You are on page 1of 20

Topics in Catalysis 1 (1994) 233-252 233

Ammonia synthesis kinetics: surface chemistry,


rate expressions, and kinetic analysis
Luis M. Aparicio and James A. Dumesic
Department of Chemical Engineering, Universityof Wisconsin,
Madison, WI53706, USA

The results of surface science studies of nitrogen adsorption and desorption on iron single
crystals are summarized with respect to ammonia synthesis reaction kinetics. Analytical rate
expressions for ammonia synthesis on high surface iron catalysts at industrial reaction condi-
tions are presented and compared to the behavior of microkinetic models based on the surface
science results. Microkinetic models based on uniform surfaces are successfulin extrapolating
"results from surface science studies at low surface coverages to describe the performance of
iron catalysts at industrial ammonia synthesisreaction conditions. We suggestthat the value of
rate constant for recombination of nitrogen atoms on the surface is a weak function of surface
coverage. Furthermore, modest variations in the heat of N2 adsorption are sufficient to
achieve fractional ammonia orders that are relativelyinsensitiveto the ammonia pressure.
Keywords: ammonia synthesis; microkinetic modeling; reaction kinetics; surface science
studies; adsorption isotherms

1. I n t r o d u c t i o n

The synthesis of a m m o n i a over iron catalysts has been studied for over 70
years. During the past 20 years, experiments involving surface science techniques
such as Auger electron spectroscopy, X-ray photoelectron spectroscopy, work
function measurements and temperature-programmed desorption have been con-
ducted to probe the adsorbed species and the kinetics o f nitrogen adsorption a n d
desorption. M o s t o f these experiments have been conducted at sub-atmospheric
pressures on iron single crystals. Furthermore, several investigators have
a t t e m p t e d to incorporate these surface science data into microkinetic models a n d
to use these models to describe reaction kinetics data collected at industrial reac-
tion conditions (e.g., temperatures near 430°C and pressures near 100 atm). The
results o f the modeling efforts, through their successes and shortcomings, have con-
tributed further to our understanding of the reaction mechanism, as described
later in this review.
In the following sections, we review the surface science work of the last
20 years relevant to elucidation of the reaction mechanism of a m m o n i a synthesis

© J.C. Baltzer AG, SciencePublishers


234 L.M. Aparicio, J.A. Dumesic / Ammonia synthesis kinetics

over iron catalysts and the kinetics of the elementary steps in the mechanism. We
also review efforts to extrapolate these data to reaction conditions using microki-
netic modeling. Finally, we review proposed analytical rate expressions and their
relationships to the microkinetic models. For more complete reviews, readers are
referred to articles byNielsen [1] and Ertl [2].

2. K n o w l e d g e prior to surface science studies

Before the extensive surface science studies were undertaken, much was already
known about the mechanism of ammonia synthesis over iron. It was generally
believed that ammonia synthesis proceeded through elementary steps in which
both nitrogen and hydrogen adsorb and dissociate on the surface, atomic hydrogen
adds stepwise to atomic nitrogen, and ammonia desorbs from the surface, as
shown below:
1 N2 + * ~ *N2
2 *N2 ~ 2 , N
3 ,N + ,H ~ ,NH + •
4 , N H + , H ~ ,NH2 + •
5 *NH2 + , H ~ ,NH3 + •
6 *NH3 ~ NH3 + •
7 H2 + 2, ~ 2 , H
where • signifies a vacant site on the surface.
Early studies established that under typical reaction conditions the rate deter-
mining step involves nitrogen adsorption and/or dissociation. This conclusion was
reached after observing that nitrogen adsorption on iron catalysts proceeded at a
rate approximately equal to that of ammonia synthesis [3,4]. Further confirmation
was obtained in studies showing that hydrogenation proceeded much faster than
nitrogen adsorption [5].
Evidence for the stepwise hydrogenation of surface nitrogen atomic species was
also found. Infrared spectroscopy experiments identified N H and NH2 species on
the surface [6]. In addition, N H + species were detected by secondary ion-ion emis-
sion studies during ammonia synthesis, suggesting the presence of N H species on
the surface [7].
It was known that nitrogen adsorbed molecularly at low temperatures but could
dissociate at higher temperatures. Various experiments showed that different spe-
cies result from nitrogen adsorption at different temperatures. For example, it was
found that nitrogen adsorbed at temperatures from 130 to 300°C inhibits the
adsorption of carbon monoxide, whereas nitrogen adsorbed at 450°C does not
[8,9]. In another study, nitrogen adsorbed at 210°C had different reactivity toward
L.M. Aparicio, J.A. Dumesic / Ammonia synthesis kinetics 235

hydrogenation than nitrogen adsorbed at 440°C [10]. Isotopicexperiments showed


that the species resulting from low temperature adsorption was molecular,
whereas the one resulting from high temperature adsorption was atomic. When a
catalyst was exposed to a 28N2/3°N2 mixture at 350°C, no evidence for dissociation
was found; however, the same experiment performed at 380 or 400°C produced
29N2, indicating dissociation [11].
It was further established that the (111) surface of iron was the most reactive.
For example, nitrogen adsorbed preferentially on the (111) surface [12]. In addi-
tion, there was evidence that the synthesis reaction was faster over the (111) sur-
face, since iron whiskers consisting exclusively of (100) and (110) surfaces were
100 times less effective for the reaction than conventional catalysts [13]. This obser-
vation was attributed to structure sensitivity, as the ammonia synthesis turnover
frequency increased with particle size over iron supported on MgO [14]. It was pro-
posed that surface sites containing seven neighboring iron atoms (C7 sites) were
most active, since such sites exist on the (111) plane but not on the (100) or (110)
planes [15].
Because the dissociative adsorption of nitrogen was rate-determining, volu-
metric experiments aimed at measuring the kinetics of this process were conducted.
These experiments showed that the sticking coefficient for dissociative adsorption
on an iron catalyst is very low, i.e., 10-6 [4]. The experiments also provided esti-
mates for the activation energy of dissociative adsorption (67 kJ/mol) and the heat
of adsorption (155 kJ/mol) [3]. Later, gravimetric experiments showed that the
activation energy for dissociative adsorption is not constant, but increases with
coverage. The experiments suggested that this activation energy starts at ca. 22 kJ/
mol at zero coverage and increases to ca. 96 kJ/mol at higher coverage. The activa-
tion energy for desorption was found to decrease with coverage, starting at
230 kJ/mol and decreasing to 127 kJ/mol [16].
Calorimetry was used to study the heat of adsorption of molecular nitrogen. A
value of 38 kJ/mol was measured for adsorption on an iron-based catalyst [17],
while a value of 21 kJ/mol was measured over an evaporated iron film [18].

