrogenation of the methoxy to the nol±steam reaction and the water-gas shift reaction
adsorbed oxymethylene, which in turn rapidly reacts during the methanol±steam reforming process is pro-
with a methoxy to yield methyl formate. They further posed. The majority of the elementary reactions are
showed that, in the absence of water on pure copper, this the same as those proposed by Jiang et al. [13]. The
intermediate methyl formate decomposes to yield CO. main differences involve the assumptions concerning
Earlier work by Minachev et al. [22] showed that the nature of the active sites and the incorporation of
adsorbed formate decomposes to yield only CO2 and H2. the water-gas shift reaction
Jiang et al. [13], as part of the development of their kCH …1†
3O
dehydrogenation±hydrolysis mechanism for metha- S1 ‡ S1a ‡ CH3 OH…g† @ CH3 O…1† ‡ H…1a† (6)
nol±steam reforming, showed that the initial rate of kÿCH
3O
…1†

formation of methyl formate from methanol had an kOH…1†

activation energy equal to that observed for methanol± S1 ‡ S1a ‡ H2 O…g† @ OH…1† ‡ H…1a† (7)
kÿOH…1†
steam reforming. This observation indicates that either
the adsorption of methanol or the dehydrogenation of k
CO
…1†
2 …1†
the methoxy group is the RDS in methanol±steam S1 ‡ CO2 …g† @ CO2 (8)
k
reforming. Combining the observations of Jiang et al. ÿCO
…1†
2
[13] and Millar et al. [14] leads to the conclusion that kCO…1†
methoxy dehydrogenation is the RDS in methanol± S1 ‡ CO…g† @ CO…1† (9)
kÿCO…1†
steam reforming. In situ FTIR spectroscopy studies of
kH…1a†
methanol adsorption on oxide-supported copper also 2S1a ‡ H2 …g† @ 2H…1a† (10)
support the hypothesis that methoxy dehydrogenation kÿH…1a†
is the RDS in methanol decomposition [14]. kCH …1†
2O
CH3 O…1† ‡ S1a @ CH2 O…1† ‡ H…1a† …RDS-I†
3.3. Unique mode of adsorption of hydrogen kÿCH
2 O…1†

(11)
As pointed out in Table 1, adsorption studies have k …1†
C2 H5 O
shown that hydrogen does not compete for the same CH3 O…1† ‡ CH2 O…1† @
2
CH3 OCH2 O…1† ‡ S1
sites that adsorb the oxygen or carbon-containing k
ÿC2 H5 O
…1†
2
species. It is likely that the hydrogen adsorption sites
are interstitial and may also involve spillover onto a (12)
k
ZnO/Al2O3 phase [23]. In a review paper, Burch et al. C2 H4 O
…1†
2

[24] report that the concept of H2 spillover in Cu/ZnO/ CH3 OCH2 O…1† ‡ S1a @ CH3 OCHO…1† ‡ H…1a†
k …1†
Al2O3 catalysts during methanol synthesis has been ÿC2 H4 O
2

proposed by various investigators. This observation is (13)

also consistent with the concept of surface restructur- kHCOOH…1†
ing suggested by Campbell and Ernst [7] during the CH3 OCHO…1† ‡ OH…1† @ HCOOH…1† ‡ CH3 O…1†
kÿHCOOH…1†
water-gas shift reaction on Cu/ZnO/Al2O3 catalyst.
(14)
Active sites exclusively for hydrogen adsorption were
kHCOO…1†
therefore incorporated into the mechanism to account HCOOH…1† ‡ S1a @ H…1a† ‡ HCOO…1† (15)
for these observations. The H2 adsorbing site asso- kÿHCOO…1†
ciated with the active phase for the methanol±steam kHCOO…1† ;b
reaction and the WGS reaction is designated as a Type OH…1† ‡ CO…1† @ HCOO…1† ‡ S1 …RDS-II†
kÿHCOO…1† ;b
1a site and the H2 adsorbing site for the second active
phase is designated as a Type 2a site. (16)
38 B.A. Peppley et al. / Applied Catalysis A: General 179 (1999) 31±49

k …1† k …2†
CO ;a C2 H4 O
…1† 2
…1a† …1† …2† 2
HCOO ‡ S1a @ H ‡ CO2 (17) CH3 OCH2 O ‡ S2a @ CH3 OCHO…2† ‡ H…2a†
k …1† k …2†
ÿCO ;a ÿC2 H4 O
2 2

(23)
3.5. Elementary reactions occurring on Type 2 and k
CH2 O
…2†
2
Type 2a sites CH3 OCHO…2† @ CH3 OCHO…g† ‡ S2 (24)
k …2†
ÿCH2 O
2
As stated elsewhere, based on the observations from kCHO…2†
the earlier kinetic studies, it has been shown that the CH3 OCHO…2† ‡ S2 @ CH3 O…2† ‡ CHO…2† (25)
decomposition reaction is occurring on a different kÿCHO…2†

