Precursor Nuclearity and Ligand Effects in Atomically-Dispersed Heterogeneous Iron Catalysts for Alkyne Semi-Hydrogenation

Article abstract of DOI:10.1002/cctc.202100235

Nanostructuring earth-abundant metals as single atoms or clusters of controlled size on suitable carriers opens new routes to develop high-performing heterogeneous catalysts, but resolving speciation trends remains challenging. Here, we investigate the potential of low-nuclearity iron catalysts in the continuous liquid-phase semi-hydrogenation of various alkynes. The activity depends on multiple factors, including the nuclearity and ligand sphere of the metal precursor and their evolution upon interaction with the mesoporous graphitic carbon nitride scaffold. Density functional theory predicts the favorable adsorption of the metal precursors on the scaffold without altering the nuclearity and preserving some ligands. Contrary to previous observations for palladium catalysts, single atoms of iron exhibit higher activity than larger clusters. Atomistic simulations suggest a central role of residual carbonyl species in permitting low-energy paths over these isolated metal centers.

Full text of DOI:10.1002/cctc.202100235

The European Society Journal for Catalysis
Supported by
Accepted Article
Title: Precursor Nuclearity and Ligand Effects in Atomically-Dispersed
Heterogeneous Iron Catalysts for Alkyne Semi-Hydrogenation
Authors: Dario Faust Akl, Andrea Ruiz-Ferrando, Edvin Fako, Roland
Hauert, Olga Safonova, Sharon Mitchell, Núria López, and
Javier Pérez-Ramírez
This manuscript has been accepted after peer review and appears as an
Accepted Article online prior to editing, proofing, and formal publication
of the final Version of Record (VoR). This work is currently citable by
using the Digital Object Identifier (DOI) given below. The VoR will be
published online in Early View as soon as possible and may be different
to this Accepted Article as a result of editing. Readers should obtain
the VoR from the journal website shown below when it is published
to ensure accuracy of information. The authors are responsible for the
content of this Accepted Article.
To be cited as: ChemCatChem 10.1002/cctc.202100235
Link to VoR: https://doi.org/10.1002/cctc.202100235
01/2020

10.1002/cctc.202100235
ChemCatChem
FULL PAPER
Precursor Nuclearity and Ligand Effects in Atomically-
Dispersed Heterogeneous Iron Catalysts for Alkyne Semi-
Hydrogenation
Dario Faust Akl,[a] Andrea Ruiz-Ferrando,[b] Edvin Fako,[b] Roland Hauert,[c] Olga Safonova,[d]
Sharon Mitchell,*[a] Núria López*[b] and Javier Pérez-Ramírez*[a]
[a]
D. Faust Akl, Dr. S. Mitchell, Prof. Dr. Pérez-Ramírez
Institute of Chemical and Bioengineering
Department of Chemistry and Applied Biosciences, ETH Zurich
Vladimir-Prelog-Weg 1, 8093 Zurich (Switzerland)
E-mail: msharon@chem.ethz.ch, jpr@chem.ethz.ch
A. Ruiz-Ferrando, E. Fako, Prof. N. López
Institute of Chemical Research of Catalonia (ICIQ)
The Barcelona Institute of Science and Technology
Av. Països Catalans 16, 43007 Tarragona (Spain)
E-mail: nlopez@iciq.es
[b]
[c]
[d]
Dr. R. Hauert
Empa–Swiss Federal Laboratories for Materials Science and Technology
Überlandstrasse 129, 8600 Dübendorf (Switzerland)
Dr. O. V. Safonova
Paul Scherrer Institute
Forschungsstrasse 111, 5232 Villigen (Switzerland)
Supporting information for this article is given via a link at the end of the document.
Abstract Nanostructuring earth-abundant metals as single atoms or
clusters of controlled size on suitable carriers opens new routes to
develop high-performing heterogeneous catalysts, but resolving
speciation trends remains challenging. Here, we investigate the
potential of low-nuclearity iron catalysts in the continuous liquid-
phase semi-hydrogenation of various alkynes. The activity depends
on multiple factors, including the nuclearity and ligand sphere of the
metal precursor and their evolution upon interaction with the
mesoporous graphitic carbon nitride scaffold. Density functional
theory predicts the favorable adsorption of the metal precursors on
the scaffold without altering the nuclearity and preserving some
ligands. Contrary to previous observations for palladium catalysts,
single atoms of iron exhibit higher activity than larger clusters.
Atomistic simulations suggest a central role of residual carbonyl
species in permitting low-energy paths over these isolated metal
centers.
The remarkable selectivity of homogeneous iron hydrogenation
catalysts[8] prompted the investigation of heterogeneous
counterparts. Colloidal metallic nanoparticles prepared by
reducing iron salts with organolithium or Grignard reagents or
thermal decomposition of labile complexes could efficiently
catalyze the full hydrogenation of a broad range of alkynes and
olefins to the respective alkanes.[9] In situ X-ray absorption
spectroscopy and deconvolution of light-element scattering
contributions in the extended X-ray absorption fine structure
(EXAFS) ) evidenced the difficulty to retain surface reduced
phases of iron by identifying a metallic iron core and surface-
coordinated solvent molecules.[9c] Following the same trends,
nanoparticle surface passivation by exposure to oxygen was
found detrimental to activity,[10] but protected the metallic core
from further oxidation by water or air,[11] as did embedding the
nanoparticles in hydrophobic polymers or ionic liquids.[12] Indeed,
in a comparative study of different silica-supported iron species
for phenylacetylene semi-hydrogenation, activity decreased in
the order Fe0>Fe2+>Fe3+ while preserving similar selectivity.[13]
Interestingly, purely iron oxide-based systems were found to
enhance the reduction of nitroarenes to anilines.[14] Comparison
of distinct unsupported iron oxides showed that the formation of
oxygen vacancies promotes hydrogen activation at adjacent
metal centers.[15] For the same transformation, interfacial FeNx
species originating from high-temperature annealing of iron
oxide composites with nitrogen-doped carbon correlated with the
catalytic activity.[16] The creation of planar metal oxide
ensembles on reducible supports such as zirconia or titania
yielded high selectivity in acetylene semi-hydrogenation. Isotopic
marker experiments showed that neighboring hydroxyl groups
on the support could adsorb the alkyne and facilitate heterolytic
splitting of hydrogen over the iron moieties.[17] The study of
similar catalyst architectures with uniform speciation revealed
that Fe3+ is the active cation.[18] Single-atom catalysts (SACs)
attract attention for catalytic hydrogenations due to the potential
to improve precious metal utilization.[19] Nevertheless, higher
barriers for hydrogen activation and stronger substrate
Introduction
Fine chemicals production relies on palladium-based catalysts
for the semi-hydrogenation of alkyne building blocks.[1] The poor
utilization of the precious metal and the presence of hazardous
lead in commercially applied Lindlar catalysts drives the search
for alternatives such as ultra-dilute palladium-copper alloys,[2]
selective palladium sulfide phases,[3] or supported nickel-copper
alloys.[4] Iron complexes are known to be active homogeneous
catalysts in various selective hydrogenation reactions. Chirik and
coworkers demonstrated that tridentate pyridinediimine
complexes could achieve turnover frequencies comparable to
precious metal catalysts in reducing substituted olefins.[5]
Parallel developments focusing on tridentate phosphinoborate
complexes highlighted the need for polydentate ligands to
stabilize unsaturated iron intermediates;[6] which is a common
strategy in homogeneous catalysis using earth-abundant
metals.[7] Both approaches involved the use of dinitrogen- or
hydride-based pre-catalysts.
