1 en 8 Chapter Author

See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/311919476
Catalysis Engineering: From the Catalytic Material to the Catalytic Reactor
Chapter · December 2017

DOI: 10.1007/978-3-319-44439-0_8
CITATIONS READS
0 280
4 authors, including:
Mauro Bracconi Alberto Cuoci

Politecnico di Milano Politecnico di Milano
36 PUBLICATIONS 758 CITATIONS 187 PUBLICATIONS 7,821 CITATIONS
SEE PROFILE SEE PROFILE
All content following this page was uploaded by Alberto Cuoci on 26 August 2021.
The user has requested enhancement of the downloaded file.

Metadata of the chapter that will be visualized in
SpringerLink
Book Title Operando Studies in Heterogeneous Catalysis
Series Title
Chapter Title Catalysis Engineering: From the Catalytic Material to the Catalytic Reactor
Copyright Year 2017
Copyright HolderName Springer International Publishing Switzerland
Author Family Name Rebughini
Particle
Given Name Stefano
Prefix
Suffix
Division Laboratory of Catalysis and Catalytic Processes - Dipartimento di Energia
Organization Politecnico di Milano
Address Via La Masa 34, 20156, Milan, Italy
Email
Author Family Name Bracconi
Particle
Given Name Mauro
Prefix
Suffix
Email
Author Family Name Cuoci
Particle
Given Name Alberto
Prefix
Suffix
Division Dipartimento di Chimica, Materiali e Ingegneria Chimica “G. Natta”
Address Piazza Leonardo da Vinci 32, 20133, Milan, Italy
Email
Corresponding Author Family Name Maestri
Particle
Given Name Matteo
Prefix
Suffix
Email matteo.maestri@polimi.it
Abstract This chapter deals with the application of chemical reaction engineering and computational fluid dynamics
(CFD) for the analysis and assessment of the interactions between mass and heat transport and chemical
reactions. In the first part of the Chapter, we review fundamental concepts of chemical reaction
engineering, by showing the potential impact of transport phenomena at the macroscale on the observed
functionality of the catalytic material. This includes both the effect of the distribution of the residence
times in the reactor and the impact of internal and external transport phenomena. In the second part, we
illustrate modern approaches to catalytic reaction engineering based on CFD simulations. In particular, we
present the algorithms to couple microkinetic models and kinetic Monte Carlo (kMC) simulations with
CFD. The potentialities of the method are assessed by means of a showcase of the CFD-based analysis of a
spectroscopic cell for operando experiments. This example clearly shows that transport artifacts in
standard equipment may lead to an erroneous interpretation of the experiments if not properly accounted
for.
Layout: T1 Standard STIX Book ID: 320723_1_En Book ISBN: 978-3-319-44437-6
Chapter No.: 8 Date: 31-8-2016 Time: 8:31 pm Page: 1/30
Author Proof
1 Chapter 8
2 Catalysis Engineering: From the Catalytic
F
3 Material to the Catalytic Reactor
OO
4 Stefano Rebughini, Mauro Bracconi, Alberto Cuoci
5 and Matteo Maestri
PR
6 Abstract This chapter deals with the application of chemical reaction engineering
7 and computational fluid dynamics (CFD) for the analysis and assessment of the
8 interactions between mass and heat transport and chemical reactions. In the first part
9 of the Chapter, we review fundamental concepts of chemical reaction engineering,
10 by showing the potential impact of transport phenomena at the macroscale on the
11 observed functionality of the catalytic material. This includes both the effect of the
12
13
D
distribution of the residence times in the reactor and the impact of internal and
external transport phenomena. In the second part, we illustrate modern approaches
TE
14 to catalytic reaction engineering based on CFD simulations. In particular, we pre-
15 sent the algorithms to couple microkinetic models and kinetic Monte Carlo
16 (kMC) simulations with CFD. The potentialities of the method are assessed by
17 means of a showcase of the CFD-based analysis of a spectroscopic cell for oper-
18 ando experiments. This example clearly shows that transport artifacts in standard
EC
19 equipment may lead to an erroneous interpretation of the experiments if not

21
20 properly accounted for.
R
OR
Book chapter to appear in.

“Operando Studies in Heterogeneous Catalysis”.
Edited by Irene M.N. Groot and Joost W.M. Frenken, Leiden University, The Netherlands.
Springer.
C
S. Rebughini ⋅ M. Bracconi ⋅ M. Maestri (✉)

Laboratory of Catalysis and Catalytic Processes - Dipartimento di Energia, Politecnico di
Milano, Via La Masa 34, 20156 Milan, Italy
UN
e-mail: matteo.maestri@polimi.it
A. Cuoci
Dipartimento di Chimica, Materiali e Ingegneria Chimica “G. Natta”, Politecnico di Milano,
Piazza Leonardo da Vinci 32, 20133 Milan, Italy
© Springer International Publishing Switzerland 2017 1

J. Frenken and I. Groot (eds.), Operando Studies in Heterogeneous Catalysis,
Springer Series in Chemical Physics 114, DOI 10.1007/978-3-319-44439-0_8
2 S. Rebughini et al.
Author Proof
22 8.1 Introduction
Catalysts are functional materials or molecules able to provide ‘active sites’ which
F
23
24 allow for a stabilization of the Gibbs free energy of the transition state of the
elementary reactions, resulting in an enhanced reaction rate. In heterogeneous
OO
25
26 gas-solid catalysis, such ‘active sites’ are, by nature, present in a separate solid
27 phase. Thus, the reactants in the gas-phase have to reach the ‘active sites’ and this
28 involves several physical transport processes [1]. These phenomena are character-
29 ized by a wide range of characteristic time and length scales, which can cover many
30 orders of magnitudes, as shown in Fig. 8.1. At the microscale, the electronic
PR
31 interaction between the active sites and the reactants drives the capability of the
32 catalyst in cleaving/forming specific bonds, which results in different atomistic
33 events (elementary steps) at the catalyst surface. The interplay between the rates of
34 different steps along with coverage and morphological effects (e.g. different active
35 sites and their relative distribution on the surfaces) is the key event at the mesoscale.
36 Here, the different rates of the elementary steps result in different catalytic cycles
37
38
39
D
and reaction mechanisms, with a strong impact on the observed performances of the
material. Finally, transport properties from the bulk fluid flow to the catalyst
interface are characterized by longer time and length scales and represent the
TE
40 macroscale. All these steps occur at a finite rate determining that concentration and
41 temperature gradients may arise between the bulk fluid phase and the active sites.
42 Thus, the physical steps may have a strong impact on the rate of the overall process
43 and, as a consequence, the observable reaction rate may differ substantially from the
44 intrinsic reaction rate of the chemical transformation under bulk fluid-phase con-
EC
45 ditions. Therefore, it is of crucial importance to distinguish between the intrinsic

46 functionality of the catalyst material and its macroscopically observed behavior.
47 The former is related to the presence of the ‘active sites’ and their ability of
48 making/breaking the chemical bonds when interacting with molecules and it is
49 determined primarily by the electronic structure at the atomic scale. On the
R
Fig. 8.1 Sketch of different

OR
scales involved in a chemical

process
C
UN
8 Catalysis Engineering: From the Catalytic Material … 3

Author Proof
50 contrary, the latter is the actual observed behavior of the catalyst in the chemical
51 reactor and it is the result of the interplay among all the phenomena at the different
52 scales. Thus, the macroscopically observed behavior is intrinsically a multiscale
F
53 property of the system and its analysis and design require to properly link simu-
54 lations and models at the different scales [2]. Typically, at the microscale the
OO
55 calculation of the kinetic parameters of the elementary reactions is performed within
56 the validity of the harmonic Transition State Theory (h-TST). In this context, the
57 most important information for the determination of a rate constant is the location of
58 the transition state, i.e. the calculation of the activation energy. Electronic structure
59 theories explicitly treat the electronic degrees of freedom and are the natural basis
60 for the calculation of the potential energy surface. In particular, for the extended
PR
61 metal surfaces typical in catalytic problems, Density Functional Theory (DFT) is
62 currently the most adopted method among the electronic-structure theories [3].
63 Moreover, in view of the huge computational cost connected to such calculations,
64 semi-empirical and less demanding methods are also used, especially for the
65 exploration of complex reaction networks [4].
66 At the mesoscale, the interplay between the rates of the different elementary
67
68
D
steps is accounted for. This interplay is crucial in determining the dominant paths,
which take part in the catalytic cycle at the given operating conditions. Statistical
simulations are used to explicitly account for the interplay between all the chemical
TE
69
70 events. In particular, master-equation-based kinetic Monte Carlo (kMC) algorithms

71 are employed in order to account for the correct and site-resolved statistical inter-
72 play among the elementary steps of the reaction network [3]. A special case is
73 represented by the “mean-field” approximation, which relies on the assumption of
EC
74 very fast diffusion of the species at the surface [5].