type of active site than the methanol±steam reaction kCO…2† ;b

and the water-gas shift reaction [18]. The majority of CHO…2† ‡ S2a @ CO…2† ‡ H…2a† (26)
kÿCO…2† ;b
the elementary reactions, however, were assumed to
be the same as those on Type 1 sites. The main Although adsorbed hydroxyls, formates and carbon
difference between the Type 1 and Type 2 sites is dioxide do not participate in the reaction mechanism,
that the latter do not support the reaction between they are known to adsorb competitively on copper
hydroxyls and methyl formate; hence, a decarbonyla- with those species involved in the mechanism; there-
tion via a formyl occurs instead of hydroxylation as in fore, adsorption of these species will be included in the
steam reforming. Furthermore, because this second site balance for Type 2 and Type 2a sites
phase is structurally different from the phase contain-
kOH…2†
ing Type 1 and Type 1a sites, it is assumed that the
S2 ‡ S2a ‡ H2 O…g† @ OH…2† ‡ H…2a† (27)
entropy and enthalpy for the adsorption of hydrogen kÿOH…2†
will be somewhat different. A Type 2a site for hydro- k …2†
gen adsorption on the second phase involving copper CO
2 …2†
S2 ‡ CO2 …g† @ CO2 (28)
is therefore required for the decomposition reaction k …2†
ÿCO
mechanism 2

k …2†
kCH …2† CO ;a
…2† 2 …2†
S2 ‡ S2a ‡ CH3 OH…g† @
3O
CH3 O …2† …2a†
‡H (18) HCOO ‡ S2a @ H…2a† ‡ CO2 (29)
kÿCH k …2†
…2† ÿCO ;a
3O 2

kCO…2†
S2 ‡ CO…g† @ CO…2† (19)
kÿCO…2† 3.6. Rate expressions and adsorption terms
kH…2a†
2S2a ‡ H2 …g† @ 2H…2a† (20) A kinetic analysis of the surface reaction mechan-
kÿH…2a†
isms was done to eliminate the unknown surface
kCH …2† concentrations. The overall equilibrium constants
2O
CH3 O…2† ‡ S2a @ CH2 O…2† ‡ H…2a† …RDS-III† for each reaction were also factored out and a site
kÿCH …2†
2O
balance for each type of active site was done. Details
(21) of these derivations are reported elsewhere [25]. The
k
C2 H5 O
…2†
2
following are the ®nal rate expressions and the adsorp-
CH3 O…2† ‡ CH2 O…2† @ CH3 OCH2 O…2† ‡ S2 tion terms for each reaction.
k …2†
ÿC2 H5 O
2

(22) Rate expression for methanol±steam reforming:

 .  .   
1=2 1=2
kR KCH3 O…1† KH…1a† pCH3 OH pH2 1 ÿ …1=KR † p3H2 pCO2 =pCH3 OH pH2 O
rR ˆ CST1 CST1 a : (30)
1 1a
 B.A. Peppley et al. / Applied Catalysis A: General 179 (1999) 31±49 39

Eq. (30) is derived directly from the rate expression where 2 and 2a are de®ned as
for the RDS for the methanol±steam reaction. KCH3 O…2† pCH3 OH
However, it is now written as the rate of the overall 2 ˆ 1 ‡ 1=2 1=2
methanol±steam reaction, rR, instead of rCH2 O…1† KH…2a† pH2
…methoxy†
while the rate constant kCH2 O…1† is now written as kR.
1 and 1a in Eq. (30) are given by the following KCO…2† KH2 …2a† pCO pH2
‡
expressions: KC H O…2† KC H O…2† KCHO…2† KCO…2† ;b
2 5 2 2 4 2 2
…oxymethylene†
KCH3 O…1† pCH3 OH
1 ˆ 1 ‡ 1=2 1=2 KCH3 O…2† KCO…2† KH…2a† pCH3 OH pCO pH2
1=2 1=2
KH…1a† pH2 ‡
…methoxy† KC H O…2† KCHO…2† KCO…2† ;b
2 4 2
KCO…1† KH2 …1a† pCO2 p2H2 …methyl biformate†
‡ 2
1=2 1=2
KC2 H5 O…1† KC2 H4 O…1† KHCOOH…1† KHCOO…1† KCO…1† ;a KOH…1† pH2 O KCH3 O…2† KCO…2† pCH3 OH pCO KCO…2† KH…2a† pCO pH2
2 2 2 ‡ ‡
…oxymethylene† KCHO…2† KCO…2† ;b KCO…2† ;b
3=2 3=2 …methyl formate† …formyl species†
KCH3 O…1† KCO…1† KH…1a† pCH3 OH pCO2 pH2
‡ 2 1=2 1=2
KCO…2† KH…2a† pCO2 pH2
KC …1† KHCOOH…1† KHCOO…1† KCO…1† ;a KOH…1† pH2 O
2 H4 O2 2 ‡ KCO…2† pCO ‡ 2
…methyl biformate†
…CO† KCO…2† ;a
2
KCH3 O…1† KCO…1† KH…1a† pCH3 OH pCO2 pH2 …formate†
‡ 2

KHCOOH…1† KHCOO…1† KCO…1† ;a KOH…1† pH2 O KOH…2† pH2 O

2 ‡ KCO…2† pCO2 ‡ 1=2 1=2
; (35)
…methyl formate† 2
…CO2 †
KH…2a† pH2
1=2 1=2 …hydroxyl†
KCO…1† KH…1a† pCO2 pH2 KCO…1† KH…1a† pCO2 pH2
‡ 2
‡ 2  
KHCOO…1† KCO…1† ;a KCO…1† ;a 1=2 1=2
2 2
2a ˆ 1 ‡ KH…2a† pH2 : (36)
…formic acid† …formate†
KOH…1† pH2 O 0
Referring Eqs. (33) and (34) back to the rate expres-
‡ 1=2 1=2
‡ KCO…1† pCO ‡ KCO p ;
2 CO2
(31) sion for the RDS of the water-gas shift mechanism and
KH…1a† pH2 …CO† …CO2 †
…hydroxyl† the decomposition reaction, respectively, it is seen that
rW and kW have been substituted for rHCOO…1† ;b and
 
