A naphtha reforming process development methodology

based on the identification of catalytic reactivity

descriptors
Olivier Said-Aizpuru, Florent Allain, Fabrice Diehl, David Farrusseng,
Jean-François Joly, Aurélie Dandeu

To cite this version:

Olivier Said-Aizpuru, Florent Allain, Fabrice Diehl, David Farrusseng, Jean-François Joly, et al.. A
naphtha reforming process development methodology based on the identification of catalytic reactivity
descriptors. New Journal of Chemistry, 2020, 44 (18), pp.7243-7260. �10.1039/c9nj05349b�. �hal-
02864677�

HAL Id: hal-02864677

https://hal.science/hal-02864677
Submitted on 7 Jul 2020

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents
entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non,
lished or not. The documents may come from émanant des établissements d’enseignement et de
teaching and research institutions in France or recherche français ou étrangers, des laboratoires
abroad, or from public or private research centers. publics ou privés.
Jour
nal
Name

Naphtha reforming process development methodol-

ogy based on the identification of catalytic reactivity
descriptors
Olivier Said-Aizpurua,b , Florent Allaina , Fabrice Diehla , David Farrussengb , Jean-
François Jolya , Aurélie Dandeua∗

Several major refining catalytic processes show a strong dependence of overall performances
upon catalyst formulation and structure. This observation is particularly true in naphtha catalytic
reforming for which product distribution is sensitive to slight changes in active phase formulation.
This review presents different issues encountered in naphtha reforming that are related to active
phase formulation change.
Two research and development approaches are usually proposed. (i) Rationalise the development
of new catalytic formulations in order to increase valuable products selectivity (ii) Design simula-
tors that would guide operation in order to maintain process performances. Current limitations to
a faster process development come from the disjunction between kinetic modelling and catalyst
optimisation.
This critical review is not an extensive analysis of existing naphtha reforming kinetic modelling
methodologies or advance in catalytic behaviour elucidation. Rather, it proposes an original and
pragmatic process development approach aimed at merging catalyst development with kinetic
modelling through the identification of “effective” and “measurable” catalytic descriptors. The
specificities of the naphtha reforming catalysts structure/property relationship are reviewed and
taken into account in order to list potential catalytic descriptors for this process. A focus is made
on a current bottleneck faced in the description of the active phase which lies in understanding the
role of the proximity and the interaction between acid and metallic sites over the balance between
different bi-functional pathways. A variety of experimental approaches that can be used to mea-
sure these naphtha reforming active phase catalytic descriptors are presented and compared.

1 Introduction
Simulation tools are required to guide operation decisions and
Catalysts play a key role in the refining industry and new commer- the influence of the catalyst over the chemical transformation is
cial formulations are frequently introduced in order to increase described through the kinetic model. Hence, there is a need for an
the performances of existing processes. Given the scale of refin- improvement of simulators robustness towards catalytic formula-
ing industrial units, significant overall yields and selectivity gains tion change and for a better structural description of the catalyst
might arise from tiny catalytic structural adjustments. Depend- within the kinetic model.
ing on the severity of the operating conditions, stability issues
The intricate link between the physico-chemical properties of
are responsible for a degradation of initial performances over the
the catalyst and the overall process performances is particularly
time on stream and catalytic structure is also engineered in order
striking in naphtha reforming catalysis. Since the development
to address deactivation. As a consequence, operating a catalytic
of first units in the 1940’s, engineers struggled for the slightest
unit leads to two questions: Which catalyst is required to fulfil
rise in selectivity and productivity. Naphtha reforming is oper-
the performance objectives? How to adjust process conditions in
ated in a narrow range of operating conditions, close to the op-
order to maintain the performances over time?
timum. 1–3 Temperature and partial pressures are carefully set in
order to satisfy several requirements: thermodynamically favour
a
dehydrogenation reactions, compensate endothermicity and han-
IFP Énergies nouvelles, Rond-Point de l’échangeur de Solaize - BP 3, 69360 Solaize,
dle deactivation by coke deposition. However, the selection of the
France. E-mail: aurelie.dandeu@ifpen.fr
b
Univ Lyon, Université Claude Bernard Lyon 1, CNRS, IRCELYON, F-69626, Villeur- fresh catalyst properties as well as the catalyst regeneration con-
banne, France. ditions offer some flexibility in process operation. The demand

J
our
nal
Name,
[yea
r][
,vol
.
],1–19 | 1
for higher reformate yields and for a better control of process se- hydrocracking, toluene dismutation or dealkylation of branched
lectivity is met by a frequent renewal of the commercial catalysts. aromatics.
This results in the introduction of more than tens of new com-
mercial catalytic formulations every five years. 1,4–6 The catalyst
structural properties can also be tuned during process operation Figure 2 illustrates the variety of reactions expected in n-
thanks to the adjustment of chlorination and hydration regener- heptane reforming. Reactions leading to aromatics are strongly
ation conditions. Given the specificities of naphtha reforming, a endothermic and equilibrated, and they are associated to a rise
better description of catalyst structure within process simulators in octane number. Isomerisation reactions are slightly exother-
paves the way for an optimised and more flexible operation of mic and also afford the rise in octane number. On the contrary,
reformers. In this study, a focus is made on the description of cracking reactions are exothermic. This differences of thermal
the active phase of a reforming catalyst. Based on a literature behaviour have a strong impact on process operation.
review, we show that the effect of an active phase formulation
change on reforming selectivity is not fully understood and that
the structural investigation of the active phase remains challeng- 2.1 Catalyst performance requirements
ing. This review does not provide an extensive description of
the state-of-the-art advancement in structural investigation of the The development of radial bed technologies in the 1970’s allowed
catalyst nor a complete report of intrinsic naphtha reforming ki- for the reduction of process pressure. Until then, the increase in
netic modelling methodologies. Here, we propose a pragmatic the production of aromatics by reformers was inhibited by ther-
approach driven by industrial considerations in order to unravel modynamic limitations. Now that reformers can be operated in
the key physico-chemical properties of the catalyst that control optimal dehydrogenation thermodynamic conditions, the biggest
the product distribution at the process scale. These properties are hurdle to higher valuable product yields lies in the design of the
referred to as reactivity descriptors. Then, a method is suggested active phase itself. The role of the catalyst is paramount in con-
in order to implement catalytic descriptors within reforming sim- trolling the balance between these two sets of reactions. 9 Reform-
ulators. This review is divided in three parts. Part one explains ing reactions require a bi-functional catalysis through hydrogen
the need to unravel selectivity descriptors in naphtha reforming. and proton transfer reactions. Naphtha reforming catalysts are
Potential descriptors as well as their measurement methods are therefore usually constituted of a metallic phase (platinum asso-
then listed in part two. Part three proposes a way to integrate ciated or not with other metals) affording hydrogen transfer re-
those descriptors into predictive tools and discusses the advan- actions and dispersed at the surface of a mild acidic solid (usually
tages of such an approach in process development. chlorinated alumina). 10,11 A fine tuning of the reaction network
is needed to adjust product distribution to the needs of the re-
2 The problem of selectivity control and finery. Some units are optimised to reach the maximal produc-
prediction in naphtha catalytic reforming tivity of gasoline bases (C+ 5 products) whereas other units are
aimed at producing aromatics (which allows lower C+ 5 yields and
Naphtha reforming process operation as well as catalyst develop- higher aromatic selectivity). Depending on the targeted selectiv-
ment is mainly about controlling the selectivity of a broad reac- ity, different commercial active phase formulations are available.
tion network. As shown in Figure 1 and Figure 2, naphtha re- Around tens of new industrial reforming catalysts are developed
forming consists in converting a hydrocarbon cut with a boiling every five years 4 , which leads to a broad variety of commercially
point inferior to 180–200◦ C (a naphtha) into a hydrocarbon cut available active phase formulations. Catalyst development en-
with roughly the same distribution in carbon atoms but enriched abled a gradual increase of reformate RON and C+ 5 yields. De-
with high Research Octane Number (RON) compounds such as spite 70 years of continuous optimisation, there is still a tiny but
aromatics and isoparaffins (the reformate). Nowadays, almost all significant theoretical performance increase margin. As an exam-
refineries in the world are equipped with at least one catalytic ple, today, a reformer can be operated at RON values between
naphtha reforming unit as this process contributes to three cru- 90-105, C+ 5 yield up to 90%wt and H2 yields of 3-4%wt . The re-
cial functions in these plants: (i) producing bases for gasoline maining 6%wt is composed of C1 -C2 and C3 -C4 undesired prod-
production (ii) producing aromatic compounds and (iii) gener- ucts. Therefore, the objective is to decrease this 6%wt amount of
ating dihydrogen that can be sent towards hydrogen consuming by-products. These values typically depend on feedstock compo-
units in the refinery. 2,7,8 sition. This theoretical potential improvement is very significant
Naphtha reforming chemical transformations can be sorted out in naphtha reforming and would arise from the optimisation of
in two main categories: the catalyst itself. In spite of the achievements in the structural
- Reactions that are targeted as they permit the increase in characterisation of the catalyst active phase 12 , the identification
octane number such as naphthenes dehydrogenation and alkanes of selectivity descriptors is still lacking to a rational design of new
dehydrocyclisation (strongly endothermic, equilibrated and catalytic systems. Up to now, conventional screening methods
generating H2) or isomerisation of paraffins and naphthenes and trial and error approaches were conducted in naphtha re-
(equilibrated and slightly exothermic). forming catalyst optimisation. Those approaches might not be
- Parasite reactions, usually exothermic and thermodynamically best suited to meet the current theoretical performance increase
enhanced. This group includes reactions such as hydrogenolysis, margin.

2| J
our
nal
Name,
[yea
r][
,vol
.
],
1–19
 160
Aromatics
Naphthenes
Isoparaffins 140
Paraffins
Naphtha

Research Octane Number

120
Dehydrogenation
Proportions (%wt)

Reformate

100

80

60
Dehydrocyclisation

Isomerisation
40
Paraffins
20 Isoparaffins
Naphthenes
Cracking Aromatics
0
3 4 5 6 7 8 9 10 11
Number of carbon atoms Number of carbon atoms

Fig. 1 Composition of a naphtha cut before and after catalytic reforming.

Fig. 2 RON of different hydrocarbons present in a naphtha cut.

