Journal of Industrial and Engineering Chemistry 66 (2018) 176–186

Contents lists available at ScienceDirect

Journal of Industrial and Engineering Chemistry

journal homepage: www.elsevier.com/locate/jiec

Process simulation of ammonia synthesis over optimized Ru/C catalyst

and multibed Fe + Ru configurations
Antonio Tripodi, Matteo Compagnoni, Elnaz Bahadori, Ilenia Rossetti*
Chemical Plants and Industrial Chemistry Group, Dip. Chimica, Università degli Studi di Milano, INSTM Unit Milano-Università and CNR-ISTM, via C. Golgi, 19,
I-20133 Milano, Italy

A R T I C L E I N F O A B S T R A C T

Article history:
Received 20 February 2018 Ammonia synthesis over different iron- and ruthenium-based catalysts was modelled with appropriate
Received in revised form 10 May 2018 rate models, used for the simulation of the process under different configurations and conditions. The
Accepted 19 May 2018 kinetic models have been simulated and validated against experimental data. A scaled up reactor has
Available online 26 May 2018 been designed, at first with a once through configuration. On this model reactor we performed a
sensitivity analysis to optimise the reaction conditions. Then, the sizing of an ammonia separation unit
Keywords: and the optimisation of the recycle loop allowed to compare different possible configurations.
Ammonia synthesis A multibed catalytic reactor with intercooling was then designed, using the same catalyst or different
Reactor modelling
catalyst types properly to maximise the ammonia productivity. In particular, Fe-based catalysts were
Multibed catalytic reactors
followed by the Ru/C one, in order to push the ammonia productivity towards the equilibrium value.
Kinetic modelling
Temkin equation The goal of the work is the design of an ammonia synthesis loop which couples different catalysts
towards the optimisation of productivity and cost of operation and installation.
© 2018 The Korean Society of Industrial and Engineering Chemistry. Published by Elsevier B.V. All rights
reserved.

Introduction surface are fundamental steps. Although the underlying chemistry

is relatively well known [7,8], very demanding operating con-
The production of sufficient food to sustain the huge growth of ditions for pressure and temperature are needed. For this reason,
the world population, mainly in developing countries and newly the technology is very sophisticated and its study continued during
industrialized ones, represents one of the priorities of the the last century, both in the scientific and industrial communities.
international community. This trend relates to the rising expect- Iron based catalyst can be used at 150–250 bar and 380–520
C [9],
ations of better living standards, that lead to a higher food demand. they are sufficiently active and not expensive, but the maximum
A study on food production relative to the most populated country, conversion achievable is limited by inhibition by ammonia. The
China, established a linear proportionality between food produc- research to decrease the operating pressure is still in progress [10–
tion and chemical fertilizer production [1] and a similar trend can 14], relying basically on the search of more active materials. The
be observed for India [2]. Generally, chemical fertilizers market process includes specifically designed reactors, where a multibed
dominates also in the developed northern countries, especially in configuration is the most employed. For instance, with traditional
central United States, Canada and Europe [3]. Nowadays, fertilizers Fe-magnetite 4 beds of catalyst, with inter-cooling after each pass
production is centralized and the core of the process is constituted lead to a conversion rate up to 15% and a recycling mechanism to
by ammonia synthesis. The latter is mainly accomplished through achieve up to 98% conversion.
the well known Haber-Bosch process [4] on a plant scale of Ruthenium showed higher activity than Fe for ammonia
1000 ton/day (100 million tons per year worldwide). This technol- production and it is not inhibited by the product, is as sensitive
ogy represents also the major success of heterogeneous catalytic as Fe to oxygen-containing impurities, but it is less sensitive to
process used on a large scale [5,6]. sulphur [15,16].
Ammonia synthesis is a complex heterogeneous catalytic The industrial application of Ru on graphitized carbon was
process, where N2, H2 and NH3 adsorption/desorption on catalyst achieved in the Kellogg’s Advanced Ammonia Process (KAAP). This
process is active in seven world-scale plants, each producing ca.
2000 tonnes per day of ammonia [17] using a “low pressure”
* Corresponding author. synthesis loop operating down to 90 bar. The Ru catalyst is placed
E-mail address: [email protected] (I. Rossetti). in different fixed beds and the higher activity of Ru allows lower

https://doi.org/10.1016/j.jiec.2018.05.027
1226-086X/© 2018 The Korean Society of Industrial and Engineering Chemistry. Published by Elsevier B.V. All rights reserved.
 A. Tripodi et al. / Journal of Industrial and Engineering Chemistry 66 (2018) 176–186 177

operating pressure than Fe. The main drawbacks of Ru/C catalysts The reaction proceeds through several adsorption and dissocia-
are (i) their inhibition by hydrogen, which prompts for the use of tion steps [36]. The related kinetic parameters are complex functions
understoichiometric reaction conditions and (ii) the poor carbon of the surface coverage and structure. The best kinetic equation, as
resistance towards methanation, which is catalysed by Ru itself determined by comparison between various rival models, was based
and may corrode the support promoting catalyst sintering and on the Temkin one, modified on purpose considering the different
deactivation [18]. features of Ru with respect to Fe. This point is crucial because, as
Aspen Plus© process simulator has been effectively used to already mentioned, Fe-based catalysts are kinetically inhibited by
investigate high pressure processes, such as methanol synthesis the strong NH3 adsorption, while Ru is inhibited by H2 due to its
[19,20], coal gasification [21], biomass gasification [22], Fischer– competitive adsorption with N2. For this reason, in the denominator
Tropsch synthesis [23], partial oxidation of natural gas [24]. A basic of the rate-equation we added an additional term to account for
requirement for a reliable simulation of reactive systems, is the hydrogen adsorption (KH2(aH2)0.3, as reported in Eq. (2) [33]:
availability of appropriate kinetic equations [25]. Surprisingly, only    