1
This article is protected by copyright. All rights reserved.

10.1002/cctc.202100235
ChemCatChem
FULL PAPER
Figure 1. a) AC-ADF-STEM images of Fex/MCN catalysts. Insets depict the molecular structure of the employed iron precursors. b) k2-weighted Fourier transform
of the experimental (solid) and simulated (dotted) EXAFS spectra of Fex/MCN, and Fe, FeO, and Fe2O3 reference materials. Gray boxes highlight the ranges,
where Fe-N/O and Fe-Fe contributions are expected. c) Gibbs energy profiles (T = 573 K) describing the successive removal of CO and cyclopentadienyl (Cp)
ligands of the deposited iron complexes in the preparation of Fex/MCN catalysts with respect to an empty MCN cavity and the isolated precursor (reference state,
gray dotted line). Ligand-free iron ensembles are modeled to occupy sites on the surface (s), the interlayer (i), or under the surface (u) of the MCN cavity.
Selected calculated structures are depicted as insets. Color code for all panels: Fe-orange; N-blue; C-gray; O-red; H-white.
adsorption render Pd, Rh, or Ir SACs less active than catalysts
integrating larger metal ensembles.[20] On the other hand, earth-
abundant metals exhibit intrinsically different speciation trends
due to their propensity to stabilize non-metallic phases.[21]
Therefore, it remains to be seen if similar nuclearity features will
govern activity trends. The main challenge in elucidating
nuclearity effects is the limited atomic-level control of most
shows that the iron species are atomically-dispersed in all cases
while molecular-level simulations evidence favorable adsorption
of the iron precursors with partial retention of the ligand sphere.
The catalytic evaluation shows that iron is among the few
reported metals, where single atoms yield higher metal-specific
rates than larger metal species in part due to the different
degree of removal of the ligands.
synthesis techniques.
A surface organometallic chemistry
approach to prepare silica-grafted dimeric complexes revealed a
superior activity for the complete hydrogenation of olefins
compared to monomers. However, bulky mesitylene ligands
complexed to the metal sites could have altered the 3D local
environment significantly.[22]
Aiming to gain further insight into iron-based alkyne semi-
hydrogenation catalysts, we explore the behavior of supported
clusters (atoms, dimers, and trimers) on a mesoporous form of
graphitic carbon nitride (MCN). The deposition follows a versatile
strategy, based on the use of commercially available iron
precursors.[23] Characterization by AC-ADF-STEM and EXAFS
Results and Discussion
Graphitic carbon nitride possesses abundant anchoring sites
that stabilize single metal atoms of different elements.
Modifications of the carrier morphology to increase the specific
surface area can enhance metal dispersion and accessibility.
Here,
a
mesoporous form of graphitic carbon nitride
(mesoporous carbon nitride MCN), synthesized via a hard-
templating approach, was used as a carrier. Elemental analysis
2
This article is protected by copyright. All rights reserved.

10.1002/cctc.202100235
ChemCatChem
FULL PAPER
mass contrast between iron and carbon nitride and the irregular
morphology of the carrier add to known challenges of accurate
cluster visualization.[23b]
A
reference sample containing
supported Fe nanoparticles (denoted FeNP/MCN) was obtained
by annealing the Fe-containing carrier in. The particles are
visualized in the form of larger aggregates (Figure 1a), which
are clearly identified by energy-dispersive X-ray spectroscopy
(EDX, Figure S4d). EDX further reveals the elemental uniformity
of the materials and the homogenous distribution of iron across
the carrier (Figure S4).
Analysis of the Fourier transformed EXAFS signal provides
insight into the local environment of the supported metal species
(Figure 1b). The Fe-N(O) scattering paths (1.5 Å) comprise the
major contribution in all materials, coinciding with the consistent
presence of isolated iron centers in the nitrogen-containing MCN
cavity. The similarity between oxidic references (FeO, Fe2O3)
and the FeNP/MCN sample indicate the prevalence of oxidic
metal nanoparticles. A metallic Fe-Fe contribution (2.1 Å) in this
material is visible as a small shoulder, but is absent in the
Fex/MCN catalysts.
Density functional theory (DFT) simulations were conducted to
provide further insight into the evolution of the supported iron
precursors in the Fex/MCN catalysts. Starting from the mono-
and multinuclear complexes (Fe(CO)5, (C5H5)2Fe2(CO)4, and
Fe3(CO)12) and an empty MCN cavity and gas-phase CO as
reference states, the binding Gibbs free energies of successive
ligand removal steps at the annealing temperature (573 K) were
obtained (Figure 1c).
The results reveal appreciable thermodynamic penalties for the
removal of carbonyl groups, suggesting that in Fe1/MCN the iron
atom is likely stabilized in the form of Fe(CO)3. Changes in the
coordination environment of Fe atoms due to the loss of two CO
ligands strengthen their interaction with the scaffold, leading to a
core-level shift of 0.75 eV compared to the physisorbed complex
(Table S1). A similar scenario is found for the Fe2/MCN catalyst,
where upon cyclopentadienyl ligand removal, the resting
complex is Fe2(CO)3. For the trimeric complex, the removal of all
the CO ligands presents a high energetic cost and is unlikely to
proceed to completion (>3 eV). In effect, partially CO-ligated
intermediates such as Fe3(CO)5 with a core-level shift of ~1 eV
with respect to Fe-i are likely stabilized. Similar observations
were previously made for systems based on clusters of Pt and
Rh.[26]
Thermogravimetric analysis of the iron precursors indicates
virtually complete removal of the ligands at the annealing
temperature (573 K), as expected from literature values.[27] Only
for the dimeric complex, a second decomposition feature above
that temperature corresponding to ca. 5 wt% associated with
CO2 and H2O formation is observed. Although complete
decomposition of the pentacarbonyl precursor is anticipated,
Figure 2. a) Fe K-edge X-ray absorption spectra of the Fex/MCN catalysts and
Fe, FeO, and Fe2O3 references. b) Fe 2p3/2 photoelectron spectra of the
Fex/MCN catalysts. Raw data (open circles) and individual components
(shaded areas) indicating the assignments of distinct Fe species.
of the resulting material (C3N5.8O0.8H0.02) reveals slight deviations
from the expected composition of graphitic carbon nitride (C3N4),
with an excess of nitrogen stemming from unpolymerized nitrile
groups. Oxygen and hydrogen, present in minor amounts as OH
or NHx functionalities, are partly formed during polymerization or
de-templating steps.[24] Powder X-ray diffraction analysis (XRD)
reveals characteristic reflections from long-range stacking of
graphitic sheets and reduced crystallinity compared to bulk
forms of graphitic carbon nitride (Figure S1a).[25] Nitrogen
isotherms (Figure S1), secondary electron microscopy (SEM),
and transmission electron microscopy (TEM, Figure S2) confirm
the (meso)porous nature of the MCN (Figure S1). Further
descriptors of porosity resemble results of previously reported
MCN,[24] with
a significantly higher specific surface area
(210 m2 g−1) originating from the ball-milling procedure to mix the
precursors preceding thermal polymerization. Assessment of the
N 1s X-ray photoelectron spectra (XPS) evidences uniform
speciation across all materials. Fitted contributions, assigned to
tertiary and ring nitrogen functionalities, are characteristic of the
tris-triazine units of the carrier (Figure S3). The O 1s spectra
contain features corresponding to C=O and C-O groups and
adsorbed water (H2Oads), with total surface oxygen contents
ranging between 1 and 4 wt.% (Figure S3).
