You are on page 1of 14

PCCP

View Article Online


PAPER View Journal

A Langmuir–Hinshelwood approach to the kinetic


Cite this: DOI: 10.1039/c3cp50715g
modelling of catalytic ammonia decomposition in
an integral reactor
Published on 11 April 2013 on http://pubs.rsc.org | doi:10.1039/C3CP50715G

S. Armenise,ab E. Garcı́a-Bordejé,b J. L. Valverde,c E. Romeoa and A. Monzón*a

The increasing interest in ammonia decomposition is due to the fact that this compound can be used
Downloaded by Ball State University on 20/05/2013 10:36:48.

advantageously as a hydrogen carrier, allowing the development of single-step hydrogen generation


systems. With the aim of developing efficient reactors for ammonia decomposition, e.g. for fuel cell
applications, it is imperative to investigate the kinetics and reaction mechanism in depth. The main goal
of this work is to develop reliable kinetic models that are able to predict the performance obtained
using integral reactors, e.g. monoliths. In this case, an almost complete NH3 conversion is obtained, with
a high H2 concentration at the exit of the reactor. The operating conditions, mainly the gas
composition, are very different along the reactor. In addition, the temperatures needed to attain such
large conversions are usually high. The kinetic models developed in this contribution are based on the
Langmuir isotherm, considering that all the adsorbed species can be kinetically relevant, that the slow
Received 15th February 2013, step or steps can be partially reversible, and that the surface can be considered as energetically
Accepted 10th April 2013 uniform, i.e. ideal. Among other conclusions, the results obtained indicate that the variable kinetic
DOI: 10.1039/c3cp50715g orders and apparent activation energies frequently reported in the literature can be direct
consequences of the data analysis and can therefore also be explained without considering any change
www.rsc.org/pccp in the controlling step with the reaction temperature or in the hydrogen or ammonia concentration.

1. Introduction terms of energy capacity and distribution infrastructure, but


suffer from their intrinsic carbon content causing end-user CO2
Currently, the technical difficulties involved in the use of emissions, since in situ capture is not feasible. However,
hydrogen as an energy carrier are mainly related to transporta- ammonia combines an easy way to store hydrogen and supply
tion and storage, which make the incorporation of this carbon-free stream to the fuel cell stack. In addition, ammonia
technology slower than initially expected.1 To mitigate these has other desirable characteristics that make it potentially
problems, various chemical and physical systems have been attractive. It is non-flammable and non-explosive, it can be
proposed for storing hydrogen safely and in an economically liquefied under mild conditions, and it has a large weight
feasible way. Some of these technologies involve complex metal hydrogen fraction (17.65% of the mass of ammonia). Ammonia
hydrides, such as NaAlH42 or LiAlH4;3 metal–organic frame- actually has a volumetric hydrogen density of about 45% higher
works (MOFs);4 and metallic amines.5 All these systems have than that of liquid hydrogen.6 The thermodynamic properties
high potential storage capacities. Nevertheless, they also have of ammonia are also advantageous. Hence, it can easily be
various handicaps such as low reversibility of hydrogen storage, stored as a liquid at room temperature and at a pressure of
high operation temperatures and large regeneration costs. about 8.6 bars, while in the case of hydrogen, an extremely low
Among the main competitors of hydrogen as an energy carrier, temperature of 20 K and very expensive tanks are required to
methane/natural gas and methanol offer clear advantages in keep it in the liquid form. The usable amount of hydrogen per
kilogram of ammonia is relatively high compared to other
a
Department of Chemical and Environmental Engineering, hydrogen storage technologies. Furthermore, only 16% of the
Institute of Nanoscience of Aragon, University of Zaragoza, Zaragoza, 50018,
energy stored in ammonia is needed to break gaseous ammonia
Spain. E-mail: amonon@unizar.es; Fax: +34 976761879; Tel: +34 976761157
b
Instituto de Carboquimica (ICB-CSIC), Zaragoza, Spain
into nitrogen and hydrogen gases.7 All these advantages sug-
c
Department of Chemical Engineering, University of Castilla La Mancha, gest that catalytic ammonia decomposition is a key stage in the
Ciudad Real, Spain implementation of this technology.

This journal is c the Owner Societies 2013 Phys. Chem. Chem. Phys.
View Article Online

Paper PCCP

Catalytic ammonia decomposition has been studied since desorption of nitrogen, the rest of the steps being quasi-
the early days of heterogeneous catalysis.8 One century later, equilibrated. Therefore, the inhibitory effect of hydrogen is
the ammonia decomposition reaction was still being studied, due to the rehydrogenation of nitride adsorbed species in
in this case to obtain information about the kinetic mechanism ammonia.25 At high temperatures and low hydrogen partial
of the ammonia synthesis, due to easier operating conditions pressures, the reaction does not show any dependence with
when compared to the Haber–Bosch process.9 The activity on respect to the partial pressure of hydrogen. This behaviour has
different pure metals shows the typical volcano plot, related to been referred to as Tamaru’s model.22,25 In this model it is
the strength of adsorption of the relevant species in the reac- assumed that both the cleavage of the N–H bond in adsorbed
tion mechanism.10 At the beginning of the 19th century, NH3 and the recombinative desorption of adsorbed nitrogen
Thenard and Dulong,8 using different metals, obtained the are slow and far from equilibrium.18,24,25 Tsai et al.,26 explained
following sequence: Fe > Cu > Ag > Au > Pt. Choudhary that inhibition of hydrogen can be explained considering that
et al.11 found that Ru is more active than Ir and Ni. In addition, chemisorbed hydrogen may block surface sites that are neces-
as regards the effect of the support, these authors have also sary for ammonia decomposition, or react chemically and
Published on 11 April 2013 on http://pubs.rsc.org | doi:10.1039/C3CP50715G

reported that the intrinsic activity expressed in terms of TOF hydrogenate NHX intermediates generated during the ammonia
(molecules per site per s) decreases in the order Ni/SiO2 D decomposition.
Ni/HY > Ni/HZSM-5 > Ni/SiO2–AlO3. As they claimed, these It is worth noting that most of the experimental studies have
Downloaded by Ball State University on 20/05/2013 10:36:48.

differences indicate that the differences in the metal–support been carried out under high vacuum conditions using model
interactions probably play a decisive role in determining the catalysts. These gaps in the operating conditions make it
rate determining step.11 Yin et al. have recently reviewed the difficult to extrapolate the results of conditions that are appro-
diverse effects of active components, support and promoters on priate to fuel cell applications, i.e. high ammonia concen-
the catalyst performance during ammonia decomposition for tration, atmospheric pressure and high temperatures, to
on-site hydrogen generation.12 These authors concluded that attain high ammonia conversion.13,23
Ru is the most active catalyst, carbon nanotubes the most Another aspect thoroughly considered in the kinetic description
effective support, and KOH the best promoter for this reaction. of this reaction is the effect of the surface non-uniformity.22,27–29
However, it is also claimed that Ni based catalysts are a valuable In this case, the kinetic models are developed under Temkin’s
alternative as far as the cost of the operation is concerned. formalism, considering that the adsorption of the reactants
In addition, large metal (Ru) dispersion and support basicity follows the Frumkin–Temkin isotherm instead of the Langmuir
and conductivity are important factors for attaining highly isotherm.30,31 The reaction rate is given by the Temkin–Pyzhev
efficient catalysts.12 model:22,27–29
In order to explain these results, several reaction mechan-  m
  pNH32
isms have been suggested, but there is still no consensus about rNH3 ¼ k0 (1)
the rate determinant step, or about the most abundant reactive pH23
intermediate (MARI).13 Examining the effect of the operating
The Temkin–Pyzhev model is equivalent to the power law rate
conditions, i.e. pNH3 and pH2, two limiting cases, depending on
expression:23,32
the reaction temperature, are usually encountered:
 