0:5 ðaH2 Þ0:375 ðaNH3 Þ
0:75

few authors tried to simulate the ammonia synthesis process aN 2  K1a

dh ðaNH3 Þ
0:25
ðaH2 Þ
1:125

implementing detailed and up-to-date kinetic expressions. Yu et al. ¼ klðqÞ
0:3
0:2 ð2Þ
tried to implement the kinetic parameters of ammonia synthesis for dt 1 þ K H2 aH2 þ K NH3 aNH3
the evaluation of a coal-based polygeneration process to coproduce
where Ka represent the equilibrium constant, dh/dt represented
synthetic natural gas and ammonia [26], but with insufficient detail
the rate of consumption of the defective reactant in mol h1
on kinetics to be reproduced. The control structure design for an
dm3cat, k is the kinetic constant of the direct reaction and ai the
ammonia synthesis process was reported by Araujo and Skogestad
activities of reactants and product. l(q) is a stoichiometric
[27], the reaction kinetics being based on the Temkin–Pyzhev
parameter set to 1 or 1.2 when the H2/N2 feeding ratio is 3 or
expression. On the other hand, Arora et al. investigated small scale
1.5 Adsorption equilibrium constants at the denominator were
ammonia production from biomass [28], but using Gibbs and
obtained considering the DHads and DSads for H2 and NH3 as
equilibrium reactors were used to model the ammonia converter.
reported in Eqs. (3)–(4) [33]:
Ammonia production via integrated biomass gasification was also
studied by Andersson and Lundgren [29]. Also in this case the reactor 56:9024 37656
loge K H2 ¼  þ ð3Þ
was simulated using a Gibbs reactor, with an iron catalyst. R RT
Neglecting the kinetics, the overall process can still be designed
relying on the mass-balances, only, but optimal operating
conditions cannot be safely identified, nor the reactors sized, 34:7272 29228
loge K NH3 ¼  þ ð4Þ
since for this reaction kinetics and thermodynamics obey to R RT
opposite requirements. The behaviour of the already mentioned iron catalysts was
Therefore this paper proposes the simulation of a relatively low instead modelled without resorting to adsorption terms, with
pressure ammonia synthesis reactor (below 120 bar). The kinetic simpler power-law rate laws (Eqs. (5) and (6), respectively from
description of the process was based on an original kinetic model, Refs. [31,33] and [35]):
derived on a patented promoted Ru/C catalyst, obtained as a result     
of 20 years of research in our laboratory: the model was developed dh aH2 2:25 aNH3 0:5
¼ klðqÞ ðKa Þ2 aN2  ð5Þ
in a previous investigation, modifying the Temkin equation with dt aNH3 1:5 aH 2 0:75

the addition of H2 and NH3 adsorption terms, in order to consider

their possible concurrent inhibiting effect [30]. The full details on
  1:5   
catalyst properties and its performance under industrially relevant dh aH 2 aNH3 1
¼ k ðKa Þ2 aN2  ð6Þ
conditions can be found elsewhere [31–33]. Temperatures below dt aNH3 1 aH2 1:5
500
C are suitable for carbon supported catalysts, preventing the
methanation of the graphitised carbon used as support [30,33,34].
Finally, the preliminary process design was here carried out by Thermodynamic analysis
comparing a commercial catalyst [35] with advanced, second
generation Fe-based catalysts tested previously in our laboratories The equilibrium constant was originally calculated according to
[31] and both these with our home-developed Ru sample. Different Gillespie and Beattie as reported elsewhere [8]:
multi-bed arrangements were compared and discussed. To better 59:9024 37656
focus the scope of the paper, we have studied the synthesis section log10 K eq ¼  þ  2:691122  log10 T
R RT
only, since the rest of the ammonia process is consolidated.  5:519265  105 T þ 1:848863  107 T 2 ð7Þ
The goal of the work is then the design of an ammonia synthesis
loop which couples different catalysts towards the optimisation of The equilibrium ammonia concentration in stoichiometric mix-
productivity and cost of operation and installation The body of the tures (i.e. containing 3 moles of hydrogen per mole of nitrogen) are
paper includes the applicative results whereas the details on the reported in Fig. S2. These preliminary calculation represent the
selection of the best thermodynamic package to model the system basis in order to overcome the activation energy without depressing
and computational/convergence issues are reported in the the equilibrium condition. The pressure range investigated is
Appendix. relatively narrow and shifted to low values, as in the scope of this
work.
Modelling and simulation Although we focus on the reactor loop in this paper, besides the
reactor, at least the gas–ammonia separation stage deserves a
Kinetic formulations similar equilibrium analysis, since also the removal efficiency of
the liquefied product affects deeply the overall mass balances. The
The ammonia synthesis process follows the stoichiometry charts mapping the previsions of a single-stage equilibrium
reported in equation (1). separation are shown in Fig. 1. Notice that the maximum ammonia
concentration in the recycle stream cannot be greater than the
N2 + 3H2 ? 2NH3 DH0 = –46 kJ mol1 [17] (1) value determined by the reactor outlet conditions.
178 A. Tripodi et al. / Journal of Industrial and Engineering Chemistry 66 (2018) 176–186

Fig. 1. Vapor–liquid split fraction of ammonia in a stoichiometric H:N = 3:1 (mol/mol) mixtures at equilibrium, calculated via the RKS-BM thermodynamic model. The dashed
lines represents recognized power-law trends of the calculation.