Based on a procedure reporting the deposition of iron dimers on
graphitic carbon nitride,[23] mono- (iron pentacarbonyl), di-
(cyclopentadienyl iron dicarbonyl dimer), and trinuclear (triiron
do-decacarbonyl) iron complexes with similar ligands were
supported on MCN in a DMF solution and subjected to thermal
treatment under flowing nitrogen at 573 K. The obtained
materials are named Fex/MCN, where x indicates the iron
precursor nuclearity. Small variations in metal content
(±0.2 wt.%) compared to the target value (0.5 wt.%) arise
because of the specific deposition procedure (Table 1).
Aberration-corrected annular dark-field scanning transmission
electron microscopy (AC-ADF-STEM) confirms the absence of
aggregates over the Fex/MCN catalysts (Figure 1a). However,
the nuclearity of the iron species derived from different
precursors cannot be distinguished from the images. The low
Table 1. Sample notation, iron precursor applied, bulk and surface metal
contents, and percentage of Fe3+
.
[a]
[b]
wFe,bulk
wFe,surface
[wt.%]
Fe3+[c]
[%]
Catalyst
Fe precursor
[wt.%]
Fe1/MCN
Fe2/MCN
Fe3/MCN
FeNP/MCN
Fe(CO)5
0.3
0.16
0.43
0.40
0.47
90
62
64
44
(C5H5)2Fe2(CO)4
Fe3(CO)12
0.5
0.4
Fe(NO3)3·9H2O
0.7
[a] From ICP-OES. [b] From XPS. [c] Estimated from XPS.
3
This article is protected by copyright. All rights reserved.

10.1002/cctc.202100235
ChemCatChem
FULL PAPER
atom catalyst (0.1 wt.%) with respect to bulk contents and is
therefore in agreement with DFT results. Samples from cluster
precursors, and with nanoparticles, in turn, show smaller
differences between bulk and surface iron concentration (0.4–
0.47 wt.%) (Table 1).
The diversity of iron species over the corresponding Fex/MCN
catalysts (Figure 1c) can be rationalized through the relative
populations (Boltzmann distribution, Table S2) over the
considered intermediates formed through the successive ligand
removal steps. This representation of each of the states at the
thermodynamic equilibrium depends on the temperature and the
Gibbs energy of each intermediate (Table S1). Removed ligands
were transferred to gas-phase reservoirs and conformational
contributions of the adsorbed organometallic were not taken into
account (Computational Methods).
Apart from the most stable tricarbonyl species, the Fe1/MCN is
composed of largely Fe1(CO)2/MCN (25.9 %) and Fe-i (15.6 %).
In the dimeric instance, a significant share of the metal occupies
interlayer
positions
(Fe-i,
16.4 %),
followed
by
Fe2(Cp)(CO)3/MCN (14.9 %) as the third most abundant state.
The surface of the Fe3/MCN catalyst is populated by species
bound with 5, 4 and 3 carbonyl ligands in decreasing order,
displaying the direct impact of the energetic penalty associated
with ligand abstraction (Table S2).
EXAFS spectra for each structure in the formation of the
Fex/MCN catalyst were weighted with the Boltzmann
populations, yielding representative spectra (Figure 1b).
Consistent with experiments revealing the cationic nature of Fe,
predominant Fe-O(N) (ca. 1.7 Å) and Fe-O-Fe (2.7 Å) scattering
features are observed. However, small shifts with respect to
experiments become noticeable at upper radial distances. The
precise determination of the coordination sphere through the
analysis of EXAFS remains problematic because C, N, and O
have similar scattering paths.
Analysis of the X-ray absorption near-edge structure (XANES)
corroborates the strong interaction leading to a positive charge
of the iron (Figure 2a). The whiteline positions close to both
oxide reference materials (FeO, Fe2O3), are consistent with the
cationic (Fe2+, Fe3+) nature of iron for the Fex/MCN catalysts.
Next to the Fe3+ reference (7123.5 eV), Fe1/MCN possesses the
highest content of Fe3+ and therefore a higher averaged
oxidation state than the supported dimers and trimers. Their
whiteline is found at lower energies (Figure 2a, inset). The
surface iron species identified by the spectral fitting of the Fe3p3/2
XPS, are exclusively cationic (Fe2+ and Fe3+), in agreement with
the XAS results (Figure 2b).[29] Threefold positively charged
species make out a share of 90 % of the total peak area in
Fe1/MCN, which decreases to around 60 % for Fe2/MCN and
Fe3/MCN (Table 1). The average oxidation state shifts to a
certain extent with precursor nuclearity, which is in agreement
with XANES and the magnitude of core-level shifts simulated
with DFT (Table S1). Notable differences in signal intensity can
be partly ascribed to variations in metal content.
Figure 3. a) Rate of 2-methyl-3-buten-2-ol formation as a function of pressure
(T = 323 K) in the flow semi-hydrogenation of MBY. The corresponding rates
based on the surface iron concentration are as shown in the gray box
(P = 20 bar). b) The rate of alkene formation during the flow
semihydrogenation of MBY (P = 8 bar and T = 323 K) over the single-atom
catalyst Fe1/MCN remains stable over 10 h on stream. c) Rates of alkene
formation in the hydrogenation of various substituted alkynes (P = 8 bar and
T = 323 K) over the Fex/MCN catalysts. The legend colors apply to all panels.
minor features in the expected infrared region of diffuse-
reflectance infrared spectroscopy (DRIFTS) of the as-prepared
Fe1/MCN catalyst point towards a small fraction of residual CO
(Figure S5).The large background absorption assigned to
unpolymerized surface groups of the MCN carrier, however
makes accurate quantification challenging.
The stabilizing effect of the Fe-O bond on Cp and CO removal
was exemplarily investigated by including oxygen defects in the
scaffold models of the dimeric species. The surface oxygen
abundance evidenced by XPS was represented by replacing a N
vacancy in the MCN cavity with a carbonyl functionality. For
comparison, Gibbs free energies of the dimeric iron complex
were reevaluated on the modified anchoring site, revealing an
overall enhanced stabilization through the Fe-O bond
(Figure S6). Despite this observation, the general shape of the
energy profile of the supported iron dimer is essentially retained.
Dissociation into bare Fe-i seems thermodynamically favored,
lowering the stability of Fe2-s as active ensemble. Therefore, the
presence of oxygen does not appear to impact the formation of
potentially active free coordination sites in such systems. The
computational results indicate that the complete removal of
ligands leading to bare Fe ensembles as active surface species,
either in the form of single atom, dimers, or trimers (Fe-s, Fe2-s,
Fe3-s, insets c, g, j Figure 1c) is unlikely. Even in instances,
where the barriers for surface ensemble formation are low
(Fe1/MCN and Fe2/MCN), this configuration is furthermore
compromised in terms of stability, due to the thermodynamic
tendency for the metal to percolate to inactive interlayer
positions (Fe-i, inset d Figure 1c).[28] The surface iron content
determined through XPS reveals metal depletion over the single-
All catalysts were first evaluated in the semi-hydrogenation of
the vitamin building block 2-methyl-3-butyn-2-ol (MBY)[30] in flow.