75 At the macroscale, transport phenomena are responsible for the macroscopic
76 distribution of the velocity of the fluid (hydrodynamics), temperature, and compo-
77 sition in the reactor. At the length of the characteristic dispersion of the reactor (e.g.
78 solid particles), they are responsible for the mixing, temperature, and concentrations
at the interface between the phases. The modeling of the reactor device is based on
R
79
80 either indefinite or macroscopic balance equations for the conservation of mass,

81 energy, and momentum. The behavior of the macroscale is mainly determined by
OR
82 fluid-dynamics. By increasing the complexity and (in principle) the accuracy of the
83 model, one can move to two-dimensional (2D) or three-dimensional (3D) macro-
84 scopic models up to the computational solution of the Navier-Stokes equations for
85 the real reactor geometry by means of Computational Fluid Dynamics (CFD).
86 This chapter introduces the application of chemical reaction engineering and
CFD to obtain a better understanding of the interactions between mass and heat
C
87
88 transport and chemical reactions in catalytic reactors. We mainly focus on the

89 difference between the intrinsic behavior of the catalytic material and its observed
UN
90 behavior in reacting conditions (catalytic reactor). First, we review some basic

91 concepts of chemical reaction engineering, in order to clearly demonstrate the
92 impact of transport phenomena at the macroscale on the observed functionality of
93 the material. This comprises both the effect of the distribution of the residence times
94 in the reactor and the impact of internal and external transport phenomena. Then we
Author Proof
95 show modern approaches to catalytic reaction engineering based on CFD simula-

96 tions. In particular, we present the algorithms to couple microkinetic models and
97 kMC simulations with CFD. To conclude, we report as a showcase the CFD-based
F
98 analysis of a spectroscopic cell for operando experiments.
OO
99 8.2 Intrinsic Versus Observed Reaction Rate
100 The chemical reactor is the device where the physicochemical transformations are
101 caused [6]. There are a huge number of different reactor types (e.g. fixed bed
PR
102 reactors, fluidized bed reactors, stirred tank reactors, structured reactors) and the
103 geometry of the reactor is specifically conceived and designed in order to ensure a
104 proper interaction between the catalyst and the reactants. In this section, we provide
105 an overview of the main factors at the macroscale, which may strongly affect the
106 observed reaction rate in the reactor. In particular, the geometry and the operating
107 conditions of the chemical reactor affect two crucial aspects:
108
109
(a) the distribution of the residence time in the reactor;
(b) the mechanisms and the rates of transport properties.
D
TE
110
111 Both these features may strongly influence the observed functionality of the
112 catalytic material.
EC
113 8.2.1 Residence Time Distribution (RTD)
114 The Residence Time Distribution (RTD) can be defined as the information how
115 long an infinitesimal volume of fluid resides in the reactor [6]. Assuming a
116 Lagrangian perspective, the residence time of a fluid element is defined as the time
R
117 elapsed since it entered the reactor. Depending on the mixing rate characterizing the
118 device, we may experience that a portion of the reactant fluid is held in the reactor
OR
119 for a longer time, whereas another portion is held for a shorter time. As such, the
120 hydrodynamics in the reactor strongly influence the time spent by the molecules in
121 the reactor. To illustrate this issue, let us consider a simple experiment, where at a
122 given time t0, a pulse of a fluid tracer is fed to a reactor and the composition of the
123 outlet stream is being monitored. As shown in Fig. 8.2, we can identify two
124 asymptotic behaviors. In case of fully segregated flow (i.e. no mixing), all the
C
125 molecules of the tracer will experience the same residence time. Therefore, no
126 marked elements will be observed at the outlet of the reactor until the residence time
UN
127 of the reactor is elapsed. In case of very fast and perfect mixing, instead, at the
128 injection time all the volume of the reactor will reach a uniform concentration of the
129 tracer. The elements of fluid leaving the reactor will reflect such instantaneous
130 mixing of the tracer at the injection time, thus resulting in an exponential decay of
131 the initial concentration in response to the inflow of the unmarked fluid. In case of

Author Proof
F
OO
PR
Fig. 8.2 Concentration of tracer species using an impulse input in a PFR (top) and in a CSTR
(bottom)
132 intermediate level of mixing, a portion of the fluid will exit at a time less than the
133
134
D
mean residence time, another portion at the mean residence time, and the remaining
part will stay in the reactor for a time longer than the mean residence time.
TE
135 The two asymptotic behaviors of fully mixed and fully segregated flow represent
136 ideal and extreme situations of RTD and can be easily modeled in terms of
137 macroscopic conservation equations. These two extreme situations are usually
138 referred to as “ideal reactor models” and they are of primary importance in the
139 interpretation and analysis of experimental data.
EC
140 8.2.1.1 Fully Mixed Reactor: Continuous Stirred Tank Reactor

141 (CSTR)
R
142 The Continuous Stirred Tank Reactor is assumed to be perfectly mixed. Due to the
143 very fast mixing, the composition of each species is homogeneous in all the por-
144 tions of the reactor volume. Thus, the chemical reactions occur at uniform operating
OR
145 conditions throughout all the reactor volume. Under these assumptions, the
146 steady-state macroscopic mass balance for the generic i-th species can be written as:
147
0 = FiIN − Fi + Ri V ð8:1Þ
149
where Fi is the stream of the i-th component, expressed in mole per unit of time, Ri
C
150
151 is the reaction term, and V is the whole reactor volume.

UN
152 8.2.1.2 Fully Segregated Reactor: Plug Flow Reactor (PFR)
153 In case of absence of mixing, each volume of the fluid entering the reactor will
154 travel along the reactor volume without any interaction with the other parts of the
Author Proof
155 reactor volume. Consequently, the concentration in the reactor will be different
156 along the reactor volume. An example of such a reactor is a circular pipe with an
157 axial length much longer than the cross diameter. If we assume a perfect mixing in
F
158 the radial direction, the macroscopic steady-state mass balance can be written in
159 differential form as follows:
OO
160
dFi
= Ri ð8:2Þ
dV
162
163 where Fi is the stream of the i-th component, expressed in mole per unit of time, Ri
164 is the reaction term, and dV is the generic infinitesimal volume. At the inlet of the
PR
165 reactor, the concentration of each species is known:
166
Fi ðV = 0Þ = Fi0 ð8:3Þ
168
169 To illustrate the effect of mixing on the observed conversion, let us consider the
170 following simple example. We assume to perform the same reaction
171
173
A→B
D ð8:4Þ
TE
174 in two isothermal and isobaric reactors. A and B are two generic species and their
175 consumption and production rates are described by a global first-order reaction rate:
176
r = kCA ð8:5Þ
178
where k is the kinetic constant, expressed in s−1, and CA is the reactant concen-
EC
179
180 tration, expressed in mole per unit of volume.

181 The two reactors have the same volume and all the operating conditions are
182 assumed identical. The only difference is that the first reactor operates with no
183 mixing (identical to a PFR reactor) and the second one is fully mixed (identical to a
R
184 CSTR reactor). The concentration of A at the end of the two reactors can be derived
185 analytically by the previous equations. In particular, for the CSTR reactor the outlet
186 concentration of A turns out to be:
OR
187
CA0
CA = ð8:6Þ
1 + kτ
189
190 whereas the integration of (8.2) and (8.3) for the PFR reactor leads to:
191
C
CA = CA0 expð − kτÞ ð8:7Þ

193
where τ is the residence time. The two outlet concentrations are different and, in
UN
194
195 particular, we observe that the PFR reactor shows a higher conversion than the
196 CSTR reactor for the same operating conditions. This different observed behavior
197 depends only on the different distribution of the residence time in the reactors.

Author Proof
198 8.2.2 Interaction Between Chemical Rates and Transport

199 Rates
F
200 In heterogeneous catalytic systems, the active sites are by nature present in a dif-
ferent phase with respect to the flow of the reactants. Figure 8.3 shows a schematic
OO
201
202 representation of the main physical and chemical steps involved in the catalytic
203 process inside a reactor: (1) inter-phase diffusion of the reactants from the bulk to the
204 catalyst surface through the boundary layer; (2) intra-phase diffusion of the reactants
205 through the porous structures of the catalyst; (3–5) adsorption, surface reaction, and
206 desorption steps; (6) intra-phase counter-diffusion of the products by the pores; and
PR
207 (7) intra-phase counter diffusion of the products from the catalyst surface to the bulk
208 of the gas phase through the boundary layer. All these steps occur at a finite rate and
209 thus concentration (and temperature) gradients may arise between the bulk fluid
210 phase and the proximity of the active site. Therefore, depending on the relative
211 velocity of physical and chemical processes, the observable reaction rate may differ
212 substantially from the intrinsic reaction rate of the chemical transformation under
213
214
215
D
bulk fluid-phase conditions. In this paragraph, a simplified description of the effect
of the physical steps of Fig. 8.3 on the observable reaction rates is described and
discussed. First, we will analyze the effects of the transport from the bulk to the
TE
216 catalyst surface. Then, we will show how the transport inside the catalyst may
217 substantially affect the observed reaction rate.
EC
B & W IN PRINT
1/ Boundary layer 7/ Film diffusion
R
2/ Pore diffusion
6/ Pore diffusion
OR
3/ Adsorption 5/ Desorption
C
4/ Chemical reaction
UN
Fig. 8.3 Main physical and chemical steps involved in the catalytic process
Author Proof
218 8.2.3 External Transport Effects
F
219 The physical phenomena of transport from the bulk of the gas phase to the catalyst
220 surface are described by steps (1) and (7) of Fig. 8.3, which are purely diffusive and
follow the Stefan-Maxwell equations:
OO
221
222
NS 1
∇xi = ∑ ðxi Nj − xj Ni Þ ð8:8Þ
j = 1 Ctot Γi, j
j≠i
224
PR
225 where xi is the mole fraction of the i-th component, Ctot is the total concentration, Ni
226 is the molar flux of the i-th component, and Γi, j is the binary diffusion coefficient
227 between the i-th and the j-th components. This equation in the case of a
228 two-component mixture at constant pressure and with equimolar counter-diffusion
229 is reduced to the First Fick’s Law:
230
232
233
Ni = − Γi, j ∇CiD
The flux of the i-th component is obtained by solving (8.9) with the proper
ð8:9Þ
TE
234 boundary conditions. If the thickness of the boundary layer δ is small compared to
235 the curvature of the catalyst, the problem reduces to:
236
dCi
Ni = − Γi, j ð8:10Þ
dx
EC
238
239 Due to the conservation of mass, the flux through the stagnant film must be
240 constant, thus the derivative of the flux is zero:
241
dNi
=0 ð8:11Þ
R
dx
243
244 By combining (8.10) and (8.11) and by assuming constant diffusivity, the
OR
245 transport through the boundary layer is described by the following equation:
246
d2 Ci
=0 ð8:12Þ
dx2
248
249 The solution of (8.12) requires the following boundary conditions:

C
250
(
Ci ðx = 0Þ = CiS
ð8:13Þ
UN
Ci ðx = δÞ = CiB
252

Author Proof
253 The second boundary condition is based on the assumption that the fluid is well
254 mixed, thus the concentration of the i-th component at the edge of the stagnant film
255 is equivalent to the concentration in the bulk. The solution of (8.12) leads to:
F
256
x
Ci = CiS + ðCiB − CiS Þ ð8:14Þ
δ
OO
258
259 and the molar flux is:

260
Γi, j B
Ni = − ðCi − CiS Þ ð8:15Þ
δ
262
PR
263 However, the thickness δ of the boundary layer is unknown. Thus the molar flux
264 is written in terms of a mass-transfer coefficient kC :
265
Ni = kC ðCiB − CiS Þ ð8:16Þ

267
268 In the case of a generic first-order kinetics A → B, where the kinetic constant is
269
270 at steady state conditions:

D
kS , the flux of A is equal to the reaction rate to prevent the accumulation or depletion
TE
271
r = kS CAS = kC ðCAB − CAS Þ ð8:17Þ

273
274 By estimating CAS from (8.17), the rate expression in terms of the measurable
275 concentration in the bulk, CAB , is obtained:
EC
276
1
r= 1 1
CAB ð8:18Þ
kS + kC
278
279 An observed reaction rate constant can be defined in terms of kS and kC :

R
280
1 1 1
= + ð8:19Þ
kOBS kS kC
OR
282
283 A reaction rate expressed in terms of observable quantities is simply given by:
284
rOBS = kOBS CAB ð8:20Þ

286
287 and the Damkholer dimensionless number can be defined as:

C
288
kS L
Da = ð8:21Þ
kC
UN
290
291 where L is the characteristic length of the catalyst geometry (e.g. the diameter for
292 spherical catalysts). The Damkholer number and the observable kinetic constant can
293 be used to investigate the influence of diffusional resistance on the observed
Author Proof
294 reaction rate in the reactor. In particular, two asymptotic conditions can be
295 identified:
F
296 1. Chemical regime: this regime occurs when the reaction is slower than the
297 transport in the boundary layer. In this case the kinetic constant is smaller than
the mass-transfer coefficient ðkS ≪ kC Þ, thus the observed reaction rate is:
OO
298
299
rOBS = kS CiB
ð8:22Þ
kOBS → kS
301
302
303
304 This results in a very small Damkholer number ðDa → 0Þ. Thus, the observed
PR
305 reaction rate is controlled only by the reaction that occurs at the catalyst. In these
306 conditions, the intrinsic and the observed reaction rates are identical.
307 2. Fully external mass-transfer regime: this regime appears for very fast surface
308 reactions. In this case, the kinetic constant is higher than the mass-transfer
309 coefficient ðkS ≫ kC Þ and the Damkholer number is high ðDa → ∞Þ. In these
310 conditions, the observed reaction rate turns out to be a first-order reaction (ir-
311
312
313
equal to the mass-transfer coefficient:
D
respective of the kinetic order of the rate equation) and the kinetic constant is
TE
rOBS = kC CiB
ð8:23Þ
kOBS → kC
315
316
317
318 Thus, in these conditions, external diffusion fully masks the intrinsic kinetics
specific of the catalyst material.
EC
319
320 8.2.4 Internal Transport Effects

R
321 Many solid catalysts contain pores in order to increase the specific surface area
322 available for adsorption and reaction. The ‘active sites’ are thus located in the pore
OR
323 network and the diffusion of molecules in confined spaces (steps (2) and (6) of
324 Fig. 8.3) can become limiting as well. This phenomenon of diffusion of gas species
325 inside a pore structure is described by the Knudsen diffusion, which assumes that
326 the molecules interact more often with the pore walls than with other molecules.
327 However, in order to properly describe the diffusion inside the catalyst pores, also
C
328 the molecule-molecule collision has to be considered. Therefore, only with a

329 coupling of the molecular diffusion with the Knudsen diffusion, the transport
330 phenomena in a porous catalyst can be adequately described. In particular, in case
UN
331 of equimolar counter-diffusion of a binary mixture, the transition diffusivity ΓTi of

332 the i-th component, can be approximated by the Bosanquet equation:

Author Proof
1 1 1
333
= B+ K ð8:24Þ
ΓTi Γi Γi
F
335
336 where ΓBi is the molecular diffusion and ΓKi is the Knudsen diffusion.
337 Now, we consider the idealized situation of a cylindrical pore in a catalytic slab.
OO
338 For an isothermal, isobaric, first-order reaction that occurs on the pore wall A → B,
339 the mole balance on a slice of Δx thickness can be written as:
340
πR2pore NA jx − πR2pore NA jx + Δx − ks CAS ð2πRpore ÞΔx = 0 ð8:25Þ

342
PR
343 where NA is the molar flux of A estimated at both sides of the slice, kS is the kinetic
344 constant, expressed in terms of catalytic area and ð2πRpore ÞΔx is the area of the pore
345 wall in the catalyst slice. By changing (8.25) and taking the limit as Δx → 0, the
346 mole balance can be re-written as:
347
dNA 2kS
− = CA ð8:26Þ
349
350
dx Rpore D
Fick’s Law (8.9) relates the molar fluxes to the concentration gradients and its
TE
351 substitution in (8.26) yields the following second-order differential equation:
352
d2 CA 2kS
− T CA = 0 ð8:27Þ
dx 2 ΓA Rpore
354
EC
355 where ΓTA is assumed constant. The surface rate constant can be re-written on a
356 volume basis by using the surface-to-volume ratio:
357
A 2
= ð8:28Þ
V Rpore
R
359
0
360 and the mole balance in terms of kinetic constant on volume basis kS :
OR
361
0
d2 CA k
− ST CA = 0 ð8:29Þ
dx2 ΓA
363
364 Equation (8.29) can be simplified by considering the dimensionless length:

365
C
x
x* = ð8:30Þ
L
367
UN
368 and the generalized Thiele modulus:

Author Proof
sffiffiffiffiffiffi
369 0
kS
ϕ=L ð8:31Þ
ΓTA
F
371
372 where L is the characteristic length of the catalyst geometry, which for general
OO
373 geometries is taken as the volume-to-surface ratio. The substitution of (8.30) and
374 (8.31) in (8.29) leads to:
375
d2 CA
− ϕ2 CA = 0 ð8:32Þ
dx*2
377
PR
378 The boundary conditions required for the solution of this differential equation
379 are:
380
8
< CA ðx* = 0Þ = CAS
ð8:33Þ
: dCA ðx* = 1Þ = 0
dx*
382
383
384
The analytical solution of (8.32) gives:
D
TE
cosh½ϕð1 − x*Þ
CA = CAS ð8:34Þ
cosh½ϕ
386
387 Figure 8.4 shows that—depending on the values of the Thiele modulus—con-
388 centration gradients can arise in the catalytic slab, thus affecting the overall reaction
EC
389 rate. In particular, when the Thiele modulus is small, transport is faster than reaction
390 and consequently the profile is flat. On the other hand, when the Thiele modulus is
391 large, a significant diffusional resistance prevents a constant concentration profile
392 inside the catalyst. This effect of the intra-porous diffusion on the observed reaction
393 rate is usually expressed in terms of the effectiveness factor η, defined as:
R
394
RV
rOBS rðCA ÞdV
η= = 0
ð8:35Þ
OR
rMAX rðCA ÞV
396
397 where V is the catalyst volume. The effectiveness factor is a useful parameter to
398 quantify and represent the internal mass-transfer regime. When the concentration
399 profile is flat, it is equal to one and all the active sites work at the same conditions of
400 concentrations. When concentration gradients establish, the reaction rate is not
C
401 uniform in the solid and thus the efficiency is lower than 1. Integrating (8.35) over
402 the catalyst volume for first-order kinetics leads to the generic definition of the
UN
403 effectiveness factor as a function of the Thiele modulus:


Author Proof
Fig. 8.4 Effect of the Thiele

modulus on the normalized
concentration profile in a
F
catalyst pore with generic
first-order surface reaction
OO
PR
404 tanhðϕÞ
η= ð8:36Þ
ϕ
406
407
408
D
It can be demonstrated that for sufficiently high values of the Thiele modulus, the
effectiveness factor has the following functional form, irrespective of the specific
TE
409 geometry of the solid:
410
1
η= ð8:37Þ
ϕ
412
EC
413 The generalized Thiele modulus, defined in (8.31), can be extended to more
414 complex reactions. For instance, the generalized Thiele modulus for an irreversible
415 reaction of order n is:
416
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
n−1
V n + 1 kS CA,
R
S
ϕ= n> −1 ð8:38Þ
A 2 ΓTA
418
OR
419 However, despite the larger complexity of the Thiele-modulus definition, the
420 asymptotic value of the effectiveness factor (at large values of the Thiele modulus)
421 is still ϕ1 .
422 We now aim to quantify the effect of such transport limitations on the observable
423 parameters. In this scope, the observed reaction rate can be written in terms of the
C
424 intrinsic rate expression and the effectiveness factor:

425
rOBS = ηkS CA,

n
S ð8:39Þ
UN
427
428 and, assuming a large value of the Thiele modulus, we can write:
Author Proof
1
429
rOBS = n
kS CA, ð8:40Þ
ϕ S
431
F
432 The substitution of (8.37) in (8.40) leads to the following expression for the
433 observed reaction rate:
OO
434
1 ̸2
A 2 ðn + 1Þ ̸2
rOBS = Γ kS
T
CA, S = kOBS CAnOBS ð8:41Þ
V n+1 A
436
437 The order of the reaction observed under conditions of severe diffusional limi-
tations is nOBS = ðn +2 1Þ instead of n. Moreover, if an Arrhenius expression for the
PR
438
439 kinetic constant is considered:

440

E
k = k0 exp − ð8:42Þ
RT
442
443
444
the activation energy for the observed kinetic constant is: D
ES + EΓ
EOBS = ð8:43Þ
TE
2
446
447 where ES is the true activation energy for the reaction and EΓ is the activation
448 energy for the diffusion. Diffusional processes, however, are weakly activated
449 compared to the chemical reactions, thus the contribution of EΓ can be considered
EC
450 negligible and the observed activation energy for a severely diffusion-limited
451 process can be estimated half of the true value.
8.2.5 Temperature Dependence of the Observed Reaction

R
452
453 Rate
OR
454 Let us summarize the different regimes that we have identified in the analysis as a
455 function of temperature. Figure 8.5 shows the Arrhenius plot for the observed
456 activation energy at different temperatures. At low temperature, kinetics is slower
457 than transport and thus it is the controlling step. Thus, the activation energy is not
458 affected by the presence of the diffusion. By increasing the temperature, gradients
C
459 start to establish in the solid and the efficiency drops to values lower than 1.
460 According to (8.43), the observed activation energy becomes half that of the in-
461 trinsic activation energy. At very high temperature, reaction is so fast that the
UN
462 system is under external mass-transfer limitations. As shown in Sect. 8.2.3, in these
463 conditions the observed reaction constant is only dominated by transport, which is
464 not an activated process. Therefore, the apparent activation energies approach very
465 small values.

Author Proof
Fig. 8.5 Temperature

dependence of the observed
rate constant of a reaction
F
occurring in a porous catalyst
OO
PR
466 Overall, in this section, we have shown by means of simple and basic concepts
467 of chemical reaction engineering how the physical steps which are required for
468
469
470
D
interaction between the active sites and the reactants can considerably affect the
observed reaction rates. These issues have to be always clearly kept in mind both
during the design of the equipment and during the interpretation of experiments.
TE
471 8.3 Computational Fluid Dynamics of Gas-Solid Catalytic
472 Reactors
EC
473 8.3.1 Multiscale Modeling
474 In the previous section, we have shown how phenomena at the macroscale can
475 strongly affect the observed macroscopic functionality of a catalyst material by
R
476 means of simplified approaches. Here, we concentrate on the description of the

477 methodologies that have been recently proposed to model in detail the interaction
between the macroscale and the micro- and the mesoscales [2]. On one side, the
OR
478
479 most natural approach would be to proceed according to a bottom-up philosophy,

480 from the atomic scale to the reactor scale. In particular, a fully “first-principles”
481 approach (i.e. by solving the fundamental governing equations at each scale) would
482 be desirable to allow for a very detailed and fundamental analysis of the prediction
483 in view of a rational design of the observed functionality. However, this approach is
C
484 hampered by several issues, which mainly concern with the coupling of very dif-
485 ferent time scales. In fact, the resolution of the macroscopic scale with the small
time step required for the microscale would lead to an enormous waste of com-
UN
486
487 putational time, since the continuum dynamics usually evolves on a much larger
488 time scale. Therefore, bottom-up approaches based on direct coupling, even if
489 desirable (especially in view of a detailed and rational understanding of the phe-
490 nomena) are not of practical relevance and their application is mainly related to
Author Proof
491 very simple and ideal cases, where the focus is more on the method development
492 rather than on the application [7].
493 On the contrary, most of the methods proposed and applied for the simulation of
F
494 gas-solid catalytic reactors are based on the effective decoupling of the interde-
495 pendencies among the scales. The main and general feature of these methods is to
OO
496 treat each scale separately, either by using specific numerical procedure or by
497 introducing specific approximations in order to effectively decouple the time scales
498 of the different scales. In doing so, each scale is simulated with different time steps,
499 thus avoiding the direct coupling of the phenomena occurring at very different
500 characteristic times.
501 In particular, our focus is on the coupling between the computational solution of
PR
502 the transport equations at the macroscale with kMC simulations and microkinetic
503 modeling at the meso- and microscales, respectively.
504 The governing equations at the macroscale are summarized in the following.
505 8.3.2 Governing Equations D

TE
506 8.3.2.1 Gas Phase
507 The reactive flows under investigation in the present work are mathematically
508 described by the conservation equations for continuous, multicomponent, com-
509 pressible, thermally-perfect mixtures of gases [8]. The conservation equations of
EC
510 total mass, momentum, individual-species mass fractions and energy for a New-
511 tonian fluid are reported in the following:
512
∂ρ
+ ∇ ⋅ ðρvÞ = 0 ð8:44Þ
∂t

R
514 ∂ 2
ðρvÞ + ∇ ⋅ ðρv⊗vÞ = − ∇ ⋅ ðpIÞ + ∇ ⋅ μ ∇v + ð∇vÞT − μð∇ ⋅ vÞI + ρg
∂t 3
OR
ð8:45Þ
∂
516
ðρωk Þ + ∇ ⋅ ðρωk vÞ = − ∇ ⋅ ðρωk Vk Þ k = 1, . . . , NCG ð8:46Þ
∂t
∂T NCG
ρC ̂P + ρC ̂P v ⋅ ∇T = ∇ ⋅ ðλ∇T Þ − ρ∇T ⋅ ∑ ĈP, k ωk Vk
518
ð8:47Þ
C
∂t k=1
520
In the equations reported above, t is the time, p and T are the pressure and the
UN
521
522 temperature, respectively, ρ the density, λ the thermal conductivity, μ the dynamic
523 viscosity, and ĈP the specific heat at constant pressure of the gas phase mixture, v
524 the velocity vector, g the acceleration vector due to the gravity, and I the identity
525 tensor. NGC is the number of gas-phase species. The subscript k refers to the

Author Proof
526 individual gas-phase species and ωk is the corresponding mass fraction, Vk is the
527 diffusion velocity, and C ̂P, k the specific heat at constant pressure.
528 The density of the mixture is calculated using the equation of state of ideal gases.
F
529 Both Fickian and thermal diffusion are taken into account for evaluating the dif-
530 fusion velocities, according to the approach suggested in [9]:
OO
531
Γk, mix Γk, mix ϑk 1
Vk = − ∇ωk − ∇T ð8:48Þ
ωk Xk T
533
534 where Γk, mix is the mixture diffusion coefficient, Xk the mole fraction, and ϑk the
535 thermal diffusion ratio of species k-th. Mass conservation is enforced by employing
PR
536 the approach proposed by [10], based on the definition of a conservation diffusion
537 velocity. In particular, in this approach the corrected diffusion velocity vector VCk
538 (to be used in (8.46) and (8.47)) is given as:
539
VCk = Vk + VC ð8:49Þ
541
542
543
D
where VC is a constant correction factor (independent of species, but varying in
space and time), introduced to satisfy the mass conservation, and evaluated as:
TE
544
NC
VC = − ∑ Vk ð8:50Þ
k=1
546
547 The mixture diffusion coefficient Γk, mix for species k-th is calculated using the
EC
548 following expression [11]:

549
∑Nj ≠S k Xj Wj
Γk, mix = X ð8:51Þ
Wmix ∑Nj ≠S k Γjkj
551
R
552 where Wmix and Wj are the molecular weights of mixture and species j-th, respec-
553 tively. Γjk is the binary mass diffusion coefficient between species j-th and k-th. The
remaining mixture-averaged transport properties are estimated from the corre-
OR
554
555 sponding pure-species properties through the application of proper mixing rules. In
556 case of dynamic viscosity, the Wilke formula [12] is adopted:
557
NS Xi μi
μ= ∑ NS ð8:52Þ
i = 1 ∑j = 1 Xj ϕi, j
C
559
560 where:
UN
561
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi" rffiffiffiffi 1 ̸4 #2
1 Wj μi Wj
ϕi, j = pffiffiffi 1+ ð8:53Þ
8 Wi + Wj μj Wi
563
Author Proof
564 For the thermal conductivity the combination averaging formula proposed by
565 Mathur et al. [13] is used:
566
" N − 1#
F
1 NS S X
i
λ= ∑ Xi λi + ∑ ð8:54Þ
2 i=1 i = 1 λi
OO
568
569 8.3.2.2 Boundary Conditions
570 The conservation equations of total mass, momentum, individual species mass
PR
571 fractions, and energy require boundary conditions to be specified for pressure,
572 velocity, species mass fractions, and temperature. The usual boundary conditions
573 for pressure and velocity are imposed. In particular, the pressure value is fixed at the
574 outlet boundaries and a zero-gradient condition is imposed at the inlets and at the
575 walls. For the velocity field, no-slip conditions are assumed at the walls. At the
576 inlet, the velocity profile is assigned, whereas at the outlet boundaries the flow is
577
578
579
D
assumed to be fully developed and zero-gradient conditions are then imposed.
Boundary conditions for the gas-phase-species mass fractions and temperature
are summarized in the following.
TE
580 Inert walls: The mass flux of the individual species k-th is set equal to zero:
581
∇ωk jinert = 0 ð8:55Þ

583
584 According to the physics of the particular problem under investigation, on inert
EC
585 walls the temperature can be prescribed through a generic function of time Tinert ðt Þ
586 or calculated to take into account the heat transfer with the external environment (at
587 temperature Tenv ðt Þ):
588
Tjinert = Tinert ðt Þ ð8:56Þ

R
590 ðλ ∇T Þjinert = U Tjinert − Tenv ðt Þ ð8:57Þ

592
OR
593 where U is the global heat-transfer coefficient.

594 Catalytic walls: The mass flux of the individual species k-th is assumed equal to
595 the formation rate due to the heterogeneous reaction occurring on the catalytic wall:
596
Aeff
cat ̇het
ρ Γk, mix ð∇ωk Þjcatalytic = Ω k = 1, . . . , NCG ð8:58Þ
C
A k
598
599 Analogously, the heat flux is set equal to the heat Qhet released by the hetero-
UN
600 geneous reactions occurring on the catalytic wall:


Author Proof
601 Aeff
cat
ðλ ∇T Þjcatalytic = Qhet ð8:59Þ
A
F
603
604 In the expressions reported above Aeff

cat is the effective catalytic area and A is the
geometric area [14].
OO
605
606 Inlet boundaries: At inlet boundaries, Danckwerts’ conditions are set for
607 gas-phase species, i.e. the total mass flux for each species k-th (accounting for
608 diffusion and convection) is specified. If composition gradients exist at the
609 boundary, these conditions allow diffusion into the computational domain, therefore
610 giving a more accurate description than classic Dirichlet conditions:
PR
611
ðρvωk Þjinlet − ðρ Γk ∇ωk Þjinlet = ðρvωk Þ0 ð8:60Þ

613
614 where the term on the right side is the prescribed total mass flux. In analogy, for
615 temperature:
616

ρvH ̂ − ðλ∇T Þjinlet = ρvH ̂ 0 ð8:61Þ
618
619
inlet D
Outlet boundaries: At the outlet boundaries, the mass flux of gas-phase species
TE
620 and the heat flux are assumed equal to zero:
621
∇ωk jinert = 0 ð8:62Þ
623 ∇Tjoutlet = 0 ð8:63Þ

EC
625
626 8.3.2.3 Adsorbed (Surface) Species and Evaluation of the Reaction

627 Term
R
628 In order to determine the formation rate due to the heterogeneous reactions
629 occurring on the catalyst surface (8.58), a conservation equation which balances the
630 rate of change of each surface species with the net formation rate due to hetero-
OR
631 geneous chemical reactions at the surface has to be accounted for. In this scope, two
632 different approaches can be adopted: (i) the mean-field approximation and (ii) ki-
633 netic Monte Carlo simulations.
C
634 8.3.3 Coupling CFD and Mean-Field Microkinetic

635 Modeling
UN
636 The mean-field approximation assumes that due to the strong mobility of the
637 intermediates at the catalyst surface, the distribution of adsorbates and active sites is
638 uniform. Thus, the local state of the active surface is well described by mean values
Author Proof
639 of species coverage. Thus, the reaction rate can be locally expressed as a function of
640 the mean surface molar concentration:
641
F
NRhet
ν
ri = ki ∏ Xsur,
i, j
j ð8:64Þ
j=1
OO
643
644 where ki is the rate constant of the i-th step, Xsur,j are the surface species mole
645 concentrations, and νi, j is the stoichiometric coefficient of the j-th species in the i-th
646 reaction [15]. Then, the evolution of the surface coverage is described by:
647
∂θi
= Ωi̇
PR
het
σ cat i = 1, . . . , NCS ð8:65Þ
∂t
649
where θi is the site fraction of the species i-th, σ cat is the site density, and Ω̇i the
het
650
651 formation rate of species i-th due to the heterogeneous reactions. NCS is the number
652 of absorbed surface species.
653 The use of detailed microkinetic models in the framework of CFD simulations
654
655
D
poses important issues from a numerical point of view and requires specific algo-
rithms. In particular, the classical fully segregated approaches usually employed in
TE
656 most of the non-reactive CFD simulations are not applicable because of the com-
657 bination of the stiffness of the elementary reactions with the typical large dimen-
658 sions of a microkinetic model [16]. Two alternatives are then possible. The first
659 possibility is to employ fully coupled methods [17], in which all the processes are
660 considered simultaneously, thus all the physical interactions among them are taken
EC
661 into account together. However, due to the prohibitive computational cost, these
662 are strongly limited for complex geometries and for extremely large governing
663 equations (especially when detailed kinetics are considered). The alternative solu-
664 tion is the application of operator-splitting algorithms, which allow the best
665 numerical method to be used for each operator of the equations (typically reaction
R
666 and transport). The resulting algorithms can be very complex and usually differ
667 from term to term. An example of this implementation is given by Maestri and
668 Cuoci [16], where the operator-splitting algorithm is implemented within the
OR
669 OpenFOAM framework [16, 18].
670 8.3.4 Coupling CFD and Kinetic Monte Carlo Simulations

C
671 When statistical simulations are used to explicitly account for the interplay between
672 all the chemical events, master-equation-based kinetic Monte Carlo algorithms are
UN
673 employed in order to account directly for the spatial arrangements of the species at
674 the surfaces and allow for a sound evaluation of the reaction term. The inclusion of
675 kinetic Monte Carlo simulations in the framework of real reactor models requires
676 dealing with very different time scales between the macroscale and mesoscale. In
677 fact, the calculation of the reaction rate at each time step and in each part of the

Author Proof
678 domain needs the explicit run of several kMC simulations. Since the characteristic
679 time for kMC simulations is an order of magnitude lower than the one of the
680 macroscale, this limits dramatically the resolution of the time step of the macro-
F
681 scale, thus leading to unbearable computational costs [7, 19–21]. Key in this respect
682 is the development of methodologies to minimize the number of kMC simulations
OO
683 that have to be run “on the fly” during the simulation of the macroscale. The two
684 main important strategies reported to solve this problem are reported below.
685 1. Instantaneous steady state approximation
686
687 This approximation exploits the typically different time scales of surface kinetics
688 and gas flow. It is assumed that upon a change of the local gas-phase conditions
PR
689 (temperature, pressure, species mass fractions), the surface adapts instantaneously
690 to the corresponding steady state. Thus, it is sufficient to have a continuous rep-
691 resentation of the steady-state reaction rates as a function of the gas-phase condi-
692 tions (temperature, pressure, species mass fractions) for coupling the kMC with the
693 macroscale models. Due to the high computational time required for the kMC
694 simulations in different conditions, the steady-state reaction rates can be
695
696
D
pre-computed once before the whole reactor simulation is performed and be used to
determine a numerically efficient continuous representation either using tabulation
techniques or suitable polynomial fitting. Then, during the simulation, the species
TE
697
698 formation and consumption rates are evaluated by making use of the stored results
699 [19–22].
700 2. Gap-tooth methods
701
EC
702 In this approach, the kMC simulations are run only in some points of the surface
703 domain and for the other points the reaction rates are interpolated. The converged
704 solution is reached with iterative simulation of the whole domain [23–32]. This
705 approach is used by Schaefer and Jansen [33] to couple the kMC simulations to a
706 1D finite-difference model for the case of CO oxidation. This scheme is used to
reduce the number of kMC simulations carried out to estimate the reaction rates. In
R
707
708 particular, due to the choice of describing the macroscale with a 1D finite-difference
709 model, Schaefer and Jansen [33] select some preferential points where the kMC
OR
710 simulations are done and they estimate the reaction rates between these points with
711 a linear interpolation. The correct description of the reactor behavior can be
712 achieved only if the solution of the whole reactor simulations does not depend on
713 the position of the points where the kMC simulations are done. On the contrary,
714 Broadbelt and co-workers [34–36] use a methodology that updates the concentra-
tion and the reaction rate at the same time in all the discretization points. The main
C
715
716 advantage of this approach is the possibility of describing also non-linear profiles
717 between the points where the kMC simulations are carried out.
UN
Author Proof
718 8.4 Case Study: Detailed Analysis of a Spectroscopic Cell