3. Overview o f selected surface science studies

Among the most informative experiments performed on single crystals was an


investigation of the kinetics of nitrogen dissociative adsorption using surface
science techniques [19-21]. These experiments confirmed that the sticking coeffi-
cient on both Fe(111) and Fe(100) is low, i.e., of the order of 10-7-10 -6. They also
confirmed that F e ( l l l ) adsorbs nitrogen faster than Fe(100), with the ratio
between the two rates being approximately 20. In agreement with the previous
gravimetric experiments, they showed that the activation energy for dissociative
adsorption increases with coverage; however, the measured activation energies
were substantially lower for the single crystal measurements. For Fe(100), the acti-
236 L.M. Aparicio, J.A. Dumesic / Ammonia synthesis kinetics

vation energy was found to be ca. 8 kJ/mol at zero coverage and 42 kJ/mol at a
coverage of 0.2. For Fe(111), the activation energy was - 3 kJ/mol. It must be
pointed out that, although the activation energy is negative at zero coverage, the
final uptake at the end of each run increased with increasing temperature, suggest-
ing that the activation energy increases with coverage and reached positive values.
Similar experiments were performed on F e ( l l l ) and Fe(100) surfaces onto
which potassium had been deposited [22]. These experiments demonstrated that
one of the major roles of a potassium promoter in ammonia synthesis catalysts is to
increase the sticking coefficient for dissociative nitrogen adsorption. On the
Fe(111) surface, the presence of potassium increased the sticking coefficient by a
factor of 8 at reaction temperature and lowered the activation energy of dissocia-
tive adsorption to - 1 2 kJ/mol. On the Fe(100) surface, the presence of potassium
increased the sticking coefficient by a factor of 280 at reaction temperature and
lowered the activation energy to ca. - 8 kJ/mol. These studies found that the pres-
ence of potassium not only increases the rates of adsorption but also significantly
diminishes the difference in activity between the two surfaces.
To complement the investigation of the kinetics of adsorption, the kinetics of
desorption of atomic nitrogen were investigated by temperature-programmed de-
sorption [19]. These experiments suggested that, although the process involves
recombination of a pair of atoms, it was first order in the coverage of atomic nitro-
gen. For the Fe(111) surface, a desorption activation energy of 213 kJ/mol was
estimated by assuming a desorption preexponential factor of 1013 s-l; for the
Fe(100) surface, the activation energy was estimated to be 243 kJ/mol.
In the same studies, the kinetics of adsorption of molecular nitrogen were investi-
gated by working at low temperatures where dissociative adsorption is negligible
[21-22]. For the Fe(111) surface, the sticking coefficient was estimated to be 10-2.
The desorption of surface molecular nitrogen was followed by temperature-pro-
grammed desorption. For the Fe(111) surface, the desorption preexponential fac-
tor was estimated to be 2 x 1010 s -1 and the activation energy 31 kJ/mol. The
presence of potassium on the surface was found to increase this activation energy to
44 kJ/mol. For the Fe(100) surface, molecular adsorption was found to be weaker
than on F e ( l l 1). These results were in agreement with the earlier calorimetry
results on polycrystalline surfaces.
Work function measurements and temperature-programmed desorption were
used to study the dissociative adsorption and desorption of hydrogen. Dissociative
adsorption was found to be non-activated, with a sticking coefficient of 0.1-0.2.
The heat of adsorption over both Fe(111) and Fe(100) was found to be between 88
and 108 kJ/mol at zero coverage, but it decreased substantially with coverage
[23]. The presence of potassium was found to increase the heat of adsorption by
8-12 kJ / mol and to lower the sticking coefficient by a factor of 2 [24].
The surface geometry after nitrogen adsorption was monitored in several stud-
ies by low energy electron diffraction [19,20]. The experiments showed that an
ordered structure forms over Fe(100), without surface reconstruction, and that this
L.M. Aparicio, J.A. Dumesic / Ammonia synthesis kinetics 237

structure is complete at a coverage of half a monolayer. This result suggested that


nitrogen atoms were located in fourfold sites on the surface. Over Fe(111), the evi-
dence suggested that the surface reconstructs, and the data were interpreted in
terms of hexagonal layers similar to the (111) plane of Fe4N.
The a m m o n i a synthesis reaction was studied over various single crystal planes
at 20 atm [27]. The relative rates of reaction were found to be 4 1 8 : 2 5 : 1 for
Fe(111), Fe(100) and Fe(110), respectively. The activation energy for a m m o n i a
synthesis over Fe(111) was determined to be 81 kJ/mol, close to the value observed
for an industrial catalyst.
Both molecular and atomic nitrogen were detected on iron surfaces by using
X-ray photoelectron spectroscopy [26-30]. On a polycrystalline surface [26,27]
two peaks were observed after molecular adsorption, at 400 and 405 eV. The peak
at 400 eV was initially assigned to a bridge-bonded species, and the peak at
405 eV was initially assigned to a terminally bonded species. After dissociative
adsorption, a single peak was detected at 397 eV and was assigned to an atomic spe-
cies. It was shown that the peaks at 400 and 405 eV could be converted to the
peak at 397 eV, suggesting that surface molecular nitrogen can be a precursor to
the a.tomic species.
More recent work has investigated nitrogen species on the surface of Fe(111),
using X-ray photoelectron spectroscopy [28-30]. It was found that the two peaks at
400 and 405 eV were first converted to a peak at 399 eV, and that this latter peak
was subsequently converted to the atomic state at 397 eV. The peaks at 400 and 405
eV were reassigned to a single molecular species denoted as the y state. The peak
at 399 eV was assigned to another molecular species, denoted as the a state, which
can dissociate to form atomic nitrogen. The peak at 397 eV was assigned to atomic
nitrogen and was denoted as the 13state. Based on high resolution electron energy
loss spectroscopic experiments, the a state was interpreted as a bridge-bonded spe-
cies with electron donation from the surface to the antibonding n levels of N2. In a
similar manner, the y state was interpreted as a terminally-bonded species. The fol-
lowing picture for nitrogen adsorption thus emerged:
N2 ~ *-N2 (y state) ~ * - N 2 - * (a state) --+2 . - N (13 state).
Because an activation barrier is expected between the y state and the a state, this
picture explains the somewhat low sticking coefficient of 10 -2 found for molecular
adsorption.
By repeating the above experiments over a potassium-promoted Fe(111) sur-
face [28-30], it was found that potassium increases the sticking coefficient into the a
state. This result was interpreted in terms of a decrease in the barrier between the
y state and the a state. Temperature-programmed desorption studies were com-
bined with these results to estimate various kinetic parameters in the presence and
absence of potassium. The sticking coefficient of the y state, for example, was esti-
mated to be 0.7. The heat of adsorption for the y state was estimated to be 24 kJ/
mol in the absence of potassium and 22 k J / m o l in the presence of potassium. For
238 L.M. Aparicio, J.A. Dumesic / Ammonia synthesis kinetics