Simulation of the micropilot plant with a standardized NaOH solution. The flowsheet of the
micropilot plant is reported in Fig. S1. Kinetic tests were carried
The kinetic model was validated by simulating the micropilot out on the Ru-based catalyst, on the magnetite and wustite Fe-
plant used to derive it. The experimental data for reactor outlet based ones, according to a full experimental design, as detailed
under different operating conditions were reported in the elsewhere [31,33].
literature and were derived as summarised in the following. Aspen Plus© provides a built-in ‘LHHW kinetics’ for calculating
Details about the preparation route of the support, catalyst and the rate of reactions described with a Langmuir–Hinshelwood–
activation can be found in the related patent [9] and papers Hougen–Watson (LHHW) model as in our approach. This kinetic
published elsewhere [30]. Briefly, a graphitised carbon was used as model consists of a (dimensioned) ‘kinetic factor’, a ‘driving force’
support and promoters were added by impregnation from aqueous expression (with such dimensions as required to get those of r to
solutions of hydroxides (K and Cs) or nitrates (Ba), in the optimal match those of k0) and an adsorption term (dimensionless) as in
amount for obtaining: Ba/Ru = 0.6 (mol/mol), Cs/Ru = 1 (mol/mol), Eq. (8). The reaction occurs in the vapour phase and the reaction
and K/Ru = 3.5 (mol/mol). The Ru content, referred to the final rate is based on catalyst weight. The chemical species concur to the
catalyst weight, was 3.2 wt%. driving force with their fugacity f powered by the proper
To extend the comparison, a Fe-based benchmark catalyst from exponents for forward and backward reactions (terms ‘for’ and
magnetite [35] was taken as a reference, and newer commercial ‘rev’ respectively). This approach was followed, instead of
materials of magnetite and wustite [31] were considered too. implementing a dedicated reaction calculation algorithm, because
When compared under the same operating conditions as for the aim of this work is to provide a comparison between the
temperature (430
C) and pressure (100 bar), GHSV = 30,000 h1, intrinsic kinetics of different catalysts (for whom the built-in
but with optimised H2/N2 molar ratio of 1.5 for Ru and 3 for Fe, the Aspen Plus© module is enough), rather than to provide a detailed
Ru/C catalyst overperformed (ca. 15.5 vol% NH3) both the Fe-based reactor model (which would require customized rate expressions
catalyst from magnetite (ca. 8 vol% NH3) and from wustite (ca. 12.5 to address the mass-transport phenomena and axial/radial
vol% NH3). Thia meand that it is possible to decrease by more than dispersion models).
40% the operating pressure or to achieve more than 40% higher   Y n Y
Ea n
conversion per pass under the same operating conditions. Similar k0 eRT Kf or r f r r  Krev p f p p
r¼ X Y ð8Þ
advantages are reported for the commercial KAAP process based ni e
i
K i i
f i
on Ru/C catalyst.
A continuous tubular reactor of 9 mm internal diameter, The consistency of the optimised parameters was checked by
400 mm long, was adopted to collect the kinetic data, by feeding the Arrhenius and Van’t Hoff equations, leading to the values
downflow a reactant gas mixture consisting of hydrogen and reported in Table 1. In order to implement the data in Aspen
nitrogen in different volumetric ratio (3:1 or 1.5:1 v/v), with Plus©, the pre-exponential factor was corrected considering the
different Gas Hourly Space Velocity (GHSV) through a catalyst bed catalyst density equal to: 0.59 g cm3 (as experimentally
of 0.15–0.25 mm particles. The GHSV was defined as the ratio of determined for the Ru/C material), 2.8–2.9 g cm3 (magnetite)
the inlet volumetric gas flow rate at normal conditions and the and 3.25 g cm3 (wustite). This expression was applied with
volume of the catalyst bed, Nm3 mcat3 h1. The catalyst was e = 1 for model (2), zero otherwise. Table 1 gives a quick reference
diluted with quartz of the same particle size in ratio catalyst/ to the kinetic and adsorption parameters used during reactor
quartz = 1/22 (v/v), to limit the hot spot across the bed. Before the simulation.
run the catalyst was activated in situ in a flow of the same The ‘RPlug’ model was adopted in both isothermal and
reactant gas mixture at 30 bar, 450
C, GHSV = 20,000 hl for 5 h in adiabatic conditions. The Peng–Robinson and Redlich–Kwong
the case of the Ru/C catalyst. For Fe-based samples, the equation of states were adopted as thermodynamic models to
temperature was increased by 50
C/h from 25 to 350
C, then compute the non-ideality of the gaseous mixture, yielding results
by 10
C up to 500
C. The whole activation lasted for 2 days. in very good agreement for every single-pass calculations of the
Activity has been determined by evaluating the volumetric ammonia yield. For the calculations involving the closed-cycle
concentration of ammonia in the effluent gas, by bubbling it in an synthesis, however, the (modified) Redlich–Kwong–Soave EOS
excess of sulphuric acid (0.1 M) and back-titrating the excess acid was used (see Supporting information).
 A. Tripodi et al. / Journal of Industrial and Engineering Chemistry 66 (2018) 176–186 179

Simulation of the full scale reactor

7.8  104
(kmol s1 kg1cat)
NH3

NH3
To perform a screening of the different catalysts performances

1

D
1

0
1.72  108
under similar reaction conditions, the ammonia synthesis was

5.42
1.5 modelled according to the following criteria (Fig. 2):

1.5
1.5
H2

H2
k0

0
Eq. (6) (‘KM’ Iron-based)

 reactants conversion is carried out into 3 beds with intercooling

9218
stages;

0
 the beds thermal conditions – either at constant temperature or
Thermodynamic and kinetic parameters for the reaction models according to the Aspen Plus© formulation (8), where the equilibrium constants are expanded as: ln(K) = A + B/T + C  ln(T) + D  T.

duty – were fixed;

(kcal mol1)

 only the reactants loop was modelled, while the synthesis gas
40.76

feed and the separation sections were omitted, assumed as

0.5

4.9
5.76
N2

N2

na
Ea

A
1
0

standardised.
7.8  104

The target yield was set on a ton/day scale (55 kg/h of ammonia
every 60 kg/h of fresh stoichiometric gases, plus a small quantity of
D

methane). The gases enter the flowsheet at atmospheric pressure,

so to have a rough evaluation of the compression work required by
5.42
1.5
NH3

NH3
(kmol s1 kg1cat)

0.5

the plant sections upstream, the nominal cycle pressure was set to
C
1

100 bar, and the separation section was lumped into a single
equilibrium stage at 30
C, in order to approximately foresee its
3.23  109

gross cooling duty. The former assumption on the gas inlet at

1.5
H2

k0

atmospheric pressure is very conservative, since the steam

reforming of natural gas provides compressed hydrogen, but it
0.75

9218

set as internal comparison and reference.

