DOI: 10.1002/cctc.201801271
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57

Minireviews

Developments in the Atomistic Modelling of Catalytic
Processes for the Production of Platform Chemicals from
Biomass
´
´
´
˜
´
Rodrigo Garcıa-Muelas,[a] Marcos Rellan-Pineiro,[a] Qiang Li,[a] and Nuria Lopez*[a]

ChemCatChem 2019, 11, 357 – 367

Wiley VCH
1901 / 122291

357

 2019 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Donnerstag, 03.01.2019
[S. 357/367]

1

Minireviews
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57

The transformation of biomass-derived molecules into platform
chemicals that can directly be employed by the chemical
industry is one of the challenges in catalysis in this century.
While some processes are cost-effective and industrially available, the molecular insight has been advancing at a slow pace
when compared to other areas (like energy conversion). In the
present review we describe the main challenges imposed on
the theoretical simulations for the study of these complex
molecules and how the most crucial issues are typically

addressed. In particular, we focus on technical aspects like the
need for London dispersion and solvation contributions. We
also deal with the complexity of reaction networks, which
requires new approaches and ways to compile the results in the
form of databases. This allows the study of large reaction
networks for the decomposition of C2 alcohols, or the plethora
of functional groups of cyclic molecules such as lignin and
sugars. Finally, we put forward a few applications that show the
potential of atomistic simulations in the field.

Introduction

have a similar degree of rate control for the overall product
distribution.[8] For these structures, small energy differences
may lead to different product distributions to a large extent[9–10]
and therefore, they should be computed accurately. Fundamental electronic structure questions arise from the previous points.
London dispersion contributions are key to address the
adsorption of large molecules on catalytic surfaces accurately.
The seminal theoretical studies on the reactivity of small
alcohols neglected these contributions as they were lacking in
the computational codes.[11–14] Nowadays, we know that these
contributions are important to describe the adsorption energy
of large molecules, especially when a carbon tail is present.[15,16]
They may change the configuration of the target compounds
as well.[17–19] In addition, biomass-derived molecules are rich in
oxygenated functional groups, which interact with the ubiquitous water molecules.[20] This interaction is crucial to understand the reactivity, thus the solvation effects need to be
included either explicitly,[21] at a very high computational cost,
or implicitly.[18] However, the solvation models, which were
widely used for homogeneous catalysis, were seldom applied
to heterogeneous studies, and thus had to be implemented
considering periodic boundary conditions.[22–24] The addition of
water implies additional challenges to the simulations as some
of the active molecules are acids and new equilibria are
established between the catalyst and the surface. Finally, the
effect of the reaction conditions should be considered so as to
develop suitable models directly comparable to experiments.
The contributions can come from pressure, concentration,
temperature, pH, and electric potential effects,[25] which may
have non-trivial consequences like inducing changes in the
state of the catalyst surface.[26]
In the present review, we aim to describe the challenges in
the theoretical study of biomass conversion into platform
chemicals as well as in the technological challenges that we
and others have faced. In particular, we focus on the
fundamental issues that were preventing the routine use of
computational techniques based on Density Functional Theory
to upgrade biomass-derived compounds. As the molecular size
increases, so does the difficulty to implement refined techniques such as microkinetics or DFT characterization over the full
reaction network, Scheme 1. To foster scientific discussions, all
our DFT results are stored and can be retrieved from the
ioChem-BD repository,[27] an open database that fulfils the FAIR
principles[28] of findability, accessibility, interoperability, and
reusability.

The non-edible fraction of biomass is the most abundant and
environmentally-friendly raw material.[1] It can be transformed
into platform chemicals by heterogeneous catalysts, which are
key processes for the sustainability of the chemical industry.[2] In
2004, the US Department of Energy (DOE) identified twelve
target compounds that can be directly obtained from biomass
and upgraded into high-value chemicals.[3] This fostered
experimental screening for new catalysts and processes. The list
was expanded in 2010.[4] Yet, due to the complexity of the
systems, the insights about these catalytic routes at a molecular
level were lacking. The atomistic simulation of biomass-derived
processes has mainly three sources of complexity: the size of
the molecules, the non-negligible effects of London dispersion
and solvation; and the behavior of the system under operation
conditions. These hurdles have been partially addressed in the
last years, and the field has reached enough maturity to ensure
that reactions can be described accurately for medium sized
molecules.
Firstly, even small molecules exhibit an increasing number
of conformations. For short-chain C1ÀC2 alcohols this effect is
minor, but for larger compounds (> C3) the number of possible
rotations increases exponentially, having an important effect in
their binding to oxide or metal surfaces. At this point, the
chirality and rigidity of the molecules are crucial to understand
the catalytic path.[5] Besides, the reaction network that connects
reactants and products becomes intricate, as the number of
intermediates and transition states increases exponentially,[6]
thus limiting the use of the available techniques. For instance,
glycerol decomposition on metals consists of 250 intermediates
connected by roughly 2000 transition states, while for a typical
C6 the total number is more than two orders of magnitude
larger.[6] For such large molecules, the product distribution
typically shows poor selectivity and this is one of the challenges
of biomass conversion.[7] Moreover, the identification of key
transition states is not straightforward as many of them may

[a] Dr. R. García-Muelas, Dr. M. Rellµn-PiÇeiro, Dr. Q. Li, Dr. N. López
Institute of Chemical Research of Catalonia, ICIQ
The Barcelona Institute of Science and Technology
Av. Països Catalans 16
Tarragona 43007 (Spain)
E-mail: nlopez@iciq.es
This manuscript is part of the Anniversary Issue in celebration of 10 years of
ChemCatChem.

ChemCatChem 2019, 11, 357 – 367

Wiley VCH
1901 / 122291

www.chemcatchem.org

358

 2019 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Donnerstag, 03.01.2019
[S. 358/367]

1

Minireviews
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57

Fundamental Challenges
London Dispersion

Scheme 1. Summary of the work performed in the Bio2Chem-d project. The
studies are divided by the catalyst type (black) and by the molecular size
(blue). The analysis performed for each molecular size and catalyst is shown
in red. Rx stands for reaction and MK for microkinetics. The arrows on the
right side indicate the growth in complexity.

Density Functional Theory (DFT) techniques have been very
successful in addressing the reactivity of small molecules on
heterogeneous catalysts. And yet, when the Bio2Chem-d
project started in 2010, several hurdles were still preventing the
consideration of large molecules as those from biomass.[29] The
earliest studies on metal catalysts did not include van der Waals
interactions.[11,13] Although this is not a problem for small
molecules containing one or two carbon atoms,[30] the vdW
contribution of each additional methyl group is exothermic by
À0.1 eV,[15,31] so this term becomes significant for C3 or larger
molecules. Traditional density functionals did not introduce
London dispersion contributions and thus alternatives were
sought to mitigate this problem. The first semi-empirical
contributions were included in the activation of methane on
Ir.[32] Later on, extensive developments in self-consistent functionals, like that of Lunqvist,[33,34] included the non-locality of
the London dispersion contributions. However, these functionals were too property-dependent and thus their use was
limited. A simpler approach was suggested by Grimme[35,36]
employing the C6 coefficients already used in molecular
mechanics. However, when the D2 set of parameters was
applied to estimate the adsorption on metals, the results were
discouraging as they led to severe overbinding. Simultaneously
to our work, Tkatchenko and coworkers derived the C6
coefficients for atoms and metallic surfaces; however the latter

´
Rodrigo Garcıa-Muelas was born in Valencia
(Venezuela) in 1987. He studied Chemical
Engineering (2003–2008) at the University of
Carabobo, Venezuela, and received his M.S in
Environmental Engineering (2011–2012) from
the Rovira i Virgili University. From 2012–2017
he conducted his PhD research on the reactivity of polyalcohols on metals under the
´
supervision of Prof. N. Lopez at the Institute of
Chemical Research of Catalonia. His current
research focuses in electrochemistry and statistical learning algorithms applied to large
reaction networks.