the conversion from the 7 state to the a state, the preexponential factor was esti-
mated to be 7 x 107 s -I . The activation energy for this conversion was estimated to
be 18 k J / m o l in the absence of potassium and 10 k J / m o l in the presence of potas-
sium. The heat of adsorption of the ct state had been previously proposed to be
31 k J / m o l in the absence of potassium; based on this value, the heat of adsorption
in the presence of potassium was estimated as 40 kJ/mol, in good agreement with
the previously proposed value of 44 kJ/mol. Finally, the pre-exponential factor for
conversion from the a state to the ), state was estimated to be 2 x 108 s-1.
In other studies, it was found that the transformation from the 7 state to the a
state results in an increase in the work function [31] and a decrease in the surface
potential [32]. By monitoring the work function during adsorption, it was found
that while the 7 state can be a precursor to the a state, there also exists a route to the
state directly from the gas phase:
N2 ~ * - N 2 - * (a state) --~ 2 . - N (13 state).
The sticking coefficient for this route is low (,,~ 10-3); however, it could become
important at the relatively high temperatures of reaction conditions, where the cov-
erage of the 7 state is low.
Molecular beam studies and Auger electron spectroscopic measurements on an
Fe(111) surface indicated that the dissociative chemisorption of nitrogen on iron
can be enhanced by increasing the kinetic energy of the impinging nitrogen mole-
cules [33,34]. The sticking coefficient was found to increase from 10 -6 at a kinetic
energy of 0.09 eV (typical of a gas at 0 ° C) to over 10-1 at 4.3 eV. By monitoring the
coverage of the a state during the experiment, it was determined that the kinetic
energy does not provide a direct route to the atomic state, but rather it increases
trapping into either the tx state or a similar intermediate species. It is therefore pos-
sible that the direct route from gaseous N2 to the molecular a state becomes impor-
tant at high temperatures.
Several recent studies have addressed the state of potassium and alumina promo-
ters on the catalyst [35,36]. It was shown that metallic potassium on Fe(100) sur-
faces is not stable under reaction conditions. When coadsorbed with oxygen,
however, a small amount of potassium can be stabilized at reaction conditions.
When coadsorbed with alumina, substantially more potassium can be stabilized on
Fe(111), Fe(100) and Fe(110) surfaces and the ratio between stabilized potassium
and alumina is approximately 1 : 1, suggesting compound formation. The promo-
ter studies also found that water vapor can reconstruct the less-active Fe(100) and
Fe(110) surfaces and make them almost as active as the Fe(111) surface. However,
after switching to ammonia synthesis conditions the surfaces revert to their origi-
nal states. The presence of alumina enhances this reconstruction and increases the
stability of the reconstructed surfaces.
The origin for the greater activity of the Fe(111) surface for a m m o n i a synthesis
has been revisited in recent studies. In particular, the reaction was studied over the
Fe(210) surface, which has no high-coordination sites but is quite " o p e n " (it
L.M. Aparicio, J.A. Dumesic / Ammonia synthesis kinetics 239

exposes second and third layer atoms), and the (211) surface, which is not an open
surface but has sites of sevenfold coordination. It was thei'eby shown that the
high-coordination (C7) sites are related to the high activity of the surface [37].
Insight into the nitrogen adsorption process has also been provided by quantum
mechanical studies. In one calculation, potential energy surfaces for N - M and
N2-M, where M is either Fe or Re, were constructed from Morse potentials. The
time-dependent Schrfdinger equation was then solved for an incident wave packet
to study the dissociation of adsorbed molecular nitrogen into atomic nitrogen
[38]. The results suggested that a tunneling mechanism may be important in this
reaction.

4. Analytical rate expressions

The most widely used rate expression for ammonia synthesis, known as the
Temkin-Pyzhev equation, was first proposed in 1940 [39]:

This expression can be derived assuming that nitrogen dissociative adsorption is


rate-determining and that the surface coverage by atomic nitrogen is high. In the
derivation, it is assumed that nitrogen adsorption is described by an isotherm in
which the heat of adsorption changes linearly with coverage. If a equals 0, then the
change in the heat of nitrogen adsorption with coverage affects only the activation
energy for desorption. If a equals 1, then the change in the heat of nitrogen adsorp-
tion with coverage affects only the activation energy for adsorption. In this latter
case, the rate constant for desorption remains constant with coverage.
The Temkin-Pyzhev equation has been used successfully by various investiga-
tors to fit experimental data for both ammonia synthesis and ammonia decomposi-
tion. In these studies, it has been found that a is equal to 0.5-0.8 [39-45]. Notice
that the relatively high values of a that have been found suggest that the activation
energy for the reverse of step 2 is not significantly affected by changes in the heat
of adsorption of atomic nitrogen. Because the rate constant for this step is most
responsible for the measured rates, as will be shown later, this observation may
explain why the microkinetic models have been relatively successful in explaining
experimental data despite their failure to take into account changes in the ener-
getics with coverage.
Some investigators have proposed other rate expressions that are extensions of
the original Temkin-Pyzhev equation. For example, Ozaki et al. showed if the sur-
face coverage by atomic nitrogen is no longer high, then the rate expression
becomes [41 ]:
k~APN2 - k~B(PNH3)2/(pH2) 3
r= (2)
[1 + K c ( P N H 3 ) / ( P H 2 ) V 2 ] 2~' "
240 L.M. Aparicio, J.A. Dumesic / Ammonia synthesis kinetics