2.25
H2

Despite the configuration simplicity, the single loop contains

two highly non-linear blocks (the reactor and the separator) which
Ea (kcal mol1)

may compromise the calculation convergence at too low purge

(Magnetite)

fractions. On the other hand, a manual short-cut calculation needs

Eq. (5)b

assumption on the single-pass conversion that have to be checked

0.5

7.8
2.88
47.5
N2

N2

na

against the actual behaviour of the kinetic model. A purge fraction

A
1
0

as high as 1% introduces only minor convergence issues (setting a

7.8  104

‘Broyden’ algorithm, tolerance: 1 105), provided the initial

k0 (kmol s1 kg1cat)

recycle mass flow and composition are properly guessed. The

1.5
NH3

NH3

0.5
D
1

reported results are achieved under this latter condition.

7.47  108

Results and discussion

5.42
0.75
1.5

2.25
H2

H2

Validation of the kinetic model

9218
B

The validation of the kinetic model was carried out using directly
Aspen Plus©, in order to check in advance the computational
(kcal mol1)

consistency of the equation. In general, a good agreement between

(Wustite)
Eq. (5)

the experimental and simulation results was obtained for a H2/N2

0.5

7.8
2.88
N2

N2
45

feeding ratio equal to 3, in the whole range of temperature, pressure

na
Ea

A
1
0

and space velocity experimentally explored (GHSV = 50,000–

1.27  104

400,000 h1, T = 360–430
C, P = 70–100 bar). Fig. S3 exemplifies

the results for the test at T = 430
C, P = 70 bar and H2/N2 = 3 (v/v).
(kmol s1 kg1cat)

0.25
0.75
NH3

NH3

NH3

0.2

For the tests with H2/N2 ratio equal to 1.5 (v/v) only the
D

D
1

0
0

0
0
0

simulation at temperature below 430
C revealed a satisfactory

1.125

agreement between the simulated ammonia fraction in the reactor

0.375
1.5

2.69
426

0.3

outlet gas and the experimental values. In Figs. 3 and 4 two

H2

H2

H2
k0

C
0

0
0
0

examples are reported considering the best and the worst cases
4609

respectively.
4529
3523

The maximum discrepancy between the calculated and the

B

B
0

experimental values were observed at the highest temperature and

Eq. (2) (Ru/C)

(kcal mol1)

lowest space velocity (i.e. at highest contact time). This occurs for
the most favourable understoichiometric feeding ratios.
1.876

6.48
10.3
7.19
0.5

23.0

The use of a very active catalyst such as the one adopted in this
0.5
N2

N2

N2
Ea

A
0

0
0
0

work can significantly reduce the contact time maintaining high

ammonia yield. Therefore, a trade off with productivity has to be
Forward term (nr)
Reverse term (np)

Adsorption term
Kinetic constant
Rate expression

Rate expression

newly searched for this and the other operating parameters, as

Stoichiometry

Term 2 (Krev)
Term 1 (Kfor)
Coefficients

Coefficients
Term 1 (Ki)
Term 2 (Ki)
Term 3 (Ki)
Term 1 (ni)
Term 2 (ni)
Term 3 (ni)

follows. A similar validation of the kinetic models for the Fe-based

Exponents

Exponents

catalysts was also done, not reported for the sake of brevity.
Table 1

A comparative table of the performances of Ru-based catalysts

is also provided (Table 2). The comparison evidences that very high
180 A. Tripodi et al. / Journal of Industrial and Engineering Chemistry 66 (2018) 176–186

Fig. 2. Basic layout employed for the simulation of an ammonia synthesis cycle [38,46]. The data are relative to a run of case 4 (Table 2) with a lower reactor inlet temperature.

Fig. 3. A) Example of data simulation for test at T = 430
C, P = 70 bar and H2/N2 = 1.5 Fig. 4. A) Example of data simulation for test at T = 460
C, P = 70 bar and H2/N2 = 1.5
(v/v). Experimental points (orange circles), simulated values (green squares). B) (v/v). Experimental points (orange circles), simulated values (green squares). B)
Parity plot. (For interpretation of the references to colour in this figure legend, the Parity plot. (For interpretation of the references to colour in this figure legend, the
reader is referred to the web version of this article.) reader is referred to the web version of this article.)

conversion per pass is achievable with this kind of active phase, optimum value of space velocity is calculated for every specified
with very high productivity attained when high conversion can be process according to the pressure of the system, the structure of
maintained at high space velocity. Of course, given the decrease of reactor and energy costs. For a low-pressure ammonia synthesis
ammonia concentration at reactor outlet when increasing the loop, a space velocity in the range of 5000–10,000 h1 is commonly
space velocity, this point should be matched with the sizing of the chosen, while values between 15,000 and 30,000 h1 or
recycle and the relative separation and compression duties. 60,000 h1 are commonly selected for middle and high-pressure
systems, respectively, using traditional iron catalysts [37].
Space velocity and temperature effect Fig. 5 compares the ammonia production predicted by the
Examples of the relationship between activity and space kinetic model for low and high residence times (GHSV = 2.0  105
velocity are shown in Fig. S3, Figs. 3 and 4. Increasing space h1 and 0.5  105 h1, respectively) with the predictions of the
velocity will normally decrease outlet ammonia concentration per thermodynamic model calculated using the ‘RGibbs’ reactor
pass, but it increases total ammonia production flow rate, thus model. This kind of reactor model is used to evaluate the
increasing the recycled flow. The decrease of ammonia concentra- conversion at chemical equilibrium through minimization of the
tion in the outlet stream led to a decrease of ammonia in the Gibbs free energy of the mixture. The comparison was done for a
recycle stream. Moreover, the outlet temperature of the converter range of reactor temperatures under isothermal conditions
was lower due to a lower temperature raise under adiabatic choosing the H2/N2 ratio equal to 1.5 (v/v). Ru catalyst, as
condition. Another important issue was the reduction of energy aforementioned, is preferably operated with understoichiometric
consumption for cooling the recycle gas. Generally, an economic ratios because of the inhibition effect of H2.
 A. Tripodi et al. / Journal of Industrial and Engineering Chemistry 66 (2018) 176–186 181

Table 2
Comparison of different Ru-based catalysts performance.