1=2 1=2 kHCOO…1† ;b , and rD and kD have been substituted for
1a ˆ 1 ‡ KH…1a† pH2 : (32)
rCH2 O…2† and kCH2 O…2† .
Similarly, for the WGS reaction and the decom-
position reaction the following rate expressions were 3.7. Summary
derived.
Reversible Langmuir±Hinshelwood rate expres-
Rate expression for the water-gas shift reaction: sions for each of the overall reactions involved in

 .  . 
1=2 1=2
kW KCO…1† KOH…1† KH …1† pCO pH2 O pH2 …1 ÿ …1=KW †…pCO2 pH2 =pCO pH2 O ††  2
rW ˆ CST1 : (33)
21
Rate expression for the methanol decomposition
reaction:
 .  .   
1=2 1=2
kD KCH3 O…2† KH…2a† pCH3 OH pH2 1 ÿ …1=KD † p2H2 pCO =pCH3 OH
rD ˆ CST2 CST2a ; (34)
2 2a
40 B.A. Peppley et al. / Applied Catalysis A: General 179 (1999) 31±49

the process of methanol±steam reforming on Cu/ZnO/ Based on the stoichiometry of Reactions (I)±(III)
Al2O3 catalysts have been derived by a kinetic ana- the rates of the other three components in the reaction
lysis of the elementary surface reactions occurring on mixture can also be modelled as follows:
the catalyst. Parallel reaction mechanisms occurring
rH2 ˆ …3rR ‡ 2rD ‡ rW †SA ˆ 3rCO2 ‡ 2rCO
on two different phases of the catalyst were developed.
For both reaction mechanisms involving methanol, the …mol sÿ1 kgÿ1 †; (39)
dehydrogenation of the adsorbed methoxy intermedi-
ÿ rCH3 OH ˆ …rR ‡ rD †SA ˆ rCO2 ‡ rCO
ate was considered to be the RDS. The reaction path
for the water-gas shift reaction occurs on the same …mol sÿ1 kgÿ1 †; (40)
phase as the methanol±steam reaction. The RDS for ÿrH2 O ˆ …rR ‡ rW †SA ˆ rCO2 …mol sÿ1 kgÿ1 †:
the water-gas shift reaction was taken to be the for-
(41)
mation of an adsorbed formate from adsorbed hydro-
xyls and adsorbed CO according to the mechanism of However, as was stated above, only two of the three
DuÈmpelmann [12]. reactions are linearly independent and there is, there-
Eqs. (30), (33) and (34) are general forms of the rate fore, no justi®cation in using more than two responses
expressions based on the surface mechanisms which in the non-linear regression.
have been proposed. There are terms for every inter- In many kinetic modelling studies the surface site
mediate surface species and there is an equilibrium concentrations are either combined with the rate con-
constant for the concentration of each of these species stants or conversely the rate is expressed as a per site
on the catalyst surface. Terms have also been included turnover number. In this work, however, the site
for adsorbed species which are not involved in the concentrations were kept separate and explicit in
reaction mechanism but could occupy a signi®cant the rate expressions. There are several reasons for
number of active sites under certain conditions. this approach. Firstly, it is hoped that eventually the
variation in site concentrations can be used to explain
the variation in the catalyst selectivity with time-on-
line. Secondly, at this point there is no way to ade-
4. Development of a comprehensive model of the
quately distinguish between the site concentrations for
kinetics of methanol±steam reforming on Cu/ZnO/
Type 1 sites versus Type 2 sites and therefore a turn-
Al2O3
over number cannot be calculated for each reaction.
In order to allow for these future objectives and to
4.1. Formulation of a kinetic model from reaction
obtain reasonable parameter estimates for the model,
rate expressions
total site concentrations were set at 7.510ÿ6
mol mÿ2 for Type 1 and Type 2 and 1.510ÿ5
In each kinetic experiment the rates of production of
mol mÿ2 for Type 1a and Type 2a sites based on
CO2 and of CO per mass of catalyst were determined
the ab initio model for the water-gas shift reaction
for a given gas composition and reactor temperature.
on the (1 1 1) face of metallic copper developed by
In order to model these experimental responses, it was
Oveson et al. [26]. Since, statistically, the parameter
necessary to combine the rate expressions for each
for the site concentration and the rate constant are
individual reaction and multiply by the surface area
perfectly correlated the least squares estimate of the
per unit mass of fresh catalyst SA (m2 kgÿ1):
rate constant will simply vary to compensate for any
rCO2 ˆ …rR ‡ rW †SA …mol sÿ1 kgÿ1 †; (37) error in the estimate of the site concentration and
hence will not effect the quality-of-®t of the model.
rCO ˆ …rD ÿ rW †SA …mol sÿ1 kgÿ1 †: (38)
Because the rate measurements were done by oper- 4.2. Simplification of the expressions for fractional
ating the reactor in differential mode, Eqs. (37) and surface coverages
(38) could be used directly and did not need to be
integrated in a design equation. This greatly simpli®ed The adsorption term for Type 1 sites, 1, is given by
the analysis. Eq. (31) and for Type 2 sites, 2 is given by Eq. (35).
 B.A. Peppley et al. / Applied Catalysis A: General 179 (1999) 31±49 41