2.2 Kinetic modelling based predictions coke deposit. One is referred to the work made by the team of
Controlling naphtha reforming product distribution goes well be- Rumschitzki 18–20 or to dedicated book chapters 1 for examples
yond the choice of the fresh catalyst formulation and the optimal of naphtha reforming models taking into account deactivation by
thermodynamic conditions. In addition to the transformation of carbonaceous deposition.
reagents on active sites, the transport of species within the cat-
alyst pores as well as deactivation phenomena have a deep in- 2.2.3 Complex feedstock description
fluence on the process. Kinetic models, by giving an insight into
these diverse but entangled phenomena play a key role in the The development of naphtha reforming kinetic models has been
elaboration of better simulation tools that would guide operation. driven by a need for more simulation details as well as more ac-
curate reformer performance prediction. Naphtha cuts consist of
2.2.1 Diffusion in catalyst pores
several thousands of compounds and present a huge composi-
When it comes to the description of the influence of textural prop- tion diversity depending on its origin. Most of proposed kinetic
erties on catalytic performance, the challenges lie on the one hand schemes rely on the very first metal/acid bi-functional reforming
in characterising the actual porous structure of the catalyst and mechanism proposed by Mills in 1953. 21
on the other hand in designing a diffusion model that matches
Lumped kinetic models are the most widespread in naphtha
the level of detail in the description of the catalyst. Once the dif-
reforming. The level of sophistication of these models increased
fusion model established, it is of prime importance to couple it
from four lumps 22 to more than twenty pseudo-compounds 23 .
with the kinetic model in order to predict the influence of textu-
Such models are compatible with the description of surface ad-
ral properties on the overall catalytic performances. The reader is
sorption and Langmuir-Hinshelwood kinetics 24 , the inclusion of
referred to comprehensive book chapters and reviews regarding
deactivation kinetics 25 or the integration of transport model
topics such as porous diffusion in naphtha reforming catalysis, its
within the pores 13 . Detailed lumped strategies were found to
coupling with full reforming kinetic models as well as naphtha
be successful approaches for the design of industrial reactors 26,27
reforming catalytic pore network optimisation. 13–16
as well as the description feedstock composition change effects 28 .
2.2.2 Catalyst deactivation
The single event methodology is an example of molecular de-
Reforming catalysts are sensitive to deactivation phenomena such tailed kinetic modelling used in naphtha reforming. Single event
as support specific surface loss, metallic particles sintering, poi- models are built on the description of reaction intermediate (car-
soning of active phase by impurities (heavy metals, Fe, Si. . . ), bocation) reactivity at the surface of the catalyst and compute the
acidity loss by dechlorination as well as carbonaceous deposit at rate constants of the whole reaction network from the kinetics of
the surface of the catalyst. 1 Coke deactivation is the main deacti- intermediates transformations. 29–32 Mixed approaches conciliat-
vation cause. The kinetics of coke formation and its influence on ing the flexibility of a lump model with the level of accuracy of
the kinetics of other reactions should also be integrated into pro- a single event model were also introduced. 33–36 State of the art
cess simulators. Dating back from the pioneering work of Myers methodologies are therefore satisfactory when it comes to predict
in this field 17 , kinetic deactivation models based on coke forma- reformate properties and unit performances with different naph-
tion are built on a mechanistic description of site poisoning by tha feed compositions.

J
our
nal
Name,
[yea
r][
,vol
.
],1–19 | 3
 Process Development
netic parameters such as apparent orders, activation energies and
-Industrial catalyst n -Industrial catalyst n+1 rate constants. These parameters are not often robust to catalyst
-Simulator suitable for catalyst n -Simulator suitable for n+1
change and therefore, the fitting procedure is to be repeated each
Pilot plant time a new catalytic formulation is developed. This makes ki-
experimentation
Catalyst n+1 netic model development a cumbersome step and a bottleneck in
Kinetic parameters fitting
the case of processes known for a high frequency of commercial
Kinetic scheme adjustment
catalyst formulation renewal.

Fig. 3 Traditional catalytic process development scheme. 3 Research and measurement of naphtha
reforming catalytic descriptors
2.3 Current limits and challenges in reforming performance Classical reforming catalysts consist of a reduced metallic phase
prediction (Pt sub-nanometric particles usually associated with other met-
Naphtha reforming is lacking reliable methodologies to model the als) that is supported on a weak acidic support (chlorinated γ-
issues related to active phase formulation modification. Yet, these Al2 O3 ). 1,8 Given the very high dispersion of the metallic phase,
issues are significant for both process operation and reforming its characterisation requires advanced techniques. 38,39 Chlorina-
development. tion has deep consequences on active phase properties. First, this
leads to an overall increase of alumina surface acidity. Chlorine
2.3.1 Operation issues linked to performance prediction
also interacts with the metallic phase. It contributes to the stabil-
The technical and economic requirements of the refinery should isation of Pt particles 40 and might affect the intrinsic activity of
be taken into account by the operators when it comes to choose reduced metal particles in hydro-dehydrogenation. 40,41 As a re-
the nature of the fresh catalyst and to set the operating condi- sult of reforming catalyst peculiarities, formulation changes might
tions. As long as a recommended fresh commercial catalyst is affect the properties of both acid and metallic functionalities as
charged, corresponding kinetic models are perfectly able to iden- well as their interaction. This section reviews the reactivity asso-
tify a set of operating conditions matching the targeted reformate ciated at each phase and reviews potential reforming selectivity
product distribution. However, in order to face deactivation phe- descriptors.
nomena or feed property modifications during the run, operation
also consists in adjusting process and regeneration conditions to
3.1 Naphtha reforming principal reaction families
maintain catalyst performance over the time. 6,37 Different regen-
eration solutions remediate the deactivation issues. Regeneration One can distinguish three groups of reforming reactions: those
usually consists in burning the coke deposits, re-dispersing the ac- catalysed by metal sites, the one occurring on acid sites and the
tive phase, compensating chlorine loss and reducing the catalyst. 1 reactions that require an interaction between the two kinds of
This actually corresponds to a real time active phase formulation sites. The main reforming reactions are listed as follows and
adjustment of the loaded catalyst. The amount of added chlo- sorted out by family as shown in Table 1. The transformation
rine, the introduction of water or the regeneration conditions are of linear paraffins into aromatics through the formation of naph-
therefore example of manifolds used by the operator to control thenes plays a key-role in process operation insofar as it leads
overall performance. Similarly, adjusting chlorination or hydra- to the formation of high value compounds and generates a high
tion of a reforming catalyst to the nature of a feedstock is a com- quantity of hydrogen. Dehydrocyclisation, that is to say the trans-
mon practise. It should be pointed out that nowadays, practical formation of linear paraffins into naphthenes, is considered as
and empirical know-how is more valuable than modelling abili- the first step of aromatisation. N-heptane reforming is a system
ties when it comes to such issues. that can be proposed to illustrate paraffin aromatisation transfor-
mations. N-heptane dehydrocyclisation mechanisms haven’t been
2.3.2 Constraints faced in process development fully unravelled yet. Three potential mechanisms are advanced,
It has been previously pointed out that the identification of cat- out of which two are believed to require acid and metallic site
alytic selectivity descriptors is required for the development of cooperation. These pathways are presented hereafter:
new catalysts as well as more reliable predicting tools (especially Bi-functional path with cyclisation on an acid site:
tools that regeneration operation would benefit from). Another One can imagine the formation of an olefin on a metallic site fol-
aspect is related to the necessity to shorten the time to market lowed by an isomerisation on a Brønsted acid site and a last step
design of both catalysts and kinetic models. Reforming catalytic of dehydrogenation on a metallic site.
process development follows the frame represented in Figure 3 Bi-functional path with cyclisation on metallic sites:
and usually consists in two distinct steps. Optimisation of the cat- A second bi-functional mechanism was suggested for paraf-
alyst is first afforded thanks to dedicated methodologies such as fin dehydrocyclisation. It assumes the direct transformation
catalyst screening combined with scale up studies. Afterwards, of a paraffin into a five atoms ring naphthene on a single
once the final structure of the new catalyst is obtained, specific metallic site. At that point, a ring expansion catalysed by
experiments are conducted to provide the data required for the a Brønsted site would lead to a 6 atoms ring naphthene.
elaboration of a suitable kinetic model. Lumped kinetic models Mono-functional dehydrocyclisation:
implemented in naphtha reforming simulators handle overall ki- Studies carried on non-acid catalysts suggested that dehydrocy-