The reaction rate increases with the temperature (Figure S7)
and pressure (Figure 3a), but the higher activity does not
significantly
alter
the
selectivity
to
the
desired
2-methyl-3-buten-2-ol product (>88 % at 48 % conversion). The
main side reaction observed is the over-hydrogenation to
2-methyl-3-butan-2-ol. Increasing the substrate or hydrogen flow
rates have the expected effects, resulting in an overall
4
This article is protected by copyright. All rights reserved.

10.1002/cctc.202100235
ChemCatChem
FULL PAPER
distribution remains uniform (Figure S8). Exposure to
hydrogenation conditions affects the metal oxidation state only
slightly, resulting in a minor reduction in the share of Fe3+ in XPS
(Figure 4), which could relate to changes in the ligand sphere
(vide infra).
For the Fe1/MCN and Fe3/MCN catalysts, DFT simulations of
acetylene semi-hydrogenation at reaction temperature
(T = 323 K), provide further molecular-level insights into the
observed activity and selectivity patterns (Figure 5). Since the
structures of reaction intermediates of supported iron dimers and
trimers are alike and the presence of residual ligands impacts
both similarly, the Fe2/MCN catalyst was omitted from the
following analysis. The reaction profiles for the different reaction
networks (Tables S3 and S4) considered both (i) the potential
coexistence of different coordination environments in the
clusters and (ii) the distinct configurations of adsorbed H2 and
C2H2. Starting from CO (gas), H2 (gas), C2H2 (gas) and the bare
ensembles as the reference states, catalysts with and without
CO ligands were studied under semi-hydrogenation conditions.
Fe(CO)3/MCN and Fe3(CO)3/MCN catalysts were chosen as
active ensembles as a compromise between (i) the stability of
the ensemble (which determines their relative abundance) and
(ii) its potential catalytic activity (in terms of availability of empty
coordination sites). The latter is of particular importance in the
case of the trimer, since the active site needs to accommodate
C2H2 and H2 by forming bridged configurations, which are
notably hindered by the presence of CO ligands. Despite
Fe3(CO)5 being the most representative structure of Fe3/MCN in
terms of stability, Fe3(CO)3 is expected to exhibit better catalytic
performance due to improved substrate adsorption.
Consequently, this system is chosen as the active ensemble
despite its relatively low expected concentration when compared
to the system of choice in Fe1/MCN. The reaction starts with the
coordination of H2 to the Fe ensemble to give the dihydrogen
complex (H2*), which potentially dissociates (2H*) if the
accommodation of the H atoms is spatially allowed. Two
potential hydrogen dissociation paths are conceivable: (i)
heterolytic dissociation, where the MCN lattice acts as a
reservoir of protons in a process controlled by electrostatic
interactions, and (ii) homolytic dissociation, in which both
hydrides adsorb on the same Fe site with a bond angle of 90º.[32]
In Fe1/MCN, H2 dissociates heterolytically, forming a tertiary NH
center in the scaffold and a hydride on the Fe atom, which
results in a filled coordination sphere of the metal center. As
previously proposed, such heterolytic H2 splitting could occur on
other N-doped carbon systems promoted by the basicity of the
cavity[33] and particularly in other single-atom systems on
MCN.[23b] Acetylene adsorption on this complex partially
compensates for the energetic cost of the required CO cleavage
Figure 4. a) Fe 2p3/2 photoelectron spectra of the used Fex/MCN catalysts
(circles) and individual components (shaded areas) indicating the assignments
of distinct Fe species. b) Share of Fe3+ in fresh and used Fex/MCN catalysts.
improvement of the activity (Figure S7). In both cases, the
catalysts retain a high alkene selectivity. The Fe2/MCN catalysts
alkene formation rate improves at high substrate concentrations
when compared to the rate at standard conditions (ca. 25 %
improvement). However, in absolute terms, the single atom
catalyst persistently outperforms the other studied materials
across all substrate concentrations.
23b]
Contrary to palladium and most other precious metals,[20c,
Fe1/MCN consistently outperforms the catalysts derived from
dimeric and trimeric complexes throughout the pressure range of
3 to 20 bar in terms of metal-specific rate (Figure 3a). When
considering the effective surface metal loading, the single-atom
catalyst displays a three times higher reaction rate. Therefore,
iron is one of few metals,[31] for which reducing the nuclearity to
1
enhances reactivity in the selective hydrogenation of
acetylenic bonds. Evaluation of the single-atom catalyst stability
for 10 h on stream in MBY hydrogenation confirmed its stable
performance Figure 3b). In the transformation of other related
alkynes, Fe1/MCN again displays superior activity. Here, the
detrimental formation of interlayer species from supported iron
dimers becomes apparent. Effects of steric hindrance become
visible, in converting the less accessible triple bond in 3-hexyne
compared to 1-hexyne, resulting in reduced activity. Comparing
branched substrates, replacing the methyl substituents in MBY
by phenyl groups significantly hinders conversion, while the ethyl
group in 1-methyl-3-pentyn-1-ol only has a marginal impact on
reactivity. Notably, 4-pentyn-1-ol remains unconverted on the
low nuclearity materials. As reported in a previous study of Au-
SACs,[27] the superior reactivity of some alkynols compared to
the aliphatic alkynes can result from a directing effect of the
hydroxyl group. Consistent with the stable performance in flow
semi-hydrogenation, no aggregate formation is visible in ADF-
STEM images of the used catalysts and the elemental
(~0.5 eV at experimental conditions), leading to
a
hexacoordinated structure of the Fe ensemble (C2H2*). The main
product then forms by a two-step process consisting of the
insertion of the hydride coming from the Fe coordination sphere
in the C2H2 molecule (C2H3*), and the subsequent incorporation
of the H from the scaffold on the C2H3 motif (C2H4*). This step is
favorable due to the high stability of the C2H4 moiety (Table S3).
In Fe3/MCN, the presence of three Fe atoms coordinated to the
heptazinic N hinders the transfer of H atoms to the host. As
observed for trimeric ensembles of palladium on MCN,[23b] H2
adsorption by single electron transfer (homolytic dissociation) is
favorable over Fe trimers. However, the formation of two
hydrides in the presence of three carbonyl ligands would imply
5
This article is protected by copyright. All rights reserved.

10.1002/cctc.202100235
ChemCatChem
FULL PAPER
Figure 5. Gibbs energy profiles (T = 323 K) of the semi-hydrogenation of acetylene on a) CO-containing and b) CO-free iron single atoms (Fe1/MCN) and trimers
(Fe3/MCN). TS denotes transition state. The labels on top of the penultimate intermediate in a) correspond to the number of CO ligands bound to the iron site. c)
Structural models of key intermediates on CO-containing and naked iron catalysts corresponding to the states in a) and b). Color code for all panels: Fe-brown; N-
blue; C-gray; O-red; H-white.
their energetically unfavorable (>1 eV) rearrangement into a
terminal configuration. Thus, acetylene directly adsorbs on the
molecular hydrogen complex, and the formation of a bridged
reaction barriers for the complete hydrogenation are lowered,
diminishing the chemoselectivity (Figure S8).