1 and 2 are the inverses of the fractions of vacant 4.3. Simplification by definition of composite
sites of Types 1 and 2, respectively. 1a and 2a are the parameters
analogous adsorption terms for hydrogen on Type 1a
and Type 2a sites. The more strongly an intermediate Further simpli®cation to the algebraic form of the
or a spectator species adsorbs on an active site, the model can be achieved by converting ratios and
fewer the number of sites available for reaction, and products of parameters into lumped parameters. This
hence slower the rate of reaction when all else is can be justi®ed based on the reduction of the total
constant. Each term in Eqs. (31) and (35) has been number of parameters in the model. The degree of
labelled according to the adsorbed surface species correlation between parameters in Langmuir±Hinshel-
which it represents. As a ®rst step in simplifying wood kinetic models is well known and reducing the
the rate expressions, a number of the terms in number of parameters ®nally results in a more reliable
Eqs. (31) and (35) will be eliminated based on obser- model [27].
vations from surface studies. The following substitutions were made in Eqs. (30),
(33), (34) and (43):
4.2.1. DRIFTS studies of surface processes on Cu/ KCH3 O…1†

ZnO/Al2O3 catalysts KCH3O
…1† ˆ
1=2
; (44)
DRIFTS measures the wavelength and intensity of KH…1a†
IR radiation absorption by various chemical bonds  KOH…1†
associated with speci®c chemical species. By taking KOH …1† ˆ
1=2
; (45)
KH…1a†
into account the absorption coef®cient, the relative
surface concentrations of the adsorbed species can be KCO…1† KH…1a†
1=2

estimated by an analysis of the spectra. Fig. 2 shows a 

KHCOO …1† ˆ 2
; (46)
typical spectrum for the adsorption of methanol and KCO…1† ;a
2
steam on a Cu/ZnO/Al2O3 catalyst. As can be seen, 
kW ˆ kW KCO…1† ; (47)
there are large bands for methoxy groups and formate
groups. Bands for methyl formate and methyl bifor-  KCH3 O…2†
KCH3O
…2† ˆ
1=2
; (48)
mate (expected at 1674 cmÿ1), however, are absent, as KH…2a†
are bands for adsorbed CO (expected at 2150 cmÿ1).
 KOH…2†
A signi®cant simpli®cation in Eqs. (31) and (35) is KOH …2† ˆ
1=2
; (49)
obtained by eliminating the site concentrations for the KH…2a†
species which do not appear in the FTIR spectra. The KCO…2† KH…2a†
adsorption terms for Site 1 and Site 2 can then be 
KHCOO …2† ˆ
2
: (50)
reduced to KCO…2† ;a
2

0 1
1=2 1=2
B KCH3 O…1† pCH3 OH KOH…1† pH2 O KCO…1† KH…1a† pCO2 pH2 C
1 ˆ B
@1 ‡ ‡ 1=2 1=2 ‡ 2
‡ KCO…1† pCO2 C
A; (42)
1=2 1=2
KH…1a† pH2 KH…1a† pH2 KCO…1† ;a 2
2 …CO2 †
…methoxy† …hydroxyl† …formate†
0 1
1=2 1=2
B KCH3 O…2† pCH3 OH KOH…2† pH2 O KCO…2† KH…2† pCO2 pH2 C
2 ˆ B
@1 ‡ ‡ 1=2 1=2 ‡ 2
‡ KCO…2† pCO2 C
A: (43)
1=2 1=2
KH…2a† pH2 KH…2a† pH2 KCO…2† ;a 2
2 …CO2 †
…methoxy† …hydroxyl† …formate†

The adsorption terms for hydrogen adsorption, 1a With these substitutions the forms of rR, rW and rD,
and 2a remain unchanged. described previously as Eqs. (30), (33) and (34),
42 B.A. Peppley et al. / Applied Catalysis A: General 179 (1999) 31±49

become considerably simpler.

 .   
 1=2 3
kR KCH 3O
…1† p CH 3 OH p H2 1 ÿ p H2 p CO 2
= kR p CH 3 OH p H 2 O CST1 CST1a
rR ˆ   ;
 1=2  1=2  1=2 1=2 1=2
1 ‡ KCH 3O
…1† …p CH3 OH =p H2 † ‡ K p p
HCOO…1† CO2 H2
‡ K …p
OH…1† H2 O
=p H2 † ‡ K CO
…1† p CO2 1 ‡ K p
H…1a† H2
2

(51)
 .
  1=2 2
kW KOH …1† pCO pH2 O …1 ÿ pH2 pCO2 =kW pCO pH2 O †CST1
pH 2
rW ˆ   .   .  2 ; (52)
 1=2  1=2  1=2
1 ‡ KCH 3O
…1† p CH 3 OH p H2 ‡ KHCOO…1†
p CO p
2 H2
‡ K OH…1†
p H 2 O p H2 ‡ K …1† pCO2
CO2
 .  
 1=2 2
kD KCH O…2† p CH 3 OH p H2 1 ÿ p H2 p CO = k D p CH 3 OH CST2 CST2a
rD ˆ   . 3  .   :
 1=2  1=2  1=2 1=2 1=2
1 ‡ KCH 3O
…2† p CH 3 OH p H2 ‡ K HCOO…2†
p CO p
2 H2
‡ K OH…2†
p H 2 O p H2 ‡ K CO
…2† p CO2
1 ‡ K p
H…2a† H2
2

(53)

   
The units of rR, rW and rD are still (mol sÿ1 mÿ2). 