(i) Effect of ammonia concentration. At low temperatures rNH3 ¼ kp pNH3a pH2b (2)
(e.g. T o 700 K for Pt14–16 or Ru,13,17,18 and T o 1000 K for
Ni19 or Fe20) and low ammonia concentrations (pNH3 o 1 Torr) In the above equations, k 0 and kp follow an Arrhenius depen-
the reaction rate shows a zero-order dependence with respect to dence with temperature, m is a constant related to the non-
ammonia. However, at high temperatures the reaction becomes uniformity of the surface,26,31 and a and b are the kinetic orders
first-order with respect to ammonia. Besides, the apparent with respect to ammonia and hydrogen, respectively.
energy of activation measured decreased from 180 kJ mol1 at The above considerations indicate that the rate determining
low temperatures to only 21 kJ mol1 at high temperatures.17,21 step depends on the operating conditions and on the catalyst
Tamaru22 has developed a theoretical expression to calculate composition. Thus, during the development of the kinetics
the transition temperature and Chellappa et al.23 have tabulated, expressions directly derived from the reaction mechanism,
for several catalysts, the experimental values of the transition the following cases have been considered:
temperatures, and energy of activation for zero and first kinetic Case A: the reaction rate is controlled only by the desorption
regimes. These changes have been explained considering that of adsorbed nitrogen atoms, and the remaining steps are in
below about 650 K desorption of adsorbed nitrogen atoms is rate equilibrium;25,33
limiting, while above 750 K the cleavage of the N–H bond in Case B: the first dissociation of adsorbed NH3 with scission
adsorbed NH3 is rate limiting.17,21 of the N–H bond is now the controlling step;17,21,34,35
(ii) Effect of hydrogen concentration. At low temperatures Case C: both stages, N–H bond cleavage and recombinative
and high hydrogen partial pressures, the reaction is found to desorption of surface nitrogen atoms are slow and irreversible
be inhibited by hydrogen.24 This case is explained by the steps;24,25,30
Temkin–Pyzhev mechanism.22,25 This mechanism assumes In the three cases it is usually considered that the adsorbed
that the rate determining step is only the recombinative N* is the most abundant reactive intermediate (MARI).18,24,33

Phys. Chem. Chem. Phys. This journal is c the Owner Societies 2013
View Article Online

PCCP Paper

However, in some cases it is considered that NH3* and H* Table 1 Experimental conditions used to determine the kinetic parameter
species are the MARI.13,36,37
Experiment W/FNH3,0 (g s mol1) pNH3,0 (atm) pH2,0 (atm) pAr,0 (atm)
Extending Case C that considers that both steps are slow and
irreversible, Bradford et al.,18 assumed that the activation of 1 13 921.9 0.031 0.494 0.475
2 3333.4 0.108 0.786 0.107
adsorbed ammonia and scission of the N–H bond is a slow but 3 1342.5 0.219 0.664 0.117
partially reversible step. This new case generalizes the previous 4 1407.3 0.233 0.767 0.000
models, but consequently has a more complex mathematical 5 729.2 0.302 0.456 0.242
6 1029.5 0.303 0.697 0.000
solution due to the appearance of quadratic expressions in the 7 739.1 0.376 0.624 0.000
solution of the reaction rate equation.36,38 It has been clearly 8 568.0 0.439 0.561 0.000
9 831.6 0.515 0.000 0.485
pointed out by Vilekar et al.,33 that in spite of the research
10 431.0 0.507 0.493 0.000
developed over decades into this reaction, there is no agree- 11 380.0 0.737 0.263 0.000
ment about which step(s) are the RDSs, or even whether there is 12 420.3 0.849 0.151 0.000
a single RDS, and which surface species are kinetically relevant. 13 444.3 1.000 0.000 0.000
Published on 11 April 2013 on http://pubs.rsc.org | doi:10.1039/C3CP50715G

The main goal of this work is to develop a reliable kinetic


model able to predict the performance obtained using integral with 50 mL min1 of a H2(50%)–N2(50%) mixture. Exit gases
reactors, e.g. monoliths.39,40 In this case, almost complete NH3 were analysed by GC using an Agilent Micro GC 3000A. H2 and
Downloaded by Ball State University on 20/05/2013 10:36:48.

conversion and high hydrogen concentrations at the exit of the N2 were separated in a molsieve column, and ammonia in a Plot-Q
reactor are obtained. The operating conditions, mainly the gas column. To ensure repeatability of the analysis, 2–3 separate GC
composition, are very different along the reactor, and the tem- samples were taken and averaged for each experimental data
peratures needed to attain such large conversions are usually high. point. Results of the analyses were typically within 3% of each
The models developed here generalize the cases described other. Ammonia conversion calculations were made taking into
above, considering that all the adsorbed species can be kinetically account the mole increase in the reaction.
relevant, that the slow step or steps can be partially reversible,
and that the surface can be considered as energetically uniform,
i.e. ideal. Among other conclusions, the results obtained indicate 3. Reaction mechanism: a kinetic modelling
that the variable kinetic orders and apparent activation energies of ammonia decomposition
obtained in many cases are direct consequences of the data
analysis and can therefore also be explained without considering The commonly accepted reaction mechanism for ammonia
any change in the controlling step with the reaction temperature, decomposition12,17,18,36,41 includes ammonia chemisorption
or hydrogen or ammonia concentration. on the catalyst surface; successive ammonia dehydrogenation
and associative nitrogen and hydrogen desorption. It is assumed
that the surface sites are energetically homogeneous, i.e. the
2. Experimental adsorption of all the species follows the Langmuir isotherm.
The catalyst used in the kinetic study is a Ni/Al2O3/monolith This mechanism can be described according to the following
prepared by a previously described procedure.39 After coating sequence of elementary reactions:
k00 ;k00
the monolith with a g-alumina layer, a NiNO3 precursor was (i) NH3 þ  ! NH3 
impregnated by equilibrium adsorption and subsequently k01 ;k01
(ii) NH3  þ  ! NH2  þH
rinsed with abundant distilled water. After drying, it was
k02 ;k02
calcined under N2 atmosphere (100 mL min1) at 873 K for (iii) NH2  þ  ! NH  þH
2 h (1 K min1). The calcined monolith was crushed and sieved k1 ;k1
(iv) NH þ  ! N þ H
to obtain particles with a diameter ranged between 125 to " #
3 k0H ;k0H
180 mm. The Ni loading deposited on the catalyst was 15 wt% 2 2
(v) 2H ! H2 þ 2
related to the alumina weight as measured by ICP-OES. The 2
catalytic tests were carried out at atmospheric pressure in a " #
1 k0N ;k0N
tubular quartz reactor with 6 mm inner diameter. Gas compo- 2 2
(vi) 2N ! N2 þ 2
sition at the inlet of the reactor was changed in order to study 2
the effect of ammonia, hydrogen and nitrogen on the reaction
The term * denotes a vacant site, and NH3*, NH2*, NH*, N*
rate. Ammonia flow rates used were varied in order to obtain spatial
and H* are the adsorbed species. If all the above reactions are
time velocities, W/FNH3,0, between 380 and 13 920 g s1 mol1
equilibrated, the constants of equilibrium can be expressed in
NH3. Under these conditions, it was experimentally verified
terms of concentration of adsorbed species as:
that the possible influence of mass or heat transfer limitations
(both internal and external) was minimum. k00 ½NH3 
K0 ¼ ¼ (3)
Table 1 shows the experimental conditions used in the k00 pNH3 ½
study. The reaction temperature was varied between 573
and 973 K. Previous to reaction, the catalyst was reduced k01 ½NH2  ½H
K1 ¼ ¼ (4)
‘‘in situ’’ at 823 K during 1 h, at a heating rate of 10 K min1, k01 ½NH3  ½

This journal is c the Owner Societies 2013 Phys. Chem. Chem. Phys.
View Article Online