Catalyst T (
C) P (bar) GHSV (h1) NH3 vol% NH3 productivity (kg h1 dm3cat) Reference
Ba-K-Ru/C 400 100 10,000 20 1.5 [43]
Mg-Ba-Ru/C 400 100 10,000 21 1.6 [44]
K-Cs-Ba-Ru/C 430 100 50,000 14.1 5.3 [33]
Ba-Ru/MgO 400 90 – 11 – [45]

Fig. 5. Comparison of ammonia production predicted by the here proposed kinetic

model at low and high residence times (GHSV = 2.0  105 h1 and 0.5  105 h1,
respectively) and thermodynamic model under isothermal condition (P = 70 bar, H2/ Fig. 6. Comparison of ammonia production calculated by kinetic model (2) at low
N2 = 1.5 v/v), calculated using Gibbs reactor or the equilibrium constant adopted for and high residence times (GHSV = 2.0  105 h1 and 0.5  105 h1 respectively)
the kinetic expression (Gillespie). under adiabatic condition (P = 70 bar, H2/N2 = 1.5 v/v), with respect to the reactor
inlet temperature. The maxima of the isothermal analysis are anticipated since
those thermal conditions are met, in this case, downstream along the reactor axis.
The ammonia production increases as the reactor temperature
increases, due to kinetic reasons, reaches a maximum at a
temperature strictly depending on GHSV (430
C and 465
C for
the high and low contact time, respectively) and then decreases with
temperature following the equilibrium conversion. The reaction
shows then a transition from a kinetic-limited to an equilibrium-
limited regime of operation due to the exothermicity of the reaction
(DH0 = 46 kJ mol1) [17]. Suitable reactor configurations (mod-
elled, in principle, as interconnected isothermal and adiabatic
stages), aiming then at keeping a temperature as high as to favour the
kinetic, but as low as to remain in a convenient equilibrium regime.
The isothermal choice implies the direct heat exchange within the
bed, ensuring the proper temperature at the inlet of the reactor and a
homogeneous thermal profile along the reactor length. This choice
implies the use of multitubular reactor configuration to achieve
optimal heat exchange. The second choice is usually more common
for the first stages of ammonia reactors [38].
Ammonia volume fraction at the outlet of the reactor predicted
by the kinetic model under adiabatic conditions is shown in Fig. 6.
The threshold temperatures at which the transition from kinetic to Fig. 7. Comparison of ammonia production predicted by the kinetic model at
thermodynamic condition occurs, are 355
C and 400
C for the different H2/N2 ratios GHSV = 0.5  105 h1, P = 70 bar).
high and low contact time respectively.
It should be noticed that the Gillespie model predicts a lower
ammonia equilibrium conversion than those obtained through of H2/N2 ratio as expected. On the other side, the opposite trend in
minimisation of the Gibbs free energy. The Gillespie equation for the kinetic regime confirmed the consistency of the kinetic
Keq was applied by us during the development of our kinetic model, expression for ruthenium catalyst. Indeed, Ru is inhibited by H2
thus it correctly represents the limit towards which our reactor and, thus, low H2/N2 ratios are more advantageous in terms of
prevision converges. ammonia conversion with respect to the stoichiometric one. The
kinetic expression considers the competitive adsorption between
Influence of feed composition H2 and N2 on the Ru-based catalyst (Eq. (2)).
The effect of H2/N2 ratio on the activity of the novel Ru/C
catalyst is shown in Fig. 7. Temperature was set at 430
C. After the Effect of reaction pressure
maximum conversion, where a thermodynamic regime governs In order to study the pressure effect, the lowest value of GHSV
the process, the ammonia production increases with the increasing considered (0.5  105 h1) was used under isothermal conditions.
182 A. Tripodi et al. / Journal of Industrial and Engineering Chemistry 66 (2018) 176–186

Full scale reactor: single-pass performances

The first analysis was made feeding a scaled-up syngas flow to 3

adiabatic reaction beds loaded with commercial iron catalyst, whose
kinetics are represented by Eqs. (5) (Wustite) and (6) (‘KM’ promoted
Iron, as per [35] and references therein). The configuration included
intercooling exchangers, which imposed the bed inlet temperatures
and, at first, was without any recycle (i.e. the vapor exiting the
separator was entirely purged). As expected, a lower temperature
granted higher product yields (Fig. 9), but required more catalyst to
reach the maximum conversion. The replacement of the iron-based
catalyst with the ruthenium-based one, i.e. of expression (5) with
expression (2) for the last bed, appreciably improved the reactor
performance at low catalytic loads. The same yield of 15 kg/h at low
temperature (380
C) was reached with 20 kg of the Wustite
material and 10 kg of the ruthenium-based one, with respect to
90 kg of iron alone. By contrast, operating at 430
C half of these
loads were needed to reach the maximum conversion allowed by
Fig. 8. Comparison of ammonia concentration in the outlet gas predicted by the the different equilibrium condition.
used kinetic model at different pressures (GHSV = 0.5  105 h1, H2/N2 = 1.5 v/v).
Recirculated reactor performances

The combination of the kinetic models with the overall mass

balances is shown in Figs. 10 and 11 for the same catalyst
combination tested above. It can be noticed that, at intermediate
catalyst load (>5 kg in each bed), the properties of the Ru-based
material can improve the ammonia yield even at low temper-
atures. For example, 20 kg of Ru-based catalyst replacing 20 kg of
the Fe-based one in the last two stages, at an inlet temperature of
380
C and 100 bar, give 58 kg/h of ammonia instead of 52 kg/h, ca.
10% higher yield. On the other hand, a quarter of this load achieved
the same performance in the temperature range 400–450
C.
Overall, while thermodynamics predicts that higher conversions
are expected for low temperature and high pressure, with a
monotonous decrease of conversion with increasing temperature,
the addition of kinetics clearly defines two distinct operating regimes.
Under kinetic regime (before the maximum evidenced in Figs. 5, 6, 10
and 11), an increase of temperature improves conversion, whereas
Fig. 9. Open-loop comparison of the NH3 yield for 3 adiabatic reaction beds, loaded after the maximum the system operates under thermodynamic
each with the same amount of different catalysts and kept at the specified inlet regime, with decreasing conversion at increasing temperature. This
temperature. ‘KM’ catalyst is described by kinetic (6), while in simulation ‘Wus’ (for transition temperature (that corresponding to the maximum, strictly
Wustite) and ‘Ru’ (Ruthenium) models (5) – (2) respectively are adopted.
depends on the operating pressure and reactants molar ratio.