Overall, the observed speciation and catalytic performance
patterns are akin to CO-containing reaction paths. Importantly,
carbonyl ligands ensure the stability of active configurations of
iron single atoms by preventing the percolation of iron into the
host scaffold, reflected in the respective Boltzmann populations.
Concurrently, the thermodynamic penalty associated with
carbonyl removal and the dynamic configurations of remaining
ligands limit the reactant adsorption on the metal trimers, further
reducing the share of active sites. Consequently, iron trimers
exhibit low activity despite presenting reduced activation barriers
for the semi-hydrogenation of acetylene. In the dimeric instance,
the energetically favored formation of interlayer species results
in inhibited activity.
Although ligand-free clusters are predicted to be unstable on
graphitic carbon nitride, the investigation of ultra-small
ensembles of iron on less-interacting carriers could be
potentially promising,[18] as the present analysis of reaction paths
shows. Furthermore, the beneficial effect of CO on the single
atom catalyst could be of practical relevance for impure
species with C2H2 coordinated to
2 Fe atoms partially
compensates the energetic cost of releasing a CO ligand
(~1.5 eV), which is much higher than for Fe1/MCN. The alkene is
formed more favorably than over the single-atom system.
Although the reaction path over ligand-free systems is virtually
the same, significant barriers to form the bare ensembles from
thermodynamically favored configurations (Fe(CO)3 and
Fe3(CO)5) have to be overcome (~2.5 eV for Fe1/MCN and
~5 eV for Fe3/MCN at 323 K). The highly activated C2H4*
moieties on Fe-s and Fe3-s potentially could potentially indicate
poor chemoselectivity of ligand-free configurations. Investigation
of the paths for the full hydrogenation of acetylene to ethane
shows that the desorption of the alkene is preferred over the
complete hydrogenation in the CO-containing systems, because
of the associated penalties of incorporating H into C2H4* and
C2H5*. The slightly favored stabilization of C2H5* on trimeric
ensembles (Fe3–s and Fe3(CO)2) compared to single-atom
species is caused by the bridging configuration. Hence, the
6
This article is protected by copyright. All rights reserved.

10.1002/cctc.202100235
ChemCatChem
FULL PAPER
further purification. In a typical procedure, 500 mg of MCN was dispersed
in 10 cm3 of N,N-dimethylformamide (99.5 %, Sigma Aldrich) and
sonicated for 10 min. The metal precursor was dissolved in 15 cm3 of
solvent and slowly added to the MCN dispersion targeting an iron loading
of 0.5 wt.%. The samples are denoted Fex/MCN with the precursor
nuclearity indicated as a subscript. After stirring overnight, the resulting
materials were filtered, washed with ethanol, and dried at 338 K. Finally,
the dry powders were thermally annealed at 573 K (ramp 2.3 K min−1) for
industrial hydrogen feeds and thus alleviate poor CO tolerance
of benchmark palladium catalysts.[34]
The catalytic performance of the Fex/MCN catalysts in the alkyne
semi-hydrogenation reaction is the complex outcome depending
on multiple factors including the scaffold nature, the metal center
and nuclearity, and its immediate coordination environment,
which relies on the presence of ligands, thus putting non-noble
single-atom reactivity even closer to that of homogeneous
organometallic catalysts.
5 h in nitrogen flow (20 cm3 min−1
Fex/MCN, where x = 1-3 reflecting the nuclearity of the iron precursor.
Following the procedure, reference sample containing oxide
) resulting in samples denoted
a
nanoparticles (denominated FeNP/MCN) was prepared using iron nitrate
nonahydrate (99 %, Sigma Aldrich) as the iron source, deionized water
as the solvent, and changing the atmosphere to static air.
Conclusion
Catalyst Characterization
The downsizing of iron to low-nuclearity species was
investigated as an avenue to nanostructuring earth-abundant
metal catalysts for the selective hydrogenation of alkynes. The
controlled wet deposition of mono, di, or trimeric carbonyl-based
complexes led to highly dispersed metal species with distinct
oxidation states in the Fex/MCN catalysts. Although it was not
possible to directly confirm, DFT simulations supported the
preserved nuclearity of metal species following stabilization of
the complexes on the host scaffold. In a rare example, the
single-atom catalyst outperformed larger metal ensembles in the
liquid-phase semi-hydrogenation of various substituted alkynes.
Despite the favorable hydrogen activation path over metal
trimers, analysis of Boltzmann-distributed populations revealed a
higher proportion of unsaturated metal centers in materials
containing single atoms than dimers and trimers. This is
because of the energetic cost of ligand abstraction to expose the
bare metal was significantly higher for the trimeric ensembles. In
the case of iron dimers redispersion into single atoms in the
interlayer of the carbon nitride scaffold was preferred. This result
highlights the importance of analyzing multiple factors when
describing the performance of non-noble atomically-disperse
catalysts, where ligand, metal center(s), and scaffold are
brought together with behavior that is closer to that of
homogeneous catalysts. The findings further highlight the need
for improved characterization techniques able to discriminate
catalytically active environments in detail and consideration of
the atomic-scale evolution in the design of low-nuclearity metal
catalysts.
Elemental analysis of C, H, N, and O of the carbon nitride supports was
performed in a LECO CHN-900 combustion furnace with an infrared
spectrometer. Powder X-ray diffraction (XRD) was performed in
a
PANalytical X’Pert PRO-MPD instrument. The diffractometer employs a
Bragg-Brentano geometry and Ni-filtered Cu K radiation. Data were
collected over a 2 range of 3-60° with a step size of 0.08° and a
counting time of 163 s per step. The surface area was determined via N2
sorption at 77 K in a Micromeritics TriStar instrument. The samples were
degassed at 423 K for around 6 h prior to measurement. The total
surface area was determined using the Brunauer–Emmett–Teller (BET)
method. X-ray photoelectron spectroscopy (XPS) was carried out in a
Physical Electronics Instruments Quantum 2000 instrument with
monochromatic Al K radiation. The beam was operated at 15 kV and
32.3 W. Spectral acquisition occurred under ultrahigh vacuum conditions
with a pass energy of 46.95 eV. All spectra were calibrated by the N 1s
signal of ring nitrogen at 398.7 eV and fitted with mixed Gaussian-
Lorentzian components after Shirley background subtraction. Main and
satellite peaks corresponding to Fe2+ and Fe3+ were narrowly constrained
to 710.2 ±0.2 eV (Fe3+), 708.3 ±0.2 eV (Fe2+), and 714.5 ±1.25 eV (Fe2+
satellite). Additional components (Fe3+
,
three, and Fe2+, two) were
35]
constrained in dependence of the main peak.[29,
The share of Fe3+
-
components from the total fitted area is used as a measure of the
average oxidation state. Thermogravimetric analysis was carried out in a
Linseis STA PT 1600 instrument using an alumina crucible. The sample
(ca. 20 mg) was dried (373 K, 1 h) in an Ar atmosphere (300 cm3 min−1
)
and heated to 1273 K (ramp 5 K min−1). On-line effluent analysis of likely
gaseous products (H2O, CO, and CO2) was conducted using a mass
spectrometer (MS, Pfeiffer Vacuum ThermoStar GSD T1). For mass loss
calculations, removal of all non-iron components of the precursors is
assumed.