SCH3 O…2†
HCH3O
…2†
KCH ˆ exp ÿ ; (62)
The adsorption terms 1, 2, 1a and 2a have 3O
…2†
R RT
now been written out in full in the equations. There
 
are three rate constants and 10 equilibrium con- 

SHCOO…2†
HHCOO

…2†
stants in Eqs. (51)±(53). The temperature dependence KHCOO …2† ˆ exp ÿ ; (63)
R RT
of each of these constants can be expressed either
using the Arrhenius expression or the van't Hoff    


SOH…2†
HOH …2†
expression: KOH …2† ˆ exp ÿ ; (64)
R RT
 
1 ÿER !
kR ˆ kR exp ; (54)
SCO…2†
HCO…2†
RT
  KCO…2† ˆ exp 2
ÿ 2
; (65)
2 R RT
1 ÿED
kD ˆ kD exp ; (55)
RT  
 
SH…2a†
HH…2a†
KH…2a† ˆ exp ÿ : (66)
 1 ÿEW R RT
kW ˆ kW exp ; (56)
RT
   At this point, after considerable simpli®cation, the



SCH3 O…1†
HCH 3O
…1† kinetic model still contained 26 adjustable parameters.
KCH3 O…1† ˆ exp ÿ ; (57)
R RT Obtaining models with large numbers of parameters is
    a regrettable characteristic of the Langmuir±Hinshel-


SHCOO…1†
HHCOO …1†
wood method of analysis. The high degree of correla-
KHCOO …1† ˆ exp ÿ ; (58)
R RT tion between many of the parameters means that there
    will be a large uncertainty associated with many of the


SOH…1†
HOH …1†
KOH…1† ˆ exp ÿ ; (59) parameter estimates [28]. The correlation between
R RT least squares estimates of heats of adsorption and
 ! entropies of adsorption is known to be particularly

SCO…1†
HCO…1†
KCO…1† ˆ exp 2
ÿ 2
; (60) bad [27], therefore it was decided to arbitrarily set the
2 R RT heats of adsorption at values consistent with values
  reported in the literature from independent adsorption

SH…1a†
HH…1a† surface studies rather than including them in the
KH…1a† ˆ exp ÿ ; (61)
R RT regression.
 B.A. Peppley et al. / Applied Catalysis A: General 179 (1999) 31±49 43

4.4. Experimental planning and results total of 43 useable runs for parameter estimation. A
complete listing of the rate data used for the non-linear
An extensive experimental program was conducted regression, with experimental conditions, can be
to collect suf®cient kinetic data to obtain parameter obtained by contacting the principal author (pepley-b
estimates for the ®nal rate model. Rate measurements @rmc.ca).
were obtained by the differential method using a ®xed-
bed plug ¯ow reactor. The change in the partial
4.5. Parameter estimation
pressure of any gaseous component was never more
than 5±10% and the average partial pressure of each
Eqs. (51)±(66) were substituted into Eqs. (37) and
component was used in the kinetic analysis. The
(38) and parameter estimates were obtained by non-
catalyst used for these experiments was BASF K3-
linear regression using a multi-response Bayesian
110 low temperature shift catalyst. Details of the
criterion [29]. Preliminary regressions revealed
properties of this catalyst, the experimental equipment
that the parameters for the adsorption of CO2 on both
and the experimental procedure used have been
the Type 1 and Type 2 sites were not statistically
described elsewhere [18,25].
different from zero and hence could be removed
Five different components are involved in Reac-
from the regression with no loss in goodness-in-®t.
tions (I)±(III) and the partial pressure of each could be
Also, the adsorption term for the formate species on
varied independently during differential rate measure-
Type 2 sites was not statistically different from
ments. Temperature could also be varied resulting in a
zero and was removed from the model. In the case
total of six experimental variables. The maximum
of Type 1 sites the CO2 surface concentration is
operating pressure of the apparatus was 35 bar and
negligible because CO2 either desorbs or is converted
the maximum recommended operating temperature of
to formate, hence the term for CO2 in the site balance
the catalyst is 533 K. Six inequalities:
can be removed. The CO2 surface concentration term
 433 Kreactor temperature533 K,
for Type 2 sites is negligible because CO2 neither
 1 barreactor pressure35 bar,
readily adsorbs on these sites nor readily forms
 0 bar<partial pressure of CO0.5 bar,
intermediate formate. The ®nal form of the model
 0.29ratio of partial pressure of CO2 to
therefore contained 20 parameters of which seven
H20.370,
were ®xed and 13 were determined by non-linear
 0<ratio of partial pressure of CH3OH to H2O1.2,
regression.
 0.6ratio partial pressure of H2 to H2O130,
The values and the con®dence intervals for all the
were used to define the experimental operating region
parameters are presented in Table 2. The con®dence
and based on these constraints, experiments were
intervals were obtained using a simple method based
designed using the D-optimal criterion for a multi-
on the concept of the pro®le t-plot proposed by Bates
response model [29]. This criterion chooses the
and Watts [28]. The con®dence interval for each
experimental conditions that result in the greatest
parameter was determined by ®nding the high and
improvement in the accuracy of the parameter esti-
low value which resulted in the sum of squared
mates for the model that is being tested.
residuals being [1‡P/(NÿP)F(P, NÿP, 0.05)] times
Fifty-eight experiments were designed and per-
the minimum value of the least squares while the
formed. Each data set for the non-linear regression
remaining parameters were kept at their least-squares
consisted of the rates of formation of CO and CO2, the
values. Fig. 3 and 4 are the parity plots for the rates of
temperature and the gas composition. In addition, after
CO and CO2 production, respectively.
each kinetic experiment, a baseline conversion versus
W=FCH3 OH;0 measurement was done to ensure that the
catalyst activity and selectivity had remained stable 4.6. Summary ± final form of the model
during the experiment. The baseline conditions were
arbitrarily de®ned as 513 K, S/M ratio of 1, and total The ®nal forms of the rate expressions for the
pressure of 1.01 bar. Fifteen runs were eliminated due overall reactions involved in the process of metha-
to signi®cant changes in baseline activity leaving a nol±steam are, therefore, as follows:
44 B.A. Peppley et al. / Applied Catalysis A: General 179 (1999) 31±49