719 Reactor
F
720 In order to exemplify the concepts reported in the previous sections, here we
OO
721 provide as showcase a detailed multiscale analysis of a spectroscopic chamber
722 reactor. In particular, we apply the mean-field approximation coupled with CFD to
723 study a real reactor geometry used for in situ spectroscopic studies [37]. Therefore,
724 the structure of the cell is expected to play a crucial role in the coupling between
725 chemistry and transport. The interactions between these two aspects generally
726 modify the observable catalytic function in a nontrivial manner, as outlined in the
PR
727 previous sections.
728 8.4.1 Reactor Geometry and Operating Conditions
729
730
731
D
The design of the chamber is heavily constrained by the invasive machinery and
this results in very complex flow profiles characterized by temporal and spatial
inhomogeneity, as we exemplify by analyzing the prototypical geometry shown in
TE
732 the left picture of Fig. 8.6. Without considering any specific invasive instruments or
733 pumping equipment, the geometry consists of a cube of edge length 3 cm, within a
734 cuboid sample holder with a base of 1 × 1 cm and a height of 1.5 cm mounted on
735 the bottom wall. The catalyst surface covers the entire top face of the cuboid. The
736 gas flows into the reactor through an inlet located at the upper left and exits through
EC
737 an outlet placed at the lower right. External heating keeps a constant temperature at
738 both the catalyst surface and the sample holder, whereas the remaining walls are
739 considered adiabatic. For the CFD simulations, we exploited the mirror symmetry
740 of the geometry, thus one-half of the chamber has been discretized with a regular
741 cubic mesh. The resolution is 0.5 mm and it is tightened to 0.25 mm in the
R
742 proximity of the catalyst surface to adequately describe the concentration and
743 temperature gradients, as shown in the right panel of Fig. 8.6. This results in
224,000 computational cells. Simulations were performed using the catalyticFoam
OR
744
745 solver [16] in the OpenFOAM environment [18].

746 We focused on methane steam reforming on Rh employing the detailed mi-
747 crokinetic mechanism by Maestri and co-workers and Vlachos and co-workers [38,
748 39], which consists of 21 gas species and 13 adsorbed species involved in 82
749 surface reactions. The gas mixture (97 % N2, 2 % H2O, 1 % CH4, in molar basis)
C
750 enters at the inlet section and a flowrate of 120 Nm L/min through the reactor,
751 which falls within the common range characteristic for such applications. The
reactor works at atmospheric pressure.
UN
752
753 As initial conditions, we adopted uniform values for all fields in the reactor,
754 employing the inlet value for temperature, the outlet value for pressure. The reactor
755 was assumed to be full of nitrogen at the starting time. The solution is advanced

Author Proof
F
OO
PR
Fig. 8.6 Computational domain (on the left) and mesh (on the right) adopted for the CFD
simulations. The gas mixture enters with a uniform velocity and fixed temperature from the left
side, streams over the catalytic surface (in red), leaving the domain through the outlet section on
the right side. The temperatures of the catalytic plate and the sample holder are kept fixed. All
remaining walls can be kept at the same temperature of catalytic plate and sample holder
(isothermal simulations) or can be considered perfectly adiabatic (adiabatic simulations). Because
of the symmetry of the system only one-half of the computational domain is meshed. Cells with
D
two different sizes were adopted: 0.25 and 0.50 mm. The resulting, total number of computational
cells is equal to 224,000
TE
756 using a Courant number of 0.05 until the simulation reaches a steady state in which
757 the relevant flow features are fully established, as proposed in [16, 40].
758 The gas enters the reactor at the inlet with a temperature of 300 K and both the
759 catalyst and the sample holder are kept at a fixed temperature. In particular, the
effect of different catalyst temperatures is investigated, assuming a catalyst/sample
EC
760
761 holder temperature of 973 and 673 K.
762 8.4.2 Reactor Analysis: Residence Time Distribution

R
763 To investigate the effect of both flow and mixing on the behavior of the reactor, the
OR
764 residence time distribution (RTD) has been analyzed. The nominal residence time
765 for the reactor in the actual operating conditions, i.e. the ratio between the volu-
766 metric inlet flow rate and the volume of the chamber, is equal to 12.86 s. We
767 performed an RTD analysis of the prototypical spectroscopic reactor to investigate
768 the effect of the complex flow field on the distribution of residence time. Figure 8.7
769 shows the RTD for the real reactor and for the two ideal limits described in
C
770 Sect. 8.2. The RTD profile presents a peak at low residence time showing the
771 presence of a bypass, which means that a preferential path exists, followed by the
UN
772 gas to travel from the inlet to the outlet. Indeed, the gas enters and drops down due
773 to density, reaching the bottom of the chamber and rapidly streams out of the
774 reactor. The tail of the profile shows the presence of quiescent zones, where the
775 elements of fluid last for a time larger than the nominal residence time. In the actual
776 operative conditions, the calculated mean residence time is 4.88 s and 6.46 s for the
Author Proof
Fig. 8.7 Calculated

residence time distribution
(continuous line) obtained
F
when the reactor is operated
in adiabatic conditions at
atmospheric pressure. The gas
OO
stream through the inlet
section takes place at 300 K,
with a uniform velocity of
4 cm/s, and the
catalyst/sample holder is kept
at 973 K. The RTD for ideal
behavior of plug flow (dotted
PR
line) and perfect mixing
(dashed line) are plotted for
comparison
D
TE
R EC
777 catalyst/sample holder maintained at 973 K and 673 K, respectively. Both values
778 are significantly lower than the nominal value. In this view, the approximations of
779 perfect mixing and of plug flow are not able to describe the behavior of the reactor,
OR
780 requiring a more accurate analysis of the flow pattern inside the chamber.
781 The non-idealities predicted accordingly to the RTD analysis are highlighted by
782 the streamlines shown in Fig. 8.8. The denser gas streams into the reaction chamber
783 from the cold inlet, drop down reaching the bottom of the reactor, and flow towards
784 the outlet. This preferential path is followed by most of the fluid, resulting in the
C
785 peak of the RTD profile. A small portion, after reaching the hot sample holder,
786 heats up and, due to buoyancy, rises inducing a vortex, which rotates counter-
787 clockwise around the catalyst. This phenomenon increases the residence time of
UN
788 some marked elements of gas, leading to the long tail representative of dead zones.
789 Therefore, the flow field inside the reactor is strongly influenced by a significant
790 degree of thermo-convection, raised in the pseudo-stationary state, leading to a
791 highly complex velocity pattern.