Eq. (2) reduces to eq. (1) when Kc(PNH3)/(PHz)3/2>> 1. Similarly, if the most
abundant surface intermediate is an NH species rather than atomic nitrogen, then
the rate expression becomes [41]:
k~ P N 2 - k/l~(PNH3)2/(PH2) 3
r-- [1 + KD(PNH3)/(Pr~2)]2~ (3)
In an effort to combine eqs. (2) and (3) above and to avoid problems of non-ideal-
ity, Nielsen et al. proposed the following expression [40,46]:
kt~aN2 -
tt 2/(all2) 3
r = [1 + g~t(aNH3)/(aH2)'r] 2~ ' (4)
where ax corresponds to the activity of species x. Attempts to fit experimental
data with this expression found that the best value for the parameter 7 was 3/2,
making it equivalent to eq. (2) when activities are assumed to be equal to pressures.
This provides further evidence that the most abundant surface intermediate at reac-
tion conditions is atomic nitrogen.
The Temkin-Pyzhev equation and other analytical rate expressions taking into
account surface non-uniformity have been very successful for describing experi-
mental kinetic data. Tables 1 and 2 and figs. 1-3 show predictions obtained with
eq. (2), using parameters from ref. [38]. Whereas the data in figs. 1 and 2 were used
by Nielsen et al. to determine the parameters of the rate expression, the compari-
son shown in fig. 3 represents a true extrapolation of the model to the data of Ozaki
et al. at lower temperatures and conversions. The rate expression does a remark-
ably good job, considering that Ozaki et al. used a different iron catalyst than Niel-
sen et al., while we have assumed for simplicity that both catalysts had the same
number of sites per gram. Furthermore, it can be shown [41] that the curves in fig. 3
should have slopes equal to 1/(2c~ + 1). The experimental data show a slope of
0.38, and the analytical rate expression predicts a slope of 0.4.
The analytical rate expressions have also been successful in explaining observed
rates of ammonia decomposition. In these studies, an ammonia order between 0.6
and 1.0 and a hydrogen order between -0.9 and -1.5 were found [42,47]. These
orders can be explained by values of a between 0.5 and 0.7.

5. Microkinetic modeling based on surface science results

In view of the extensive studies summarized above, it is possible to test whether


these data can be extrapolated to predict the kinetics of ammonia synthesis at
industrial reaction conditions. One of the first models used for this purpose
(denoted as model I in this paper) assumed that the reaction proceeds through the 7
elementary steps mechanism presented earlier [48-51]. The model was solved tak-
ing step 2 to be rate-determining and assuming that all other steps are in thermody-
L.M. Aparicio, J.A. Dumesic / Ammonia synthesiskinetics 241

Table 1
Experimental data for Haldor Topsoe K M I R catalyst [40] versus predictions from microkinetic
models and analytical rate expression a

T (°C) P(atm) Space Inlet % NH3 in % NI-I3 in % NH3 in % NH3 in


velocity flow rate b effluent effluent c effluent e effluent d
(h -1) (cm3 STP/min) (exp.) (microkin. (microkin. (eq. (2))
model I) model II)

490 214 27200 519 19.3 15.4 15.8 17.6


490 214 52000 992 17.4 13.1 13.9 16.6
490 214 75800 1447 15.9 11.8 12.7 15.6
450 320 16000 305 30.0 20.9 21.1 27.9
450 320 16400 313 29.9 20.8 20.9 27.8
450 320 24000 458 26.7 18.7 18.9 25.8
450 320 30100 574 27.7 17.6 17.7 24.5
450 320 80000 1527 18.4 13.1 13.1 1~.4
450 214 14000 267 22.8 17.4 17.2 22.3
450 214 16000 305 22.2 16.8 16.7 21.8
450 . 214 24000 458 20.0 15.0 14.9 20.1
450 214 27200 519 19.2 14.5 14.4 19.5
450 214 50000 954 15.3 12.1 12.0 16.4
448 214 85600 1634 13.8 10.2 10.1 13.8
450 107 11600 221 13.9 12.2 11.7 14.3
450 107 16000 305 13.2 11.2 10.8 13.6
450 107 23500 449 12.4 10.1 9.7 12.6
450 107 24000 458 12.5 10.0 9.7 12.5
450 107 47100 899 9.8 8.2 8.0 10.4
410 320 15200 290 26.6 19.0 17.1 25.7
410 323 28100 536 20.8 15.9 14.1 21.3
410 330 50600 966 15.8 13.4 11.7 17.6
410 320 72700 1388 13.2 11.7 10.1 15.2
411 214 35800 683 16.8 12.0 10.5 15.5
410 107 22800 435 10.7 9.1 7.8 11.2
410 107 43800 836 8.3 7.4 6.3 8.8
370 214 24000 458 11.4 11.3 8.4 13.2
350 214 21700 414 8.2 10.7 7.2 11.7
331 214 20000 382 4.1 9.9 6.1 10.3

a 1.25 cm 3 of catalyst, 1.25 cm 3 of diluent, H2 : N2 = 3, particle size= 0.3-0.8 mm, reduced at a parti-
cle size of 7.6 mm.
b Assumes space velocity is based on inlet flow rate at NTP and on catalyst volume (1.25 cm3).
c Assumes 1.123 x 102o sites on the surface, calculated assuming catalyst and diluent particles occu-
pied 2/3 of the bed, a catalyst particle density of 3.73 g/cm 3, and a site density of 6 x 10-5 mol/g
as estimated by CO chemisorption.
d Assumes the following values for the parameters (the units of the rate being kmol/h m 3 of
catalyst): k'A = 3.289 exp[(50.69 kJ/mol)/RT], k'B = 7.35 × 1012 exp[(-59.04 kJ/mol)/RT], Kc
= 0.0307 exp[(81.00 kJ/mol)/RT~.
242 L.M. Aparicio, J.A. Dumesic / Ammonia synthesis kinetics