4| J
our
nal
Name,
[yea
r][
,vol
.
],
1–19
clisation could happen on metallic sites only. For example, the
AlIV: octaheral uncoordinated site AlIII: tetrahedral uncoordinated site
group of Davis reported n-heptane dehydrocyclisation activity us-
ing Pt/Al2 O3 -K whereas the team of Arcoya obtained similar re- (100) Surface (110) Surface
sults on Pt/BaKL zeolite. 42,43 Still, Davis states that dehydrocy-
clisation through the mono-functional pathway is hundred times
slower than bi-functional dehydrocyclisation. 44
Bulk atoms
3.2 Acid functionality descriptors
Over the past twenty years, an increasing number of studies Fig. 4 Comparison between (100) and (110) γ-alumina dehydrated struc-
tures. 72
dedicated to the identification of solid acid reactivity descriptors
has been reported. Significant advance in unravelling key struc-
ture/activity relationships was achieved for systems of industrial
interest. So far, this advance was limited to systems exhibiting NMR hydroxyl surface groups on chlorinated alumina based on
well defined acids sites on which mechanistic studies can be con- a structural Density Functional Theory (DFT). 61 Other analyti-
ducted. For example, the combination between kinetic studies cal techniques also proved to be successful in the characterisation
and the computation of site adsorption energies brought conse- of chlorinated-gamma alumina surfaces. 1 H NMR, 27 Al NMR as
quential insight for catalysts such as polyoxometallates 51,52 or well as bi-dimensional NMR were used to assess the acidity of
zeolites 53,54 . Gamma alumina based supports are also used as in- protons borne by HO groups. 62–64 The adsorption of some probe
dustrial acid catalysts but do not range among well-defined acid molecules can also be suggested to probe the impact of chlori-
sites systems. 55,56 The alumina surface exhibits a variety of un- nation on acidity (e.g. trimethylamine adsorption followed by IR
coordinated AlIV atoms and hydroxyl groups that are respectively spectroscopy 65 ). Adsorption of trimethylphosphine followed by
31 P NMR (which requires toxic hazard management) was found
potential Lewis and Brønsted acid sites. However, the point of
zero charge of this solid is equal to 8. It means that the acidity to be very well suited to the study of the chlorination impact for
of gamma alumina is very low. The fixation of chlorine atoms reforming catalysts. 66,67
on the γ-Al2 O3 surface is a way to enhance the acidity of sur- 3.2.2 Computation of descriptors by molecular modelling.
face HO groups. Chlorine fixation is generally assumed to pro-
ceed through the exchange of a Cl atom with a surface hydroxyl As discussed above, accurate structural characterisations of chlo-
group. 40,55 This overall increase in acidity is either explained by rinated gamma-alumina relies on DFT based models. 68 Example
electron withdrawing effect 57 or by the disturbance of the H-bond of bulk and surface representations shown in Figure 4 and Figure
network at the alumina surface 58 . In naphtha reforming cataly- 5 are derived from ab initio. calculations according to Krokidis
sis, the amount of Cl is set around 1%wt to provide the desired model. 69 Nevertheless, some conclusions based on DFT studies of
moderate acidity to these systems. Cutting edge characterisation γ-alumina surfaces dynamics are debated issues. DFT calculations
and computation techniques give an atomic insight into chlori- were carried out by different groups in order to study the relative
nated gamma alumina structure. What is still poorly understood stability of gamma alumina crystallite facets. Results obtained
is the structural features underlying the moderate acidic reactivity by Raybaud and Sautet indicate a preferential stability of (100)
required for naphtha reforming reactions. facets compared to (110) and (111) facets. 70 On the contrary,
the team of Pinto lead similar studies and found that (100) and
3.2.1 Chlorinated γ-Al2 O3 surface characterisation tech-
(111) surfaces show similar energy and are more stable than the
niques.
(110) facet. 71 Molecular modelling of chlorinated gamma alu-
The combination between various spectroscopy techniques and mina surfaces is not only a powerful tool to provide their detailed
molecular modelling permits an extensive characterisation of structural representation, but it is also a way to study the dy-
gamma-alumina. These techniques are able to identify and quan- namics of these surfaces with regards to variations in operating
tify each type of hydroxyl surface moieties and uncoordinated conditions. Molecular modelling is a way to study in details the
aluminium at the surface of a model reforming catalyst. Infrared substitution of hydroxyl groups by chlorine atoms on γ-Al2 O3 sur-
spectroscopy is a classical and historical tool used for the charac- faces and its consequences on the acidity of HO groups. These
terisation of γ-Al2 O3 surface hydroxyls. The measurement of the calculations give access to the free enthalpy associated to these
H-O bond elongation frequency for a hydroxyl group is directly exchange reactions (endothermic on the whole). 58,71 Carrying on
correlated to the H-O bond strength and therefore to the hydroxyl these calculations at different temperatures and HCl partial pres-
acidic properties. The H-O elongation frequency depends both on sures conditions helps determining the most stable structure in
the geometric environment of the hydroxyl group as well as the different conditions. This leads to the drawing of surface energy
nature of the Al atoms bearing it and different historical mod- state diagrams. Figure 6 is an example of chlorinated γ-Al2 O3
els were built for the assignation of the associated IR peaks. 59,60 surface diagram obtained thanks to DFT studies led by the team
Ab initio. modelling was later used for the attribution of H-O of Raybaud and Sautet. These results are consistent with a well-
peaks obtained in IR spectroscopy of chlorinated gamma alumina known observation in naphtha catalytic reforming which is the
surfaces 58 . More recently, Batista et al. extended the investiga- strong interaction between HCl and H2 O partial pressures and
tion to crystallite edge sites by proposing a new attribution of 1 H the acido-basic properties of the surface.

J
our
nal
Name,
[yea
r][
,vol
.
],1–19 | 5
Table 1 Diversity of the catalytic naphtha reforming reaction scheme

Single site reactions Comments Studies

Catalysis by metals

Dehydrogenations Fast and strongly endothermic

Unsaturated intermediates likely to be transformed on other 45–47

sites
Hydrogenolysis Leads to C1 /C2

Direct cyclisation Rare and slow 42,44

Position isomerisation Perhaps anecdotic 48

Acid catalysis

Olefin isomerisation Likely to occur on both Lewis and Brønsted sites 49

Olefin cyclisation Carried on mild Brønsted sites

(g) (g)

Naphthene expansion Carried on mild Brønsted sites 50

(g) (g)

Cracking Unsaturated intermediates likely to be transformed on other

sites

Multiple site reactions

Hydroisomerisation Dehydrogenation of the linear paraffin on a metal site fol-

lowed by isomerisation on an acid site. Branched olefin is
hydrogenated on a second metal site.
Hydrocracking + Leads to C3 /C4
(g)

Aromatisation Several possible pathways. Requires at least one metallic site.

Coking n. + H2 The formation of coke can also be described as a catalytic

(g) (g)

(s) reaction requiring catalysis by acid and metal sites

6| J
our
nal
Name,
[yea
r][
,vol
.
],
1–19
 3.2.3 Probing acidity by model reactions.

The interpretation of data acquired by a catalytic test involving

acid catalysed reactions is also a way to study the properties of
acid sites (provided the reaction mechanism is well known and
the data are significant). One advantage presented by these meth-
Initial conditions: PHCl=0.016bar, T=800K, PH2O=0.001bar ods is that the investigation of the catalyst is made in conditions
1200
HCl and H2O OH-Cl substitution
that are close to the industrial ones. These kinds of chemical reac-
desorption H2O desorption
tions are usually referred to as “model reactions”. An ideal model
reaction satisfies the following criteria: (i) being unique (ii) hav-
1000 ing an unravelled mechanism (iii) involving chemical species that
OH-Cl substitution
Temperature (K)

HCl adsorption
Cl-OH substitution
r age
can be identified by routine analytic techniques.
HCl desorption o ve
i so - c
In acid catalysis, the kinetics of the transformation depend on
800
HCl adsorption
site nature, strength, density and reaction temperature. Apply-
OH-Cl substitution
ing these principles, Bourdillon and Gueguen managed to iden-
Cl-OH substitution
H2O adsorption tify the distribution in strength and in density of the acid sites
600 of a HY zeolite. 49 Seven model reactions were used, each one
H2O adsorption probing a specific range of acid strength. These reactions can
Cl-OH substitution
HCl and H2O
adsorption
be sorted out by increasing required acid site strength as fol-
400 lows: 3,3-dimethylbutene isomerisation, isomerisation (and dis-
-3 -2 -1 0 1 2 3
log(PHCl/P°) mutation) of 1,2,4-trimethylbenzene, isomerisation (and dismu-
tation) of orthoxylene, 2,2,4-trimethylpentane cracking, isomeri-
Fig. 5 Crystallographic γ-alumina bulk structure obtained by ab initio. sation (and cracking) of 2,4-dimethylpentane, isomerisation (and
calculations according to Krokidis et al. 69 The repartition between AlVI cracking) of 2-methylpentane as well as isomerisation and crack-
and AlIV atoms differs from what is expected for a pure spinel structure.
ing of n-hexane. In order to quantify the strength of the acid
Brønsted sites that are probed, authors use minimal pyridine des-
orption temperature for which a significant activity is detected
(noted TD). This value is calculated in each case by the follow-
ing procedure: the whole catalytic surface site is first poisoned by
pyridine at low temperature. At that point, the middle is heated
up little by little, which leads to a progressive desorption of pyri-
dine from the weakest to the strongest Brønsted sites. Finally, at
each temperature, one calculates the ratio between the measured
activity and the activity that would be measured in the same con-
ditions but in the absence of pyridine. Figure 7 represents the
Oxygen fcc node positions results obtained by Bourdillon and Gueguen and shows the reac-
Tetrahedral site tions considered. For each reaction, a kinetic law is fitted and is
Aluminium
Octahedral site
used to extrapolate an activity at the same temperature (350◦ C).
The authors are then able to associate a rate constant obtained at
O z (111)
(100) the same temperature for each reaction and use it as descriptor
(110)
Al VI (~75%) of the number of active sites for a given range of site strength.
Al IV (~25%) The work of Bourdillon and Gueguen shows that a set of model
y reactions can be used to characterise a variety of acid sites on a
x
given catalyst. Reciprocally, one can use a single reaction, target-
Krokidis bulk Spinel structure ing a given range of acid sites, in order to compare different cata-
model for cristallographic plans lysts. This second approach was tested by Morra et al. to compare
gamma alumina Miller indexation the density in mild acid sites for a set of different metallic phases
supported either on δ or αθ alumina. 73,74 3,3-dimethylbutene
Fig. 6 Surface phase diagram obtained by ab initio. calculations. (100) γ-
(3,3-DMB) isomerisation is taken as the model acid reaction. The
alumina-Cl surface evolution mechanisms in function of the temperature
and the HCl partial pressure. Water partial pressure was fixed at 0.001 reagent structure is not suitable for a π-allylic rearrangement.
bar. 73 Thus, the reaction entails a proton exchange. Pines suggested two
isomerisation mechanisms (shown in Figure 8) for this reaction
according to the Lewis or Brønsted nature of the active site. 75 Yet,
it is generally admitted that this reaction is carried on with mild
Brønsted sites only. The authors used 3,3-dimethylbutene (3,3-
DMB) as a single model reaction, which is known to target acid

J
our
nal
Name,
[yea
r][
,vol
.
],1–19 | 7
 -
1 Cl Cl
+
Al O Al + HCl = Al O Al Cl + H
AR 3,3-DMB n-C6 Cl Cl
o-Xylene
Lewis strong site Brønsted strong site
0.5
2,2,4-tmC5
2mC5 Fig. 10 Strong Lewis and Brønsted sites associated to dichlorinated Al
surface species. 42,76
2,4-dmC5
1,2,4-tmb