Paper PCCP

k02 ½NH ½H Now, the values of K0 and K1 will be given by the following
K2 ¼ ¼ (5)
k02 ½NH2  ½ expressions:

k03 ½N ½H k00 ½NH3 2 


K3 ¼ ¼ (6) K0 ¼ ¼ , ½NH3 2  ¼ K0 pNH3 ½2 (13)
k03 ½NH ½ k00 pNH3 ½2

k0H2 ½H2 k01 ½NH2  ½H ½


KH2 ¼ ¼ (7) K1 ¼ ¼ , ½NH2  ¼ K1 K0 pNH3 (14)
k0H2 pH2 ½2 k01 ½NH3 2  ½H

k0N2 ½N2 For this mechanism the balance of active sites is given by:
KN2 ¼ 0 ¼ (8)
k N2 pN2 ½2 [L] = [*] + [H*] + [N*] + [NH*] + [NH2*] + [NH3*2] (15)

If it is supposed that the desorption on N*, stage (vi), is slow but Therefore, and irrespective of the rate determining step of
Published on 11 April 2013 on http://pubs.rsc.org | doi:10.1039/C3CP50715G

reversible, and the rest of the stages attains the equilibrium the mechanism, it is clear that the balance of active sites will
(named Case A), then eqn (8) does not apply and the reaction lead to a quadratic equation, and the solution for the reaction
rate will be given by: rate is substantially more complex38,42 and different from the
 
Downloaded by Ball State University on 20/05/2013 10:36:48.

rNH3 ¼ k0N2 ½N2 k0N2 pN2 ½2 (9) usual solutions.


The following paragraphs describe the kinetics equation
In the case that stage (ii) is the controlling step (named Case B), obtained in each case.
then eqn (4) does not apply and the reaction rate is: Case A: Considering the balance of active sites for the initial
  mechanism, eqn (11), the solution of this equation allows us to
rNH3 ¼ k01 ½NH3  ½  k01 ½NH2  ½H (10)
attain the expression of the fractional coverage of each species
Finally, in Case C, it is considered that both steps, for the three cases. Thus, for Case A the fractional coverage of
(ii) and (vi), are slow and therefore eqn (4) and (8) do not vacant sites can be calculated as:

½ 1
y ¼ ¼ !!!! (16)
½L pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi K1 K2 K3
1 þ KH2 pH2 þ K0 pNH3 1 þ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 þ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 þ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
KH2 pH2 KH2 pH2 KH2 pH2

apply. Their rates will be equal, fulfilling the ensuing From this equation, the fractional coverage of all the
condition:18,25,36 adsorbed species can be easily calculated as:
 
rNH3 ¼ k01 ½NH3  ½  k01 ½NH2  ½H ½NH3 
(11) yNH3  ¼ ¼ K0 pNH3 y (17)
½L
¼ k0N2 ½N2 k0N2 pN2 ½2
In any case, considering all competitively chemisorbing species,
½NH2  K0 K1 pNH3
i.e. reactants, products, and intermediates, the balance of active yNH2  ¼ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi y (18)
½L KH2 pH2
sites will be given by:

[L] = [*] + [H*] + [N*] + [NH*] + [NH2*] + [NH3*] (12)


½NH K0 K1 K2 pNH3
yNH ¼ ¼ y (19)
The term [L] represents the total concentration of active sites ½L KH2 pH2
(mol a.s. per g cat).
Previous to obtaining the final expression for the reaction ½N K0 K1 K2 K3 pNH3
rate in these cases, it is interesting to consider an alternative yN ¼ ¼ y (20)
½L KH23=2 pH23=2
mechanism where the ammonia adsorption occurs simulta-
neously over two sites, without direct dissociation of the
adsorbed ammonia molecule. The scission of the N–H bond ½H pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
yH ¼ ¼ KH2 pH2 y (21)
occurs in the next step, according to the following sequence: ½L
k00 ;k00
(i) NH3 þ 2 ! NH3 2
Finally, substituting the above expressions in eqn (9), the
k01 ;k01
(ii) NH3 2 ! NH2 þ H rate of ammonia decomposition for this case is:
 2 2 2 2  
K0 K1 K2 K3 pNH32 1 pN2 pH23
kA 1 
K H2 3 pH23 Keq pNH32
ðrNH3 Þ ¼ !!!!2 (22)
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi K1 K2 K3
1 þ KH2 pH2 þ K0 pNH3 1 þ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 þ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 þ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
KH2 pH2 KH2 pH2 KH2 pH2

Phys. Chem. Chem. Phys. This journal is c the Owner Societies 2013
View Article Online

PCCP Paper

In this expression, the kinetics constants of the direct and surface, in accordance with the Langmuir hypothesis.30,31 In
reverse reactions follow an Arrhenius dependence with the this regard, and as an alternative model, Deshmukh et al.,36
temperature, and are defined by: have considered a linear decrease of the activation energies
with the N* coverage, assuming, in fact, the Temkin model.30,31
1
kA ¼ kA0 expðEA =RT Þ; kA0 ¼ k0N2 ;0 ½L2 (23) We have also considered that the fractional coverage of all
2
the adsorbed species is kinetically relevant. However, as has
1 been discussed before, the presence of ‘‘a most abundant
kA ¼ kA0 expðEA =RT Þ; kA0 ¼ k0N2 ;0 ½L2 (24) specie’’ or species (MARI) is usually considered. For example,
2
if it is supposed that the covered surface is mainly occupied by
The equilibrium constant, Keq, of the ammonia decomposi- N* and H*, the balance of active sites is now given by:
tion reaction, 2NH3 3 N2 + 3H2, can be calculated from the
thermodynamic data:43 L = [*] + [H*] + [N*] (36)
   
KNH32 pN2 pH23 DG This assumption is equivalent to considering that stages (i)
Published on 11 April 2013 on http://pubs.rsc.org | doi:10.1039/C3CP50715G

Keq ¼ ¼ ¼ exp  (25)


KN2 KH23 pNH32 eq RT to (iv) can be represented by the sum of the quasi-equilibrated
steps:
In the interval from 673 K to 1273 K, the value of DG can be
½N ½H3
Downloaded by Ball State University on 20/05/2013 10:36:48.

estimated using this correlation: NH3 þ 4 , N  þ 3ðHÞ; KNH3 ¼ (37)


pNH3 ½4
2 6 3
DG = 95 117  193.67T  0.035293T + 9.22  10 T (26)
The fractional coverage of the significant species, [*], [H*]
Under the operating conditions used in this study, the term
  and [N*], can be calculated as:
pN2 pH23 Keq pNH32 in eqn (22) is virtually zero, indicating that
the reaction is very far from the equilibrium, and therefore this 1
y ¼   (38)
term can be neglected. Under these conditions, the reaction pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi KNH3 pNH3
1 þ KH2 pH2 þ
rate can be re-written as: KH23=2 pH23=2

  kA KNH32 pNH32
rNH3 ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi!!!!2 (27)
 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi K H 2
pH 2
K H 2
p H 2
KH2 pH2
KH23=2 pH23=2 1 þ KH2 pH2 þ KNH3 pNH3 1 þ 1þ 1þ
K3 K2 K1

In this equation the term KNH3 is a lumped parameter


defined as: pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
yH ¼ KH2 pH2 y (39)
KNH3 = K0K1K2K3 (28)
 
and the equilibrium constants, defined in eqn (3)–(7), vary with KNH3 pNH3
yN ¼ y (40)
the temperature according to the Van’t Hoff equation: KH23=2 pH23=2

K0 ¼ K00 expðQ0 =RT Þ (29) Finally, substituting these equations in eqn (9), and
considering again that the reaction occurs far from the equili-
K1 ¼ K10 expðQ1 =RT Þ (30) brium, the reaction rate for this simplified case, named
Case A-1, is given by:
K2 ¼ K20 expðQ2 =RT Þ (31)
  kA KNH32 pNH32
rNH3 ¼   pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2 (41)
K3 ¼ K30 expðQ3 =RT Þ (32) KNH3 pNH3 þ KH23=2 pH23=2 1 þ KH2 pH2
 