Table 2
Low temperature experimental data for doubly promoted iron catalyst [41] versus predictions from
microkinetic models and analytical rate expression a
T ( ° C ) P(atm) Flowrate Flowrate %NH3in %NH3in %NI-I3in %NH3in
(£NTP/h) (cm3 STP/min) effluent effluent b effluent b effluent c'd
(exp.) (microkin. (microkin. (eq. (2))
model I) model II)
302 1 4.35 66.4 0.148 0.278 0.148 0.191
302 1 3.85 58.7 0.155 0.290 0.154 0.201
302 1 3.33 50.9 0.165 0.304 0.162 0.213
302 1 2.78 42.4 0.173 0.323 0.172 0.229
302 I 2.17 33.2 0.195 0.351 0.187 0.252
302 1 2.13 32.5 0.190 0.353 0.188 0.254
302 1 1.33 20.4 0.230 0.413 0.221 0.307
278 1 4.35 66.4 0.100 0.243 0.112 0.148
278 1 3.33 50.9 0.110 0.266 0.122 0.165
278 1 2.78 42.4 0.118 0.279 0.130 0.177
278 1 2.13 32.5 0.130 0.305 0.143 0.197
278 1 1.33 20.4 0.158 0.357 0.167 0.238
278 1 1.28 19.6 0.163 0.362 0.170 0.241
251 1 4.55 69.4 0.058 0.197 0.078 0.106
251 1 4.00 61.1 0.061 0.207 0.081 0.111
251 1 3.45 52.7 0.065 0.218 0.086 0.118
251 I 2.86 43.6 0.072 0.231 0.091 0.127
251 1 2.78 42.4 0.072 0.233 0.092 0.129
251 1 2.17 33.2 0.078 0.252 0.I00 0.142
251 1 1.33 20.4 0.095 0.301 0.118 0.173
251 1 1.28 19.6 0.098 0.305 0.119 0.176
a 0.773 g of catalyst before reduction, H2 : N2 = 3,
b Assumes 2.02 x 1019 sites on the surface, calculated assuming catlyst weight after reduction was
0.559 g and a site density of 6 × 10-5 mol/g as estimated by CO chemisorption.
c Assumes catalyst weight after reduction was 0.559 g.
d Assumes the following values for the parameters (the units of the rate being mmol/h g of
catalyst): /ch = 1.323 exp[(50.69 kJ/mol)/RT], k'B = 2.96 x 1012exp[(-59.04 kJ/mol)/RT], Kc
= 0.0307 exp[(81.00 k J/tool)/R 7~.

namic equilibrium. A Langrnuir-Hinshelwood treatment of the problem shows


that the rate should be a function of:
(1) the gaseous pressures,
(2) the product of the equilibrium constant for step 1 and the forward rate con-
stant for step 2,
(3) the equilibrium constants for steps 3-7, and
(4) the equilibrium constant for the overall reaction.
The product of the equilibrium constant for step 1 and the forward rate constant
for step 2 was calculated from the sticking coefficient and activation energy meas-
ured for dissociative nitrogen adsorption on K-containing Fe(111). The equili-
L.M. Aparicio, J.A. Dumesic / Ammonia synthesis kinetics 243

35

~
30 OD 320 aim

25 214 arm
0
.c_ 20
-r"
Z 107 aim
15
o

10

5 I I I

0 1 2 3 4
1 / Inlet Flow Rate (rain / I STP)

Fig. 1. Experimental data for Haldor Topsge KMIR catalyst at 450°C and various pressures [40]
versus predictions from eq. (2).

brium constants for steps 3-7 were calculated from estimated partition functions
and heats of reaction for each step. The most important assumption in this analysis
involk,ed the heat of adsorption of atomic nitrogen. A value of 170 kJ/mol can be
estimated from temperature-programmed desorption experiments on F e ( l l l ) ,
assuming a desorption pre-exponential factor of 1.3 x 109 s -1. The resulting model
is able to reproduce experimental ammonia yields obtained at pressures of
107 arm, 9 or more orders of magnitude higher than the pressures at which the TPD
studies were conducted. The model also predicts that the most abundant surface
intermediate under reaction conditions is atomic nitrogen, its coverage exceeding
0.80 throughout most of the reactor.
A second group of investigators constructed a similar model (denoted as model
II in this paper), replacing step 7 with two steps, in which hydrogen adsorbs molecu-

25 i i i i

20

15
O

_r.'~
Z 10

5 ,x

0
0 1 2 3 4 5
1 / Inlet Flow Rate (min / I S T P )

Fig. 2. Experimental data for Haldor Topsoe KMIR catalyst at 214 atm and various temperatures
[40] versus predictions fromeq. (2).
244 L.M. Aparicio, J.A. Dumesic / Ammonia syn thesis kinetics

0.4

c 0.3
o
302 *C
u.. 0.2
278°0
o

.1. ~
Z 0.1 0 cO 251 *C

= 0.08 0
CD
0
0.08 0
0
i i I
0.6 08 1 ; ;
1 / Flow Rate (s / cm 3 (STP))

Fig. 3. Experimental data for doubly promoted iron catalyst at 1 atm and low temperature [41]
versus predictions from eq. (2).

larly and subsequently dissociates on the surface [52-55]. Instead of performing a