The simulation results are shown in Fig. 8. The relationship Effect of cycle pressure
between the ammonia concentration at the outlet and reaction
pressure is roughly linear: the higher the pressure, the higher Restricting the analysis to the mixed catalyst layout (i.e. first bed
would be the concentration of outlet ammonia. of commercial iron-based catalyst plus one bed loaded with the

Fig. 10. Closed-loop ammonia yields for 3 adiabatic reaction beds following model (5) (left) and models (5)–(5)–(2) (right) at the specified inlet temperatures and pressures.
 A. Tripodi et al. / Journal of Industrial and Engineering Chemistry 66 (2018) 176–186 183

Fig. 11. Closed-loop recycle flows for 3 adiabatic reaction beds following model (5) (left) and models (5)–(5)–(2) (right) at the specified inlet temperatures and pressures.

Fig. 12. Closed-loop simulation results for 3 adiabatic reaction beds following the kinetic models (5)–(5)–(2) at the selected inlet temperature of 430
C. The heat duties are
negative as they represent released heat.

ruthenium-based one) at the preferred reaction temperature of Reactor thermal profile for multibed configurations
430
C, the effect of pressure on different outputs is shown in
Fig. 12. Actually, the only parameter appreciably affected by this At this point, the analysis was extended varying the bed
variation is the compression duty needed to reach the desired loadings, at the fixed working pressure of 100 bars and at
operating pressure, while the total recycle flow depends essen- variables inlet temperatures around the central value of 400
C.
tially on the catalytic load (i.e. on the capability of the reactor to The results are represented in Fig. 13, that shows the thermal
yield as much ammonia as the fresh feed at the highest possible condition and ammonia concentration along the reactor axial
concentration). The separation duty follows accordingly, since it is coordinate (stages lengths: 8 m, 8 m and 4 m), supposing to load
dominated by the latent heat of ammonia. For instance, the each bed with different amounts of a catalyst obeying to the
decrease of the operating pressure from 125 to 100 bar can different kinetics model as in Table 3. The inter-cooling
decrease the gross compressors duty by ca. 7%, computed as exchangers were not modelled rigorously, because in this phase
difference of the compressors block duty for the two cases. we were only interested in considering the fate of the reactant
184 A. Tripodi et al. / Journal of Industrial and Engineering Chemistry 66 (2018) 176–186

Fig. 13. Reactor temperature and ammonia concentration for the closed-loop simulations summarized in Table 2.

Table 3
Summary of the input specifications for the test-cases at variable catalysts mix. In any case the ammonia yield is of 56–57 kg/h and the total compression power of 69–70 kW.
Eq. (2) represent the Ru-based catalyst, Eqs. (5) and (6) the newer magnetite/wustite or the older commercial Iron respectively.

Case P Tin Bed 1 Bed 2 Bed 3 Separator NH3

duty max
fraction
Catalyst Kinetic Temperature profile Catalyst Kinetic Temperature profile Catalyst Kinetic Temperature profile
Load model Load model Load model
# (bar) (
C) (kg) (Eq. no) (kg) (Eq. no) (kg) (Eq. no) (kW) (mol/
mol)
4 100 430 10 (5) Adiabatic + intercooling 10 (5) Adiabatic + intercooling 5 (2) Adiabatic + intercooling 206 9.56%
10 100 400 15 (5) Adiabatic + intercooling 15 (5) Adiabatic + intercooling 10 (2) Adiabatic + intercooling 196 9.66%
21 100 400 20 (5) Adiabatic + intercooling 20 (5) Adiabatic + intercooling 20 (2) Adiabatic + intercooling 151 13.2%
22 100 400 20 (5) Adiabatic + intercooling 20 (5) Adiabatic + intercooling 20 (5) Adiabatic + intercooling 237 7.47%
32 100 430 10 (5) Adiabatic + intercooling 10 (5) Adiabatic + intercooling 10 (2) Constant at inlet Temp. 155 12.4%
35 100 400 5 (5) Adiabatic + intercooling 7 (2) Imposed rise 10 (2) Imposed fall 140 13.4%
38 100 400 7 (5) Adiabatic + intercooling 3 (2) Imposed rise 7 (2) Imposed fall 164 11.5%

mixture and the amount of heat to be subtracted to lower the energy production cycles (e.g. a Rankine cycle) to partially sustain
temperature to the desired set point. Further work will compare the compressors duty.
alternative options including the preheating of the reactant Cases 4–10 show how the catalyst loading must be increased,
mixtures and the recovery of the available reaction heat to sustain to achieve the same ammonia yield, when the reactor
 A. Tripodi et al. / Journal of Industrial and Engineering Chemistry 66 (2018) 176–186 185