Qiang Li was born in Shandong (China) in
1988. He is currently a postdoc researcher in
the group of Nuria Lopez at the Institute of
Chemical Research of Catalonia (ICIQ). He
received his B.S. degree in Chemistry and M.S.
degree in Physical Chemistry from Inner
Mongolia University in 2010 and Institute of
Coal Chemistry, Chinese Academy of Science
in 2013. Then, he earned his Ph. D. degree in
Chemical Science and Technology from Institute of Chemical Research of Catalonia (ICIQ)
in 2017. His research focuses on the biomass
conversion mechanisms on metal catalysts,
photocatalysis of water splitting via singleatom catalysts, and catalyst designs by tuning
the metal-support interactions and interfacial
properties.

´
˜
Marcos Rellan-Pineiro was born in Vigo
(Spain) in 1986. He obtained its graduated in
Chemistry in 2009 at Universidade de Vigo
where he also finished the M.S. in Theoretical
Chemistry and Computational Modeling. After
that, he received a one-year grant to work in
the R&D department of the pharmaceutical
company Janssen-Cilag, SA. In 2014 he joined
´
´
to the Prof. Nuria Lopez group at the Institute
of Chemical Research of Catalonia (ICIQ) to
perform his PhD thesis in theoretical heterogeneous catalysis. He obtained a Severo
Ochoa grant and received his PhD at 2018.
Currently he is working in the same group as
postdoctoral researcher.

ChemCatChem 2019, 11, 357 – 367

Wiley VCH
1901 / 122291

www.chemcatchem.org

´
´
Nuria Lopez obtained her PhD at the University of Barcelona, Spain in 1999 and realized
post-doc stage in the group of Prof. Nørskov
at the Technical University in Denmark. In
2005 she joined ICIQ as a group leader. Her
field of expertise is the use of computational
simulations to materials in particular those
employed as a heterogeneous catalyst. She
obtained an ERC-Starting grant in 2010 (Bio2Chem-d project). She is the co-author of about
180 publications and has received about 8000
citations for her work.

359

 2019 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Donnerstag, 03.01.2019
[S. 359/367]

1

Minireviews
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57

required some measurements making it difficult to extend.[37,38]
In our approach, we simplified the search for the coefficients in
the following manner: the C6 terms are related to the polarizability of the surface, thus different electric fields were used to
obtain them.[39] These models can then be employed with
enough robustness together with the values for the atoms
derived in Grimme’s D2. Since then, more advanced schemes
have been reported: (i) the D3 extension by Grimme,[40] (ii) the
non-local correlation functional vdW-DF[33] that can be combined with several functionals as BEEF-vdW,[41] and (iii) Tkatchenko’s full set of DFT-TS,[42] DFT-TS/HI,[43,44] TS + SCS[45] and
MBD@rsSCS.[46] All these terms lead to different adsorption
energies.[47,48] Besides, in some cases the orientation of the
molecule on the surface may change depending on the density
functional chosen, thus resulting in different chemical behaviors.[49] An in-depth assessment of these methods can be found
in a recent review[50] by Grimme et al.
Increasing the molecular complexity has another side effect:
flexible molecules can have multiple conformations and sometimes the most stable gas-phase configuration does not
correspond to the adsorbate ground state. This problem has
seldom been addressed in the literature. Among our first steps,
we develop some rules depending on the relative strengths on
intramolecular and metal-adsorbate interactions.[15] We found
that the conformational changes can have big impacts on the
adsorption of molecules and in their reactivity. For instance,
when glucose adsorbs on Ru, a proton is transferred from a
hydroxyl to a carbonyl group, which favors the breaking of a
CÀH bond vicinal to the donating group.[51] The presence of
vicinal methanol molecules also affects the preferred reaction
path during their decomposition on metals.[21] Thus, the metal
reactivity can only be studied provided that the right configurations are identified.
To find the most stable configuration for adsorption is even
more complicated on metal oxides due to their multiple
adsorption sites with different acid-base and redox character.[52]
The difficulty increases if different metal coordinations or
vacancies appear on the surface. The adsorption of alcohols
and polyalcohols was also studied by us on TiO2 in which
different behaviors were observed for different compounds:
primary and secondary alcohols prefer the adsorption on
vacancy positions, whereas tertiary or poly-alcohols prefer the
Ticus channels.[53]

Molecular Complexity
An additional side effect of larger molecules is that their
associated reaction networks grow exponentially by size and
complexity.[6] Therefore, the study of the reactivity for biomassderived molecules is very computationally demanding. Two
main directions have been taken to reduce this burden: (i) to
use small surrogates, for which the full reaction network can be
sampled; (ii) to search for the preferred reaction path (and
maybe one major parasite reaction). Both approaches have
significant drawbacks. Small molecules might not properly
represent the activity of larger ones since. For instance, they
ChemCatChem 2019, 11, 357 – 367

Wiley VCH
1901 / 122291

www.chemcatchem.org

lack intramolecular hydrogen bonds that stabilize certain
conformations. They also are unable to represent the stiffness
of large aromatic groups that provide rigidity and anchor them
to the catalytic surface, like in lignin. Therefore, surrogates
should be used with caution. Simplifying reaction networks has
a higher risk as the complexity of molecules derived from
biomass cannot be avoided since, in many cases, they contain
multiple functional groups in different positions of the molecule, and the desired transformation would encompass both
chemo and regioselectivities. Sampling one reaction path does
not ensure that the critical selectivity points of the reaction
network are considered. In this sense, the modular approaches
employed to obtain the contributions from the different units
that compose the desired molecule might be effective. Such
additive schemes have been used to describe the gas-phase
thermochemistry with impressive accuracy.[54] For adsorbates,
the existing rules normally span short-sized molecules;[55]
although semiempirical equations for the adsorption of large
molecules have been put forward.[56,57]
In the foreseeable future, the combined use of databases
and statistical learning algorithms will allow generating robust
thermochemical models for large molecules.[58] These databases
should fulfill the FAIR principles, which require that data is
findable, accessible, interoperable, and reusable.[28] In this
context, we have created the ioChem-BD repository,[27] in use
since 2015, to store all calculations coming from our DFT-based
projects.

Microkinetics
Even when the reaction networks are fully characterized in
terms of their thermodynamic and kinetic parameters at the
DFT level, their predictability is limited. As we have explained
above, many functional groups are present in different parts of
the molecule. In many cases, we have found that the barriers of
several parallel paths are comparable, thus they might be active
simultaneously.[77] This means that a successful analysis of the
reaction network requires the use of microkinetic models so
that the molecular insight obtained for all elementary steps can
be mapped to experimental data like reaction orders, activation
barriers, activity of catalyst, and product selectivity.[59] Many
reactivity descriptors and volcano plots are based on microkinetics.[60,61] Still, due to its simplicity, the molecular interpretation based on energy profiles is more common, at least for
simple reaction networks.
Microkinetic models can have different levels of simplification. For instance, many differential equations can be grouped
to represent a single-rate determining step,[62] as the energies
of intermediates and transition states are highly correlated.[9]
This reduction is very convenient to optimize reaction conditions. However, this approach cannot be used to screen
catalytic systems, as the potential catalysts might prefer different reaction paths, e. g., the reforming of ethanol on Pd and
Pt.[77] Another approach is to consider flow reactors to be in
quasi-equilibrium. Despite removing the dependence on time,
this simplification may yield unphysical results as flow reactors

360

 2019 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Donnerstag, 03.01.2019
[S. 360/367]