T D° 300 400 500 T(°C)

ture desorption higher than 623K. The formation of dichlorinated
aluminium surface atoms was found to be strongly dependent on
Fig. 7 Use of model reactions to probe and characterise the range of
acidity of the different surface sites of a zeolite HY. AR : reaction rate in
the method of chlorination. As an example, the use of perchlo-
presence of pyridine divided by its value without pyridine. 49 rates (CCl4 ) is known to favour the fixation of two Cl on a single
surface aluminium through the formation of an AlCl3 s intermedi-
3,3-DMB ate. N-butane isomerisation is a model reaction used to probe the
slow step catalytic activity arising from dichlorinated aluminium atoms.
2,3-DMB1
Depending on the study, Lewis or Brønsted strong activities are
considered to be at stake in n-butane isomerisation (contrary to
other acid sites at the surface of a chlorinated γ-Al2 O3 ). Primet
et al. used this reaction to investigate the presence of such sites
2,3-DMB2 on chlorinated alumina as well as naphtha reforming-like catalyst
(Pt/γ-Al2 O3 -Cl). Though chlorination of industrial naphtha re-
Fig. 8 3,3-DMB isomerisation mechanism according to Pines. 75
forming catalysts is usually afforded by oxy-chlorination (which
is less likely to form dichlorinated species than the method using
sites (TD between 200 and 400◦ C). This reaction is performed on CCl4 ), over-hydrocracking activity is sometimes suspected to arise
a set of catalysts and a ranking is based on the fitted first order from a tiny amount of dichlorinated atoms at the catalyst surface.
rate constant. Figure 9 shows the results obtained by the authors. Therefore, the injection of small and controlled quantities of wa-
In this case, the first order isomerisation rate constant is directly ter to remove these strong acid sites takes part to the operating
taken as a reactivity descriptor insofar as it enables to rank and procedures of a naphtha reforming unit.
distinguish the activity of different catalysts. On can note that
3,3-DMB isomerisation involves a limited number of species that 3.3 Metallic functionality descriptors
can easily be distinguished by gas chromatography, which make Naphtha reforming catalysts exhibit sub-nanometric platinum
this reaction very convenient to handle. particles (usually associated to other metals such as Sn or Re) at
Particular attention should be paid to dichlorinated aluminium the surface of chlorinated γ-Al2 O3 . Under reductant conditions,
atoms at the γ-Al2 O3 surface. As shown in Figure 10, these groups this metallic phase is able to perform hydro/dehydrogenation and
act as strong Lewis acid site that can be turned into superacid hydrogenolysis reactions. Platinum being expansive, high disper-
Brønsted site under high HCl partial pressures. 76–78 The group of sion levels guarantee a high active metallic surface at low Pt con-
Primet associates these strong Lewis sites to a pyridine tempera- centrations (around 0.3%wt for industrial catalyst). These plat-
inum particles present a very small diameter, usually lower than
one nanometre, which makes the direct observation and study of
BPdMo1.06
these particles difficult owing to the limits of resolution of most
APdMo1.06
analytical devices. The identification of descriptors should also
APt
A take into account the role of chlorine on platinum nanoparticles.
BPt The study of the metallic phase is even more complex in the case
B of multi-metallic systems such as PtSn/γ-Al2 O3 -Cl that are used
BPtSn0.8S with Continuous Catalytic Recycling (CCR) reforming technolo-
ANiMn0.44S gies.
BRhMn0.36S
3.3.1 Pt only catalysts.
0.01 0.1 1 10 100
Even though the share of Pt only catalysts on the current reform-
Ainitial(mmol.h-1.gcata-1) ing market is shrinking, these systems remain favoured for ad-
vance structural studies. 12 Routine electronic microscopy is not
Fig. 9 Acidity scale relying on the initial activity in 3,3-Dimethylbutene
isomerisation measured at 260◦ C for two different solids. 74 Solid A: δ - suitable with an easy observation of sub-nanometric structures.
alumina. Solid B: αθ -alumina. Thus, more sophisticated microscopy techniques are required for

8| J
our
nal
Name,
[yea
r][
,vol
.
],
1–19
 experiments. An XAS experiments provides both XANES (X-ray
Adsorption Near Edge Structure) and EXAFS (Extended X-ray Ad-
sorption Fine Structure) spectra. The nature and the oxidation
number of elements present on the sample can be inferred from
XANES data whereas the EXAFS spectrum gives indications about
the interaction between a metal and its neighbour atoms. EXAFS
study give access to the number of metal atoms within a metallic
particle or average inter-atomic distances. 81–84 To sum up, the
combination between high resolution microscopy, DFT calcula-
tions and XAS spectroscopy agrees with the following observa-
tions. Reforming platinum particles appear as systems comprising
between 10 to 13 atoms with preferential flat or bi-planar mor-
Fig. 11 Two Transmission Electronic Microscopy observation modes of phology. Size and shape rearrangement are to be expected from
the same photograph representing the surface of Pt(3%wt )/γ-Al2 O3 cat- H2 surface partial pressure shifts. Chlorine contributes to the sta-
alyst. 39 Photograph A was provided by Z contrast techniques. White bilisation of both particles and single-atoms. Therefore, particle
zones are associated by authors with platinum particles. Photograph B surface density, particle size and shape, Pt-Pt distances as well as
was taken by “Bright Field” methods and gives an atomic resolution of
particles/single-atom ratios are potential descriptors that can be
the sample. According to the authors, circled structures correspond to
sub-nanometric particles of 1 to 4 Pt atoms. provided by state-of-the-art techniques.

3.3.2 PtSn only catalysts.

a direct observation of metallic nanoparticles. High Resolution The introduction of tin into a naphtha reforming catalyst compli-
Transmission Electronic Microscopy (HR-TEM) as well as High cates its structure and the structural study of this kind of cata-
Angle Annular Dark Field Scanning Transmission Electron Mi- lyst is less documented. Owing to the electronegativity difference
croscopy (HAADF-STEM) are examples of techniques that can be between Pt and Sn, Pt catalytic activity might be influenced by
used for an accurate imaging of metallic particles. The direct ob- charge transfer. Understanding the oxidation state of Sn atoms
servation of these systems is complicated by the high intensity of as well as their interaction with Pt is required for the identifica-
the electron beam that can cause particles sintering during TEM tion of CCR catalyst descriptors. This can be achieved through
imaging. different experimental techniques.
Another issue that can be faced is the strong interaction be- Temperature Programmed Reduction data usually agree with
tween the particles and the support that can complicate the in- a strong interaction between Sn and the γ-Al2 O3 surface. Ac-
terpretation of microscopy pictures. The group of Frenkel con- cording to Carvalho, around 80% of Sn atoms are linked to sup-
sidered a model of platinum particles constituted by 15±9 Pt port oxygen atoms and an irreversible reduction of these atoms in
atoms to interpret naphtha reforming catalyst MET pictures show- Sn(II) occurs during a TPR analysis. 85 Burch et al. realised TPR
ing particles with a diameter ranging between 0.7 and 1.1 nm.39 experiment on γ-alumina successively impregnated with Pt, Sn
HAADF- STEM allows for an observation of even smaller struc- and Re. 86,87 The authors confirm that most of the tin atoms are
tures. The team of Nellist was able to identify Pt structures present as Sn(II). According to them, a fraction of the tin atoms
formed by 1 to 3 atoms as shown in Figure 11. 38 ab initio. mod- is irreversibly linked to the support while some Sn atoms could
elling of supported Pt particles and the role of chlorine atoms form a solid solution with Pt particles Mö ssbauer spectroscopy is
on the surface properties of reforming like catalysts is also car- a useful tool to determine the oxidation number but is restricted
ried on. A systematic study led by Mager-Maury et al. consisted to certain elements such as Fe or Sn. This technique was applied
in predicting the most stable configuration of Pt13/γ-Al2 O3 sys- by Olivier-Fourcade et al. to the study of PtSn/γ-Al2 O3 systems
tems at different temperatures and H2 partial pressures. Figure considering different methods of tin introduction. 88–90 Whatever
12 corresponds to the surface thermodynamic diagram provided the preparation method, authors state that Pt3 Sn and PtSn alloys
by this study. 79 The same team also simulated the influence of are present before reduction. After reduction by H2 , oxometallic
surface chlorination on the given system. One major conclu- Pt-Sn bonds are detected in small quantities and indicate that Sn
sion was that in reforming conditions at chlorine rates close to is partially linked to O atoms from the gamma alumina surface.
the value of the industrial catalysts, the migration of chlorine XANES experiments were applied by Gorczyca et al. to the
and hydrogen atoms from the support to the metallic particles study of the metallic phase of PtSn CCR reforming catalysts. 91,92
is likely to occur, thus leading to a stronger interaction between Ab initio. calculations were used for the interpretation of XANES
Pt particles and gamma alumina. Hence, chlorination of the cat- results. It was found that the Pt10 Sn3 stoichiometry was associ-
alysts participates to the stabilisation of sub-nanometric particles ated to the best description of XANES spectra and was therefore
at the surface of naphtha reforming catalysts. 40 X-ray Absorption the most likely. According to these results H2 would be preferen-
Spectroscopy (XAS), when coupled with ab initio. calculations tially adsorbed on Pt instead of Sn. The introduction of Sn within
is also advantageous for the characterisation of Pt/γ-Al2 O3 metal Pt particles was found to be favoured far from the γ-Al2 O3 sur-
phase. 80 XAS analyses can be carried on under a hydrogen at- face and would result in an electron enrichment of the Pt atoms
mosphere. They are then suited for in situ. reductant conditions inside the particle. Nevertheless, the authors were not able to de-

J
our
nal
Name,
[yea
r][
,vol
.
],1–19 | 9
 H2/HC=100
H/cluster~10-26

log(PH2/P°)
H2/HC=1-10
H/cluster~6-18

H2/HC=0.01
H/cluster~4

Temperature (K)

Fig. 12 Thermodynamic state diagram computed by DFT for Pt13 particles supported on a non-chlorinated (100) γ-Al2 O3 surface. The number of
hydrogen atoms adsorbed at the surface of the particle is given in function of the temperature and the PH2 /PC2 H6 ratio. 79

P7 Eley-Rideal are all possible mechanisms relying on different hy-

Pt Pt pothesis regarding reactant adsorption. Then, the expression of
the corresponding micro-kinetic model depends on the assumed
+ + Pt
H + mechanism. 32,46 In case naphthene dehydrogenation kinetic pa-
H rameters are selected as metallic activity descriptors, one should
C7- +
pay a very close attention to the expression of the kinetic law that
N7 iP7 Pt is used.
Pt