Table 2
Parameters for comprehensive kinetic model of methanol±steam reforming on BASF K3-110 Cu/ZnO/Al2O3 catalyst

Rate constant or
Si (J molÿ1 Kÿ1) or S(t,ÿ95%) S(t,‡95%)
Hi or E S(t,ÿ95%) S(t,‡95%)
equilibrium constant ki1 (m2 sÿ1 molÿ1) (kJ molÿ1)

kR (m2 sÿ1 molÿ1) 7.4E‡14 6.8E‡14 8.0E‡14 102.8 102.4 103.1

 ÿ0.5
KCH 3O
…1† (bar ) ÿ41.8 ÿ42.7 ÿ40.9 ÿ20.0 ± ±
 ÿ0.5
KOH…1† (bar ) ÿ44.5 ÿ47.0 ÿ42.1 ÿ20.0 ± ±
KH…1a† (barÿ0.5) ÿ100.8 ÿ103.7 ÿ97.9 ÿ50.0 ± ±
 ÿ1.5
KHCOO …1† (bar ) 179.2 174.4 182.9 100.0 ± ±
kD (m2 sÿ1 molÿ1) 3.8E‡20 1.6E‡20 6.1E‡20 170.0 168.0 173.9
 ÿ0.5
KCH 3O
…2† (bar ) 30.0 18.7 41.0 ÿ20.0 ± ±
 ÿ0.5
KOH …2† (bar ) 30.0 19.0 41.3 ÿ20.0 ± ±
KH…2a† (barÿ0.5) ÿ46.2 ÿ55.0 ÿ31.1 ÿ50.0 ± ±

kW (m2 sÿ1 molÿ1) 5.9E‡13 5.1E‡13 6.6E‡13 87.6 87.1 88.1
Note: Italicised values were set at the values indicated based on published data on the heats of adsorption. They were not included in the non-
linear regression.

Methanol±steam reaction:
 .  
 1=2 3
kR KCH 3O
…1† p CH 3 OH p H2 1 ÿ p H2 p CO 2
= k R p CH 3 OH p H2 O CST1 CST1a
rR ˆ   .   .  : (67)
 1=2  1=2  1=2 1=2 1=2
1 ‡ KCH 3O
…1† p CH 3 OH p H2 ‡ K HCOO…1†
p CO p
2 H2
‡ K OH…1†
p H2 O p H2 1 ‡ K p
H…1a† H2

Water-gas shift reaction:

 . 
  1=2 2
kW KOH …1† pCO pH2 O pH2 …1 ÿ pH2 pCO2 =kW pCO pH2 O †CST1
rW ˆ   .   . 2 : (68)
 1=2  1=2  1=2
1 ‡ KCH 3O
…1† p CH3 OH p H2 ‡ K p p
HCOO…1† CO2 H2
‡ KOH…1†
p H2 O p H2

Decomposition reaction:
 .  
 1=2 2 T T
kD KCH 3O
…2† p CH3 OH p H2 1 ÿ p H2 CO D CH3 OH CS2 CS2a
p = k p
rD ˆ   .   .  : (69)
 1=2  1=2 1=2 1=2
1 ‡ KCH 3 O …2† pCH3 OH p H2
‡ K OH …2† pH2 O p H 2
1 ‡ K H…2a† pH2

By substituting these expressions into Eqs. (37) and atmosphere and time-on-line [5,16,18] and therefore
(38) a comprehensive model of methanol±steam the validation of the ®nal model for ``real operating
reforming is obtained which predicts the rates of conditions'' was not a trivial task. Accumulating
production of both CO and CO2. Furthermore, by suf®cient data for which the variation in the activity
using Eqs. (39)±(41) the rate of production of hydro- of the catalyst did not confound the validation test was
gen and the rates of disappearance of methanol and dif®cult. Conversely, developing a method of correct-
steam can also be predicted. ing data from a ``real reactor'' so that the catalyst
behaviour could be normalised also presented pro-
blems. Several methods were, however, developed for
5. Validation of the final model testing the model that minimised the effect of changes
in the state of the catalyst.
The surface area and copper dispersion of Cu/Zn/ Fig. 5 shows the variation in the parameter W with
Al2O3 catalysts are known to vary with both operating conversion. W, which is de®ned as
 B.A. Peppley et al. / Applied Catalysis A: General 179 (1999) 31±49 45

Fig. 3. Parity plot for rate of CO2 production after terms involving negligible surface coverages were removed. The slope is 0.98 and the
intercept is 890 mmol sÿ1 kgÿ1. These values are not statistically different from unity and zero, respectively. The correlation coefficient
r2ˆ0.97.

Fig. 4. Parity plot for rate of CO production after terms involving negligible surface coverages were removed. The slope is 1.0 and the
intercept is 6.7 mmol sÿ1 kgÿ1. The value of the intercept is not statistically different from zero. Correlation coefficient r2ˆ0.97.
46 B.A. Peppley et al. / Applied Catalysis A: General 179 (1999) 31±49

Fig. 5. Model prediction of the exit composition of a plug flow reactor at 513 K, 1 bar and steam-to-methanol ratioˆ1.0 M. The plotted points
are based on baseline data from kinetic studies on various catalyst beds.