Diffuse-reflectance
infrared
spectroscopy
(DRIFTS)
measurements were performed in a Harricks DRIFT cell and a Bruker
Optics Vertex 70 spectrometer with a liquid nitrogen-cooled mercury
cadmium telluride (MCT) detector. The catalyst was loaded and pressed
to obtain
a flat surface and degassed at 423 K under Ar flow
Experimental Section
(30 cm3 min−1) for 1 h. Finally, the spectrum was acquired at 303 K. KBr
was used for measurement of the background signal. The metal content
determination was performed via inductively coupled plasma optical
Catalyst Synthesis
emission spectroscopy (ICP-OES) in
(photomultiplier tube detector). The solid (approximately 5 mg) was
loaded in polytetrafluoroethylene tube containing an HNO3:H2O2
mixture (3:1, 2 cm3). Subsequently, the tubes were heated to 533 K via
microwave irradiation under autogenous pressure for 20 min. Following
digestion, the obtained clear solutions were diluted to 25 cm3 with
Millipore water and filtered before analysis. High-angle annular dark field-
scanning transmission electron microscopy (ADF-STEM) and energy-
dispersive X-ray spectroscopy (EDX) elemental maps were measured in
a Horiba Ultra 2 instrument
Mesoporous graphitic carbon nitride (MCN) was prepared using a hard-
templating approach reported elsewhere.[24] Cyanamide (20 g, 99 %,
Sigma Aldrich) was dissolved in colloidal silica (50 g, 40 wt% in water,
Ludox HS-40, Sigma Aldrich). After evaporation of the water under
stirring at 363 K, the white powder was subjected to ball-milling for
15 min (500 rpm, 3 min resting time between intervals). The ground
powder was transferred to crucibles, covered, and heated to 873 K (ramp
2.3 K min −1) for 4 h in a static nitrogen atmosphere. The silica template
was removed by treatment in a stirred solution of ammonium hydrogen
fluoride (4 M, 40 cm3 gSilica−1, Alfa Aesar) for 48 h. The resulting MCN
power was collected by filtration, washed with water followed by ethanol,
and dried at 338 K. Iron pentacarbonyl (99.99 %, Sigma Aldrich),
cyclopentadienyliron dicarbonyl dimer (99 %, Fluka), and triiron
dodecacarbonyl (99 %, Acros) were used as iron precursors without
a
a
Talos F200X instrument with a FEI SuperX detector at 200 kV
acceleration potential. The catalysts were dusted on standard copper or
nickel mesh holey carbon films (EMresolutions). Aberration-corrected
(AC)-ADF-STEM measurements were performed on an HD2700-CS
Hitachi microscope with beam diameters of 0.1 and 0.2 nm equipped with
secondary-electron and energy dispersive X-ray detectors. Image frame
7
This article is protected by copyright. All rights reserved.

10.1002/cctc.202100235
ChemCatChem
FULL PAPER
times ranged from 20 to 40 s (1024×1024 pixel). X-ray absorption
spectroscopy (XAS) was conducted at the X10DA (SuperXAS) beamline
of the Swiss Light Source. The X-ray beam from the 2.9 T superbend
was collimated using a Si-coated mirror, monochromatized using a
Si(111) channel-cut monochromator, and focused to a spot size of
500×100 m (horizontal×vertical) using a Rh-coated toroidal mirror. Data
were acquired from pressed pellets at the Fe Kedge in transmission
mode, using three 15 cm long Ar/N2-filled ionization chambers. The
samples were placed between the first and the second ionization
chamber. For the absolute energy calibration, an Fe foil was measured
simultaneously between the second and a third ionization chambers. The
resulting spectra were energy-calibrated, background-corrected, and
normalized using the Athena program from the Demeter software suite.
Simulated EXAFS spectra were generated using FEFF8[42] and post-
processed with Athena.[43] All inputs, outputs, and structures can be
accessed through the ioChem-BD[4344] repository using the following link:
https://www.doi.org/10.19061/iochem-bd-1-197[4445]
To compute the Boltzmann population of each system, ligands are
removed to gas-phase reservoirs and conformational contributions of the
adsorbed organometallic moieties are not taken into account The bridge
to terminal ligands in Fe3(CO)Y (Y = 4-11) gas-phase structures are
known to behave dynamically (barrier 0.25 eV for their bridge-opening
[46]
bridge-closing).
Although ensemble calculations would be ideal for
this purpose including all entropic terms,[47] the scaffold limits the
multiplicity of the structures (complicating its evaluation) and their
fluxionality due to a wall effect. Thus, only one possible structure for each
system was considered. The energy differences when removing ligands
are large enough to ensure that the considered resting states are
relevant even if these contributions were not introduced in the model.
Catalyst Evaluation
Continuous flow hydrogenation experiments were performed in
a
ThalesNano H-Cube Pro setup. The catalyst was pressed and sieved
(mesh size 200–400 m) and 100 mg of the fraction was loaded in a
stainless-steel cartridge (3.5 mm internal diameter). The remaining
volume was filled with silicon carbide (46 grit, Alfa Aesar). A 0.4 M
reaction solution of 2-methyl-3-butyn-2-ol (MBY, 98 %, Sigma Aldrich) in
toluene (99.8 %, Fischer Scientific) was continuously fed with a rate of
FL = 1 cm3 min- 1 and mixed with in situ generated hydrogen
Acknowledgments
This publication was created as part of NCCR Catalysis, a
National Centre of Competence in Research funded by the
Swiss National Science Foundation. Dr. Evgeniya Vorobyeva
and Dr. Frank Krumeich for assistance with electron microscopy
and ScopeM at ETH Zurich for access to their facilities. BSC-
RES for providing generous computational resources.
(FG = 36 cm3 min- 1
) before contact with the catalyst cartridge. The
standard conditions employed were P = 8 bar and T = 323 K, with an
equilibration time under steady conditions of 10 min prior to sample
collection. To explore the substrate scope, the semi-hydrogenation of 1-
hexyne (98 %, Acros), 3-hexyne (99 %, Acros), 4-pentyn-1-ol (97 %,
Sigma Aldrich), 3-methyl-1-pentyn-3-ol (98 %, TCI), or 1,1-diphenyl-2-
propyn-1-ol (99 %, Sigma Aldrich) in toluene (0.4 M) were also studied.
Keywords: Alkyne hydrogenation • Iron catalysts • Nuclearity
effects • Single atoms • Carbon nitride
The obtained liquid product was analyzed offline with
a gas
chromatograph (HP-6890, HP-5 capillary column) equipped with a flame
ionization detector (FID). The metal-specific rate of alkene formation r
was used as a comparative measure and is quantified from the liquid
flowrate (FL), the feed concentration (c0), the conversion (X), the alkene
selectivity (S-ene), and the bulk (nFe,bulk) and surface amounts of iron
[1]
[2]
G. Vilé, D. Albani, N. Almora-Barrios, N. López, J. Pérez-Ramírez,
ChemCatChem 2016, 8, 21-33.
a) A. J. McCue, A. Gibson, J. A. Anderson, Chem. Eng. J. 2016, 285,
384-391; b) M. J. Taylor, S. K. Beaumont, M. J. Islam, S. Tsatsos, C. A.
M. Parlett, M. A. Issacs, G. Kyriakou, Appl. Catal. B 2021, 284, 119737.
a) Y. Liu, Y. Li, J. A. Anderson, J. Feng, A. Guerrero-Ruiz, I. Rodríguez-
Ramos, A. J. McCue, D. Li, J. Catal. 2020, 383, 51-59; b) D. Albani, M.