KH2 ¼ KH0 2 exp QH2 =RT (33) In a further simplification, if it is assumed that the adsorbed
N* is the MARI, the balance of active sites can be stated as:
Consequently, KNH3 varies with the temperature as follows:
  L = [*] + [N*] (42)
0
KNH3 ¼ KNH 3
exp QNH3 =RT (34)
Now, all the single steps, except the desorption of N*, can be
0
where KNH3
and QNH3 are calculated as: expressed as a sole lumped equilibrium:
0
KNH3
¼ K00 K10 K20 K30 ; QNH3 ¼ Q0 þ Q1 þ Q2 þ Q3 (35) 3 ½NpH23=2
NH3 þ  , N  þ H2 ; KT ¼ (43)
2 pNH3 ½
The values of the Qi appearing in the above equations are
actually the difference in the activation energies of the elemen- The term KT is defined by:
tary reactions involved in each step. It has been considered that
these values are independent of the degree of coverage of the KT ¼ KT0 expðQT =RT Þ (44)

This journal is c the Owner Societies 2013 Phys. Chem. Chem. Phys.
View Article Online

Paper PCCP
. . pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
KT ¼ K0 K1 K2 K3 K H2 3=2 ¼ KNH3 K H2 3=2 (45) yH ¼ KH2 pH2 y (55)

3=2
KT0 ¼ KNH
0
KH0 2 ; QT ¼ QNH3  ð3=2ÞQH2 (46) Developing the equations as in Case A, the expression is
3
obtained for the reaction rate in this Case B:
  kB K0 pNH3
rNH3 ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi!!!2 (56)
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi KH2 pH2 KH2 pH2
1 þ K0 pNH3 þ KH2 pH2 þ KN2 pN2 1 þ 1þ
K3 K2

For this case, named Case A-2, the fractional coverages are The kinetic constant for this case, kB, is given by:
given by:
kB ¼ kB0 expðEB =RT Þ; kB0 ¼ k01 ½L2 (57)
3=2
pH
y ¼  3=2 2  (47)
pH2 þ KT pNH3
Published on 11 April 2013 on http://pubs.rsc.org | doi:10.1039/C3CP50715G

As in Case A, similar simplified cases can be developed,


assuming the existence of species acting as MARI. For example,
KT pNH3 if the surface is mainly covered by NH3*, N* and H*, the
yN ¼  3=2  (48)
pH2 þ KT pNH3
Downloaded by Ball State University on 20/05/2013 10:36:48.

equation obtained is:

Using eqn (42), (47) and (48), and remembering that the   kB K0 pNH3
rNH3 ¼  pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2 (58)
reaction occurs far from the equilibrium, the reaction rate, 1 þ K0 pNH3 þ KH2 pH2 þ KN2 pN2
eqn (9) is converted into:

  kA KT2 pNH32 Similarly to these Cases A and B, and following the same
rNH3 ¼  2 (49) methodology, several new sets of equations (i.e. new cases),
pH23=2 þ KT pNH3
corresponding to the situations in which each one of the
This equation has already been deduced previously, remaining steps is the controlling stage, could be straight-
Bradford et al.,18 Gjéga-Mariadassou et al.,25 Chellappa forwardly developed.
et al.,23 considering that the assumptions taken here are However, an alternative approach is to consider that none of
fulfilled for reactor operation at high hydrogen pressure. Under the elementary reactions is taken, a priori, as controlling. This
these conditions, rehydrogenation of nitride adsorbed species situation is considered in the development of microkinetic
in ammonia can occur, this fact explaining the inhibitory effect models.33,36,45,46
of hydrogen. However, in the previous cases, Case A and Case Case C: as has been explained before, in this case the
A-1, the inhibition by hydrogen can also be explained by the reaction rate is determined by eqn (11) which in terms of
competition of hydrogen for the active sites of the catalyst fractional coverages is:
surface, showing that the hydrogen surface coverage is not  
rNH3 ¼ k1 yNH3  y  k1 yNH2  yH ¼ kN2 yN 2  kN2 pN2 y2
negligible.44
Case B: in this case, it is assumed that stage (ii) is the (59)
controlling step. Consequently, eqn (4) does not apply and the
reaction rate is given by eqn (10). In this case, the fractional As in the previous cases, the kinetic constants are
coverages are calculated as: defined as:

1
y ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi!!! (50)
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi KH2 pH2 KH2 pH2
1 þ K0 pNH3 þ KH2 pH2 þ KN2 pN2 1 þ 1þ
K3 K2

yNH3  ¼ K0 pNH3 y (51) k1 ¼ k01 ½L2 ; k1 ¼ k01 ½L2


(60)
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 1
KH2 pH2 KN2 pN2 kN2 ¼ k0N2 ½L2 ; kN2 ¼ k0N2 ½L2
yNH2  ¼ y (52) 2 2
K2 K3
Taking into account the balance of active sites, eqn (12),
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffipffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
KH2 pH2 KN2 pN2 together with eqn (3), (5), (6) and (7), the linear relationship
yNH ¼ y (53) between y* and yN* can be deduced:
K3

pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1  ly
yN ¼ (61)
yN ¼ KN2 pN2 y (54) k

Phys. Chem. Chem. Phys. This journal is c the Owner Societies 2013
View Article Online

PCCP Paper

The chemisorption terms, l and k, of the above equation are As a first particular case, named Case C-1, step (vi) can be
defined as: considered irreversible, then kN2 = 0; d = KN2 = 0, and the
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi! reaction rate is now:
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi KH2 pH2 KH2 pH2
l ¼ 1 þ K0 pNH3 þ KH2 pH2 ; k ¼ 1 þ 1þ  
K3 K2   1  ly 2
rNH3 ¼ kN2 yN 2 ¼ kN2 (73)
k
(62)

Substituting eqn (61) in eqn (59), the following quadratic In a second simplification, it is also assumed that step (ii) is
expression is obtained for y*: also irreversible, i.e. k1 = 0, b = 0. If, in addition, it is further
considered that H* and N* are the dominant species at the catalyst
my*2  ny*  p = 0 (63) pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
surface, it is fulfilled that l ¼ 1 þ KH2 pH2 and k = 1. Under all
where the coefficients m, n and p are defined by: these assumptions, the reaction rate for Case C-2 is given by:
      k1 K0 pNH3 k1
l l2 b l c rNH3 ¼  pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2 ; KC ¼ (74)
m¼ aþd þb c 2 ; n¼  2c 2 ; p ¼ 2
Published on 11 April 2013 on http://pubs.rsc.org | doi:10.1039/C3CP50715G

(64) k N2
k k k k k 1 þ KH2 pH2 þ KC K0 pNH3

and the coefficients a, b, c and d are determined by the kinetic Finally, if the MARI is N*, Case C-3, the reaction rate is
Downloaded by Ball State University on 20/05/2013 10:36:48.

constants of the steps (ii) and (vi): simplified to:


 3=2  
KH2 pH2 k1 K0 pNH3
a ¼ k1 K0 pNH3 ; b ¼ k1 rNH3 ¼  pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2 (75)
K2 K3 (65) 1þ KC K0 pNH3
c ¼ k N2 ; d ¼ kN2 pN2
This equation has been previously used by Bradford et al.,18
Gjéga-Mariadassou et al.,25 Chellappa et al.,23 to explain the
Taking the positive root of eqn (63), the value of y* is results obtained at low hydrogen pressure. Under these condi-
calculated as: tions, the concentration of hydrogen is not sufficient to allow
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi! the reverse reaction of step (ii), and the specie NH2* cannot be
n þ n2 þ 4mp
y ¼ (66) rehydrogenated to form NH3.25 However, it is worth remember-
2m
ing that eqn (75) is a very simplified case of eqn (71), in which
both steps are considered as reversible, and all the adsorbed
Using now eqn (3), (5)–(7) and (12), the rest of the fractional
species are, in theory, kinetically relevant.
coverages can be calculated as:

yNH3* = K0pNH3y* (67)