Langmuir-Hinshelwood treatment, in which a rate-determining step must be
assumed, these investigators estimated rate constants for all elementary steps, tak-
ing as many parameters as possible from the surface science studies. In agreement
with the above model, they found that step 2 is rate-determining under reaction
conditions and that under industrial ammonia synthesis conditions the most abun-
dant surface intermediate is atomic nitrogen. They reported, however, that their
model underpredicted the rate of ammonia synthesis at reaction conditions by over
an order of magnitude. In this model, the heat of adsorption of atomic nitrogen
was assumed to be 213 kJ/mol, and they used a desorption preexponential factor
of 2.6 x 1012 s -1 . The failure of the model to predict rates at reaction conditions
was attributed to this high heat being inappropriate at the high coverages involved
under industrial reaction conditions.
Another research group tested both models to gain insight into the discrepancy
between the reported results [56]. The mathematical problem was solved without
assuming a rate-determining step, transforming the Langmuir-Hinshelwood
model into an equivalent one with rate constants for all individual steps. Both mod-
els gave essentially identical results at high pressure industrial conditions, predict-
ing ammonia yields within ca. 50% and always underpredicting the experimental
results. This behavior can be seen in table 1, which shows predictions by both mod-
els compared to experimental data collected over a Haldor Topsoe K M I R catalyst
at pressures from 107 to 320 arm and temperatures from 331 to 490°C [40]. For
clarity, the data collected at 450°C and various pressures are plotted in figs. 4 and
5, while the data collected at 214 atm and various temperatures are plotted in
figs. 6 and 7.
Because both models give identical predictions at industrial conditions, we attri-
bute the apparent discrepancies between the two reports to the following factors:
L.M. Aparicio, J.A. Dumesic / Ammonia synthesis kinetics 245

35

30
o
o
25
0
20
"1-
Z
15

10
~ t I I I
atm

5
0 1 2 3 4
1 / InletFlowRate (min/ I STP)

Fig. 4. Experimental data for Haldor Topsee KMIR catalyst at 450°C and various pressures [40]
versus predictions from microkinetic model I.

(1) differences in the assumed number of sites (one group calculated it from CO
chemisorption, the other from surface areas),
(2i differences in interpretation of flow rate and amount of catalyst in the
reported experimental data collected at industrial conditions, and
(3) differences in the presentation of the data, because calculated output ammo-
nia concentrations do not deviate from experiments as much as reaction rates since
industrial conditions are close to equilibrium.
We conclude that both of the above models can be used to extrapolate surface
science data to industrial conditions with some success. However, the predictions
do not match experiments with high precision. In some cases the deviation in rate
may be as high as a factor of 5. This agreement is still remarkable, considering that
the modeling involves an extrapolation of 9 or more orders of magnitude in the

35

30
o
o
25
--1
0
20

Z
15

10
~ I I I I
atm

5
0 1 2 3 4
1 / Inlet Flow Rate (rain / I STP)

Fig. 5. Experimental data for Haldor Topsee KMIR catalyst at 450°C and various pressures [40]
versus predictions from microkinetic model II.
246 L.M. Aparicio, J.A. Dumesic / Ammonia syn thesis kinetics

25 L I

[] [3

20 t~ [3

15 o ~ / _ . . . - - I ~ o oc
0

10
Z
o~
5 A

I I I
0
0 1 2 3 4
1 / Inlet Flow Rate (min / I STP)

Fig. 6. Experimental data for Haldor Topsoe KMIR catalyst at 214 atm and various temperatures
[40] versus predictions ofmicrokinetic model I.

pressure. Furthermore, as shown in fig. 8, the agreement between the models and
experiments can be improved considerably if the number of assumed sites is
increased by an factor of only 3.
We have investigated further why both models give similar results. This behav-
ior is related to the high coverage of atomic nitrogen. As the atomic nitrogen cover-
age approaches unity, it can be shown by a Langmuir-Hinshelwood analysis that
the turnover frequency for the reaction can be estimated by the following expres-
sion, provided that step 2 remains rate-determining:

r=k_20~(K e q ( ~PN2 (PH2)3)


)2-1 , (5)

where k-2 is the reverse rate constant of step 2 in s -1, Keq is the overall equili-

25 i i i

[3 O
20 e~ rn
0 490 °C:
--i
15
o
t-

Z
10
/
°

/ ~ 331 *(3
5

0 I L

0 1 2 3 4
1 / Inlet Flow Rate (rain / I STP)

Fig. 7. Experimental data for Haldor Tops~e KMIR catalyst at 214 atm and various temperatures
[40] versus predictions ofmicrokinetic model II.
L.M. Aparicio, J.A. Dumesic / Ammonia synthesis kinetics 247

35 i i i i

~
30 aD 320 atm

.$
25 atm

0
20
.I-~
Z
15
o~
10

I I I I
5
0 1 2 3 4 5
1 / Inlet Flow Rate (min / I STP)

Fig. 8. Experimental data for Haldor Topsoe KMIR catalyst at 450°C and various pressures [40]
versus predictions from microkinetic model I increasing number of sites by a factor of 3.

brium constant for the reaction and P0 is the standard state pressure (usually
1 atm). It is apparent that the most important kinetic parameter for predicting
ammonia synthesis rates at industrial reaction conditions in the models is the
reverse rate constant of step 2. Although the two models employed different activa-
tion energies for k-2, the preexponential factors compensated such that the reverse
rate constant for step 2 was nearly identical in both models over the temperature
range (400-500°C) where most experiments were run.
Due to the different activation energies for the reverse rate constant of step 2,
the two models deviate when they are tested over a wide temperature range. This
can be seen in figs. 6 and 7, where the calculated curves for reaction temperatures
of 331 and 370°C differ for the two models. This behavior is also evident in table 2
and figs. 9 and 10, which show the predictions of both microkinetic models corn-

• ' I

o~ 0.4

c
._o 0.3

o~ 302 *C
u. 0.2
¢1
0 0 GO 278 *C
0 0
0
-1-" -3 0
Z 0.1 C° 251 *C
o.o8 0
0 O
0.06 O O
, L i i I
o6 o8 1 ~ ; ~.
1 / Flow Rate (s / cm 3 (STP))

Fig. 9. Experimental data for doubly promoted iron catalysts at 1 atm and low temperature [41]
versus predictions from microkinotic model I.
248 L.M. Aparicio, J.A. Dumesic / Ammonia synthesis kinetics

0.4

0.3

302 *C
it 0.2
"5 ~ 278*c

Z CO 251 °C

0.08
O 0
0
0.06 1 0
, , I
0.6 0.8 1 ; ;
1 / Flow Rate (s / cm3 (STP))

Fig. 10. Experimental data for doubly promoted iron catalyst at 1 atm and low temperature [41]
versus predictions from microkinetic model II.