Shahrokhi, Z. Chen, S. Mitchell, R. Hauert, N. López, J. Pérez-
Ramírez, Nat. Commun. 2018, 9:2634.
(nFe,surf). Unconverted alkyne (c-yne) and over-hydrogenated alkane (c-ane
)
[3]
of the corresponding substrate are considered for the selectivity
calculation. These metrics are computed using equations (1-3).
c0 −c-yne
(1)
X =
S-ene
r =
[4]
[5]
Y. Liu, A. J. McCue, P. Yang, Y. He, L. Zheng, X. Cao, Y. Man, J. Feng,
J. A. Anderson, D. Li, Chem. Sci. 2019, 10, 3556-3566.
c0
c-ene
c-ene + c-ane
(2)
(3)
=
a) S. C. Bart, E. Lobkovsky, P. J. Chirik, J. Am. Chem. Soc. 2004, 126,
13794-13807; b) R. J. Trovitch, E. Lobkovsky, E. Bill, P. J. Chirik,
Organometallics 2008, 27, 1470-1478.
c0F XS-ene
L
nFe,bulk
[6]
[7]
K. Junge, K. Schröder, M. Beller, Chem. Commun. 2011, 47, 4849-
4859.
R. M. Bullock, J. G. Chen, L. Gagliardi, P. J. Chirik, O. K. Farha, C. H.
Hendon, C. W. Jones, J. A. Keith, J. Klosin, S. D. Minteer, R. H. Morris,
A. T. Radosevich, T. B. Rauchfuss, N. A. Strotman, A. Vojvodic, T. R.
Ward, J. Y. Yang, Y. Surendranath, Science 2020, 369, eabc3183.
a) P. J. Chirik, Acc. Chem. Res. 2015, 48, 1687-1695; b) S. Enthaler, K.
Junge, M. Beller, Angew. Chem. Int. Ed. 2008, 47, 3317-3321.
a) P. H. Phua, L. Lefort, J. A. Boogers, M. Tristany, J. G. de Vries,
Chem. Commun. 2009, 3747-3749; b) C. Rangheard, C. de Julian
Fernandez, P. H. Phua, J. Hoorn, L. Lefort, J. G. de Vries, Dalton
Trans. 2010, 39, 8464-8471; c) A. Welther, M. Bauer, M. Mayer, A.
Jacobi von Wangelin, ChemCatChem 2012, 4, 1088-1093; d) V.
Kelsen, B. Wendt, S. Werkmeister, K. Junge, M. Beller, B. Chaudret,
Chem. Commun. 2013, 49, 3416-3418.
Computational Methods
Spin-polarized density functional theory (DFT) simulations were
performed using the Vienna Ab Initio Simulation Package (VASP)
code.[36] A generalized gradient approximation (GGA) was employed,
expressed by the Perdew–Burke–Ernzenhof functional[37] with D3
dispersion correction[3738] to describe van der Waals interactions. Inner
electrons were described by projector augmented waves (PAW),[39] while
valence electrons were expanded in plane waves with a cut-off kinetic
energy of 450 eV. Models for the potential structures formed in situ for
each set (Fe1/MCN, Fe2/MCN, and Fe3/MCN), the reaction intermediates,
and transition states were constructed in a (2x2) heptazinic supercell with
four layers of thickness, with the bottom one frozen to the configuration of
the bulk. Slabs were separated by 14 Å of vacuum and a dipole
correction along the z-direction was used.[40] The Brillouin zone was
sampled with a gamma-centered 3×3×1 k-point grid (~0.3 Å). In all
cases, the ionic and electronic convergence criteria were 10−4 and
10−5 eV, respectively. Transition states were located using the climbing
image nudged elastic band algorithm[41] and confirmed by diagonalizing
the numerical Hessian matrix obtained by displacements of ± 0.02 Å.
[8]
[9]
[10] M. Stein, J. Wieland, P. Steurer, F. Tolle, R. Mülhaupt, B. Breit, Adv.
Synth. Catal. 2011, 353, 523-527.
[11] R. Hudson, A. Rivière, C. M. Cirtiu, K. L. Luska, A. Moores, Chem.
Commun. 2012, 48, 3360-3362.
[12] a) T. N. Gieshoff, A. Welther, M. T. Kessler, M. H. Prechtl, A. Jacobi
von Wangelin, Chem. Commun. 2014, 50, 2261-2264; b) R. Hudson, G.
Hamasaka, T. Osako, Y. M. A. Yamada, C. J. Li, Y. Uozumi, A. Moores,
Green Chem. 2013, 15, 2141-2148.
8
This article is protected by copyright. All rights reserved.

10.1002/cctc.202100235
ChemCatChem
FULL PAPER
[13] a) A. A. Shesterkina, E. V. Shuvalova, O. A. Kirichenko, L. M. Kustov,
Russ. J. Phys. Chem. A 2018, 92, 2412-2416; b) A. A. Shesterkina, E.
V. Shuvalova, E. A. Redina, O. A. Kirichenko, O. P. Tkachenko, I. V.
Mishin, L. M. Kustov, Mendeleev Commun. 2017, 27, 512-514.
[14] a) M. Benz, A. M. van der Kraan, R. Prins, Appl. Catal. A 1998, 172,
149-157; b) I. T. Papadas, S. Fountoulaki, I. N. Lykakis, G. S. Armatas,
Chem. Eur. J. 2016, 22, 4600-4607.
[31] R. Lin, D. Albani, E. Fako, S. K. Kaiser, O. V. Safonova, N. López, J.
Pérez-Ramírez, Angew. Chem. Int. Ed. 2019, 58, 504-509.
[32] D. Albani, M. Capdevila-Cortada, G. Vilé, S. Mitchell, O. Martin, N.
López, J. Pérez-Ramírez, Angew. Chem. Int. Ed. 2017, 129, 10895-
10900.
[33] J. L. Fiorio, R. V. Gonçalves, E. Teixeira-Neto, M. A. Ortuño, N. López,
L. M. Rossi, ACS Catal. 2018, 8, 3516-3524.
[15] H. L. Niu, J. H. Lu, J. J. Song, L. Pan, X. W. Zhang, L. Wang, J. J. Zou,
Ind. Eng. Chem. Res. 2016, 55, 8527-8533.
[34] M. García-Mota, B. Bridier, J. Pérez-Ramírez, N. López, J. Catal. 2010,
273, 92-102.
[16] a) R. V. Jagadeesh, A. E. Surkus, H. Junge, M. M. Pohl, J. Radnik, J.
Rabeah, H. Huan, V. Schunemann, A. Brückner, M. Beller, Science
2013, 342, 1073-1076; b) R. V. Jagadeesh, K. Natte, H. Junge, M.
Beller, ACS Catal. 2015, 5, 1526-1529; c) B. Sahoo, C. Kreyenschulte,
G. Agostini, H. Lund, S. Bachmann, M. Scalone, K. Junge, M. Beller,
Chem. Sci. 2018, 9, 8134-8141; d) Z. J. Wei, Y. X. Hou, X. M. Zhu, L.
Y. Guo, Y. X. Liu, A. Y. Zhang, ChemCatChem 2018, 10, 2009-2013.
[17] M. Tejeda-Serrano, J. R. Cabrero-Antonino, V. Mainar-Ruiz, M. López-
Haro, J. C. Hernández-Garrido, J. J. Calvino, A. Leyva-Pérez, A.