A7 C3/C4 3.4 Bi-functional descriptors

Fig. 13 Bi-functional n-heptane reforming pathways showing competition The association of a hydro/dehydrogenation function with a pro-
between different site transformations. ton transfer activity is required for various refining processes
such as paraffin hydroisomerisation or heavy feedstock hydroc-
racking. These catalysts associate a metallic phase (affording hy-
termine the relative distribution between SnO and SnO2 oxides dro/dehydrogenations) and acid sites located on the support (of-
on the basis of this XANES experiment. ten zeolites or SiO2/Al2 O3 ). These catalysts are often designed,
3.3.3 Model reactions. compared and optimised on the basis of criteria such as site bal-
ance, 97,98 site intimacy 99,100 or support porous structure 101 .
As for acid catalysis, model reactions can be used to generate
catalytic descriptors of the metallic phase. Naphthenes dehydro- Direct extrapolation of these concepts to reforming catalysts is
genation reactions satisfy several requirements to be considered not straightforward. First, γ-Al2 O3 support surface is less organ-
as model reactions. First, these reactions are important in naph- ised than what is observed with zeolites. 13 Second, as previously
tha reforming. They greatly contribute to the increase in octane mentioned, naphtha reforming transformations involve a weak
number as well as to hydrogen production and they are respon- acidity and are therefore predominantly afforded on the metallic
sible for the overall endothermic nature of the process. Second, phase. 102,103 According to Weitkamp classification 103,104 , naph-
their selectivity is high under reforming conditions. Third, cy- tha reforming catalysts can be described as (M+,A-), which
clohexane and methylcyclohexane experimental dehydrogenation means that the activity of the metallic phase is high compared to
into aromatics is well documented. 93–95 However, naphthene de- the one of the acid sites. As a consequence, cracked products are
hydrogenations are strongly endothermic and fast compared to excepted to result from hydrogenolysis reactions at the surface of
other naphtha reforming reactions, which makes the experimen- the catalyst. In this same formalism, hydrocracking and hydroi-
tal study of their intrinsic kinetics difficult. In spite of being somerisation catalyst are rather noted (M+,A+) or (M-,A+).
well studied, the precise mechanism of this reaction on the sur- 3.4.1 Describing reaction competition.
face of an industrial reforming catalyst has not been totally un-
ravelled. 96 In the case of methylcyclohexane aromatisation to In spite of the significant variety of metal/acid bi-functional re-
toluene, different mechanistic pathways could be suggested. Re- fining catalysts, there is a limited amount of possible transforma-
verse Horiuti-Polanyi, Langmuir-Hinshelwood-Hougen-Watson or tions on each kind of sites (see sections above). The succession of

10 | J
our
nal
Name,
[yea
r][
,vol
.
],1–19
 k1
k2
A B C
Site diffusion k-1
A B from C
S1 to S2
Site1 C
Site1
BS1 BS2 CS2
AS1
Catalyst surface
A B
Site2
D
Fig. 14 Bi-functional transformation mechanism at the surface of a catalyst Fig. 15 Two examples of bi-functional transformations. The second case
according to the Weisz formalism. 99 shows the possible interception of a bi-functional reaction by a double
mono-functional transformation.

active phase transformations as well as the competition between be difficult to determine if the intermediary actually diffuses from
the adsorption of different species is responsible for the variety one site to another (as in the Weisz model), or if all the transfor-
of catalytic applications. Tuning reaction competition is therefore mations occur in adsorbed phase.
the key to selectivity control. Figure 13 indicates that it is a typical Tuning of intermediary species diffusion from one site to an-
issue encountered in naphtha reforming catalysis, for which the other does not only matters in order to optimise catalytic activity,
intermediate of the aromatisation pathway could be intercepted but has also consequences on selectivity. Let it consider that the
by other reactions (cracking, coking, and isomerisation). Inter- intermediary B also reacts on site A. In this situation, one has to
mediate species distribution between active sites is often consid- face the issue presented in Figure 15. Two competitive reactions
ered as the main phenomenon ruling bi-functional transforma- can occur provided B is re-adsorbed on S1 or on S2. One can con-
tions. The traditional and general framework that is employed clude, within the Weisz framework, that the ratio between S1-S1
to describe this issue was developed by Weisz and published in and S1-S2 distances controls the probability of B to be re-adsorbed
1962. Weisz’s formalism relies on multi-step transformations at on S1 or S2 and as a consequence, to control the selectivity of the
the surface of a heterogeneous catalyst. 99 An illustration is given overall transformation.
in Figure 14. The generic transformation of a species A to C is
3.4.2 Experimental research of catalytic descriptors on bi-
represented through successive reactions on distinct surface sites,
functional catalysts.
S1 and S2. This example introduces B as a species generated on a
surface site and transformed on another. B can be desorbed and Within the Weisz formalism, inter-site distance is expected to be
re-adsorbed on the surface. a relevant activity descriptor. Depending on the catalytic system
which is studied, different authors were able to confirm or not the
One purpose of the Weisz study was to determine conditions
importance of site intimacy at the catalytic surface on the activity
for which the diffusion of the intermediate species from one
and selectivity of the reaction. 100,105,106 Studying the validity of
site to another can be neglected compared to the intrinsic kinet-
Weisz intimacy criterion with these systems is challenging inso-
ics of the site transformations. The mathematical derivation of
far as it means controlling the nanoscale distance between sites.
these conditions leads to the so-called “Weisz intimacy criterion”.
Fatty acid hydrocracking on Pt supported on an acidic supports
Consequently, according to Weisz, the diffusion of B can be ne-
was investigated be Zečević et al.. 100 In this case, the authors
glected when the Thiele modulus of the reaction B → C is lower
managed to tune the nanoscale repartition between sites by con-
than 1, that is to say when the following condition is fulfilled:
d[C] 1 L2   trolling the grafting of Pt particles on a hybrid support (nanoscale
dt . [Beq ] D < 1. In this expression, Beq refers to the concentra- mixture between acid HY zeolite and inert alumina). This study
tion of B in case the equilibrium A = B is reached. L represents
confirms the strong link between inter-site distance and product
the characteristic length associated to the diffusion of B whereas
selectivity, even below the distance defined by the Weisz inti-
D corresponds to the diffusion coefficient of B.
macy criterion. Instead of quantifying inter-site distance using
The Weisz intimacy criterion is often used in catalyst optimisa- advanced analytic techniques to estimate Weisz intimacy crite-
tion. Catalysts whose acid and metallic sites are close enough to rion, Mendes et al. have suggested the use of model reactions to
guarantee the satisfaction of the Weisz criterion are often targeted probe the inter-site distance. 105
in order to optimise the activity of the catalyst. Nevertheless, the
Weisz formalism relies on strong assumptions. It supposes a ho- 3.4.3 Probing active site interaction in naphtha reforming.
mogeneous distribution of sites, which is not necessarily guaran- The separate investigation of acid and metallic phase reactivity
teed with systems such as naphtha reforming catalysts. Moreover, (thanks to the methods suggested on the previous section) is a
it is not always easy to determine an average distance between prerequisite to the study of site interaction. A couple of issues
sites as well as a characteristic length of diffusion. In absence linked to mono-functional catalysis are to be addressed first. For
of detailed information about the reaction mechanism, it can also instance, bi-functional hydrocracking might be ruled out depend-

J
our
nal
Name,
[yea
r][
,vol
.
],
1–19 | 11
ing on the ability of the acid of phase to catalyse or not branched deposition.
paraffin cracking. As pointed out, direct dehydrocyclisation of
paraffins into aromatics on the metallic phase only has been re-
This hurdle can be lowered through the choice of an adequate
ported in the literature. The rates of mono and bi-functional de-
testing methodology. The aim is to acquire experimental reac-
hydrocyclisation paths should therefore be compared. The frac-
tion data that are sensitive to active phase descriptor variations.
tion of the acid sites that is actually involved in bi-functional
Dedicated experimental methodologies address this point. The
transformations is still unclear. For a given metallic site, coopera-
choice of the reactor system can be adapted to lower the impact
tion with acid sites might be limited to neighbouring OH groups.
of heat and mass transfer. Testing powders in a miniature isother-
Hence the interest of probing the properties of the acid phase only
mal reactor is an example of conventional methodology suited
in presence or not of the metallic phase.
for intrinsic kinetics studies. Using a set of model reactions (see
To our knowledge, experimental investigation of Weisz crite-
2), preferably not prone to coke formation, is also recommended
rion relevance in naphtha reforming chemistry hasn’t been re-
to probe the reactivity of each kind of sites separately and in in-
ported in the literature yet. Though technically challenging, tun-
teraction. Therefore, the kinetic parameters of model reactions
ing the location of active sites at the crystallite scale thanks to
obtained in these specific testing conditions are considered as the
dedicated preparation techniques is possible. Then, carrying on
first gate into which reactivity descriptors can be integrated (see
reactions that target both acid and metallic sites can be suggested
Figure ??).
to test Weisz’s approach in naphtha reforming catalysis. If not,
inter-site distance might not be a relevant bi-functional descrip-
tor contrary to the overall balance between acid and metallic sites
or to the intrinsic properties of individual sites. 4.2 Model reaction descriptor based models
Examples of naphtha reforming model reactions are compared
in Table 2. N-heptane reforming is a common simplified system
which is used to assess the reactivity of a naphtha reforming cata- A quantitative expression of model reaction intrinsic kinetic pa-
lyst. First, n-heptane is one of the main components of a naphtha rameters in function of descriptors is sought in the proposed
and represents between 5 and 15%wt of the feedstock (C7 cut pro- method. Dedicated statistical correlation tools exist and could
portion being around 35-40%wt ). Moreover, n-heptane is likely to be applied to the identification of the descriptors that are actually
undergo most of reforming transformations (see Figures 2 and correlated to the targeted parameters. The Quantitative Structure
13), including its aromatisation into toluene which is associated Activity Relationship (QSAR) formalism has already been applied
to a high increase in RON. Most of model reactions mentioned in to the design of heterogeneous catalytic descriptor libraries 115
Table 2 are actually encountered in n-heptane reforming. One can and to the elaboration of statistical predictive models in both
also point out that n-heptane reforming presents a good compro- acid 74 and bi-functional catalysis 116 . Figure 17 illustrates a four
mise between the diversity of its reaction network and the num- step approach aimed at building descriptor based models. A set of
ber of parameters required in order to build a detailed kinetic catalysts presenting a diversity of formulations is first gathered.
model. It is then tested on a target reaction. A set of kinetic parame-
ters corresponding to the same lumped model is then fitted for
4 Integration of active phase descriptors each catalyst. Given the variety of active phase descriptors and
within performance predictive tools possible model reactions, this approach goes along with a signif-
icant amount of experiments. Acquisition of kinetic information
The following discussion proposes an approach to integrate in-
would benefit from parallelisation and miniaturisation of experi-
trinsic reactivity descriptors into the overall naphtha reforming
ments provided by high throughput experimentation techniques.
process development scheme. The aim of this methodology is to
This reactor configuration also allows for the operation of reactor
answer the need for a better prediction of active phase formula-
in isothermal mode and closer to ideal hydrodynamics, which is
tion change in current reformer simulators stressed out in Section
a valuable asset for kinetic studies. The high amount of formula-
one.
tions that are to be tested cannot be reasonably fully characterised
by the set of sophisticated techniques (XANES, HR-TEM. . . ) re-
4.1 Identifying active phase descriptors inlets quired for a thorough structural investigation. Industrial consid-
One of the main challenges encountered in building descriptor erations would insist on descriptors that are easily measurable
based predicting tools lies in the correlation between numerous and controllable thanks to a preparation method (H2 /O2 titration,
and diverse phenomena. First, transport limitation and cata- XRF. . . ). This does not mean that the state-of-the art structural
lyst deactivation interfere with the active phase transformations. investigation of reforming catalyst is useless. On the contrary,
However, as summed up in Table 3, the corresponding descrip- recent advance provide atomic scale crystallite models that can
tors are distinct. Second, the intrinsic reactivity of the catalyst be used from the interpretation of routine analysis. For exam-
itself reflects the contribution of various active sites transforma- ple, thermodynamic state of the art DFT models can be suggested
tions. This problem is precisely faced with conventional naphtha to deduce chlorine or hydroxyl crystallite repartition from overall
reforming catalyst screening techniques that give few clues about chlorine content (measurable by XRF) and crystallite shape (de-
the intermingling between metal/acid site interaction and coke duced from TEM observation on the alumina precursor).