pared to effluent ammonia concentrations for a double-promoted iron catalyst at


1 atm in the temperature range 251-302°C [41]. At these low temperatures, the
microkinetic models tend to overpredict the experimental reaction rates. The first
model, with an activation energy of 155 kJ/mol for the reverse of step 2, overpre-
dicts the experimental data to a greater extent. The models predict that under these
conditions the coverage of atomic nitrogen is still high, exceeding 0.9 throughout
most of the reactor, and step 2 is still rate-determining. Thus, the reverse rate con-
stant of step 2 should still be the parameter to which the calculated results are
most sensitive. The results therefore suggest that both models, but in particular the
first model, have an activation energy for the reverse of step 2 that is too low. For
example, by increasing the forward and reverse activation energies of step 2 in the
first model by 67 kJ/mol and increasing the forward and reverse preexponential
factors of the step by a factor of 160 000, the predictions of the model at low
temperatures improves substantially, as shown in fig. 11. With these changes, the
behavior of the model at higher temperatures also improves, with results similar to
those in fig. 8.
The excellent success of rate expressions based on non-uniform surfaces sug-
gests that the deficiencies of the microkinetic models developed so far can be attrib-
uted to their failure to include changes in energetics with coverage. In this respect,
we note that the forward rate of ammonia synthesis is given by
r = k2K1PN202,, (6)
where k2 is the forward rate constant for step 2, K1 is the equilibrium constant for
step 1, and 0, is the fraction of the surface that is vacant. In this case, the ammonia
reaction order is given by the variation of 0, with the ammonia pressure. Since the
dissociative adsorption of ammonia to give atomic nitrogen is fast, the value of 0, is
determined by the equilibrium adsorption isotherm for the reaction
L.M. Aparicio, J.A. Dumesic / Ammonia syn thesis kinetics 249

0.4

c
0
0.3
0 302 *C
u- 0.2
0
..o.-f
I
Z 0.1 ~ 251 "C
0.0s
O
0.06
0.6
1 / Flow Rate (s / cm 3 (STP))

Fig. 11. Experimental data for doubly promoted iron catalyst at 1 atm and low temperature [41]
versus predictions from microkinetic model I increasing activation energies for step 2 by 67 kJ/mol
and increasing preexponential factors for step 2 by a factor of 160 000..

NH3 + * ~ N • + 3H 2 . (7)
We now consider the three following adsorption isotherms [57]:
Langmuir isotherm:
0
(8)
Bragg-Williams isotherm:

J~eqPNH3/pL~ = 10N-oNexp(cwO~/kT); (9)

quasi-chemical isotherm:

K~qPNH'IPL:--1 ----~h,~exp<cw/2kT) ((fl --(fl!__+2_0N)


1- --- 20N)0N'~)'~
+<1 #2/ (I0)

fl = {1 - 40N(1 -- 0N)[1 -- exp(-w/kT)]} °5 , (11)


where K~q is the adsorption equilibrium constant at zero coverage, c is the number
of nearest neighbor sites surrounding an adsorbed nitrogen atom (c = 4 for a
square lattice), and w is the interaction energy between pairs of nitrogen atoms on
nearest neighbor sites.
The three adsorption isotherms are plotted in fig. 12 for c = 4 and w = 10 kJ/
mol. The ammonia order is equal to 2 d In 0./d In PNH3. It is apparent that while the
Langmuir isotherm can give rise to fractional ammonia orders (e.g., -1.5), this
fractional kinetic order would vary with ammonia pressure. The Bragg-Williams
and the quasi-chemical isotherms can also lead to fractional ammonia orders; how-
ever, these fractional orders are more constant with respect to variations in amino-
250 L.M. Aparicio, J.A. Dumesic / Ammonia synthesis kinetics

I - - -- - Bragg-Williams
Langmuir Isotherm
Isotherm
I
......... Quasi-chemical Isotherm
o
"'"-"-.:..-2. ...-~,.,,~, ' . . . .
.1
"~~'~ "~°"o"~ ~°°

.2
¢D
r \ \ ~,...
¢_
-3

.4

-5
-5 0 5 10 15
In ( g° 1.5
~1 PNH3 / P H2 )

Fig. 12. Adsorption isotherms for atomic nitrogen from ammonia according to NH3 + * ~ N ,
+ ~H2.

nia pressures than the behavior given by the Langmuir isotherm. We suggest that
these adsorption isotherms based on lateral interactions between nitrogen atoms
should lead to microkinetic models that describe the observed reaction kinetics bet-
ter than do microkinetic models based on Langmuirian kinetics.
In closing, we note that the value of w equal to 10 kJ/mol corresponds to a
change in the heat of dissociative N2 adsorption equal to 40 kJ/mol as the coverage
varies from 0 to 1. If the heat of dissociative N2 adsorption is - 2 0 0 kJ/mol at zero
coverage, then this variation in heat with respect to coverage is only 20%. Thus,
we suggest that two phenomena are responsible for the success of microkinetic
models based on uniform surfaces to extrapolate results from surface science stud-
ies to describe the performance of iron catalysts at industrial ammonia synthesis
reaction conditions: (i) the value of k-2 is a weak function of surface coverage, and
(ii) modest variations in the heat of N2 adsorption are sufficient to achieve frac-
tional ammonia orders that are relatively insensitive to the ammonia pressure.

Acknowledgement

We wish to thank Randy Cortright and Brian Spiewak for their help and sugges-
tions during the preparation of this paper.

References

[1] A. Nielsen, Catal. Rev. 4 (1970) 1.