1

Minireviews
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57

may operate in a pseudostationary state.[63,77] Several simplifications are common during the obtention of the free energies;
therefore, the accuracy can be increased with a correct
description of the entropy terms.[64] Finally, it is common to
assume a constant adsorption energy for each intermediate, by
neglecting coverage effects and lateral interactions. For some
intermediates, like water or carbon monoxide, this simplification may lead to large deviations. For water, cooperative effects
mediated by hydrogen bonds appear at medium concentrations, which in turn vary the relative binding energies of the
different substrates to the active sites. For carbon monoxide,
the repulsion is very high, so the binding energies should be
lower at large coverages. Repulsive effects may be included as
a linear function from the coverage, such as in Ref. [65], at least
for the common poisons CO, C, and O. As these contributions
are repulsive, a refinement of the technique is to include lateral
interactions above a threshold surface concentration, i. e. 1/3
for hexagonal surfaces.[66] Below this value, the adsorbates do
not “feel” each other. In any case, the repulsive interactions
mediated by metal surfaces are inversely proportional to the
cube of the distance,[67] thus next-neighbor interactions are
seldom included.[68] Finally, the full set of lateral interactions can
be considered altogether through Kinetic Monte Carlo (KMC)
simulations. KMC boosts the accuracy of kinetic measurements,
but it is considerably more expensive than microkinetic
modeling. As other groups,[69–73] we have also been active in
reporting KMC codes and applications.[74] Refinements in the
accuracy of DFT and the way lateral interactions are treated will
be the object of future research, where again statistical learning
techniques can improve our models.
For biomass conversion, microkinetics has been typically
applied on the full reaction network of small alcohols, such as
methanol,[75] ethanol,[9] or ethylene glycol.[76] Particularly in our
group we computed the full set of reactions encompassing the
decomposition of all these species on four different metals (Cu,
Ru, Pd, and Pt), comprising 75 intermediates and 250 transition
states for each metal. This allowed us to unveil the activity of
these materials for hydrogen production under four reaction
conditions.[21,77] We focused on the direct decomposition to
compare our results with surface science data, as well as three
emerging technologies: autothermal reforming, steam reforming, and aqueous phase reforming. For such a complex reaction
network, a genomic notation was implemented to allow the
systematic identification of intermediates, as described below.
The 1000 transition states of our reaction database are openly
available. They can be easily extended to include larger
molecules or lateral paths like oxidation reactions via the linear
scaling methods based on Brønsted-Evans-Polanyi (BEP)[78–80]
and Transition State Scalings (TSS).[81–83] In our study, the TSS
were used to estimate the 1697 C3 related transition states of
glycerol decomposition on a Ru surface which are used in the
microkinetic study to predict the most suitable technology for
H2 production. By combining the linear scalings for the
intermediates and the BEP relationships for the transition states,
Greeley et al. investigated the thermochemistry and kinetics in
the free energy diagram of glycerol decomposition on: (i)
Pt(111),[84] (ii) a series close packed transition metal surfaces,[85]
ChemCatChem 2019, 11, 357 – 367

Wiley VCH
1901 / 122291

www.chemcatchem.org

and (iii) on a PtÀMo alloy.[86] The OÀH, CÀH, CÀO and CÀC
competition in different dehydrogenation stages has been well
described. For oxides, we also computed the reaction network
for glycerol hydrodeoxygenation (HDO) to propylene catalyzed
by MoO3[87,88] and analyzed it through a microkinetic model. The
main reaction path, the apparent activation energy, and the
reaction order of H2 were obtained from the microkinetic
analysis.[89]

Solvation
Another important particularity of biomass is the need to
include the effects of water in the simulations. Water is
ubiquitous in raw biomass. It is also generated from bondbreaking reactions that eliminate oxygen in the form of water
from sugars and other biomass compounds. Metal-water
interactions have been extensively investigated in the contexts
of heterogeneous catalysis and electrochemistry, both by
experimental and computational techniques. Detailed reviews
on the state-of-the-art of this extensive topic are available in
Ref. [90] and [91]. The main results concern the wetting/nonwetting character of the layers;[92,93] the stability of different
water arrangements on the surface,[91,93] as distinctive patterns
appear at sub-monolayer coverages depending on the metal;
and the possibility of fitting new potentials that could be used
in electrochemical simulations.[94] There are two ways of
including the effect of solvation on theoretical simulations,
either by explicit water molecules, or by replacing the vacuum
region with an implicit solvation, Scheme 2.

Scheme 2. (a) Explicit solvation model, where the liquid water molecules are
placed in the simulation box.[96] (b) In the implicit model, the vacuum is
replaced by a continuum with electric permittivity e delimited by a smooth
frontier.[19]

Most of the studies with explicit water molecules were
based on low water coverages (1 bilayer)[14,95] or in small cells
where lateral interactions were non-negligible.[13] Extended
dynamics on metal systems have been possible just recently.[96]
Neural network potentials based on DFT may further expand
the complexity of the systems that can be studied at a
reasonable computational cost.[97] Our early simulations on the
dynamics of the metal-water interface identified that, depending on the nature of the metal (reactive or unreactive), the

361

 2019 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Donnerstag, 03.01.2019
[S. 361/367]

1

Minireviews
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57

metal-water interfaces differ. For instance, the wetting layer of
Pd and Pt is denser than the bulk liquid. Besides, the interface
shows low ordering as five- and seven-membered rings appear,
in good agreement with experiments done at lower coverages.[98] On Ru, half of the water molecules dissociate, and form
a very rigid structure containing mostly hexagonal patters.[99,100]
However, there is no explicit pH variation at the interface as
protons and hydroxyls are trapped on the surface.[96] On oxide
surfaces the situation is more complex. Wetting and dewetting
behaviors have been found to correlate with the acid-base
properties of the surface, such as the presence of hydroxyl
groups,[101] and geometric factors. The acid strength promoted
the wettings of the surface, but the geometric match between
the oxygens of the ice structure and the adsorption sites on the
surface were shown to be the most important factor that
controls wettability.[102]
However, water-containing simulations are complex as the
dynamic and cooperative effects with solvent are very
important. In homogeneous catalysis, implicit water solvation
models[103] have been applied extensively. Typically, the best
schemes are the explicit consideration of one or two water
molecules surrounded by a continuum of constant polarizability. Including more than two explicit waters induces
additional problems related to the configurational entropy (that
is blocked) whereas not including them implies considering
very specific and directional bonds.[104] However, continuum
models are only recently being implemented in periodic
boundary conditions codes.[22–24] The solution to the Poisson
equation can be done in different ways, and in our particular
case we used a multigrid algorithm to solve the underlying
Poison equation,[23] an approach that was later used by other
groups.[24] Our tests demonstrated that solvation values are in
good agreement with the molecular implementations. Besides,
the same scheme can be used to investigate transition states.[19]
However, many developments are needed in this area so that
such approaches become the standard in the field.

Complex Electronic Structure
In parallel, the study of the reactivity of biomass on oxide
surfaces has many more fundamental problems arising from
the complexity of the electronic structure of these materials[105]
but also for the much wider variability in the types of
elementary steps that they allow. For instance, ceria and
molybdenum oxides have been described as active in the
transformation of methanol (a byproduct of several biomass
conversion processes)[106] into formaldehyde.[107–114] For CeO2,
the structure sensitivity drives the selectivity: the closed-packed
(111) and (110) surfaces produce formaldehyde, which on the
(100) surface is further oxidized to CO.[115] For molybdenum
oxides, the influence of Mo oxidation state and the degree of
surface reduction were analyzed. It was shown that only the
fully oxidized MoO3 is selective for formaldehyde production. In
addition, the role of Fe as dopant, which was experimentally
found to increase the catalytic activity,[116] was unraveled. Iron
atoms act as electron buffer in redox steps decreasing the
ChemCatChem 2019, 11, 357 – 367