Author Proof
792 8.4.3 Reactor Analysis: Species Concentration

793 and Temperature Fields
F
794 The complex flow field within the chamber, especially close to the sample holder,
OO
795 influences the concentration of the reactants in proximity of the catalyst surface.
796 Figure 8.9a shows the predicted CH4 mass fraction distribution within the catalyst
797 chamber obtained with the catalyst/sample holder at 673 K. Strong
798 non-homogeneities in the spatial distribution of methane experiences the important
799 effect of the interactions between transport and chemistry. For the higher sample
800 holder temperature, the catalyst ends up in a highly reactive state, leading to strong
PR
801 mass-transfer limitations, as apparent from the calculated mass fraction of CH4
802 inside the reactor. The controlling phenomenon is the species diffusion from the
803 bulk of the gas phase to the catalyst surface, since the reaction rates are extremely
804 fast. The heat conduction and convection in the reactor are not sufficiently strong to
805 make the temperature uniform, with the exception of a small region close to the
806 catalyst/sample holder. When the cold inlet gas enters the chamber, it drops down
807
808
809
D
due to the difference of density, without having enough time to heat up. Thus, the
region of the reactor below the inlet section shows a temperature of ∼250 K lower
with respect to the sample holder temperature. The species and temperature con-
TE
810 centration inside the chamber is non-uniform, and strong gradients exist between
811 the bulk and the catalyst surface, due to the combined effect of complex flow field
812 along with high reactivity of the catalyst determining external mass-transfer control.
813 The gas-phase composition in the proximity of the catalyst strongly relies on
814 transport phenomena, making the assumption of perfect mixing within the chamber
EC
815 unfeasible. On the other hand, the hypothesis of plug flow is incorrect due to the
816 recirculation induced by the difference of density, which produces an additional
R
C OR
UN
Fig. 8.8 Streamlines arising when the reactor is operated in adiabatic conditions at atmospheric
pressure. The gas stream (1 % CH4, 2 % O2, 97 % N2, on a molar basis) enters at the inlet section
at 300 K, with uniform velocity of 4 cm/s, the catalyst/sample holder is maintained at 973 K. The
streamlines are colored using the velocity field magnitude
Author Proof
817 mixing. Nevertheless, even if the catalyst should operate under less reacting con-
818 ditions (e.g. by decreasing the sample holder temperature), the complex flow field
819 would render it impossible to assume a uniform distribution of species and tem-
F
820 perature. Due to the lower reaction rates, the characteristic time of the diffusion
821 process from the gas phase to the surface becomes at most of the same order of
OO
822 magnitude of the reaction rates. The controlling process is the rate of the hetero-
823 geneous chemistry and, even in this context, a concentration gradient rises from the
824 surface to the bulk of the gas phase. The predicted temperature distribution,
825 reported in Fig. 8.9b, highlights the presence of a colder region in the left side of
826 the chamber due to the cold inlet stream. Once again, the predicted distributions of
827 species and temperature are inhomogeneous. In particular, the concentration near
PR
828 the catalyst surface is different from the outlet concentration, making the infor-
829 mation recovered from the outlet mass-flow rate not totally representative of the
830 state of the reactor.
831 The complex flow field induced by the interactions between the reactor geom-
832 etry, the transport phenomena, and the surface reactivity alters the species distri-
833 bution on the catalyst surface. The calculated mass fractions along the main axes,
834
835
D
presented in Fig. 8.10a, b, highlight the presence of spatial gradients in both
directions. In particular, the recirculation moves the gas toward the left side of the
catalyst, altering the transport rate of the species to the catalytic surface. Overall, the
TE
836
837 species concentrations are non-uniform, pointing out the importance of taking into
838 account the macro-scale phenomena to obtain an accurate description of the reactor.
839 In particular, an analysis by means of CFD allows for the clarification of the
840 coupling between macroscale transport processes and surface chemistry, leading to
EC
841 a better understanding and design of the experiments. The non-uniform gas com-
842 position on the catalyst surface affects the site-coverage distribution, which shows
843 the same trends of the gas phase species, as shown in Fig. 8.10c, d.
844 The interaction between transport and chemistry strongly affects the behavior of
845 the spectroscopic chamber, especially in case of fast reactions that establish
mass-transport limitations between the gas phase and the catalyst. Moreover, the
R
846
847 complex flow field, experienced by the non-ideal behavior of the RTD, influences
C OR
UN
Fig. 8.9 Examples of calculated maps of gas-phase species and temperature (on the middle
section) when the reactor is operated in adiabatic conditions at atmospheric pressure, inlet
temperature of 300 K, and sample holder/catalyst temperature of 673 K

Author Proof
F
OO
PR
D
TE
Fig. 8.10 Examples of calculated mass fraction profiles of selected gas-phase species along the
axial direction of catalytic plate (x coordinate) for catalyst sample holder at 673 K, panels (a) and
(b). Example of calculated site coverage profiles of selected adsorbed species along the axial
EC
direction of catalytic plate (x coordinate) for catalyst sample holder at 673 K, panels (c) and (d)
848 the distribution of species across the reactor, leading to non-homogeneities, which
849 have to be accounted for in order to derive a detailed understanding of the catalyst
850 material in reacting conditions. These effects have to be considered when aiming to
R
851 relate operando spectroscopic data obtained in the cell with the activity of the
852 catalyst. In this context, CFD is a useful tool to obtain a deep insight into the
853 behavior of the spectroscopic reactor, with the aim of understanding the real
OR
854 interactions between chemical and transport processes.
855 8.5 Summary and Outlook

C
856 In heterogeneous catalysis ‘active sites’ are, by nature, present in a separate solid
857 phase. Thus, the reactants in the gas phase have to reach the ‘active sites’ and this
UN
858 involves several physical transport processes. All these steps occur at a finite rate
859 and consequently concentration and temperature gradients may arise between the
860 bulk fluid phase and the areas in proximity of the active site. Therefore, the physical
861 steps may have a strong impact on the rate of the overall process and the observable
Author Proof
862 reaction rate may differ substantially from the intrinsic reaction rate of the chemical
863 transformation under bulk fluid-phase conditions.
864 Thus, it is of crucial importance to distinguish between the intrinsic functionality
F
865 of the catalyst material and its observed macroscopical behavior in the reactor, i.e.
866 the device where the physicochemical transformations occur. By reviewing basic
OO
867 concepts of chemical reaction engineering, we have shown the multiscale character
868 of the observed macroscopic behavior of the catalytic process. The modes of
869 operation of the chemical reactor strongly influence the performance of a given
870 material, both in terms of hydrodynamics (and its immediate effect on the residence
871 time distribution) and in terms of transport properties. Therefore, a thorough
872 understanding of chemical reaction engineering is a crucial requirement in order to
PR
873 properly design reaction equipment or to correctly interpret the experimental evi-
874 dence. In this view, fundamental multiscale modeling is appearing in the last decade
875 as one of the main frontiers in chemical reaction engineering, especially con-
876 tributing to the quest of advancing our scientific understanding of multiscale kinetic
877 transport interactions to enable better reactor choice and to ensure higher reactor
878 and process efficiencies. Specifically, the main task is to provide an improved
879
880 information from the atoms to the reactor.

D
scientific basis for designing the chemical transformation in a seamless flow of
From a reaction engineering perspective, CFD simulations have matured into a

TE
881
882 powerful tool for understanding mass and heat transport in catalytic reactors, by
883 coupling the flow field and heat-transport models with models for heterogeneous
884 chemical reactions. The most natural approach would be to proceed according to a
885 bottom-up philosophy, from the atomic scale to the reactor scale. However, this
EC
886 approach is hampered by several issues, which mainly concern with the coupling of
887 very different time scales. Indeed, the resolution of the macroscopic scale with the
888 small time step required for the microscale would lead to an enormous effort in
889 computational time since the continuum dynamics usually evolves on a much larger
890 time scale. In this respect, we have shown that most of the methods proposed and
applied in the literature for the simulation of gas-solid catalytic reactors are based
R
891
892 on an effective decoupling of the interdependencies among the scales. The main and
893 general feature of these methods is to treat separately each scale either by using
OR
894 specific numerical procedures or by introducing specific approximations in order to

895 effectively decouple the time scales of the different scales. In doing so, each scale is
896 simulated with different time steps, thus avoiding the direct coupling of phenomena
897 occurring at very different characteristic times. Such methods enable the incorpo-
898 ration of both complex mean field microkinetic models and kinetic Monte Carlo
Simulations in the fundamental modeling of real reactor geometries. We have
C
899
900 exemplified the applications of these methods through the CFD analysis of a
901 spectroscopic cell. This study has clearly shown that transport artifacts may arise
UN
902 due to the complicated geometries usually employed, and that this can lead to an
903 erroneous interpretation of the experiments, if not properly accounted for.
904 As a whole, fundamental multiscale chemical reaction engineering represents
905 one of the most important progress areas for catalysis engineering. By bridging
906 between chemistry and engineering science, physics and material science, it draws

Author Proof
907 from hitherto rather disjoint communities and therefore it is ideally suited for broad
908 interdisciplinary collaborations with research groups with specific and comple-
909 mentary expertise (e.g. material synthesis, operando and in situ characterization,
F
910 kinetic experiments, transport phenomena and chemical engineering). AQ1
OO
911 References
912 1. O. Deutschmann, Modeling and Simulation of Heterogeneous Catalytic Reactions: From the
913 Molecular Process to the Technical System (Wiley, 2013)
PR
914 2. M.P. Dudukovic, Science 325, 698 (2009)
915 3. M.K. Sabbe, M.-F. Reyniers, K. Reuter, Catal. Sci. Technol. 2, 2010 (2012)
916 4. M. Maestri, K. Reuter, Angew. Chem. Int. Ed. 50, 1194 (2011)
917 5. M. Salciccioli, M. Stamatakis, S. Caratzoulas, D.G. Vlachos, Chem. Eng. Sci. 66, 4319
918 (2011)
919 6. G.F. Froment, K.B. Bischoff, J. De Wilde, Chemical Reactor Analysis and Design, vol.
920 2 (Wiley, New York, 1990)
7. K. Reuter, D. Frenkel, M. Scheffler, arXiv:cond-mat/0408080 (2004)
921
922
923
924
D
8. M.R. Charest, C.P. Groth, Ö.L. Gülder, Combust. Theor. Model. 14, 793 (2010)
9. S. Chapman, T.G. Cowling, The Mathematical Theory of Non-uniform Gases: An Account of
the Kinetic Theory of Viscosity, Thermal Conduction and Diffusion in Gases (Cambridge
TE
925 University Press, 1970)
926 10. T. Coffee, J. Heimerl, Combust. Flame 43, 273 (1981)
927 11. R.B. Bird, W.E. Stewart, E.N. Lightfoot, Transport Phenomena, 2nd edn. (Wiley, New York,
928 2002)
929 12. C.R. Wilke, J. Chem. Phys. 18, 517 (1950)
930 13. S. Mathur, P.K. Tondo, S.C. Saxena, Mol. Phys. 12, 569 (1967)
EC
931 14. M. Maestri, A. Beretta, T. Faravelli, G. Groppi, E. Tronconi, D.G. Vlachos, Chem. Eng. Sci.
932 63, 2657 (2008)
933 15. M. Maestri, D.G. Vlachos, A. Beretta, G. Groppi, E. Tronconi, A.I.Ch.E. J. 55, 993 (2009)
934 16. M. Maestri, A. Cuoci, Chem. Eng. Sci. 96, 106 (2013)
935 17. M.D. Smooke, R.E. Mitchell, D.E. Keyes, Combust. Sci. Technol. 67, 85 (1986)
936 18. H. Jasak, A. Jemcov, Z. Tukovic, International Workshop on Coupled Methods in Numerical
R
937 Dynamics (2007), p. 1