4. Results and discussion
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Taking into account that the experimental data have been obtained
yH ¼ KH2 pH2 y (68)
in an integral reactor, the kinetic parameters of all the models
  tested have been estimated by non-linear multivariable regression
KH2 pH2 KH2 pH2 1  ly
yNH2  ¼ yN ¼ (69) using the Levenberg–Mardquardt algorithm, coupled to a Runge–
K2 K3 K2 K3 k
Kutta–Fehlberg routine. This integration routine is used to solve
 3=2  3=2   the steady-state ammonia mass balance in a plug-flow reactor:
KH2 pH2 KH2 pH2 1  ly
yNH ¼ yN ¼ (70) dXNH3  
K3 K3 k .
¼ rNH3 (76)
0
d Wcat FNH3
Finally, the general expression for the reaction rate in
0
Case C is: XNH3 is the ammonia conversion, Wcat/FNH 3
is the spatial
  time (g cat s per mol NH3) and rNH3 is given by the different
  1 pH23=2 yN
rNH3 ¼ k1 K0 pNH3 y2 1  expressions deduced for each model to be tested. Knowing the
KT pNH3 y feed composition, the numerical solution of eqn (76) allows the
 2 ! (71)
calculation of the ammonia conversion, and therefore the partial
2 y
¼ kN2 yN 1  KN2 pN2 pressures of NH3, H2 and N2, which can then be compared with
yN
the experimental values. The minimized objective function was
the sum of the squared residuals (SSR):
In this expression, the terms K1, KN2 and KT are calculated, i¼n
2
X exp
respectively, as: O:F: ¼ SSR ¼ min XNH calc
 XNH (77)
3 3
i¼1
k1 kN2 K0 K1 K2 K3 KNH3
K1 ¼ 4 1; KN2 ¼ o 1; KT ¼ ¼
k1 k N2 KH23=2 KH23=2 With the aim of selecting the best model, and considering
(72) that the models proposed have a different number of parameters,

This journal is c the Owner Societies 2013 Phys. Chem. Chem. Phys.
View Article Online

Paper PCCP

the statistical discrimination between them has been made must be for the model to be deemed more appropriate. Thus,
using the Model Selection Criterion (MSC), defined as:47,48 the model with the higher MSC is selected as the most appro-
    priate. In contrast, the best model will be the one that provides
SST 2p
MSC ¼ ln  (78) the lower value of the Variance of error.
SSR n
Tables 2–7 show the kinetic parameters obtained with all
where p and n are the number of parameters and of experi- the models tested, including the power-law model. All these
mental points, respectively. The term SST is the sum of total tables include the values of the parameters, the standard
squares, defined as: error of each parameter, and the lower (L.L.) and upper limits
i¼n
X
2 (U.L.) of the confidence interval at the 95% probability level.
exp exp
SST ¼ XNH  XNH (79) Table 8 shows a comparison of the main statistical parameters
3 3
i¼1 used for the discrimination of the models used. The results
shown in these tables indicate that the best model is
In addition, the variance of the error has been calculated
that named as Case A (see Table 2) which considers that
according to the following expression:
Published on 11 April 2013 on http://pubs.rsc.org | doi:10.1039/C3CP50715G

vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi the RDS is the desorption of adsorbed nitrogen atoms. How-


uiP
2 ever, this case also assumes that all the species present at the
u ¼n exp calc
u XNH  X NH catalyst surface are kinetically relevant, a key difference with
t 3 3
VarðSSRÞ ¼ i¼1 (80)
Downloaded by Ball State University on 20/05/2013 10:36:48.

np the usually assumed hypothesis that the only one significant
specie is N*.
This parameter also takes into account the different number Tables 3 and 4 include the results obtained with the parti-
of parameters of each model. When comparing two models cular Cases A-1 and A-2. Furthermore, as can be seen in Tables 5
with different numbers of parameters, the Model Selection and 6, the results obtained with the models corresponding to
Criterion not only places a burden on the model with more Cases B and C clearly show that both models are substantially less
parameters to have a better coefficient of determination, i.e. a appropriate than that of Case A. Both models give several para-
lower value of SSR, but it also quantifies how much better it meters without statistical validity, and therefore must be rejected.

Table 2 Kinetic parameters corresponding to Case A

Parameter Value Standard error L.L. (95%) U.L. (95%)


1 4 4 4
kAm (mol per g cat s ) 7.062  10 1.128  10 5.935  10 8.190  104
EA (J mol1) 86 369.4 10 716.8 75 652.5 97 086.2
K0m (atm1) 3.230  101 1.117  101 2.113  101 4.346  101
Q0 (J mol1) 111 003.1 30 733.7 80 269.5 141 736.8
K1m (atm1) 7.125  102 4.025  102 3.100  102 1.115  101
Q1 (J mol1) 17 674.9 2087.9 15 587.0 19 762.8
K2m (atm1) 3.642  102 1.353  102 2.289  102 4.995  102
Q2 (J mol1) 3212.2 1318.3 1893.9 4530.5
K3m (atm1) 4.648  101 2.421  101 2.227  101 7.068  101
Q3 (J mol1) 1256.1 93.0 1163.2 1349.1
KH2m (atm1) 1.437  102 2.151  103 1.222  102 1.653  102
QH2 (J mol1) 136 909.5 16 728.7 120 180.8 153 638.2

Table 3 Kinetic parameters corresponding to Case A-1

Parameter Value Standard error L.L. (95%) U.L. (95%)


1 4 6 4
kAm (mol per g cat s ) 3.615  10 1.014  10 3.595  10 3.635  104
Ea2 (J mol1) 113 419.1 1803.0 109 855.7 116 982.5
KNH3m (atm1) 6.293  104 2.075  105 5.883  104 6.704  104
QNH3 (J mol1) 67 943.3 4958.4 58 143.8 77 742.8
KH2m (atm1) 3.937  103 1.766  104 3.588  103 4.286  103
QH3 (J mol1) 124 019.1 4409.4 115 304.7 132 733.5

Table 4 Kinetic parameters corresponding to Case A-2

Parameter Value Standard error L.L. (95%) U.L. (95%)


k2,m (mol g1 s1) 3.712  104 7.232  106 3.569  104 3.855  104
(Ea)ap (kJ mol1) 111 743.2 2176.6 107 442.0 116 044.4
KT,m (atm1/2) 2.370 0.115 2.143 2.598
QT (kJ mol1) 120 370.2 5812.0 131 855.5 108 884.9

Phys. Chem. Chem. Phys. This journal is c the Owner Societies 2013
View Article Online

PCCP Paper

Table 5 Kinetic parameters corresponding to Case B

Parameter Value Standard error L.L. (95%) U.L. (95%)


1 1
kBm (mol g s ) 0.449 41.716 41.267 42.165
EB (J mol1) 23 237.9 80 237.9 57 000.0 103 475.8
K0m (atm1) 172.4 16 058.9 15 886.5 16 231.3
Q0 (J mol1) 0.0 6278.4 6278.4 6278.4
K2m (atm1) 8.33  103 5603.2 5603.2 5603.2
Q2 (J mol1) 69 421.0 1.353  1010 1.353  1010 1.353  1010
K3m (atm1) 3.5 3.007  106 3.007  106 3.007  106
Q3 (J mol1) 65 133.6 6.217  109 6.217  109 6.217  109
KH2m (atm1) 2.793  105 5.152  107 5.124  107 5.180  107
QH2 (J mol1) 254 517.8 148 870.9 105 646.9 403 388.7
KN2m (atm1) 1.097  108 4.146  103 4.146  103 4.146  103
QN2 (J mol1) 27 619.3 1.714  1010 1.714  1010 1.714  1010
Published on 11 April 2013 on http://pubs.rsc.org | doi:10.1039/C3CP50715G

Table 6 Kinetic parameters corresponding to Case C

Parameter Value Standard error L.L. (95%) U.L. (95%)


Downloaded by Ball State University on 20/05/2013 10:36:48.