inlet temperature is relatively low respect to the activation RGibbs Reactor for minimum Gibbs energy
energy of the Wustite-based material. Notice that the Ru-loaded EOS Equation of state
section, instead, works essentially in the same way since the
activation energy for this material is the half than for Fe-based
samples. Symbols
Cases 21–22 represent a complementary comparison, showing a Activity
how the substitution of Wustite with Ruthenium in the last stage D Matrix determinant
effectively increases ammonia concentration in the recycled Ea Activation energy
stream: in this case, the higher temperature reached after the f Fugacity
first stage is due essentially to the decreased gross heat capacity of F Mass flow
the flowing mixture, while in the last stage this phenomenon is h Conversion
also enhanced by the higher activity of the Ru-material when the k0 Kinetic prefactor
hydrogen content is lower. The second stage works more or less in K Equilibrium constant
the same way because, for Iron-based catalysts, it is a higher P Pressure
ammonia content that limits the reaction rate, and then the effect r Reaction rate
of a reduced gas flow is roughly compensated. R Gas constant
More generally, in order to produce ammonia gradually along s Split fraction
all the reactor profile, lower inlet temperatures are preferable if the T Temperature
heat released by the reaction can be exploited to overcome the t Contact time
activation energy, and the activity of the Ruthenium catalyst
Appendix A. Supplementary data
towards ammonia-enriched mixtures can keep the reaction going
on in the last stage.
Supplementary data associated with this article can be found, in
On this basis, also a strategy for the catalyst economization can be
the online version, at https://doi.org/10.1016/j.jiec.2018.05.027.
sketched, relaxing the assumption of working with strictly adiabatic
beds. If, for example, a first Wustite bed is operated adiabatically with
an inlet temperature and catalyst load that grant a high activity (case References
#32), then a following bed of the same material would experience a
[1] Z.L. Zhu, D.L. Chen, Nutr. Cycl. Agroecosyst. 63 (2–3) (2002) 117, doi:http://dx.
too high ammonia content, while its exiting temperature would yet doi.org/10.1023/A:1021107026067.
be enough to keep an isothermal Ru-loaded bed active. Case #35 [2] P. Potter, N. Ramankutty, E.M. Bennett, S.D. Donner, Earth Interact. 14 (2)
(2010), doi:http://dx.doi.org/10.1175/2009EI288.1.
shows how a similar performance can be reached, with much less
[3] M. Reese, C. Marquart, M. Malmali, K. Wagner, E. Buchanan, A. McCormick, E.L.
Wustite and more Ruthenium-based material, if the temperature Cussler, Ind. Eng. Chem. Res. 55 (13) (2016) 3742, doi:http://dx.doi.org/
profile of the reactor is properly managed: in case 38 this strategy is 10.1021/acs.iecr.5b04909.
furtherly worked out in order to achieve a still reasonable ammonia [4] T. Kandemir, M.E. Schuster, A. Senyshyn, M. Behrens, R. Schlögl, Angew. Chem.
Int. Ed. 52 (48) (2013) 12723, doi:http://dx.doi.org/10.1002/anie.201305812.
concentration with (ideally) as little Ruthenium-based catalyst as [5] N. Cherkasov, A.O. Ibhadon, P. Fitzpatrick, Chem. Eng. Process. Process Intensif.
possible. Then, only the final stages have to be loaded with 90 (2015) 24, doi:http://dx.doi.org/10.1016/j.cep.2015.02.004.
Ruthenium, and their performance maximized if a temperature [6] R. Lan, J.T.S. Irvine, S. Tao, Int. J. Hydrogen Energy 37 (2) (2011) 1482, doi:http://
dx.doi.org/10.1016/j.ijhydene.2011.10.004.
decrease (leading towards better equilibrium conditions, once the [7] M.I. Temkin, V. Pyzhev, Acta Physicochim. URSS 12 (1940) 327.
great part of the ammonia was produced), rather than adiabatic [8] L.J. Gillespie, J.A. Beattie, Phys. Rev. 36 (1930) 743.
operation, can be assured. This assumption is probably the one that [9] L. Forni, N. Pernicone, Catalysts for ammonia synthesis. PCT Int. Appl. (2003)
WO 2002-EP11707 20021018, 2002.
needs more development in further works. Nevertheless a similar [10] B. Lin, Y. Qi, Y. Guo, J. Lin, J. Ni, Catal. Sci. Technol. 5 (2015) 2829, doi:http://dx.
management of ammonia reactor is not entirely new nor unfeasible doi.org/10.1039/C5CY00014A.
[39–41], while other studies report at least a flat temperature profile [11] M. Karolewska, E. Truszkiewicz, B. Mierzwa, L. Keopin  ski, W. Raróg-Pilecka,
Appl. Catal. A Gen. 445–446 (2012) 280, doi:http://dx.doi.org/10.1016/j.
as a better choice for the final stages [42]. apcata.2012.08.028.
Some computational details are reported for the interested [12] C. Fernández, C. Pezzotta, E.M. Gaigneaux, N. Bion, D. Duprez, P. Ruiz, Catal.
reader in the Supplementary information file. Today 251 (2015) 88, doi:http://dx.doi.org/10.1016/j.cattod.2014.11.010.
[13] K. Narasimharao, P. Seetharamulu, K.S. Rama Rao, S.N. Basahel, J. Mol. Catal. A
Chem. 411 (2015) 157, doi:http://dx.doi.org/10.1016/j.molcata.2015.10.019.
Conclusions [14] J. Roman, Appl. Catal. A Gen. 400 (2011) 48, doi:http://dx.doi.org/10.1016/j.
apcata.2011.04.010.
A detailed study of the ammonia reaction yield with newly [15] D. Carpenter, K. Maloney, System and method for ammonia synthesis.
US9272920B2, 2012.
available kinetic and thermodynamic parameters was performed.
[16] B. Lin, K. Wei, J. Ni, J. Lin, ChemCatChem 5 (7) (2013) 1941, doi:http://dx.doi.
The previously developed kinetic models for Ru/C, Fe-magnetite org/10.1002/cctc.201200889.
and Fe-wustite based catalysts were adapted for use in the Aspen [17] D.E. Brown, T. Edmonds, R.W. Joyner, J.J. McCarroll, S.R. Tennison, Catal Lett.
Plus1 plug flow reactor model. The preliminary validation of the 144 (4) (2014) 545, doi:http://dx.doi.org/10.1007/s10562-014-1226-4.
[18] Z. Kowalczyk, S. Jodzis, W. Raróg, J. Zielin  ski, J. Pielaszek, Appl. Catal. A Gen 173
calculations against the available experimental data was done to (1998) 153, doi:http://dx.doi.org/10.1016/S0926-860X(98)00175-6.
check consistency and to assess the best operating conditions for [19] W.L. Luyben, Ind. Eng. Chem. Res. 49 (2010) 6150.
each material. The best conditions for the single-pass operation of [20] S.É Van-Dal, C. Bouallou, J. Clean Prod. 57 (2013) 38, doi:http://dx.doi.org/
10.1016/j.jclepro.2013.06.008.
the Ru/C material were selected in the 400–450
C range. [21] D.-H. Jang, H.-T. Kim, C. Lee, S.-H. Kim, Int. J. Hydrogen Energy 38 (14) (2013)
A multibed reactor configuration, with intercooling, was selected, 6021, doi:http://dx.doi.org/10.1016/j.ijhydene.2013.01.167.
holding different catalyst amounts and types. Mixing initial iron- [22] M.B. Nikoo, N. Mahinpey, Biomass Bioenergy 32 (12) (2008) 1245, doi:http://
dx.doi.org/10.1016/j.biombioe.2008.02.020.
based catalysts in the first bed(s) with Ru-based catalyst in the last [23] S. Srinivas, S.M. Mahajani, R.K. Malik, Ind. Eng. Chem. Res. 49 (20) (2010) 9673,
one(s) revealed the best option to improve the yield with optimised doi:http://dx.doi.org/10.1021/ie100108p.
catalyst loading. Accordingly, less demanding operating conditions [24] M. Khoshnoodi, Y.S. Lim, Fuel Process. Technol. 50 (2–3) (1997) 275, doi:http://
dx.doi.org/10.1016/S0378-3820(96)01079-X.
can be envisaged for mixed multibed configurations. [25] A. Tripodi, M. Compagnoni, R. Martinazzo, G. Ramis, I. Rossetti, Catalysts 7 (5)
Acronyms (2017), doi:http://dx.doi.org/10.3390/catal7050159.
GHSV Gas Hourly Space Velocity [26] B.Y. Yu, I.L. Chien, Ind. Eng. Chem. Res. 54 (41) (2015) 10073, doi:http://dx.doi.
org/10.1021/acs.iecr.5b02345.
RPlug Plug-flow reactor
186 A. Tripodi et al. / Journal of Industrial and Engineering Chemistry 66 (2018) 176–186