[2] G. Ertl, Catal. Rev.-Sci. Eng. 21 (1980) 201.
[3] P.H. Emmett and S. Brunauer, J. Am. Chem. Soc. 56 (1934) 35.
L.M. Aparicio, £ A. Dumesic / Ammonia synthesis kinetics 251

[4] P.H. Emmett and S. Brunauer, J. Am. Chem. Soc. 55 (1933) 1738.
[5] K.T. Kozhenova and M.Ya. Kagan, J. Phys. Chem. USSR (1940) 1250.
[6] T. Nakata and S. Matsushita, J. Phys. Chem. 72 (1968) 458.
[7] V.I. Shvachko, Ya.M. Fogel and Ya. Kolot, K_inet. Katal. 7 (1966) 834.
[8] N. Takezawa and P.H. Emmett, J. Catal. 11 (1968) 131.
[9] Y.-Y. Huang and P.H. Emmett, J. Catal. 24 (1972) 101.
[1(3] N. Takezawa, J. Catal. 24 (1972) 417.
[11] Y. Morikawa and A. Ozaki, J. Catal. 12 (1968) 145.
[12] R. Brill, E.-L Richter and E. Ruch, Ang. Chem. Int. Ed. 6 (1967) 882.
[13] R. Brill, J. Catal. 16 (1970) 16.
[14] J.A. Dumesic, H. Topsoe, S. Khammouma and M. Boudart, J. Catal. 37 (1975) 503.
[15] J.A. Dumesic, H. Topsoe and M. Boudart, J. Catal. 37 (1975) 513.
[16] J.J.F. Scholten, P. Zwietering, J.A. Konvalinka and J.H. de Boer, Trans. Faraday Soc. 55
(1959) 2166.
[17] O. Beeck, W.A. Cole and A. Wheeler, Discussions Faraday Soc. 52 (1956) 86.
[18] G. Wedler, D. Borgmann and K.P. Geuss, Surf. Sci. 47 (1975) 592.
[19] F. Boszo, G. Ertl, M. Grunze and M. Weiss, J. Catal. 49 (1977) 18.
[20] G. Ertl, M. Grunze and M. Weiss, J. Vac. Sci. Technol. 13 (1976) 314.
[21] G. Ertl, S.B. Lee and M. Weiss, Surf. Sci. 114 (1982) 515.
[22] G. Ertl, S.B. Lee and M. Weiss, Surf. Sci. 114 (1982) 527.
[23] F. Boszo, G. Ertl, M. Grunze and M. Weiss, Appl. Surf. Sci. 1 (1977) 103.
[24] G. Ertl, S.B. Lee and M. Weiss, Surf. Sci. 111 (1981) L711.
[25] N.D. Spencer, R.C. Schoonmaker and G.A. Somorjai, J. CataL 74 (1982) 129.
[26] K. Kishi and M.W. Roberts, Surf. Sci. 62 (1977) 252.
[27] D.W. Johnson and M.W. Roberts, Surf. Sci. 87 (1979) L255.
[28] M. Grunze, M. Golze, W. Hirschwald, H.-J. Freund, H. Pulm, U. Seip, M.C. Tsai, G. Ertl and
J. Kfippers, Phys. Rev. Lett. 53 (1984) 850.
[29] L.J. Whitman, C.E. Bartosch, W. Ho, G. Strasser and M. Grunze, Phys. Rev. Lett. 56 (1986)
1984.
[30] L.J. Whitman, C.E. Bartosch andW. Ho, J. Chem. Phys. 85 (1986) 3688.
[31] M. Grunze, G. Strasser and M. Golze, Appl. Phys. A 44 (1987) 19.
[32] R. Dfis, E. Nowicka, Z. Wolfram and W. Gubernator, Surf. Sci. 247 (1991) 257.
[33] C.T. Rettner and H. Stein, Phys. Rev. Lett. 59 (1987) 2768.
[34] C.T. Rettner and H. Stein, J. Chem. Phys. 87 (1987) 770.
[35] S.R. Bare, D.R. Strongin and G.A. Somorjai, J. Phys. Chem. 90 (1986) 4726.
[36] D.R. Strongin, S.R. Bare and G.A. Somorjai, J. Catal. 103 (1987) 289.
[37] D.R. Strongin, J. Carraza, S.R. Bare and G.A. Somorjai, J. Catal. 103 (1987) 213.
[38] M. Asscher, O.M. Becker, G. Haase andR. Kosloff, Surf. Sci. 206 (1988) L880.
[39] M. Temkin andV. Pyzhev, Acta Physicochim. URSS 12 (1940) 327.
[40] A. Nielsen, An Investigation on Promoted Iron Catalysts for the Synthesis of Ammonia
(Gjellerup, Copenhagen, 1967).
[41] A. Ozaki, H. Taylor and M. Boudart, Proc. Roy. Soc. A 258 (1960) 47.
[42] K.S. Love andP.H. Emmett, J. Am. Chem. Soc. 63 (1941) 3297.
[43] R. BriU, J. Chem. Phys. 19 (1951) 1047.
[44] V.D. Livshits and I.P. Sidorov, Zh. Fiz. Khim. 21 (1947) 1177.
[45] V.D. Livshits and I.P. Sidorov, Zh. Fiz. Khim. 26 (1952) 538.
[46] A. Nielsen, J. Kj =r and B. Hansen, J. Catal. 3 (1964) 68.
[47] E. Winter, Z. Phys. Chem. B 13 (1931 ) 401.
[48] P. Stoltze and J.K. Nerskov, Phys. Rev. Lett. 55 (1985) 2502.
[49] J.K. Nerskov and P. Stolze, Surf. Sci. 189/190 (1987) 91.
[50] P. Stoltze and J.K. Nerskov, J. Catal. 110 (1988) 1.
252 L.M. Aparicio, J.A. Dumesic / Ammonia synthesis kinetics

[51] P. Stoltze and J.K. Nzrskov, Surf. Sci. 197 (1988) L230.
[52] M. Bowker, I.B. Parker and K.C. Waugh, Appl. Catal. 14 (1985) 101.
[53] I.B. Parker, K.C. Waugh and M. Bowker, J. Catal. 114 (1988) 457.
[54] M. Bowker, I.B. Parker and K.C. Waugh, Surf. Sci. 197 (1988) L223.
[55] M. Bowker, Catal. Today 12 (1992) 153.
[56] J.A. Dumesic and A.A. Trevifio, J. Catal. 116 (1989) 119.
[57] T.L. Hill, An Introduction to Statistical Mechnics (Addison-Wesley, Reading, 1960)
pp. 235-260.

You might also like