1 1
kCm (mol g s ) 1.295 1.144 0.151 2.439
E1 (J mol1) 104 429.5 270 937.9 166 508.4 375 367.5
k1rm (mol g1 s1) 0.301 0.345 0.044 0.646
E1r (J mol1) 1.2 3.9 2.6 5.1
k6m (mol g1 s1) 5.183  104 1.496  104 3.687  104 6.680  104
E6 (J mol1) 101 426.9 19 749.9 81 677.0 121 176.9
K0m (atm1) 0.459 0.633 0.174 1.092
Q0 (J mol1) 72 005.7 147 915.9 75 910.2 219 921.6
K2m (atm1) 7.396  102 6.539  102 8.569  103 1.394  101
Q2 (J mol1) 1.5 3.1 1.6 4.6
K3m (atm1) 0.659 1.958 1.299 2.617
Q3 (J mol1) 2.6 1.6 1.0 4.1
KH2m (atm1) 8.403  103 1.869  102 1.029  102 2.709  102
QH2 (J mol1) 99 271.6 165 648.9 66 377.2 264 920.5

Table 7 Kinetic parameters corresponding to the power-law model Fig. 1 and 2 present the results of the fitting obtained with
the model corresponding to Case A. The parity-plot in Fig. 1
Parameter Value Standard error L.L. (95%) U.L. (95%)
indicates that Model A gives a homoscedastic distribution of
kPm a 1.556  104 8.817  106 1.381  104 1.730  104
the error along the whole range on ammonia conversions.
EP (J mol1) 202 835.3 3461.3 195 996.0 209 674.5
a 0.73 0.02 0.69 0.76 The excellent correspondence between the experimental and
b 0.64 0.02 0.67 0.60 predicted values clearly shows that the Model A is very robust
a
The units of kPm are (mol per g cat s atm(a+b)). to predict conversions higher that 95% over a wide range of
0
Wcat/FNH 3
values and operating temperatures (see Table 1).

Table 8 Statistical comparison between the models

Model p Variance error SSR MSC R2


Case A 12 0.0239 0.0816 5.0725 0.9946
Case A-1 6 0.0351 0.3783 3.6159 0.9751
Case A-2 4 0.0500 0.3805 3.6621 0.9749
Case B 12 0.0320 0.1457 4.4921 0.9904
Case C 14 0.0635 0.5646 3.1116 0.9629
Power-law 4 0.0477 0.3409 3.7468 0.9776

In principle, these results indicate that the N–H bond scission


of the adsorbed ammonia is not the controlling stage, attaining
this stage the equilibrium. The intrinsic energy of activation
calculated with Model A is 86.4 kJ mol1. This result is quite
close to that recently presented by Vilekar et al.,33 who also
assumed the same controlling step as that of Case A. With
respect to the values of Qi included in Table 2, these results are
in agreement with estimations carried out in the development
of microkinetic models.33 Fig. 1 Parity-plot of the calculated vs. experimental ammonia conversion.

This journal is c the Owner Societies 2013 Phys. Chem. Chem. Phys.
View Article Online

Paper PCCP
Published on 11 April 2013 on http://pubs.rsc.org | doi:10.1039/C3CP50715G

Fig. 2 Influence of the hydrogen concentration on the evolution of ammonia


conversions. The data correspond to experiments 10 to 13 (Table 1) and the solid
Downloaded by Ball State University on 20/05/2013 10:36:48.

lines correspond to Case A.

Fig. 2 shows several examples of XNH3 vs. temperature data fittings,


and how the model satisfactorily accounts for the inhibitory effect
of hydrogen in the whole range of temperatures studied.
With the kinetic parameters presented in Tables 2–4, i.e.
those corresponding to Cases A, A-1 and A-2, it is possible to
calculate the fractional coverages of each species at the surface.
Thus, for Case A-1, Fig. 3a–c display the evolution of yN*, yH*
and y* with the temperature for several conversions in the
reactor. In these figures, each vertical line that can be drawn at
a given temperature will represent an isothermal operation.
These results indicate that, at low temperatures, the fraction of
sites covered by hydrogen is even higher than that covered by
nitrogen atoms. As the temperature increases, the hydrogen
coverage decreases and the catalyst surface becomes covered by
nitrogen. Thus, results presented in Fig. 3c, clearly indicate that
at high conversions, ca. 99.5%, the surface is quite bare
attaining values of y* higher than 0.9, particularly at tempera-
tures above 800 1C. In addition, results in Fig. 3b indicate that
at low temperature range (o750 1C), and at ammonia conver-
sions, above 40%, hydrogen coverage was greater than 20%.
These results show that the assumption that N* is the MARI can
be an erroneous oversimplification.
Fig. 4a and b present, on a logarithmic scale, the results
obtained with the general Case A for two levels of ammonia
conversion (ca. 60% and 99.5%, respectively). The case takes into
account all the species and therefore the information attained is Fig. 3 Influence of temperature and ammonia conversion on the evolution of
yN* (a), yH* (b), y* (c). Calculations have been done for Case A-1.
more relevant than that obtained with the simplified cases. In
fact, it can be seen that depending on the reaction temperature
or on the conversion attained, i.e. on the position inside the the kinetic data. In this regard, Table 7 shows the results
reactor, the distribution of surface coverages changes dramati- obtained with the power-law rate model. These results can be
cally. Thus, at 60% of conversion, most of the surface is written in terms of a rate equation as:
unoccupied, and hydrogen is the most abundant of the adsorbed
  
species. On the other hand, at 99.5% of conversion, NH3* is the   8 202835 pNH30:73
rNH3 ¼ 4:859  10 exp  (81)
dominant specie at low temperatures, but at high temperatures RT pH20:64
the most abundant specie is the adsorbed nitrogen.
The above results indicate the necessity of using mecha- The apparent activation energy shown in this equation,
nistically based models to obtain a realistic interpretation of 202.3 kJ mol1, is in very good accordance with the value

Phys. Chem. Chem. Phys. This journal is c the Owner Societies 2013
View Article Online

PCCP Paper
Published on 11 April 2013 on http://pubs.rsc.org | doi:10.1039/C3CP50715G

Fig. 4 Influence of temperature on the evolution of the fractional coverages calculated with Case A. (a) NH3 conversion = 0.6; (b) NH3 conversion = 0.995.
Downloaded by Ball State University on 20/05/2013 10:36:48.

reported by Gjéga-Mariadassou et al.25 However, the ratio of the step of the reaction.17 Nevertheless, many of these changes can
kinetic orders, a/b, is not in agreement with the prediction of be ascribed to the method of analysis of the data usually used,
the Temkin formalism.29 As has been discussed previously, the implying that they can be explained directly by the equations
Temkin–Pyzhev model has been widely used to study the previously derived for each mechanism without considering
mechanism and the kinetic behaviour of numerous catalysts used any intrinsic variation in the reaction mechanism.
in this reaction. As a consequence of these studies, it has been stated The kinetic parameters of the power-law model, kP (i.e. Eapp),
that at high temperatures and low ammonia partial pressures, the a and b are generally calculated from the Arrhenius plot, and
reaction is of first-order with respect to ammonia and of zero-order from the log–log plots of rate versus ammonia and hydrogen
with respect to hydrogen.23 Likewise, working at high hydrogen concentration. Subsequently, to formally derive these orders,
partial pressures and low temperatures, i.e. when hydrogen inhibi- the following logarithmic-derivatives must be applied:49,50
tion is substantial, the kinetic order with respect to hydrogen, b in     
d ln rNH3 d r p
eqn (2) is negative. Depending on the operating conditions, a and b a¼    ¼  NH3  NH3 ; pH2 ¼ const: (82)
d ln pNH3 d pNH3 rNH3
can vary from 0 to 1, and from 0 to 2 respectively.18 Thus, Table 9
presents the results obtained with the power-law model for each     
d ln rNH3 d r p
individual experiment. The variability observed in all the parameters b¼    ¼  NH3  H2 ; pNH3 ¼ const: (83)
is a consequence of the different experimental conditions used in d ln pH2 d pH2 rNH3
each experiment. The results in Table 7, where all the experiments
are fitted simultaneously, are averages. Furthermore, associated with Likewise, and taking into account the Arrhenius equation,
the changes in the kinetic order, variations in the apparent activation a similar approach may be applied to calculate the apparent
energy of the reaction have also been reported. activation energy:
    