[27] A. Aráujo, S. Skogestad, Comput. Chem. Eng. 32 (2008) 2920, doi:http://dx.doi. [38] B. Evans, S. Hawkins, G. Schulz (Eds.), Ullmann’s Encyclopedy of Industrial
org/10.1016/j.compchemeng.2008.03.001. Chemistry, VCH, Weinheim, 1991.
[28] P. Arora, A.F.A. Hoadley, S.M. Mahajani, A. Ganesh, Ind. Eng. Chem. Res. 55 (22) [39] L.M. Patnaik, I.G. Sarma, N. Viswanadham, Int. J. Syst. Sci. 10 (2) (1979) 225.
(2016) 6422, doi:http://dx.doi.org/10.1021/acs.iecr.5b04937. [40] B.V. Babu, R. Angira, Comput. Chem. Eng. 29 (5) (2005) 1041, doi:http://dx.doi.
[29] J. Andersson, J. Lundgren, Appl. Energy 130 (2014) 484, doi:http://dx.doi.org/ org/10.1016/j.compchemeng.2004.11.010.
10.1016/j.apenergy.2014.02.029. [41] D. Yancy-Caballero, L.T. Biegler, R. Guirardello, Chem. Eng. Trans. 43 (2015)
[30] I. Rossetti, N. Pernicone, L. Forni, Catalysis Today 102–103 (2005) 219. 1297, doi:http://dx.doi.org/10.3303/CET1543217.
[31] N. Pernicone, F. Ferrero, I. Rossetti, L. Forni, P. Canton, P. Riello, G. Fagherazzi, M. [42] N. Nika9 cevic, M. Jovanovic, M. Petkovska, Chem. Eng. Res. Des. 89 (4) (2011)
Signoretto, F. Pinna, Appl. Catal. A Gen. 251 (1) (2003) 121, doi:http://dx.doi. 398, doi:http://dx.doi.org/10.1016/j.cherd.2010.08.011.
org/10.1016/S0926-860X(03)00313-2. [43] J. Ni, J. Lin, X. Wang, B. Lin, J. Lin, L. Jiang, ChemistrySelect 2 (21) (2017) 6040,
[32] I. Rossetti, L. Forni, Appl. Catal. A Gen. 282 (1–2) (2005) 315–320. doi:http://dx.doi.org/10.1002/slct.201700859.
[33] I. Rossetti, N. Pernicone, F. Ferrero, L. Forni, Ind. Eng. Chem. Res. 45 (12) (2006) [44] J. Ni, R. Wang, F. Kong, T. Zhang, J. Lin, B. Lin, K. Wei, Cuihua Xuebao 32 (3)
4150. (2011) 436, doi:http://dx.doi.org/10.1016/S1872-2067(10)60179-9.
[34] I. Rossetti, F. Mangiarini, L. Forni, Appl. Catal. A Gen. 323 (2007) 219. [45] D. Szmigiel, W. Raróg-Pilecka, E. Miskiewicz, M. Glin  ski, M. Kielak, Z. Kaszkur,
[35] D.C. Dyson, J.M. Simon, Ind. Eng. Chem. Fundam. 7 (4) (1968) 605, doi:http:// Z. Kowalczyk, Appl. Catal., A 273 (1-2) (2004) 105, doi:http://dx.doi.org/
dx.doi.org/10.1021/i160028a013. 10.1016/j.apcata.2004.06.020.
[36] J.R. Jennings, Catalytic Ammonia Synthesis, Springer, Boston, 1991. [46] C.A. Vancini, La sintesi dell’Ammoniaca, Hoepli, Milan, 1961.
[37] H. Liu, Ammonia Synthesis Catalysts: Innovation and Practice, Chemical
Industry Press, 2013.