12 | J
our
nal
Name,
[yea
r][
,vol
.
],1–19
Table 2 Example of reactions that could be used to probe surface properties of a naphtha reforming catalyst

Reaction Competitive reactions Comments Studies

Acid catalysis

+ None (under mild conditions) Targets mild Brønsted sites 49

(g) (g) (g)

+ None (under mild conditions) Isomers ratio is a function of acid strength and targets mild 107–109

(g) (g) (g) Brønsted sites

Cracking Targets strong acid sites (dichlorinated Al species) 77,78

(g) (g)

+
(g) (g) 50
Cracking (and dehydrogenation on metal sites) Poor selectivity in presence of metal sites
(g)

+ +
(g) (g)

Catalysis by metals

Cracking (severe conditions) and contraction (on acid sites) Well studied, kinetic studies available but mechanism still de- 45–47

bated

+3H2 Ring expansion and cracking (under severe conditions) Reactions sensitive to deactivation 110

(g) (g) (g)

Isomerisation and cracking Poor selectivity in presence of acid sites 111

(g) (g)

Bi-functional catalysis

Complex reactive scheme Unknown mechanism 20,44,112

+3H2 Complex reactive scheme Well studied 44,113,114

(g) (g) (g)

+ + Cracking under severe conditions Does not represent the overall diversity of reforming reac- 115,116
tions
(g) (g) (g) (g)

J
our
nal
Name,
[yea
r][
,vol
.
],
1–19 | 13
 Issues
Simulator

Suitable for a reference catalyst and a set of feedstocks

Next catalyst generation property adjustment

Ex-regenerator catalyst performance prediction

Unconventional catalyst performance prediction Reactor model

Kinetic model
Pore plugging model
Pore diffusion model
=f(textural descriptors,
= f(textural descriptors)
deactivation descriptors)

Intrinsic kinetics scheme

Catalyst characterisation

mono-functional reaction Deactivation model

= f(active phase descriptor) =f(deactivation descriptors)
Textural descriptors

bi-functional descriptor Coking

Active phase descriptors
= f(active phase descriptors) Dechlorination

Deactivation descriptors

Fig. 16 Different descriptors inlets into complete reforming simulators.

Table 3 Examples of parameters suggested to describe the performances of naphtha reforming catalysts

Issues Possible descriptors Quantitative integration

Specific surface (SBET) Enables mass balance in a grain
Porous volume (N2 adsorption isotherm) De f f = f (ε, τ)
Textural properties Modelling intra-particular diffusion Crystallite size (TEM) (effective
Porosity (ε ,Hg porosimetry) Tortuosity (τ) p diffusion coefficient)
φ = L ( Dekf f ) (Thiele modulus)
Fractal dimension (N2 adsorption isotherm)
Determining the nature and the strength of an acid site %wt Cl (FX analysis)
T OF = f (nbsites) (Turn Over Frequency)
Acid functionality Assigning a type of reactions to a range of acid sites ν-OH (FTIR)
(k, Ea , n) (kinetic parameters of acid model reactions)
Quantifying the acid sites covering rates given by DFT calculation(ΘOH , ΘCl )
%wt Pt (FX)
Metal dispersion (H2 chemisorption)
Identifying the metallic sites and their structure Singlet atom/Particle ratio
T OF = f (nbsites) (Turn Over Frequency)
Metallic functionality Quantifying the metallic sites Oxidation number (XANES)
(k, Ea , n) (kinetic parameters of acid model reactions)
Assessing hydrogen transfer efficiency Shape of nanometric particles (HAADF-TEM/DFT)
Size of sub-nanometric particles (HAADF-TEM)
ΘH (covering rate from DFT calculations)
Determining the interaction and Pt/H+ proximity (inferred from 1 H 2D
the proximity between acid and metallic sites NMR+EXAFS+HAADF-TEM+DFT calculations)
Bi-functional catalysis (k, Ea , n) (bi-functional reaction)
Identifying the fraction of acid site Number of OH at the Pt neighbouring
actually involved in bi-functional transformations Acid/Metal overall site ratio
Estimating catalytic deactivation over the time %wtCl (Overall amount of
Estimating the amount of coke produced coke from elemental analysis or TPO)
in function of the operating conditions Tc (temperature of combustion
Coking Deactivation function parameters
Assessing the ability to regenerate the catalyst obtained from TPO)
in function of the Deposit crystallographic structure (XRD, TEM)
nature of the carbonaceous deposit H/C coke ratio

14 | J
our
nal
Name,
[yea
r][
,vol
.
],1–19
 kinetic studies
New catalysts Initial guess of Final set of
parameters parameters
parameter
Acquisition of Extrapolation to a family optimisation
Set of catalyst 1 Catalytic tests descriptors of reactions by LFER
Complete kinetic model
formulations
Descriptors Kinetic parameters
3 for a model system

Characterisations, 2 Kinetic modelling Descriptor based model

model reactions,
etc Fig. 18 Use of descriptor based model for a faster kinetic model fitting
procedure.
Catalytic Sets of kinetic
descriptors parameters(k,Ea,n)
erty degrees in kinetic modelling. One can legitimately assume a
common dependence of kinetic parameters for a given family of
reactions. For instance, the impact of a formulation change on
4 Statistical correlations isomerisation kinetics should be extrapolated between different
kinds of paraffins. This principle has been successfully applied by
Descriptor based model: (k,Ea,n) = f(descriptors) the group of Klein on the basis of Linear Free Energy Relationships
(LFER). 117,118 This methodology enables the author to extrapo-
late the kinetics of complex refining process schemes from a lim-
Fig. 17 Method suggested to build descriptor based kinetic models ited number a reactions. This method was successfully applied to
thanks to the identification of statistical correlations between kinetic pa- the identification of catalytic structural descriptors. 33 From this
rameters and descriptors. perspective, a similar approach is suitable to naphtha reforming
issues.
4.3 Extrapolation of naphtha reforming kinetic parameters
from model reaction testing 4.4 Prospects and benefits of descriptor integration into
The nature of the kinetic models fitted on model reaction data models
should be discussed. The use of model reactions allows for the use The establishment of descriptor based models is particularly
of detailed kinetic models based on a mechanistic description of promising with regards to catalyst process development. The-
the transformation. One can also suggest simpler rate expressions oretically, it could circumvent the difficulties mentioned above
(for instance power laws) that will be comparable to the actual by shifting from a two phases process development scheme illus-
expressions handled in most of naphtha reforming lumped kinetic trated in Figure 3 to a continuous and simultaneous development
models. Integrating catalytic descriptors into simple power laws of both catalysts and models as shown in Figure 19. This alter-
seems to bring less information on mechanisms but presents the native process improvement methodology consists in coupling the
advantage of handling a limited number of parameters. As illus- development of new catalysts with the elaboration of their kinetic
trated in Figure 18, this also permits an easier extrapolation of model and presents several advantages. First, the catalytic testing
parameters fitted on a model reaction towards parameters associ- strategy is more efficient. The usual scheme consists in carrying
ated to similar reactions in a full lumped reforming model. on two distinct families of tests (screening and kinetic studies).
When it comes to adjust a kinetic model to a new catalyst for- Here, we propose the inclusion of model reaction testing that
mulation, several issues encountered in the parameter estimation benefit both to the determination of catalyst performance and the
procedure can be distinguished. What are the reactions that are acquisition of some kind of kinetic information. De-correlating in-
the most affected by a given formulation change? Which initial trinsic kinetic studies from deactivation and heat and mass trans-
values might be reasonably suggested for the initialisation of the fer thanks to the experimental strategy eases the elaboration of
optimisation procedure? Both question are hard to solve in the models.
conventional process development scheme (see Figure 3). Di- Provided the descriptor based models are predictive, descrip-
rect extrapolation of values provided by model reaction descriptor tor based models can also be used backwards in order to opti-
based models could therefore lead to a valuable fastening of fit- mise catalyst formulation thanks to in silico. screening. Within
ting procedures. For instance, if a formulation change is found to this framework, descriptor based kinetic parameters fitting is a
affect the reactivity of metallic phase mono-functional reactions continuous process and integrates the reaction data obtained all
only, this information can be directly applied to simplify the fitting along the screening steps. This results in a significant speed up
procedure. The guess of initial parameters is equally a valuable of the full process development scheme. The integrated catalytic
asset in order to choose between several sets of parameters ob- process development methodology which is proposed is therefore
tained by optimisation procedures and to deal with the problem attractive but one can insist on the need in truly reliable and easy-
of local extrema. access effective descriptors that can be easily tailored through the
Despite the tremendous amount of species in a naphtha cut, catalyst preparation steps. Otherwise, the strategy presented in
reactions can be gathered in a limited number of families. This Figure 19 might be difficult to execute. Very similar approaches
observation paves the way for a dramatic reduction of the lib- have already been successfully implemented by several authors

J
our
nal
Name,
[yea
r][
,vol
.
],
1–19 | 15
 Scientific breakthrough, new technical
constrainsts...