These facts have been explained assuming modifications of d ln rNH3 d rNH3 1
ðEa Þap ¼  ¼   (84)
the reaction mechanism, for example, a change in the controlling dðVTÞ dðVTÞ rNH3

In the above expression, the variable VT is defined with reference


Table 9 Kinetic parameters of the power-law model for each individual to the temperature of reparametrization, Tm, as follows:
experiment ,
ðT m  T Þ X
i¼n

Experiment kP0 a Ea (kJ mol1) a b a/b VT ¼ ; Tm ¼ Ti exp n (85)


RTm T i¼1
20
1 7.41  10 380.0 2.0 2.5 0.81
2 4.73  1010 235.7 0.9 2.4 0.39 In our case, the value of Tm used in all the calculations is
3 1.47  1012 259.0 1.1 2.3 0.48
4 3.33  104 167.2 0.1 1.3 0.10 848 K. In terms of the variable VT, the Arrhenius equation can
5 2.17  1011 249.4 0.8 1.7 0.49 be written now as follows:
6 1.03  1017 338.8 1.7 1.7 0.99
7 2.74  109 215.3 1.0 0.0 — kP = kPm exp(EappVT) (86)
8 7.12  1016 347.4 1.2 2.3 0.52
9 2.09  107 176.0 0.9 0.6 1.51 where kPm is the kinetic constant evaluated at Tm:
10 1.97  109 215.3 0.7 1.0 0.70
11 4.09  1018 379.6 1.8 2.4 0.77 kPm = kP0 exp(Eapp/RTm) (87)
12 2.07  109 226.5 0.3 1.8 0.17
13 3.69  105 149.8 0.5 0.4 1.13 As an example, we present here only the equations deduced
a
The units of kP0 are (mol per g cat s atm(a+b)). for the Case A-1. The rest of the cases can be deduced in a

This journal is c the Owner Societies 2013 Phys. Chem. Chem. Phys.
View Article Online

Paper PCCP

similar manner. Thus, substituting the reaction rate for this


Case A-1, eqn (41), into eqn (80) and (81), the following
expressions are obtained for a and b, respectively:
 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 1 þ K H2 pH2
a¼   ¼ 2ðy þ yH Þ ¼ 2ð1  yN Þ
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi K NH3 pNH3
1 þ K H2 pH2 þ
KH23=2 pH23=2
(88)
 
4pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
3 1þ K H2 pH2
3
b ¼  
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi K NH3 pNH3
1 þ K H2 pH2 þ (89)
KH23=2 pH23=2

¼  ð3y þ 4yH Þ ¼ y  4ð1  yN Þ


Published on 11 April 2013 on http://pubs.rsc.org | doi:10.1039/C3CP50715G

These two expressions clearly demonstrate that the reaction


orders are not constant, but depend on the experimental
Downloaded by Ball State University on 20/05/2013 10:36:48.

conditions as can be also seen at Fig. 5. In addition, and taking


into account eqn (38)–(40), it is shown that these parameters
can be calculated as a function of surface coverages.50 Therefore,
if the reactor operates at high conversions, the gas composition
will vary greatly from the inlet to the outlet and therefore the
degree of coverage of the catalyst surface will change with the
position along the reactor. For differential operation, i.e. low
ammonia conversions, the changes in the feed composition
and/or in the reaction temperature will provide estimations of
the changing values of the kinetic orders.
The Temkin–Pyzhev model predicts that the ratio a/b is
3/2.27,29 According to eqn (87) and (88), for the model corre-
sponding to the Case A-1 this ratio is given by:
 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
a 2 1 þ K H2 pH2
¼   (90)
b 4pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
3 1þ K H2 pH2
3
Fig. 5 Influence of temperature and ammonia conversion on the evolution of
It can be easily demonstrated that for the Case A-2, i.e. when the kinetic orders calculated with Case A-1. (a) NH3/H2 = 0.99/0.01; (b) NH3/H2 =
the MARI is the specie N*, the value of a/b is exactly 3/2. 0.233/0.767.

Therefore, Case A-2 is a limiting case of Case A-1. However, it


must be recalled that both are particular cases of the general
Case A, in which all the species are considered as kinetically
relevant. In addition, it is interesting to note that Case A-2
provides the same value for the ratio a/b as the Temkin–Pyzhev
model, which also assumes that MARI is the nitrogen adsorbed.
In order to calculate the apparent activation energy, the
application of eqn (82) to eqn (41) gives the following result:
 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi  pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
QH2 3  4 K H2 pH2  2QNH3 1 þ K H2 pH2
ðE1 Þapp ¼ E1 þ  
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi K NH3 pNH3
1 þ K H2 pH2 þ
KH23=2 pH23=2
(91)

As in the calculation of a and b, the apparent activation


energy may be interpreted in terms of surface coverage as:51
   
yH yH
ðE1 Þapp ¼ E1 þ QH2 3y þ  QNH3 2y þ (92)
2 2

These equations clearly demonstrate that the apparent Fig. 6 Influence of temperature and ammonia conversion on the evolution of
activation energy measured for a given set of experiments the apparent activation energy calculated with Case A. NH3/H2 = 0.99/0.01.

Phys. Chem. Chem. Phys. This journal is c the Owner Societies 2013
View Article Online

PCCP Paper

depends directly on the operating conditions used, and References


therefore on the fractional coverages attained, see Fig. 6.
In summary, eqn (88)–(92) are examples that indicate that 1 G. W. Crabtree, M. S. Dresselhaus and M. V. Buchanan,
the variation in the kinetic orders and/or the apparent acti- Phys. Today, 2004, 57, 39–44.
vation energy can be explained without considering any change 2 D. Sun, S. S. Srinivasan, G. Chen and C. M. Jensen, J. Alloys
in the reaction mechanism, or in the controlling step of the Compd., 2004, 373, 265–269.
reaction. 3 T. Vegge, Phys. Chem. Chem. Phys., 2006, 8, 4853–4861.
4 H. Furukawa, M. A. Miller and O. M. Yaghi, J. Mater. Chem.,
2007, 17, 3197–3204.
5. Conclusions 5 C. H. Christensen, R. Z. Sorensen, T. Johannessen,
U. J. Quaade, K. Honkala, T. D. Elmoe, R. Kohler and
The growing interest in the ammonia decomposition reaction
J. K. Norskov, J. Mater. Chem., 2007, 15, 4106–4108.
resulting from the perceived advantages of ammonia as a
6 G. P. G. Thomas, A Study of Issues Related to the Use Ammonia
hydrogen carrier is boosting research into basic aspects such
Published on 11 April 2013 on http://pubs.rsc.org | doi:10.1039/C3CP50715G

as the development of reliable kinetic models and the elucida- for On-Board Vehicular Hydrogen Storage, U. S. D. o. Energy,
tion of the reaction mechanism. 2006.
The main goal of this work is to develop mechanistically 7 A. T. Raissi, Proceedings of the 2001 DOE Hydrogen
Program Review, NREL/CP-570-30535, F. S. E. Center,
Downloaded by Ball State University on 20/05/2013 10:36:48.