938 19. S. Matera, K. Reuter, Catal. Lett. 133, 156 (2009)
939 20. S. Matera, K. Reuter, arXiv:1006.0343 (2010)
OR
940 21. S. Matera, K. Reuter, Phys. Rev. B 82, 085446 (2010)

941 22. S. Matera, M. Maestri, A. Cuoci, K. Reuter, ACS Catal. 4, 4081 (2014)
942 23. A. Armaou, I.G. Kevrekidis, C. Theodoropoulos, Comput. Chem. Eng. 29, 731 (2005)
943 24. A. Bindal, M.G. Ierapetritou, S. Balakrishnan, A. Armaou, A.G. Makeev, I.G. Kevrekidis,
944 Chem. Eng. Sci. 61, 779 (2006)
945 25. C. Gear, I.G. Kevrekidis, arXiv:physics/0211043 (2002)
C
946 26. C.W. Gear, J. Li, I.G. Kevrekidis, Phys. Lett. A 316, 190 (2003)
947 27. W. Gear, I. Kevrekidis, J. Hyman, P. Kevrekidis, O. Runborg, C. Theodoropoulos, Commun.
948 Math. Sci. (2003)
949 28. I.G. Kevrekidis, C.W. Gear, G. Hummer, AlChE J. 50, 1346 (2004)
UN
950 29. I.G. Kevrekidis, C.W. Gear, J.M. Hyman, P.G. Kevrekidid, O. Runborg, C. Theodoropoulos,
951 Commun. Math. Sci. 1, 715 (2003)
952 30. A. Roberts, I. Kevrekidis, ANZIAM J. 46, 637 (2005)
953 31. G. Samaey, I.G. Kevrekidis, D. Roose, Multiscale Modelling and Simulation (Springer,
954 2004), p. 93
Author Proof
955 32. G. Samaey, D. Roose, I.G. Kevrekidis, Multiscale Model. Simul. 4, 278 (2005)
956 33. C. Schaefer, A. Jansen, J. Chem. Phys. 138, 054102 (2013)
957 34. L.J. Broadbelt, R.Q. Snurr, Appl. Catal. A 200, 23 (2000)
F
958 35. D.J. Dooling, L.J. Broadbelt, Ind. Eng. Chem. Res. 40, 522 (2001)
959 36. D. Majumder, L.J. Broadbelt, AlChE J. 52, 4214 (2006)
960 37. A. Stierle, A.M. Molenbroek, MRS Bull. 32, 1001 (2007)
OO
961 38. M. Maestri, D. Livio, A. Beretta, G. Groppi, Ind. Eng. Chem. Res. 53, 10914 (2014)
962 39. M. Maestri, D.G. Vlachos, A. Beretta, G. Groppi, E. Tronconi, AlChE J. 55, 993 (2009)
963 40. T. Maffei, S. Rebughini, G. Gentile, S. Lipp, A. Cuoci, M. Maestri, Chem. Ing. Tech. 86,
964 1099 (2014)
PR
D
TE
R EC
C OR
UN
Author Proof
Author Query Form
F
123
OO
Book ID : 320723_1_En
Chapter No : 8
the language of science
Please ensure you fill out your response to the queries raised
below and return this form along with your corrections.
PR
Dear Author,
During the process of typesetting your chapter, the following queries have
arisen. Please check your typeset proof carefully against the queries listed
below and mark the necessary changes either directly on the proof/online
grid or in the ‘Author’s response’ area provided below
D
TE
Query Refs. Details Required Author’s Response
AQ1 No queries.
R EC
C OR
UN
MARKED PROOF
Please correct and return this set
Please use the proof correction marks shown below for all alterations and corrections. If you
wish to return your proof by fax you should ensure that all amendments are written clearly
in dark ink and are made well within the page margins.
Instruction to printer Textual mark Marginal mark

Leave unchanged under matter to remain
Insert in text the matter New matter followed by
indicated in the margin or
Delete through single character, rule or underline
or or
through all characters to be deleted
Substitute character or
through letter or new character or
substitute part of one or
more word(s) through characters new characters
Change to italics under matter to be changed
Change to capitals under matter to be changed
Change to small capitals under matter to be changed
Change to bold type under matter to be changed
Change to bold italic under matter to be changed
Change to lower case Encircle matter to be changed
Change italic to upright type (As above)
Change bold to non-bold type (As above)
or
Insert ‘superior’ character through character or
under character
where required
e.g. or
Insert ‘inferior’ character (As above) over character
e.g.
Insert full stop (As above)
Insert comma (As above)
or and/or
Insert single quotation marks (As above)
or
or and/or
Insert double quotation marks (As above)
or
Insert hyphen (As above)
Start new paragraph
No new paragraph
Transpose
Close up linking characters
Insert or substitute space through character or

between characters or words where required
Reduce space between between characters or

characters or words words affected
View publication stats

1 en 8 Chapter Author

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

1 en 8 Chapter Author

Uploaded by

Copyright:

Available Formats

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

Catalysis Engineering: From the Catalytic Material to the Catalytic Reactor

Chapter · December 2017

Mauro Bracconi Alberto Cuoci

SEE PROFILE SEE PROFILE

The user has requested enhancement of the downloaded file.

19 equipment may lead to an erroneous interpretation of the experiments if not

Book chapter to appear in.

S. Rebughini ⋅ M. Bracconi ⋅ M. Maestri (✉)

© Springer International Publishing Switzerland 2017 1

45 ditions. Therefore, it is of crucial importance to distinguish between the intrinsic

Fig. 8.1 Sketch of different

scales involved in a chemical

8 Catalysis Engineering: From the Catalytic Material … 3

70 events. In particular, master-equation-based kinetic Monte Carlo (kMC) algorithms

74 very fast diffusion of the species at the surface [5].

80 either indeﬁnite or macroscopic balance equations for the conservation of mass,

88 transport and chemical reactions in catalytic reactors. We mainly focus on the

90 behavior in reacting conditions (catalytic reactor). First, we review some basic

95 show modern approaches to catalytic reaction engineering based on CFD simula-

113 8.2.1 Residence Time Distribution (RTD)

8 Catalysis Engineering: From the Catalytic Material … 5

140 8.2.1.1 Fully Mixed Reactor: Continuous Stirred Tank Reactor

151 is the reaction term, and V is the whole reactor volume.

152 8.2.1.2 Fully Segregated Reactor: Plug Flow Reactor (PFR)

180 tration, expressed in mole per unit of volume.

CA = CA0 expð − kτÞ ð8:7Þ

8 Catalysis Engineering: From the Catalytic Material … 7

198 8.2.2 Interaction Between Chemical Rates and Transport

218 8.2.3 External Transport Effects

249 The solution of (8.12) requires the following boundary conditions:

8 Catalysis Engineering: From the Catalytic Material … 9

259 and the molar flux is:

Ni = kC ðCiB − CiS Þ ð8:16Þ

270 at steady state conditions:

r = kS CAS = kC ðCAB − CAS Þ ð8:17Þ

279 An observed reaction rate constant can be deﬁned in terms of kS and kC :

rOBS = kOBS CAB ð8:20Þ

287 and the Damkholer dimensionless number can be deﬁned as:

320 8.2.4 Internal Transport Effects

328 the molecule-molecule collision has to be considered. Therefore, only with a

331 of equimolar counter-diffusion of a binary mixture, the transition diffusivity ΓTi of

8 Catalysis Engineering: From the Catalytic Material … 11

πR2pore NA jx − πR2pore NA jx + Δx − ks CAS ð2πRpore ÞΔx = 0 ð8:25Þ

364 Equation (8.29) can be simpliﬁed by considering the dimensionless length:

368 and the generalized Thiele modulus:

403 effectiveness factor as a function of the Thiele modulus:

8 Catalysis Engineering: From the Catalytic Material … 13

Fig. 8.4 Effect of the Thiele

424 intrinsic rate expression and the effectiveness factor:

rOBS = ηkS CA,

439 kinetic constant is considered:

8.2.5 Temperature Dependence of the Observed Reaction

8 Catalysis Engineering: From the Catalytic Material … 15

Fig. 8.5 Temperature

473 8.3.1 Multiscale Modeling

476 means of simpliﬁed approaches. Here, we concentrate on the description of the

479 most natural approach would be to proceed according to a bottom-up philosophy,

505 8.3.2 Governing Equations D

8 Catalysis Engineering: From the Catalytic Material … 17

548 following expression [11]:

569 8.3.2.2 Boundary Conditions