…PCO2 PH2 † shows that the comprehensive model developed in this

W ˆ (70)
…PCO PH2 O †KW work is valid for pressures considerably higher than
previous models. Furthermore, in Fig. 7 we see that
is a measure of the product composition relative to the the model accurately predicted the decrease in the
water-gas shift reaction equilibrium (i.e. when the total conversion as the pressure was increased from 1
WGS reaction is at equilibrium Wˆ1). The line in to 39 bar. This indicates that the effects of the reverse
Fig. 5 is the value of W predicted by the ®nal reaction and the reaction equilibrium are represented
model based on Eqs. (67)±(69) using the parameter well by the model.
estimates from Table 2. This curve uses an average Fig. 8 shows the effect of temperature on the frac-
value for the catalyst surface area, as used in Eqs. (37) tional conversion versus W=FCH3 OH;0 . The data at
and (38), to model data which were collected from 513 K were collected at baseline conditions. The
various beds of the BASF catalyst used in the experi- temperature was then increased to 533 K and a second
mental program. Fig. 5 shows that the model was able set of data was collected. Again the predictions of the
to accurately predict the product composition for a model are in good agreement with the experimental
wide range of methanol conversions. No previous observations.
models for methanol±steam reforming on Cu/ZnO/
Al2O3 catalysts would have been able to generate this
curve. Fig. 5 also demonstrates that the model 6. Conclusions
achieved the key objective of being able to predict
the variation in the CO content of the product stream 6.1. Surface mechanisms for overall reactions
for methanol±steam reforming on Cu/ZnO/Al2O3
catalysts. The elementary reactions which occur on the sur-
Euler's method was used to integrate Eqs. (37) and face of Cu/ZnO/Al2O3 catalysts during methanol±
(38) over the length of an isothermal catalyst bed to steam reforming were thoroughly reviewed and sur-
generate predictions of integral methanol conversion face mechanisms for all three overall reactions were
(e.g. greater than 10% conversion). Fig. 6 shows pre- proposed. The mechanisms were progressively
dicted methanol conversion versus W=FCH3 OH;0 for improved based on results from an extensive experi-
pressures ranging from 1.16 to 15.9 bar. This ®gure mental program.
 B.A. Peppley et al. / Applied Catalysis A: General 179 (1999) 31±49 47

Fig. 6. Prediction of model compared to low conversion experimental data.

Fig. 7. Model prediction of variation in methanol conversion with pressure: Lines are model prediction. Experimental data: (*) 1 bar; (~)
39 bar.

Kinetic analyses of these reaction mechanisms were 6.2. Comprehensive kinetic model
done to yield reversible Langmuir±Hinshelwood rate
expressions for each of the overall reactions. The RDS A differential ®xed-bed reactor was used to accu-
chosen for both the methanol±steam reaction and the mulate a signi®cant amount of rate data for methanol±
decomposition reaction was the dehydrogenation of steam reforming on a BASF low temperature shift
adsorbed methoxy. As proposed by DuÈmpelmann catalyst. Estimates of 13 adjustable parameters were
[12], the production of a surface formate from obtained using a multi-response non-linear regression.
adsorbed hydroxyl and CO was used as the RDS The result is a comprehensive kinetic model which is
for the water-gas shift reaction. able to predict both the rate of hydrogen production
48 B.A. Peppley et al. / Applied Catalysis A: General 179 (1999) 31±49

Fig. 8. Variation in methanol conversion with temperature: Pˆ1.01 bar, S/Mˆ1.0. Lines are model prediction. Experimental data: (~) 513 K;
(&) 533 K.

and the composition of the product gas for the process Ki equilibrium constant of reaction i or
of methanol±steam reforming on a commercial Cu/ adsorption coefficient for surface species i
ZnO/Al2O3 catalyst. In particular the rate of produc- N number of experimental runs included in
tion of CO is successfully modelled. This kinetic the regression
model is referred to as comprehensive because it is pi partial pressure of component i (bar)
the ®rst model of methanol±steam reforming which P number of parameters in model
simultaneously takes into account all three reversible ri rate of reaction i (mol sÿ1 mÿ2) or rate of
overall reactions. formation of component i (mol sÿ1 (kg of
catalyst)ÿ1)
r2 correlation coefficient of linear fit to
7. Nomenclature parity plots (Figs. 3 and 4)

Si entropy of adsorption for species i
CSTi total surface concentration of site i (J molÿ1 Kÿ1)
(mol mÿ2) SA surface area of fresh catalyst (m2 kgÿ1)
Ei activation energy for rate constant of Si active site i in reaction mechanism
reaction i (kJ molÿ1) S(t,95%) the parameter value which produces an
FCH3 OH;0 molar flow rate of methanol in feed to excess sum of squared residuals cor-
reactor (mol sÿ1) responding to the Fisher's F statistic at
F (P, Nÿ Fisher's statistic for P parameters and the 95% confidence level; columns 3, 4, 6
P, 0.05) NÿP degrees of freedom at the 95% and 7 of Table 2
confidence level S/M molar ratio of steam to methanol in feed

Hi heat of adsorption for surface species i or to reactor
heat of reaction for formation of surface W mass of catalyst (kg)
species i (kJ molÿ1) xCH3 OH conversion of methanol
ki rate constant for reaction i; units will
be specific to the form of the rate Greek symbols
expression
ki1 pre-exponential term in Arrhenius expres- W parameter relating composition of reactor
sion; this is the intercept of the Arrhenius effluent to the WGS equilibrium
equation as 1/T!0 or T!1 k adsorption term for active site k
 B.A. Peppley et al. / Applied Catalysis A: General 179 (1999) 31±49 49

 fractional coverage; used in Table 1 (1995) 821±827.