Wiley VCH
1901 / 122291

www.chemcatchem.org

energy barriers thus accelerating the process.[117] On the other
hand, for many metal oxides, fundamental questions such as
the way hydrogen is activated on the surface were only partially
understood.[118,119] As several reactions involving biomass are
dedicated to reduce the oxygen content via hydrodeoxygenation (HDO) processes,[120] which are carried out with the help of
hydrogen,[121–124] the question of how H2 interacts with the
surface remains crucial to assess the catalytic capabilities of the
surface. On CeO2(111) we observed that H2 is activated
heterolytically; thus, at the transition state, a proton and a
hydride are formed,[125] but due to the surface reducibility the
whole reaction ends up in the formation of two surface
hydroxyls and the reduction of two cations. The use of the
same surface in methanol conversion is the perfect example
that demonstrates the limitations of standard DFT when
investigating reducible oxides. The adsorption of hydrogen and
formation of a surface vacancy by water releasing was studied
for MoO3.[89] Acid-base reactions are easily described by
standard GGA methods. The main issue is that reactions that
reduce/oxidize the surface require the use of GGA + U approaches[126,127] where the U stands for a local fix of the electron
self-interaction that appears for strongly correlated materials.[128]
This problem has been reported in the literature for years but
there are further implications. The GGA + U approximation
works for static problems (i. e. electronic structure) but when
reduction (oxidation) processes take place, the best performing
U value for the total reaction energy is not the best performing
for the kinetic evaluation (transition state).[129] This is due to the
different electron localization and imposes severe constrains
when addressing the reactivity on oxides in complex reaction
networks. While acid-base reactions are stable against the U
term, the redox ones need to be benchmarked.[126–131] In
addition, we have observed that the electronic structure can be
very dynamic for reducible oxides. For oxides with more than
one consecutive oxidation state, defects like oxide vacancies
can provide two electronic structures that are dynamically
interconnected, and thus both can be active parts of the
catalytic cycle.[126] All these fundamental problems are also likely
to emerge in the context of energy production[132,133] and
storage[134,135] as many of them are based on small energy
differences between different oxidation and spin states that can
be easily masked by the inaccuracy and deficiencies that we
have described above.
As electronic structure complexity increases the use of
simple energy descriptors (like the adsorption energies of key
species employed in metals) is no longer an option and more
chemically adapted terms can be more suitable. We developed
a new series of descriptors which are not based on adsorption
energies and can be directly mapped to experimental observations. They encompass geometric, redox and acid-base terms,
leaving out second order contributions such as phase cooperation, multifunctionality of active sites, and site isolation.[136] Our
results show that elementary steps on different surfaces and in
doped materials can be traced back to these parameters,
particularly the redox terms, thus establishing a robust framework for future studies.[137,138]

362

 2019 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Donnerstag, 03.01.2019
[S. 362/367]

1

Minireviews
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57

Case Studies
In the following we present three case studies that have been
at the core of the research:

Decomposition of C1ÀC2 Alcohols on Metals
The decomposition of methanol, ethanol and ethylene glycol
was studied on different metals to analyze how the external
conditions affected the activity and selectivity of the catalysts.
Initially we searched for the most stable adsorption configurations extracting a new set of rules for adsorption on the
pristine surfaces.[15] In parallel, the adsorption of water and its
interaction with coadsorbed alcohols was investigated.[17] In
both cases, the London dispersion contributions were found to
be necessary to reproduce experiments. Taking these structures,
we considered all the possible decomposition pathways arising
from CÀH, OÀH, CÀC, and CÀO breakings, as well as the watergas shift reaction. The resulting reaction network contains 75
intermediates and 250 transition states.[21,77] Due to its complexity, a genomic notation was developed to label the intermediates and easily account for the different decomposition routes
of C1 and C2 alcohols on the metals under study. For example,
CH3OH!CH2OH!CH2O!CHO!CO was denoted as: 1411!
1312!1211!1111!1011. In each notation, from left to right,
the first three digits stand for the number of C, H, and O atoms
in the given species. The last one is used to label possible
structural isomers, which are typically 1–2 for C1, 1–5 for C2, and
1–9 for C3. These genomic strings are well suited for automation
procedures of C1ÀC3 molecules and were key to develop fast
robust Density Functional Theory inputs. However, for larger
molecules, other frameworks such as SMILES should be
considered.[139] Similarly, all the transformations between fragments can be encoded. These provide a complete matrix of
reactivity that can be downloaded as a csv file from the
ioChem-BD database[27] to be further processed and then new
elementary steps can be added if required. Since they are
stored according to the FAIR[28] database principles, the data
can be employed by other researchers in the field of Catalysis
to reduce the computational and human burden of calculating
such large networks. Besides, the database is openly available
to be further analyzed by anyone through statistical-learning
algorithms.
The database was then interrogated by a microkinetic
model, to describe the catalyst behavior for H2 production from
alcohols under four different reaction conditions. These include
steam reforming along two emerging alternatives, namely the
aqueous-phase reforming and autothermal reforming.[77] In all
cases, the reaction network for the decomposition plus the
water-gas shift reaction are sufficient to describe the reactivity.
However, in some cases, a few additional reactions were
included to address particular issues such as CO elimination by
oxidation and water assisted processes. Then, we were able to
predict the activity of Ru for the reforming of glycerol by using
linear scaling relationships (LSR). We have also found that LSR
hold for solvent environments, as shown in Figure 1.[18,77] This
ChemCatChem 2019, 11, 357 – 367

Wiley VCH
1901 / 122291

www.chemcatchem.org

Figure 1. Initial and final states can be used to predict the energy of the
transition states of O–H and C–H breakings, respectively. The corresponding
scaling lines were deducted from data taken from Ref. [21] and [77]. Note
that, these scalings are preserved for the reactions taking place in solvated
environments, [19].

means that once solvation effects are considered for the
intermediates, their effect on the activation energies can be
predicted. Recently, a thermochemical model based on group
additivity quantified the solvation effects.[94] Moreover, the
scaling relationships drawn for surrogates can be transferred to
higher alcohols provided that a few rules are preserved. Firstly,
the adsorbates should not change their orientation due to large
anchoring effects. Secondly, the molecule should have enough
flexibility so that the reaction centers can interact with the
surface without hindrance. Conjugation and aromaticity within
the adsorbate molecule should be considered apart, vide infra.
Finally, the surface-mediated interaction between functional
groups should be mild. These conditions allow the formulation
of a compact model for large molecules.

Epimerization and NanoselectTM Ru
Biomass-derived molecules can also have a large impact on the
food industry, where molecules as polysaccharides can be
converted into high-value products like sweeteners and other
food additives.[140] Nature generates sugars that differ from
interesting precursors by the orientation of only one chiral
CHOH center. The reaction that inverts the stereochemistry of a
single stereocenter is called epimerization.[141] Mo-based catalyst, either in molecular or solid forms, were reported to
catalyze the C2 epimerization of aldoses through a 1,2 Cshift.[142–144] We investigated the reaction path and the ability of
different Mo-materials to carry out the reaction[145] and found
that the same computational scheme can be used for both
types of catalysts, linking their properties and showing that a
unified theory in catalysis could be developed irrespectively the
nature of the catalyst. The solvent contributions were obtained
for the key step by applying implicit VASP-MGCM[23] code and
by adding explicit molecules, thus helping into the formulations. The most relevant catalytic properties are related to the
ability of Mo centers to cycle between different oxidation
states, which can be modulated by the lattice, or the scaffold in
molecular catalysts. Thus, reducibility, calculated as hydrogen

363

 2019 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Donnerstag, 03.01.2019
[S. 363/367]

1

Minireviews
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57

addition energy (HAE),[146,147] was found as the single descriptor
of the activity, which shows a volcano-shape dependence on
reducibility, see Figure 2. Finally, a better catalyst able to
perform the reaction four times faster was identified.

high activity and higher robustness. The reaction network
leading from LA to GVL is a complex one and encompasses
intramolecular ring esterification and hydrogenation steps. A
domino reaction was earlier proposed in which the full LA was
firstly protonated to RÀC(OH)2 + and then followed by the ring
formation and water elimination.[153] The final product, GVL,
would be produced from the hydrogenation of the unsaturated
ring intermediate. Our mechanistic investigation[154] described
that to achieve full conversions into the final products the
domino reactions are the CÀO bond scissions from the COOÀ or
COOH group on Ru surface as shown in Figure 3. In addition,

Figure 3. Mechanism of levulinic acid conversion to g-valerolactone from
DFT calculations.