Industrial catalyst Unknown solids

Suitable kinetic model

5) Scale up Obsolete set of 1) Characterisations

kinetic parameters

Promising
New set of kinetic New possible
catalysts "Pre-descriptors"
parameters catalysts (hits)
(leads)

First fit of kinetic

parameters
2) Catalytic testing
4) Catalytic testing

New possible
Rough ranking of
catalysts (hits)
3) In silico screening hits performances

Catalyst development

Fig. 19 Descriptor based approach coupling the development of new catalysts with the elaboration of their associated kinetic models.

when it comes to the modelling of transport phenomena within based kinetic models at the process scale can go beyond the op-
heterogeneous porous catalysts and their backward application timisation of new solids. Such an approach also accelerates the
to optimise the textural properties of this solids. For instance, adjustment of kinetic models to new catalysts and paves the way
the teams of Szczygieł and Coppens have already achieved meso- for a description of the structural evolution of the catalyst during
porous γ-Al2 O3 textural optimisation on the basis of a transport operation.
model. 14,15,119,120 Another example, in the case of zeolites, is
provided by the group of Marin. 121 Recently, a comprehensive 5 Conclusion
descriptor based heterogeneous catalyst development methodol- A couple of naphtha reforming industrial issues are linked to ac-
ogy was proposed by Pirro et al. 122 This methodology focuses tive phase formulation changes. The choice of the fresh catalyst
on kinetic model based catalyst design and reviews the different formulation as well as the choice of regeneration parameters are
statistic tools that can be used for the identification of catalytic important operating issues that call for a better understanding of
descriptors before being applied on methane oxidative coupling. the structure/activity relationship of the catalyst. The elaboration
Therefore, the descriptor based process development shown in of predicting tools for the next generation of catalysts hampers
Figure 19 can fully benefit from the discussion made by Pirro and the reduction of simulators time to market. This problem also
co-workers. The mutual benefit of catalyst and model develop- asks for the implementation of reactivity descriptors in the design
ment is not afforded in the traditional approach depicted in Fig- of kinetic models.
ure 3. It is shown here that the interest in handling descriptor
This review presents possible descriptors and highlights the

16 | J
our
nal
Name,
[yea
r][
,vol
.
],1–19
gaps for which descriptors are currently missing. Recent studies 13 M. Coppens and G. Froment, Chemical engineering science,
provided a deep insight on chlorinated gamma alumina surface 1996, 51, 2283–2292.
structure and reforming catalyst nanoparticle properties. How- 14 J. Szczygiel, Chemical Industries-New York-Marcel Dekker-,
ever, the physico-chemical parameters controlling the selectivity 2004, 100, 275–334.
of reforming bi-functional transformations are still unclear and 15 J. Szczygieł, Energy & Fuels, 2005, 19, 7–21.
the intricate relation between chlorination, metallic phase disper- 16 J. Szczygieł, Energy & fuels, 1999, 13, 29–39.
sion and surface acidity remains to be investigated. 17 C. G. Myers, W. H. Lang and P. B. Weisz, Industrial & Engi-
A pragmatic process development methodology involving the neering Chemistry, 1961, 53, 299–302.
use of descriptor based kinetic models is suggested. This method
18 K. Liu, S. Fung, T. Ho and D. Rumschitzki, Journal of Catal-
relies on the identification of measurable and quantifiable over-
ysis, 1997, 169, 455–468.
all catalytic descriptors. The addition of model reaction kinetic
19 K. Liu, S. C. Fung, T. C. Ho and D. S. Rumschitzki, Industrial
testing to the process development scheme is advised in order to
& Engineering Chemistry Research, 1997, 36, 3264–3274.
probe active phase properties and to get access to the valuable ki-
20 K. Liu, S. Fung, T. Ho and D. Rumschitzki, Journal of Catal-
netic information. Descriptor based kinetic modelling approaches
ysis, 2002, 206, 188–201.
require a heavy and well settled experimental strategy. However,
as discussed here with the example of naphtha reforming, these 21 G. Mills, H. Heinemann, T. Milliken and A. Oblad, Industrial
methods appear as a powerful tool to unlock significant issues & Engineering Chemistry, 1953, 45, 134–137.
encountered with processes for which operating conditions are 22 R. Smith, Chem. Eng. Prog, 1959, 55, 76–80.
dictated by catalyst handling. 23 P. Van Trimpont, G. Marin and G. Froment, Applied Catalysis,
1986, 24, 53–68.
Conflicts of interest 24 W. Kmak, AIChE Meeting, Houston, TX, 1972.
There are no conflicts to declare. 25 M. Ramage, K. Graziani and F. Krambeck, Chemical Engineer-
ing Science, 1980, 35, 41–48.
Acknowledgements
26 J. Ancheyta-Juárez and E. Villafuerte-Macías, Energy & Fuels,
The authors would like to thank colleagues from IFPEN Catalysis, 2000, 14, 1032–1037.
Biocatalysis and Separation division for fruitful discussions and 27 J. Ancheyta-Juarez, E. Villafuerte-Macias, L. Diaz-Garcia and
advice regarding the application of a descriptor based approach E. Gonzalez-Arredondo, Energy & fuels, 2001, 15, 887–893.
in naphtha reforming. 28 J. Ancheyta, E. Villafuerte, P. Schachat, R. Aguilar and
Notes and references E. Gonzalez, Chemical engineering & technology, 2002, 25,
541–546.
1 G. Antos and A. Aitani, Catalytic naphtha reforming, revised
29 P. Joshi, K. Ankush, M. Tahmid and M. Klein, Energy & fuels,
and expanded, CRC Press, 2004.
1999, 13, 1135–1144.
2 P. Leprince et al., Procédés de transformation. Le raffinage du
30 W. W., C. Bennett, R. Tanaka, G. Hou and M. Klein, Fuel
pétrole, Éditions Technip, Paris, 1998.
Processing Technology, 2008, 89, 344–349.
3 J. Boitiaux, J. Deves, B. Didillon and C. Marcilly, Marcel
31 R. Sotelo-Boyás and G. Froment, Industrial & Engineering
Dekker, New York, 1995, 79–111.
Chemistry Research, 2009, 48, 1107–1119.
4 G. Ballal, PEP Report 129C, 2015.
32 J. Verstraete, PhD thesis, Universiteit Gent, 1997.
5 P. Le Goff, F. Le Peltier, B. Domergue and J. Joly, Petroleum
33 B. Watson, M. Klein and R. Harding, Industrial & Engineering
technology quarterly, 2003, 8, 27–31.
Chemistry Research, 1996, 35, 1506–1516.
6 P. Y. Le-Goff, J. Lopez and J. Ross, Hydrocarbon processing,
34 H. Cochegrue, P. Gauthier, J. Verstraete, K. Surla, D. Guil-
2012, 91, 47.
laume, P. Galtier and J. Barbier, Oil & Gas Science and
7 V. Haensel and G. Donaldson, Industrial & Engineering Chem-
Technology–Revue d’IFP Energies nouvelles, 2011, 66, 367–
istry, 1951, 43, 2102–2104.
397.
8 G. Curtis and V. Haensel, Reforming process, 1949, US Patent
35 M. Maradei Garcia, PhD thesis, Université de Poitiers, 2002.
2,478,916.
36 H. Cochegrue, PhD thesis, Université de Poitiers, 1992.
9 D. Iranshahi, M. Karimi, S. Amiri, M. Jafari, R. Rafiei and
37 B. Domergue and P. le Goff, Petroleum technology quarterly,
M. Rahimpour, Chemical Engineering Research and Design,
2006, 11, 67–73.
2014, 92, 1704–1727.
38 P. Nellist and S. Pennycook, Science, 1996, 274, 413.
10 C. Marcilly, Acido-basic catalysis: application to refining and
39 A. Frenkel, C. Hills and R. Nuzzo, The Journal of Physical
petrochemistry, Technip Ophrys Editions, 2006, vol. 2.
Chemistry B, 2001, 105, 12689–12703.
11 J. F. Le Page, Paris: Editions Technip, 1987.
40 C. Mager-Maury, C. Chizallet, P. Sautet and P. Raybaud, ACS
12 S. Kelly, W. Sinkler, L. Xu, S. Sanchez, C. Akatay, H. Wang
catalysis, 2012, 2, 1346–1357.
and J. Chen, Chinese Journal of Catalysis, 2019, 40, 1637–
41 G. Arteaga, J. Anderson and C. Rochester, Journal of Cataly-
1654.
sis, 1999, 187, 219–229.
42 B. Davis, Journal of Catalysis, 1973, 29, 398–403.