 fraction of active sites which are [6] C. Rhodes, G.J. Hutchings, A.M. Ward, Catal. Today 23
(1995) 43±58.
available; used in Eq. (4) [7] C.T. Campbell, K.-H. Ernst, Forward and reverse water-gas
shift reactions on model copper catalysts, in: D.J. Dwyer,
Subscripts F.M. Hoffman (Eds.), Surface Science of Catalysis, American
Chemical Society Symposia Series No. 482, Chapter 8,
R direct-reforming; Reaction (I) American Chemical Society, Washington, DC, 1992.
[8] E. Santacessaria, S. CarraÁ, Riv. Combust. 32 (1978) 227±232.
W water-gas shift; Reaction (II) [9] J.C. Amphlett, M.J. Evans, R.A. Jones, R.F. Mann, R.D. Weir,
D decomposition; Reaction (III) Can. J. Chem. Eng. 59 (1981) 720±727.
1 active site 1 when on variable S [10] J.C. Amphlett, M.J. Evans, R.F. Mann, R.D. Weir, Can. J.
1a active site 1a when on variable S Chem. Eng. 63 (1985) 605±611.
[11] J.C. Amphlett, M.J. Evans, R.F. Mann, R.D. Weir, Can. J.
2 active site 2 when on variable S
Chem. Eng. 66 (1988) 950±956.
2a active site 2a when on variable S [12] R. DuÈmpelmann, Kinetische Untersuchungen des Methanol-
a, b indicates an alternate route to a surface reforming und der Wassergaskonvertierungsreaktion in einem
intermediate (as in Eqs. (16) and (17)) konsentrationgeregelten Kreislaufreaktor, Ph.D. Dissertation,
EidgenoÈssischen Technischen Hochschule, ZuÈrich, 1992.
Superscripts [13] C.J. Jiang, D.L. Trimm, M.S. Wainwright, N.W. Cant, Appl.
Catal. A 97 (1993) 145.
[14] G.J. Millar, C.H. Rochester, K.C. Waugh, J. Chem. Soc.,
(i) species adsorbed on active site i where i is Faraday Trans. 87(17) (1991) 2795.
1, 1a, 2 or 2a [15] C.G. Hill, An Introduction to Chemical Engineering Kinetics
 composite parameter as defined in and Reactor Design, Wiley, New York, 1977, p. 88.
[16] B.S. Clausen, J. Schiùtz, C.V. Lars GraÊbñk, K.W. Ovesen, J.K.
Eqs. (44)±(50)
Jacobsen, H. Nùrskov, T. Topsùe, Topics Catal. 1 (1994) 367.
T indicating total concentration of [17] M. Boudart, G. Djega-Mariadassou, Kinetics of Heteroge-
active sites neous Catalytic Reactions, ISBN 0-691-08346-0, Princeton
University Press, Princeton, NJ, 1984, pp. 77±117.
[18] B.A. Peppley, J.C. Amphlett, L.M. Kearns, R.F. Mann, Part 1,
Acknowledgements Appl. Catal. A 179 (1999) 21±30.
[19] D.J. Pritchard, D.W. Bacon, CJChE 52 (1974) 103±109.
[20] G.J. Millar, C.H. Rochester, K.C. Waugh, J. Chem. Soc.,
Support of this work by the Canadian Department Faraday Trans. 87(17) (1991) 2785±2793.
of National Defence, Chief of Research and Devel- [21] I.E. Wachs, R.J. Madix, J. Catal. 53 (1978) 208±227.
opment, is gratefully acknowledged. [22] K.M. Minachev, K.P. Kotyaev, G.I. Lin, A.Y. Rozovskii,
Catal. Lett. 3 (1989) 299±308.
[23] W.C. Conner, J.L. Falconer, Chem. Rev. 95 (1995) 759.
[24] R.J. Burch, S.E. Golunski, M.S. Spencer, Catal. Lett. 5(1)
References (1990) 55±60.
[25] B.A. Peppley, Ph.D. Dissertation, Royal Military College of
[1] K.C. Waugh, Catal. Today 15 (1992) 51±75. Canada, May 1997.
[2] J. Skrzypek, J. Sloczynski, S. Ledakowicz, Methanol [26] C.V. Ovesen, P. Stolze, J.K. Norskov, C.T. Campbell, J. Catal.
Synthesis, ISBN 83-01-11490-8, Polish Scientific Publishers, 134 (1992) 445±468.
Warsaw, 1994. [27] D.J. Pritchard, D.W. Bacon, Chem. Eng. Sci. 33 (1978) 1539±
[3] G.H. Graaf, E.J. Stamhuis, A.A.C.M. Beenackers, Kinetics of 1543.
low-pressure methanol synthesis, Chem. Eng. Sci. 43 (1988) [28] D.M. Bates, D.G. Watts, Non-linear regression analysis and
3185±3195. its applications, in: Barnett et al. (Eds.), Wiley Series in
[4] K.M. Vanden Bussche, G.F. Froment, J. Catal. 161 (1996) Probability and Mathematical Statistics, Wiley, New York,
1±10. 1988, p. 201.
[5] M.G. Kalchev, A.A. Andreev, N.S. Zotov, Kinet. Catal. 36(6) [29] G.E.P. Box, N.R. Draper, Biometrika 52 (1965) 355.
 本文献由“学霸图书馆-文献云下载”收集自网络，仅供学习交流使用。

学霸图书馆（www.xuebalib.com）是一个“整合众多图书馆数据库资源，

提供一站式文献检索和下载服务”的24 小时在线不限IP 图书馆。

图书馆致力于便利、促进学习与科研，提供最强文献下载服务。

图书馆导航：

图书馆首页 文献云下载 图书馆入口 外文数据库大全 疑难文献辅助工具