Figure 2. a) Schematic representation of the glucose C2 epimerization to
mannose though a C-shift. b) Normalized reaction rate as a function of the
hydrogen addition energy, HAE. Blue dots mark the rate calculated for the
experimental catalyst.

The epimerized sugar feedstock can then be hydrogenated
to produce the desired sugar alcohols. In this case, we studied
the sugar (mannose and glucose) hydrogenations on the Ru
catalyst.[51] Each of them has different isomers with linear and
cyclic structures (a and b).[148] The concentration and adsorption
energies of these isomers in the solution and on the surface,
combined with the different rate-limiting steps (the water
assisted ring opening and CÀH bond formation) in hydrogenation network, determine the differential hydrogenation
rates. The kinetic model based on the Langmuir-Hinshelwood
mechanism[149] was used by considering the factors above. The
reaction order of mannose and glucose hydrogenations from
this model agrees well with experimental results.

Conversion of Acids and Lignin Derivatives
Increasing the complexity of the reactants leads to several
issues. For instance, levulinic acid (LA) and g-valerolactone
(GVL) were labeled in the target list by DOE. LA can be derived
from the Levulinic acid hydration process of 5-hydroxymethylfuran (HMF) from cellulose and hemicellulose.[150–152] And GVL
can be obtained from LA conversion. Several indications
pointed out to the need for a catalyst with acidic properties to
increase the rate of the catalysts. Experimentally, it was found
that a new class of materials based on surfactant-decorated
metals, in this case Ru, were able to perform the reaction with a
ChemCatChem 2019, 11, 357 – 367

Wiley VCH
1901 / 122291

www.chemcatchem.org

our simulations showed how for the ligand-decorated Ru
nanoparticles the reaction could be promoted by the interface
high acidity, which can also be used for interpreting the similar
behavior of graphene when it is used as the support for metal
catalysts in LA conversions.[155,156] As we have mentioned before,
acidity promotes a route where the reaction is speeded up
about four times because it tips the balance in the starting
point for the reaction networks. The complexity of both the
catalyst and the reaction mechanism, as well as the of pH
effects in decorated nanoparticles discussion for the first time
present a new perspective that highlights the need to study
decorated nanoparticles by tailoring their interfaces rather that
the separated units (metal and ligand) independently. Recently,
we have been able to use a similar approach to investigate the
decarbonylation transformations of acids to alkenes on
phosphine-decorated Pd catalysts.[157] Monodentate phosphines
cover the surface and lead to the poor activity whereas
bidentate ligands would break the ice by creating the transient
cavities via their inherent dynamics, thus increasing the activity.
Besides, the presence of the ligands also prevents the surface
poisoning derived from the strong binding molecule, C, in a
similar behavior as that of the ligand that prevents Ru oxidation
by reducing oxygen adsorption.
The ultimate frontier in reactant complexity comes from
lignin.[158] Lignin is a widely available biomass resource accounting for 15–30 % of weight and up to 40 % of the energy of
lignocellulose.[159,160] Its rigidity and strength originate from the
variable composition of polymer structures formed via CÀC and
aryl-ether linkages[161] and thus model systems need to be used
to address the most fundamental questions like bond breaking
mechanisms.[162] In particular, for theoretical research on the
most common linkage b-O-4, dissociation mechanisms have
been proposed by using simple surrogate models such as 2-

364

 2019 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Donnerstag, 03.01.2019
[S. 364/367]

1

Minireviews
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57

phenoxyethanol on Pt(111)[163] and 2-phenoxy-1-phenylethanol[164] on Pd(111). We have studied the role of chirality,
conjugation and rigidity for dimers reacting on Ni and doped
Ni. We have found that the surrogate approximation that has
been so successfully applied for small molecules fails here due
to the sergeants and soldiers principle.[165] Besides, rigidity
inhibits the reactions on the CÀC and CÀH bond breakings.
Conjugation and chirality can be potentially used to design
catalysts for the b-O-4 bond selective breakings. In addition,
single-atom catalysis (in this case as dopants in metals) shows a
preeminent ability to selectively cleave and depolymerize such
complex compounds.

Acknowledgements
We would like to thank all the previous members of the group
that were funded through the Bio2chem-d ERC project Dr. P.
´
Błonski, Dr. G. Carchini, Dr. N. Almora-Barrios, Dr. L. Bellarosa, Dr.
´
´
G. Revilla-Lopez, Dr. M. Garcıa-Melchor, Dr. M. Garcia-Rates, Dr. S.
Pogodin, Dr. M. Capdevila-Cortada, and BSC-RES for providing
´
generous computational resources. We thank Prof. J. Perez´
Ramırez for fruitful discussions on the project.

Conflict of Interest
The authors declare no conflict of interest.

Conclusions
During the last years many research groups have been trying to
study by theoretical means the transformation of large
biomass-derived molecules, mainly alcohols and polyalcohols,
into molecules that can serve as chemical platforms. In this
context we have contributed to solve fundamental problems
by: (i) estimating new approximated parameters for the London
dispersion contributions, (ii) building an implicit solvation
model for periodic systems, (iii) identifying new linear scaling
relationships on metals, (iv) determining conformational contributions for large molecules, and (v) automatizing the search
of transition states. Metals and oxides were among the
materials employed as catalysts. The largest challenges were
related to: (i) the complexity of the electronic structures, like
the dynamic oxidation states found for molybdenum oxides; (ii)
the appearance of alternative non-conventional reaction paths
such as the heterolytic cleavage of hydrogen on ceria; and (iii)
the need for descriptors that address both redox and acid-base
chemistry. In many cases, microkinetics was needed to understand the behavior of large reaction networks, such the ones
arising from the decomposition of alcohols on metals. However,
the comprehensive treatment of lateral interactions is still
difficult and Kinetic Monte Carlo codes are not yet fit for this
purpose. Despite the advances made so far, there are plenty of
challenges and opportunities ahead. The intensive search for
novel materials that can serve as catalysts will require their
detailed electronic structure characterization. Biomass conversion is an area where the emerging techniques of statistical
learning could reduce the computational burden and speed-up
the discovery of new materials. However, sufficiently large and
open data is needed to get reliable results from such
algorithms. Therefore, the openness of the work performed by
our group is one of the most encouraging aspects as newer
models can be derived from our data. We hope that our studies
may be helpful for the future modeling of catalytic systems
with higher complexity. These models, combined with experimental data, shall provide valuable insights to keep building a
sustainable chemical industry.