based kinetic models able to predict the performance of


integral reactors where the key factor is to achieve almost Florida, 2002.
complete NH3 conversion with a high H2 concentration at the 8 P. L. Dulong and L. J. Thenard, Ann. Chim. Phys., 1823,
exit of the reactor. 23, 440.
The kinetic models developed here are based on the 9 F. Haber and G. Van Oordt, Z. Anorg. Chem., 1905, 43,
assumption of the Langmuir isotherm, considering that all 111–115.
the adsorbed species can be kinetically relevant, that the slow 10 S. Ichikawa, Chem. Eng. Sci., 1990, 45, 529–535.
step or steps can be partially reversible, and that the surface can 11 T. V. Choudhary, C. Sivadinarayana and D. W. Goodman,
be considered as energetically uniform, i.e. ideal. With the Catal. Lett., 2001, 72, 197–201.
results obtained it has been concluded that the controlling 12 S. F. Yin, B. Q. Xu, X. P. Zhou and C. T. Au, Appl. Catal., A,
step is the recombinative desorption on adsorbed nitrogen 2004, 277, 1–9.
atoms, and the rest of the elementary stages attain the chemical 13 V. Prasad, A. M. Karim, A. Arya and D. G. Vlachos, Ind. Eng.
equilibrium. However, the best model selected, named Case A, Chem. Res., 2009, 48, 5255–5265.
considers all the species to be kinetically relevant. In addition, 14 J. J. Vajo, W. Tsai and W. H. Weinberg, J. Phys. Chem., 1986,
this model, and its simplified cases, Case A-1 and Case A-2, are 90, 6531–6535.
also able to predict the results of the Temkin–Pyzhev model 15 D. G. Löffler and L. D. Schmidt, J. Catal., 1976, 41, 440–454.
that is based on the assumption of a non-ideal surface. 16 D. G. Löffler and L. D. Schmidt, Surf. Sci., 1976, 59, 195–204.
Among other relevant conclusions, the results obtained here 17 W. Tsai and W. H. Weinberg, J. Phys. Chem., 1987, 91,
indicate that the observed variation in kinetic orders and in the 5302–5307.
apparent activation energies frequently reported in the litera- 18 M. C. J. Bradford, P. E. Fanning and M. A. Vannice, J. Catal.,
ture, may be a consequence of the ambiguous method of data 1997, 172, 479–484.
analysis used, and not due to a true change in the controlling 19 R. W. McCabe, J. Catal., 1983, 79, 445–450.
step, but due to a variation in the operating conditions, mainly 20 D. G. Löffler and L. D. Schmidt, J. Catal., 1976, 44, 244–258.
the reaction temperature, or the hydrogen or ammonia con- 21 C. H. Kunsman, J. Am. Chem. Soc., 1929, 51, 688–695.
centrations. The equations developed, e.g. eqn (88)–(92), here 22 K. Tamaru, Acc. Chem. Res., 1988, 21, 88–94.
clearly demonstrate that the variations in the kinetic orders, 23 A. S. Chellappa, C. M. Fischer and W. J. Thomson, Appl.
and in the apparent activation energy, are direct consequences Catal., A, 2002, 227, 231–240.
of the different operating conditions used, and therefore of the 24 C. Egawa, T. Nishida, S. Naito and K. Tamaru, J. Chem. Soc.,
fractional coverages attained at the catalyst surface under these Faraday Trans. 1, 1984, 80, 1595–1604.
conditions. This does not necessarily mean that such variations 25 G. Djéga-Mariadassou, C. H. Shin and G. Bugli, J. Mol. Catal.
could not be due to any change in the reaction mechanism, but A: Chem., 1999, 141, 263–267.
the hypothesis should be carefully checked by other techniques 26 W. Tsai, J. J. Vajo and W. H. Weinberg, J. Am. Chem. Soc.,
such as, for example, spectroscopic techniques. 1985, 89, 4926.
27 P. H. Emmett and J. T. Kummer, Ind. Eng. Chem., 1943, 35,
Acknowledgements 677–683.
28 S. T. Oyama, J. Catal., 1992, 133, 358–369.
The authors acknowledge financial support from the Spanish 29 M. Temkin and V. Pyzhev, Acta Physicochim. URSS, 1940, 12,
Government (MAT 2008-02365, CTQ 2007-62545/PPQ), and 327–356.
from the regional government of Aragon (DGA-LACAIXA 30 Kinetics of Heterogeneous Catalytic Reactions, ed. M. Boudart
GA-LC-043/2010). and G. Djéga-Mariadassou, Princeton, Princeton, NJ, 1984.

This journal is c the Owner Societies 2013 Phys. Chem. Chem. Phys.
View Article Online

Paper PCCP

31 M. A. Vannice, Kinetics of Catalytic Reactions, Springer, 42 J. B. Butt and E. E. Petersen, Activation, Deactivation, and
New York, USA, 2005. Poisoning of Catalysts, Elsevier, 1988, pp. 27–61.
32 J. Zhang, H. Xu and W. Li, Appl. Catal., A, 2005, 296, 257–267. 43 Introduction to Chemical Engineering Thermodynamics, ed.
33 S. A. Vilekar, I. Fishtik and R. Datta, Chem. Eng. Sci., 2012, J. M. Smith, H. Van Ness and M. Abbott, The Mcgraw-Hill
71, 333–344. Chemical Engineering Series, United States, November 12,
34 S. Stolbov and T. S. Rahman, J. Chem. Phys., 2005, 2004 edn, 2004.
123, 204716. 44 L. O. Apel’baum and M. Temkin, Russ. J. Phys. Chem., 1959,
35 W. Huang, W. Lai and D. Xie, Surf. Sci., 2008, 602, 1288–1294. 33, 585.
36 S. R. Deshmukh, A. B. Mhadeshwar and D. G. Vlachos, Ind. 45 S. Appari, V. M. Janardhanan, S. Jayanti, L. Maier, S. Tischer
Eng. Chem. Res., 2004, 43, 2986–2999. and O. Deutschmann, Chem. Eng. Sci., 2011, 66, 5184–5191.
37 A. B. Mhadeshwar, J. R. Kitchin, M. A. Barteau and 46 J. A. Dumesic and A. A. Trevino, J. Catal., 1989, 116,
D. G. Vlachos, Catal. Lett., 2004, 96, 13–22. 119–129.
38 J. Corella, M. P. Aznar and A. Monzon, Int. Chem. Eng., 1989, 47 S. Sclove, Psychometrika, 1987, 52, 333–343.
Published on 11 April 2013 on http://pubs.rsc.org | doi:10.1039/C3CP50715G

29, 86–100. 48 M S Software, Statistical Analysis, Micromath, Salt Lake City,


39 C. Plana, S. Armenise, A. Monzón and E. Garcı́a-Bordejé, Utah (USA), 1995.
J. Catal., 2010, 275, 228–235. 49 I. Chorkendorff and J. W. Niemantsverdriet, in Concepts of
Downloaded by Ball State University on 20/05/2013 10:36:48.

40 C. Plana, S. Armenise, A. Monzón and E. Garcı́a-Bordejé, Modern Catalysis and Kinetics, ed. W.-V. V. G. Co, Strauss
Top. Catal., 2011, 54, 914–921. Offsetdruck, Morlenbach, 2003, p. 27.
41 W. L. Guthrie, J. D. Sokol and G. A. Somorjai, Surf. Sci., 1981, 50 P. Stoltze, Prog. Surf. Sci., 2000, 65, 65–150.
109, 390–418. 51 P. Stoltze and J. K. Nørskov, J. Catal., 1988, 110, 1–10.

Phys. Chem. Chem. Phys. This journal is c the Owner Societies 2013

You might also like