J
our
nal
Name,
[yea
r][
,vol
.
],
1–19 | 17
 43 A. Arcoya, X. Seoane and J. Grau, Applied Catalysis A: Gen- 67 D. Guillaume, S. Gautier, F. Alario and J. Deves, Oil & Gas
eral, 2005, 284, 85–95. Science and Technology, 1999, 54, 537–545.
44 B. Davis, Catalysis Today, 1999, 53, 443–516. 68 M. Digne, P. Sautet, P. Raybaud, P. Euzen and H. Toulhoat,
45 J. Sinfelt, H. Hurwitz and R. Shulman, The Journal of Physi- Journal of Catalysis, 2004, 226, 54–68.
cal Chemistry, 1960, 64, 1559–1562. 69 X. Krokidis, P. Raybaud, A.-E. Gobichon, B. Rebours, P. Eu-
46 P. Van Trimpont, G. Marin and G. Froment, Industrial & en- zen and H. Toulhoat, The Journal of Physical Chemistry B,
gineering chemistry fundamentals, 1986, 25, 544–553. 2001, 105, 5121–5130.
47 M. Usman, D. Cresswell and A. Garforth, Industrial & Engi- 70 M. Digne, P. Sautet, P. Raybaud, P. Euzen and H. Toulhoat,
neering Chemistry Research, 2011, 51, 158–170. Journal of Catalysis, 2002, 211, 1–5.
48 M. Guisnet, G. Perot and R. Maurel, Journal de Chimie 71 H. Pinto, R. Nieminen and S. Elliott, Physical review B, 2004,
Physique, 1972, 69, 1059–1064. 70, 125402.
49 G. Bourdillon, C. Gueguen and M. Guisnet, Applied Catalysis, 72 C. Mager-Maury, PhD thesis, École Normale Supérieure de
1990, 61, 123–139. Lyon, 2012.
50 G. McVicker, O. Feeley, J. Ziemiak, D. Vaughan, 73 G. Morra, PhD thesis, Université Claude Bernard Lyon 1,
K. Strohmaier, W. Kliewer and D. Leta, The Journal of 2008.
Physical Chemistry B, 2005, 109, 2222–2226. 74 G. Morra, D. Farrusseng, E. Guillon, S. Morin, C. Bouchy and
51 P. Deshlahra and E. Iglesia, ACS Catalysis, 2016, 6, 5386– C. Mirodatos, Topics in Catalysis, 2010, 53, 49–56.
5392. 75 H. Pines, Journal of Catalysis, 1982, 78, 1–16.
52 M. Sarazen, E. Doskocil and E. Iglesia, Journal of Catalysis, 76 P. Bernard and M. Primet, J. Chem. Soc., Faraday Trans.,
2016, 344, 553–569. 1990, 86, 567–570.
53 A. Vos, K. Nulens, F. De Proft, R. Schoonheydt and P. Geer- 77 A. Melchor, E. Garbowski, M. Mathieu and M. Primet, Jour-
lings, The Journal of Physical Chemistry B, 2002, 106, 2026– nal of the Chemical Society, Faraday Transactions 1: Physical
2034. Chemistry in Condensed Phases, 1986, 82, 1893–1901.
54 C. Wang, R. Brogaard, B. Weckhuysen, J. Nørskov and 78 A. Melchor, E. Garbowski, M. Mathieu and M. Primet, Jour-
F. Studt, The journal of physical chemistry letters, 2014, 5, nal of the Chemical Society, Faraday Transactions 1: Physical
1516–1521. Chemistry in Condensed Phases, 1986, 82, 3667–3679.
55 A. Castro, O. Scelza, G. Baronetti, M. Fritzler and J. Parera, 79 C. Mager-Maury, G. Bonnard, C. Chizallet, P. Sautet and
Applied Catalysis, 1983, 6, 347–353. P. Raybaud, ChemCatChem, 2011, 3, 200–207.
56 G. Ertl, H. Knözinger and J. Weitkamp, 1997. 80 F. Le Normand, A. Borgna, T. Garetto, C. Apesteguia and
57 A. Kytökivi, M. Lindblad and A. Root, Journal of the Chemical B. Moraweck, The Journal of Physical Chemistry, 1996, 100,
Society, Faraday Transactions, 1995, 91, 941–948. 9068–9076.
58 M. Digne, P. Raybaud, P. Sautet, D. Guillaume and H. Toul- 81 M. Vaarkamp, J. Miller, F. Modica and D. Koningsberger,
hoat, Journal of the American Chemical Society, 2008, 130, Journal of Catalysis, 1996, 163, 294–305.
11030–11039. 82 S. Bradley, W. Sinkler, D. Blom, W. Bigelow, P. Voyles and
59 A. Tsyganenko and V. Filimonov, Journal of Molecular Struc- L. Allard, Catalysis letters, 2012, 142, 176–182.
ture, 1973, 19, 579–589. 83 H. Mistry, F. Behafarid, S. R. Bare and B. Roldan Cuenya,
60 H. Knözinger and P. Ratnasamy, Catalysis Reviews, 1978, 17, ChemCatChem, 2014, 6, 348–352.
31–70. 84 W. Sinkler, S. Sanchez, S. Bradley, J. Wen, B. Mishra, S. Kelly
61 A. Batista, D. Wisser, T. Pigeon, D. Gajan, F. Diehl, M. Rival- and S. Bare, ChemCatChem, 2015, 7, 3779–3787.
lan, L. Catita, A. Gay, A. Lesage, C. Chizallet and P. Raybaud, 85 L. Carvalho, C. Pieck, M. Rangel, N. Fıgoli, J. Grau, P. Reyes
Journal of Catalysis, 2019, 378, 140–143. and J. Parera, Applied Catalysis A: General, 2004, 269, 91–
62 E. DeCanio, J. Edwards and J. Bruno, Journal of catalysis, 103.
1994, 148, 76–83. 86 R. Burch and L. Garla, Journal of Catalysis, 1981, 71, 360–
63 R. Wischert, P. Florian, C. Copéret, D. Massiot and P. Sautet, 372.
The Journal of Physical Chemistry C, 2014, 118, 15292– 87 R. Burch, Journal of Catalysis, 1981, 71, 348–359.
15299. 88 J. Olivier-Fourcade, M. Womes, J. Jumas, F. Le Peltier,
64 M. Taoufik, K. Szeto, N. Merle, I. Rosal, L. Maron, J. Tré- S. Morin and B. Didillon, ChemPhysChem, 2004, 5, 1734–
bosc, G. Tricot, R. M. Gauvin and L. Delevoye, Chemistry-A 1744.
European Journal, 2014, 20, 4038–4046. 89 A. Jahel, S. Lacombe, P. Avenier, J. Olivier-Fourcade and
65 W. Haag and H. Pines, Journal of the American Chemical So- J. Jumas, Hyperfine Interactions, 2012, 207, 37–40.
ciety, 1960, 82, 2488–2494. 90 A. Jahel, P. Avenier, S. Lacombe, J. Olivier-Fourcade and
66 D. Guillaume, PhD thesis, Université Pierre et Marie Curie, J. Jumas, Journal of Physics: Conference Series, 2010, p.
1997. 012045.
91 A. Gorczyca, PhD thesis, Université de Grenoble, 2014.

18 | J
our
nal
Name,
[yea
r][
,vol
.
],1–19
 92 A. Gorczyca, P. Raybaud, V. Moizan, Y. Joly and C. Chizallet, 108 G. Kramer and G. McVicker, Accounts of Chemical Research,
ChemCatChem, 2019, 11, 3941–3951. 1986, 19, 78–84.
93 J. K. Ali and A. Baiker, Chemical Engineering Communica- 109 G. Kramer, G. McVicker and J. Ziemiak, Journal of Catalysis,
tions, 2018, 1–10. 1985, 92, 355–363.
94 M. Usman, D. Cresswell and A. Garforth, ISRN Chemical En- 110 B. Coq and F. Figueras, Journal of Catalysis, 1984, 85, 197–
gineering, 2013, 2013,. 205.
95 J. Chaouki, A. Touzani, D. Klvana, G. Belanger and 111 Z. Paál and P. Tétényi, Nature, 1977, 267, 234–236.
J. Bournonville, Revue de l’Institut Français du Pétrole, 1988, 112 P. Van Trimpont, G. Marin and G. Froment, Ind. Eng. Chem.
43, 873–881. Res, 1988, 27, 51–57.
96 W. Zhao, C. Chizallet, P. Sautet and P. Raybaud, Journal of 113 Z. Paál, Encyclopedia of Catalysis.
Catalysis, 2019, 370, 118–129. 114 E. Christoffel, F. Fetting and H. Vierrath, Journal of Catalysis,
97 T. Degnan and C. Kennedy, AIChE journal, 1993, 39, 607– 1975, 40, 349–355.
614. 115 D. Farrusseng, C. Klanner, L. Baumes, M. Lengliz,
98 J. Thybaut, C. Laxmi Narasimhan, J. Denayer, G. Baron, C. Mirodatos and F. Schüth, Molecular Informatics, 2005, 24,
P. Jacobs, J. Martens and G. Marin, Industrial & engineering 78–93.
chemistry research, 2005, 44, 5159–5169. 116 G. Morra, D. Farrusseng, C. Bouchy and S. Morin, Oil & Gas
99 P. Weisz, Advances in Catalysis, 1962, 13, 137 – 190. Science and Technology - Revue d’IFP Energies nouvelles, 2013,
100 J. Zecevic, G. Vanbutsele, K. de Jong and J. Martens, Nature, 68, 487–504.
2015, 528, 245–248. 117 S. Horton and M. Klein, Energy & Fuels, 2013, 28, 37–40.
101 J. M. Arroyo, G. Martens, G. Froment, G. Marin, P. Jacobs 118 R. Landau, S. Korre, M. Neurock, M. Klein and R. Quann,
and J. Martens, Applied Catalysis A: General, 2000, 192, 9– Am. Chem. Soc. Div. Fuel Chem. Prepr. Pap, 1992, 37, 1871.
22. 119 S. Rao and M. Coppens, Chemical engineering science, 2012,
102 M. Guisnet, Catalysis Today, 2013, 218-219, 123–134. 83, 66–76.
103 J. Weitkamp, ChemCatChem, 2012, 4, 292–306. 120 J. Szczygieł, Computers & Chemical Engineering, 2011, 35,
104 M. Guisnet, Technique de l’Ingénieur, 2015. 985–998.
105 P. Mendes, A. Gregório, A. Daudin, C. Bouchy, J. Silva and 121 B. D. Vandegehuchte, J. W. Thybaut and G. B. Marin, Indus-
M. Ribeiro, Catalysis Communications, 2017, 89, 152–155. trial & Engineering Chemistry Research, 2014, 53, 15333–
106 E. G. Acebo, C. Leroux, C. Chizallet, Y. Schuurman and 15347.
C. Bouchy, ACS Catalysis, 2018. 122 L. Pirro, P. Mendes, S. Paret, B. Vandegehuchte, G. Marin
107 M. Guisnet and L. Pinard, Catalysis Reviews, 2018, 1–100. and J. Thybaut, Catalysis Science & Technology, 2019.

J
our
nal
Name,
[yea
r][
,vol
.
],
1–19 | 19