ChemCatChem 2019, 11, 357 – 367

Wiley VCH
1901 / 122291

www.chemcatchem.org

Keywords: DFT · Biomass conversion · ioChem-BD · van der
Waals · Solvation

[1] R. Rinaldi, F. Schüth, Energy Environ. Sci. 2009, 2, 610–626.
[2] P. Anastas, N. Eghbali, Chem. Soc. Rev. 2010, 39, 301–312.
[3] T. Werpy, G. Petersen, A. Aden, J. Bozell, J. Holladay, J. White, A.
Manheim, D. Eliot, L. Lasure, S. Jones, Top Value Added Chemicals
from Biomass. Volume I - Results of Screening for Potential.
Candidates from Sugars and Synthesis Gas. Department of Energy,
Washington DC, USA. 2004.
[4] J. J. Bozell, G. R. Petersen, Green Chem. 2010, 12, 539–551
[5] W. M. Wan, G. R. Jenness, K. Xiong, D. G. Vlachos, J. G. G. Chen,
ChemCatChem 2017, 9, 1701–1707.
[6] J. E. Sutton, D. G. Vlachos, Chem. Eng. Sci. 2015, 121, 90–199.
[7] L. I. Godina, A. V. Kirilin, A. V. Tokarev, D. Y. Murzin, ACS Catal. 2015, 5,
2989–3005.
[8] C. T. Campbell, J. Catal. 2001, 204, 520–524.
[9] J. E. Sutton, W. Guo, M. A. Katsoulakis, D. G. Vlachos, Nat. Chem. 2016,
8, 331–337.
[10] M. S. Frei. M. Capdevila-Cortada. R. García-Muelas. C. Mondelli. N.
López. J. A. Stewart. D. Curulla-FerrØ. J. PØrez-Ramírez, J. Catal. 2018,
361, 313–321.
[11] S. K. Desai, M. Neurock, K. Kourtakis, J. Phys. Chem. B 2002, 106, 2559–
2568.
[12] J. Greeley, M. Mavrikakis, J. Am. Chem. Soc. 2004, 126, 3910–3919.
[13] D. Cao, G. Q. Lu, A. Wieckowski, S. A. Wasileski, M. Neurock, J. Phys.
Chem. B 2005, 109, 11622–11633.
ˇ
[14] J. Carrasco, B. Santra, J. Klimes, A. Michaelides, Phys. Rev. Lett. 2011,
106, 026101.
[15] B. A. Sexton, K. D. Rendulic, A. E. Huges, Surf. Sci. 1982, 121, 181–198.
[16] R. García-Muelas, N. López, J. Phys. Chem. C, 2014, 118, 17531–17537.
´
[17] P. Błonski, N. López, J. Phys. Chem. C 2012, 116, 15484–15492.
[18] J. B. A. Davis, F. Baletto, R. L. Johnston, J. Phys. Chem. A 2015, 119,
9703–9709.
[19] M. Garcia-RatØs, R. García-Muelas, N. López, J. Phys. Chem. C 2017, 121,
13803–13809.
[20] C. O. Tuck, E. PØrez, I. T. Horvµth, R. A. Sheldon, M. Poliakof, Science
2012, 337, 695–699.
[21] R. García-Muelas, Q. Li, N. López, ACS Catal. 2015, 5, 1027–1036.
[22] K. Mathew, R. Sundararaman, K. Letchworth-Weaver, T. A. Arias, R. G.
Hennig, J. Chem. Phys. 2014, 140, 084106.
[23] M. Garcia-RatØs, N. López, J. Chem. Theory Comput. 2016, 12, 1331–
1341.
[24] J. C. Womack, L. Anton, J. Dziedzic, P. J. Hasnip, M. IJ Probert, C.-K.
Skylaris, J. Chem. Theory Comput. 2018, 14, 1412–1432.
[25] J. K. Nørskov, J. Rossmeisl, A. Logadóttir, L. Lindqvist, J. R. Kitchin, T.
Bligaard, H. Jónsson, J. Phys. Chem. B 2004, 108, 17886–17892.
[26] G. M. Lari, R. García-Muelas, C. Mondelli, N. López, J. PØrez-Ramírez,
Green Chem. 2016, 18, 4682–4692.
[27] M. lvarez-Moreno, C. De Graaf, N. López, F. Maseras, J. M. Poblet, C.
Bo, J. Chem. Inf. Model. 2015, 55, 95–103,

365

 2019 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Donnerstag, 03.01.2019
[S. 365/367]

1

Minireviews
[28] M. D. Wilkinson, M. Dumontier, IJ. J. Aalbersberg, G. Appleton, M.
Axton, A. Baak, N. Blomberg, J. W. Boiten, L. B. da Silva Santos, P. E.
Bourne, et al. Sci. Data, 2016, 3, 160018.
´
[29] N. López, N. Almora-Barrios, G. Carchini, P. Błonski, L. Bellarosa, R.
García-Muelas, G. Novell-Leruth, M. García-Mota, Catal. Sci. Technol.
2012, 2, 2405–2417.
[30] N. Almora-Barrios, S. Pogodin, L. Bellarosa, M. García-Melchor, G.
Revilla-López, M. García-RatØs, A. B. Vµzquez-García, P. HernµndezAriznavarreta, N. López, ChemCatChem, 2015, 7, 928–935.
[31] R. Lin, D. Albani, E. Fako, S. K. Kaiser, O. V. Safonova, N. López, J. PØrezRamírez. Submitted.
[32] G. Henkelman, H. Jónsson, Phys. Rev. Lett. 2001, 86, 664–667.
[33] M. Dion, H. Rydberg, E. Schrçder, D. C. Langreth, B. I. Lundqvist, Phys.
Rev. Lett. 2004, 92, 246401.
[34] K. Lee, E. D. Murray, L. Kong, B. I. Lundqvist, D. C. Langreth, Phys. Rev. B
2010, 82, 081101.
[35] S. Grimme, J. Comb. Chem. 2006, 27, 1787–1799.
[36] S. Grimme, J. Antony, S. Ehrlich, S. Krieg, J. Chem. Phys. 2010, 132,
154104.
[37] A. Tkatchenko, M. Scheffler, Phys. Rev. Lett. 2009, 102, 073005.
[38] A. Tkatchenko, R. A. Di Stasio, R. Car, M. Scheffler, Phys. Rev. Lett. 2012,
108, 236402.
´
[39] N. Almora-Barrios, G. Carchini, P. Błonski, N. López, J. Chem. Theory
Comput. 2014, 10, 5002–5009
[40] S. Grimme, S. Ehrlich, L. Goerigk, J. Comb. Chem. 2011, 32, 1456–1465.
[41] J. Wellendorff, K. T. Lundgaard, A. Møgelhøj, V. Petzold, D. D. Landis,
J. K. Nørskov, T. Bligaard, K. W. Jacobsen, Phys. Rev. B 2012, 85, 235149.
[42] A. Tkatchenko, M. Scheffler, Phys. Rev. Lett. 2009, 102, 073005.
ˇ
[43] T. Bucko, S. Leb›gue, J. Hafner, J. G. ngyµn, J. Chem. Theory Comput.
2013, 9, 4293–4299.
ˇ
[44] T. Bucko, S. Leb›gue, J. G. ngyµn, J. Hafner, J. Chem. Phys. 2014, 141,
034114.
[45] A. Tkatchenko, R. A. Di Stasio, R. Car, M. Scheffler, Phys. Rev. Lett. 2012,
108, 236402.
[46] A. Ambrosetti, A. M. Reilly, R. A. DiStasio, A. Tkatchenko, J. Chem. Phys.
2014, 140, 018A508.
[47] B. Liu, L. Cheng, L. Curtiss, J. Greeley, Surf. Sci. 2014, 622, 51–59.
[48] S. Gautier, S. N. Steinmann, C. Michel C, P. Fleurat-Lessard, P. Sautet,
Phys. Chem. Chem. Phys. 2015, 17, 28921–28930.
[49] R. L. H. Freire, D. Guedes-Sobrinho, A. Kiejna, J. L. F. Da Silva, J. Phys.
Chem. C 2018, 122, 1577–1588.
[50] S. Grimme, A. Hansen, J. G. Brandenburg, C. Bannwarth, Chem. Rev.
2016, 116, 5105–5154.
[51] G. M. Lari, O. G. Grçninger, Q. Li, C. Mondelli, N. López, J. PØrezRamírez, ChemSusChem. 2016, 9, 3407–3418.
[52] J. M. Vohs, Chem. Rev. 2013, 113, 4136–4163.
[53] G. Carchini, N. López, Phys. Chem. Chem. Phys. 2014, 16, 14750–14760.
[54] S. W. Benson. Thermochemical kinetics. Wiley, 1976.
[55] F. Abild-Pedersen, J. Greeley, F. Studt, J. Rossmeisl, T. R. Munter, P. G.
Moses, E. Skulason, T. Bligaard, J. K. Nørskov, Phys. Rev. Lett. 2007, 99,
016105.
[56] M. Salciccioli, D. G. Vlachos, J. Phys. Chem. C 2010, 114, 20155–20166.
[57] M. Salciccioli, S. M. Edie, D. G. Vlachos, J. Phys. Chem. C 2012, 116,
1873–1886.
[58] Z. W. Ulissi, A. J. Medford, T. Bligaard, J. K. Nørskov, Nat. Commun.
2017, 8, 14621.
[59] J. M. Lu, S. Behtash, M. Faheem, A. Heyden, Catal. 2013, 305, 56–66.
[60] A. J. Medford, A. Vojvodic, J. S. Hummelshøj, J. Voss, F. Abild-Pedersen,
F. Studt, T. Bligaard, A. Nilsson, J. K. Nørskov, J. Catal. 2015, 328, 36–42.
[61] J. K. Nørskov, T. Bligaard, J. Rossmeisl, C. H. Christensen, Nat. Chem.
2009, 1, 37–46.
[62] J. W. Thybaut, G. B. Marin, J. Catal. 2013, 308, 352–362.
[63] M. Jørgensen, H. Grçnbeck, ACS Catal. 2016, 6, 6730–6738.
[64] M. Jørgensen, H. Grçnbeck, J. Phys. Chem. C 2017, 121, 7199–7207.
[65] S. Wang, V. Vorotnikov, D. G. Vlachos, ACS Catal. 2014, 5, 104–112.
[66] A. C. Lausche, A. J. Medford, T. S. Khan, Y. Xu, T. Bligaard, F. AbildPedersen, J. K. Nørskov, F. Studt, J. Catal. 2013, 307, 275–282.
[67] W. Kohn, K.-H. Lau, Solid State Commun. 1976, 18, 553–555.
[68] J. Nielsen, M. d’Avezac, J. Hetherington, M. Stamatakis, J. Chem. Phys.
2013, 139, 224706.
[69] D. Mei, M. Neurock, C. M. Smith, J. Catal. 2009, 268, 181–195.
[70] M. Stamatakis, D. G. Vlachos, J. Chem. Phys., 2011, 134, 214115.
[71] H.-J. Chun, V. Apaja, A. Clayborne, K. Honkala, J. Greeley, ACS Catal.
2017, 7, 3869–3882.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57