Corma, ACS Catal. 2017, 7, 3721-3729.
[35] A. P. Grosvenor, B. A. Kobe, M. C. Biesinger, N. S. McIntyre, Surf.
Interface Anal. 2004, 36, 1564-1574.
[36] a) G. Kresse, J. Furthmüller, Phys. Rev. B 1996, 54, 11169-11186; b)
G. Kresse, J. Furthmüller, Comput. Mater. Sci. 1996, 6, 15-50.
[37] J. P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 1996, 77, 3865-
3868.
[38] S. Grimme, J. Antony, S. Ehrlich, H. Krieg, J. Chem. Phys. 2010, 132,
154104.
[39] a) P. E. Blöchl, Phys. Rev. B 1994, 50, 17953-17979; b) G. Kresse, D.
Joubert, Phys. Rev. B 1999, 59, 1758-1775.
[18] M. Tejeda-Serrano, M. Mon, B. Ross, F. Gonell, J. Ferrando-Soria, A.
Corma, A. Leyva-Pérez, D. Armentano, E. Pardo, J. Am. Chem. Soc.
2018, 140, 8827-8832.
[40] G. Makov, M. C. Payne, Phys. Rev. B 1995, 51, 4014-4022.
[41] G. Henkelman, H. Jónsson, J. Chem. Phys. 2000, 113, 9978-9985.
[42] M. Newville, J. Synchrotron Radiat. 2001, 8, 322-324.
[43] B. Ravel, M. Newville, J. Synchrotron Radiat. 2005, 12, 537-541.
[44] M. Álvarez-Moreno, C. de Graaf, N. López, F. Maseras, J. M. Poblet, C.
Bo, J. Chem. Inf. Model. 2015, 55, 95-103.
[19] a) S. K. Kaiser, Z. Chen, D. Faust Akl, S. Mitchell, J. Pérez-Ramírez,
Chem. Rev. 2020, 120, 11703-11809; b) S. Mitchell, E. Vorobyeva, J.
Pérez-Ramírez, Angew. Chem. Int. Ed. 2018, 57, 15316-15329; c) L.
Zhang, M. Zhou, A. Wang, T. Zhang, Chem. Rev. 2020, 120, 683-733.
d) G. Kyriakou, M. B. Boucher, A. D. Jewell, E. A. Lewis, T. J. Lawton,
A. E. Baber, H. L. Tierney, M. Flytzani-Stephanopoulos, E. C. Sykes,
Science 2012, 335, 1209-1212.
[45] A. Ruiz-Ferrando, ioChem-BD Dataset. DOI: 10.19061/iochem-bd-1-
197
[46] B. E. Mann, J. Chem. Soc., Dalton Trans. 1997, 1457-1472.
[47] B. Zandkarimi, A. N. Alexandrova, WIREs Comput. Mol. Sci. 2019, 9,
e1420
[20] a) J. Lu, P. Serna, C. Aydin, N. D. Browning, B. C. Gates, J. Am. Chem.
Soc. 2011, 133, 16186-16195; b) E. Guan, B. C. Gates, ACS Catal.
2017, 8, 482-487; c) M. Rondelli, G. Zwaschka, M. Krause, M. D.
Rötzer, M. N. Hedhili, M. P. Högerl, V. D’Elia, F. F. Schweinberger, J.-
M. Basset, U. Heiz, ACS Catal. 2017, 7, 4152-4162.
[21] C. Sun, N. Mammen, S. Kaappa, P. Yuan, G. Deng, C. Zhao, J. Yan, S.
Malola, K. Honkala, H. Hakkinen, B. K. Teo, N. Zheng, ACS Nano
2019, 13, 5975-5986.
[22] R. R. Langeslay, H. Sohn, B. Hu, J. S. Mohar, M. Ferrandon, C. Liu, H.
Kim, A. Jeremy Kropf, C. Yang, J. Niklas, O. G. Poluektov, E. Ercan
Alp, P. Ignacio-de Leon, A. P. Sattelberger, A. S. Hock, M. Delferro,
Dalton Trans. 2018, 47, 10842-10846.
[23] a) S. Tian, Q. Fu, W. Chen, Q. Feng, Z. Chen, J. Zhang, W. C. Cheong,
R. Yu, L. Gu, J. Dong, J. Luo, C. Chen, Q. Peng, C. Draxl, D. Wang, Y.
Li, Nat. Commun. 2018, 9:2353; b) E. Vorobyeva, E. Fako, Z. Chen, S.
M. Collins, D. Johnstone, P. A. Midgley, R. Hauert, O. V. Safonova, G.
Vilé, N. López, S. Mitchell, J. Pérez-Ramírez, Angew. Chem. Int. Ed.
2019, 58, 8724-8729.
[24] Z. P. Chen, S. Mitchell, E. Vorobyeva, R. K. Leary, R. Hauert, T.
Furnival, Q. M. Ramasse, J. M. Thomas, P. A. Midgley, D. Dontsova,
M. Antonietti, S. Pogodin, N. López, J. Pérez-Ramírez, Adv. Funct.
Mater. 2017, 27, 1605785.
[25] Z. P. Chen, E. Vorobyeva, S. Mitchell, E. Fako, N. López, S. M. Collins,
R. K. Leary, P. A. Midgley, R. Hauert, J. Pérez-Ramírez, Natl. Sci. Rev.
2018, 5, 642-652.
[26] M. M. Kauppinen, M. M. Melander, K. Honkala, Catal. Sci. Technol.
2020, 10, 5847-5855.
[27] a) J. Phillips, B. Clausen, J. A. Dumesic, J. Phys. Chem. 1980, 84,
1814-1822; b) K. Asakura, K. Ooi, Y. Iwasawa, J. Mol. Catal. 1992, 74,
345-351; c) T.S. Piper, F.A. Cotton, G. Wilkinson, J. Inorg. Nucl. Chem.
1955, 1, 165-174.
[28] E. Vorobyeva, V. C. Gerken, S. Mitchell, A. Sabadell-Rendón, R.
Hauert, S. Xi, A. Borgna, D. Klose, S. M. Collins, P. A. Midgley, D. M.
Kepaptsoglou, Q. M. Ramasse, A. Ruiz-Ferrando, E. Fako, M. A.
Ortuño, N. López, E. M. Carreira, J. Pérez-Ramírez, ACS Catal. 2020,
10, 11069-11080.
[29] M. C. Biesinger, B. P. Payne, A. P. Grosvenor, L. W. M. Lau, A. R.
Gerson, R. S. C. Smart, Appl. Surf. Sci. 2011, 257, 2717-2730.
[30] S. Vernuccio, R. Goy, P. R. von Rohr, J. Medlock, W. Bonrath, React.
Chem. Eng. 2016, 1, 445-453.
9
This article is protected by copyright. All rights reserved.

10.1002/cctc.202100235
ChemCatChem
FULL PAPER
TOC Figure
Friend or foe? Catalysts prepared by depositing iron carbonyl complexes of distinct nuclearity on graphitic carbon nitride exhibit
differing activity in the liquid-phase semi-hydrogenation of alkynes. Simulations reveal that the superior performance of single-atom
catalysts originates from the partial retention of carbonyl ligands, which despite introducing a high energetic cost for their removal,
ensure the accessibility of the metal center.
10
This article is protected by copyright. All rights reserved.