ChemCatChem 2019, 11, 357 – 367

Wiley VCH
1901 / 122291

www.chemcatchem.org

[72] M. J. Hoffmann, S. Matera, K. Reuter, Comput. Phys. Commun. 2014,
185, 2138–2150.
[73] S. Matera, M. Maestri, A. Cuoci, K. Reuter, ACS Catal. 2014, 4, 4081–
4092.
[74] S. Pogodin, N. López, ACS Catal. 2014, 4, 2328–2332.
[75] S. Kandoi, J. Greeley, M. A. Sanchez-Castillo, S. T. Evans, A. A. Gokhale,
J. A. Dumesic, M. Mavrikakis, Top. Catal. 2006, 37, 17–28.
[76] S. Kandoi, J. Greeley, D. Simonetti, J. Shabaker, J. A. Dumesic, M.
Mavrikakis, J. Phys. Chem. C 2010, 115, 961–971.
[77] Q. Li, R. García-Muelas, N. López, Nat. Commun. 2018, 9, 526.
[78] M. Evans, M. Polanyi, Trans. Faraday Soc. 1938, 34, 11–24.
[79] A. Logadottir, T. H. Rod, J. K. Nørskov, B. Hammer, S. Dahl, C. Jacobsen,
J. Catal. 2001, 197, 229–231.
[80] J. K. Nørskov, T. Bligaard, A. Logadottir, S. Bahn, L. B. Hansen, M.
Bollinger, H. Bengaard, B. Hammer, Z. Sljivancanin, M. Mavrikakis, J.
Catal. 2002, 209, 275–278.
[81] R. Alcalµ, M. Mavrikakis, J. A. Dumesic, J. Catal. 2003, 218, 178–190.
[82] J. Zaffran, C. Michel, F. Auneau, F. Delbecq, P. Sautet, ACS Catal. 2014,
4, 464–468.
[83] J. E. Sutton, D. G. Vlachos, ACS Catal. 2012, 2, 1624–1634.
[84] B. Liu, J. Greeley, J. Phys. Chem. C 2011, 115, 19702–19709.
[85] B. Liu, J. Greeley, Phyc. Chem. Chem. Phys. 2013, 15, 6475–6485.
[86] B. Liu, M. Zhou, M. K. Y. Chan, J. Greeley, ACS Catal. 2015, 5, 4942–
4950.
[87] V. Zacharopoulou, E. S. Vasiliadou, A. A. Lemonidou, Green Chem.
2015, 17, 903–912.
[88] V. Zacharopoulou, E. S. Vasiliadou, A. A. Lemonidou, ChemSusChem
2018, 11, 264–275.
[89] M. Rellµn-PiÇeiro, N. López, Submitted.
[90] A. Hodgson, S. Haq, Surf. Sci. Rep. 2009, 64, 381–451.
[91] J. Carrasco, A. Hodgson, A. Michaelides, Nat. Mater. 2012, 11, 667.
[92] J.-J. Velasco-Velez, C.-H. Wu, T. A. Pascal, L. F. Wan, J. Guo, D.
Prendergast, M. Salmeron Science 2014, 1, 1259437.
[93] A. Michaelides, K. Morgenstern, Nat. Mater. 2007, 6, 597.
[94] G. H. Gu, B. Schweitzer, C. Michel, S. N. Steinmann, P. Sautet, D. G.
Vlachos, J. Phys. Chem. C 2017, 121, 21510–21519.
[95] K. Tonigold, A. Groß, J. Comput. Chem. 2012, 33, 695–701.
[96] L. Bellarosa, R. García-Muelas, G. Revilla-López, N. López, ACS Cent. Sci.
2016, 2, 109–116.
[97] S. K. Natarajan, J. Behler, Phys. Chem. Chem. Phys. 2016, 18, 28704–
28725.
[98] S. Nie, P. J. Feibelman, N. C. Bartelt, K. Thürmer, Phys. Rev. Lett. 2010,
105, 026102.
[99] G. Revilla-López, N. López, Phys. Chem. Chem. Phys. 2014, 16, 18933–
18940.
´
[100] G. Revilla-López, P. Błonski, N. López, J. Phys. Chem. C, 2015, 119, 5478–
5483.
[101] M. M. Gentleman, J. A. Ruud, Langmuir 2009, 26, 1408–1411.
[102] G. Carchini, M. García-Melchor, Z Łodziana, N. López, ACS Appl. Mater.
Interfaces 2016, 8, 152–160.
[103] M. Cossi, V. Barone, R. Cammi, J. Tomasi, Chem. Phys. Lett. 1996, 255,
327–335.
[104] M. Besora, P. Vidossich, A. Lledos, G. Ujaque, F. Maseras, J. Phys. Chem.
A 2018, 122, 1392–1399.
[105] E. Dagotto, Science 2005, 309, 257–262.
[106] V. Phillips, P. Takahashi, Environ. Sci. Technol. 1990, 24, 1136–1137.
[107] D. R. Mullins, M. D. Robbins, J. Zhou, Surf. Sci. 2006, 600, 1547–1558.
[108] J. J. Uhlrich, B. Yang, S. Shaikhutdinov, Top. Catal. 2014, 57, 1229–1235.
[109] A. P. Soares, F. Montemor, M. F. Portela, A. Kiennemann, J. Mol. Catal. A
2015, 397, 93–98.
[110] T. Kropp, J. Paier, J. Phys. Chem. C, 2014, 118, 23690–23700.
ˇ
[111] V. Matolín, J. Libra, M. Skoda, N. Tsud, K. C. Prince, T. Skµla, Surf. Sci.
2009, 603, 1087–1092.
[112] M. Bowker, A. F. Carley, M. House, Catal. Lett. 2008, 120, 34–39.
[113] C. Brookes, M. Bowker, P. P. Wells, Catalysts 2016, 6, 92.
[114] T. Choksi, J. Greeley, ACS Catal. 2016, 6, 7260–7277.
[115] M. Capdevila-Cortada, M. García-Melchor, N. López, J. Catal. 2015, 327,
58–64.
[116] C. Brookes, P. P. Wells, N. Dimitratos, W. Jones, E. K. Gibson, D. J.
Morgan, G. Cibin, C. Nicklin, D. Mora-Fonz, D. O. Scanlon, C. R. A.
Catlow, M. Bowker, J. Phys. Chem. C 2014, 118, 26155–62161.
[117] M. Rellµn-PiÇeiro, N. López, ChemSusChem 2015, 8, 2231–2239.
[118] H.-Y. T. Chen, L. Giordano, G. Pacchioni, J. Phys. Chem. C 2013, 117,
10623–10629.

366

 2019 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Donnerstag, 03.01.2019
[S. 366/367]

1

Minireviews
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57

[119] H.-T. Chen, Y. M. Choi, M. Liu, M. C. Lin, ChemPhysChem 2007, 8, 849–
855.
[120] E. Furimsky, Appl. Catal. A 2000, 199, 147–190.
[121] T. Prasomsri, T. Nimmanwudipong, Y. Romµn-Leshkov , Energy Environ.
Sci. 2013, 6, 1732–1738.
[122] T. Prasomsri, M. Shetty, K. Murugappan, Y. Roman-Leshkov, Energy
Environ. Sci. 2014, 7, 2660–2669.
[123] V. Zacharopoulou, E. S. Vasiliadou, A. A. Lemonidou, Green Chem.
2015, 17, 903–912.
[124] V. Zacharopoulou, E. S. Vasiliadou, A. A. Lemonidou, ChemSusChem
2018, 11, 264–275.
[125] M. García-Melchor, N. López , J. Phys. Chem. C 2014, 118, 10921–10926.
[126] M. Rellµn-PiÇeiro, N López, J. Phys. Chem. Lett. 2018, 9, 2568–2573.
[127] M. Capdevila-Cortada, N. López, ACS Catal. 2015, 5, 6473–6480.
[128] S. L. Dudarev, G. A. Botton, S. Y. Savrasov, C. J. Humphreys, A. P. Sutton,
Phys. Rev. B 1998, 57, 1505.
[129] M. Capdevila-Cortada, Z. Łodziana, N. López, ACS Catal. 2016, 6, 8370–
8379.
[130] S. Lutfalla, V. Shapovalov, A. T. Bell, J. Chem. Theory Comput. 2011, 7,
2218–2223.
[131] A. B. Getsoian, A. T. Bell, J. Phys. Chem. C 2013, 117, 25562–25578.
[132] M. Vasilopoulou, A. M. Douvas, D. G. Georgiadou, L. C. Palilis, S.
Kennou, L. Sygellou, A. Soultati, Ioannis Kosti, G. Papadimitropoulos,
D. Davazoglou, P. Argitis, J. Am. Chem. Soc. 2012, 134, 16178.
[133] X. Yu, T. J. Marks, A. Facchetti, Nat. Mater. 2016, 15, 383–396
[134] H. S. Kim, J. B. Cook, H. Lin, J. S. Ko, S. H. Tolbert, V. Ozolins, B. Dunn,
Nat. Mater. 2017, 16, 454–460.
[135] X. Deng, Z. Wei, C. Cui, Q. Liu, C. Wang, J. Ma, J. Mater. Chem. A 2018,
6, 4013–4022.
[136] W. T. Hong, M. Risch, K. A. Stoerzinger, A. Grimaud, J. Suntivich, Y.
Shao-Horn, Energy Environ. Sci. 2015, 8, 1404–1427.
[137] M. Capdevila-Cortada, N. López, ACS Catal. 2015, 5, 6473–6480.
[138] M. Capdevila-Cortada, G. VilØ, D. Teschner, J. PØrez-Ramírez, N. López,
Appl. Catal. B 2016, 197, 299–312.
[139] D. Weininger, J. Chem. Inf. Comput. Sci. 1988, 28, 31–36.
[140] P. Mäki-Arvela, B. Holmbom, T. Salmi, D. Y. Murzin, Catal. Rev. Sci. Eng.
2007, 49, 197.
[141] I. Delidovich, R. Palkovits, ChemSusChem 2016, 9, 547–561.
[142] M. L. Hayes, N. J. Pennings, A. S. Serianni, R. Barker, J. Am. Chem. Soc.
1982, 104, 6764–6769.
[143] F. Ju, D. VanderVelde, E. Nikolla, ACS Catal. 2014, 4, 1358–1364.
[144] M. Orazov, M. E. Davis, Proc. Mont. Acad. Sci. 2015, 112, 11777–11782.

ChemCatChem 2019, 11, 357 – 367

Wiley VCH
1901 / 122291

www.chemcatchem.org

[145] M. Rellµn-PiÇeiro, M. Garcia-RatØs, N. López, Green Chem. 2017, 19,
5932–5939.
[146] P. Deshlahra, E. Iglesia, J. Phys. Chem. C 2014, 118, 26115–26129.
[147] P. Deshlahra, R. T. Carr, S. H. Chai, E. Iglesia, ACS Catal. 2015, 5, 666–
682.
[148] Y. Zhu, J. Zajicek, A. S. Serianni, J. Org. Chem. 2001, 66, 6244–6251.
[149] I. Chorkendorff, J. W. Niemantsverdriet, 2005, Concepts of modern
catalysis, kinetics. John Wiley & Sons.
[150] D. W. Rackemann, W. O. Doherty, Biofuels Bioprod. Biorefin. 2011, 5,
198–214.
[151] C. O. Tuck, E. PØrez, I. T. Horvµth, R. A. Sheldon, M. Poliakoff, Science
2012, 337, 695–699.
[152] R. J. van Putten, J. C. van der Waal, E. de Jong, C. B. Rasrendra, H. J.
Heeres, J. G. de Vries, Chem. Rev. 2013, 113, 1499–1597.
[153] K. P. C. Vollhardt, N. E. Schore, Organic Chemistry: Structure and
function, Macmillan, 2003.
[154] D. Albani, Q. Li, G. VilØ, S. Mitchell, N. Almora-Barrios, P. T. Witte, N.
López, J. PØrez-Ramírez, Green Chem. 2017, 19, 2361–2370.
[155] C. Xiao, T. W. Goh, Z. Qi, S. Goes, K. Brashler, C. Perez, W. Y. Huang, ACS
Catal. 2016, 6, 593–599.
[156] J. Tan, J. Cui, T. Deng, X. Cui, G. Ding, Y. Zhu, Y. Li, ChemCatChem 2015,
7, 508–512.
[157] M. A. OrtuÇo, N. López, ACS Catal. 2018, 8, 6138–6145.
[158] Q. Li, N. López, ACS Catal. 2018, 8, 4230–4240.
[159] C. Li, X. Zhao, A. Wang, G. W. Huber, T. Zhang, Chem. Rev. 2015, 115,
11559–11624.
[160] I. Delidovich, P. J. Hausoul, L. Deng, R. Pfützenreuter, M. Rose, R.
Palkovits, Chem. Rev. 2015, 116, 1540–1599.
[161] F. G. Calvo-Flores, J. A. Dobado, ChemSusChem 2010, 3, 1227–1235.
[162] R. Behling, S. Valange, G. Chatel, Green Chem. 2016, 18, 1839–1854.
[163] C. A. O. Hamou, R. RØocreux, P. Sautet, C. Michel, J. B. Giorgi, J. Phys.
Chem. C 2017, 121, 9889–9900.
[164] J. Lu, M. Wang, X. Zhang, A. Heyden, F. Wang, ACS Catal. 2016, 6,
5589–5598.
[165] M. M. Green, M. P. Reidy, R. D. Johnson, G. Darling, D. J. O’Leary, G.
Willson, J. Am. Chem. Soc. 1989, 111, 6452–6454

Manuscript received: April 5, 2018
Accepted Article published: September 25, 2018
Version of record online: October 30, 2018

367

 2019 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Donnerstag, 03.01.2019
[S. 367/367]

1