“This document is the Accepted Manuscript version of a Published Work that appeared in final form in Journal of
the American Chemical Society, copyright © American Chemical Society after peer review and technical editing
by the publisher. To access the final edited and published work see [insert ACS Articles on Request author- directed link to Published Work, see http://pubs.acs.org/doi/abs/10.1021/jacs.6b05680

Tracking the structural and electronic configurations of a cobalt
proton reduction catalyst in water
Dooshaye Moonshiram,*,1, # Carolina Gimbert-Suriñach,*2, # Alexander Guda,3 Antonio Picon,1 C. Stefan Lehmann,1 Xiaoyi Zhang,4 Gilles Doumy,1 Anne Marie March,1 Jordi Benet-Buchholz,2 Alexander
Soldatov,3 Antoni Llobet,*,2,5 and Stephen H. Southworth1
1

Chemical Sciences and Engineering Division, Argonne National Laboratory, 9700 S. Cass Avenue, Lemont IL 60439,
U.S.A
2
Institute of Chemical Research of Catalonia (ICIQ), Barcelona Institute of Science and Technology, Avinguda Països Catalans 16, 43007 Tarragona, Spain
3
International Research Center “Smart Materials”, Southern Federal University, 344090 Rostov-on-Don, Russia
4
X-Ray Science Division, Argonne National Laboratory, 9700 S. Cass Avenue, Lemont IL, U.S.A
5
Departament de Química, Universitat Autònoma de Barcelona, 08193 Cerdanyola del Vallès, Barcelona, Spain
ABSTRACT: Time resolved X-ray absorption spectroscopy (X-TAS) has been used to study the light induced hydrogen evolution
reaction catalyzed by a highly stable tetradentate macrocyclic cobalt complex with the formula [LCoIIICl2]+ (L = macrocyclic ligand), [Ru(bpy)3]2+ photosensitizer and an equimolar mixture of sodium ascorbate/ascorbic acid electron donor in pure water.
XANES and EXAFS analysis of a binary mixture of the octahedral Co(III) pre-catalyst and [Ru(bpy)3]2+ after illumination, revealed
in-situ formation of a Co(II) intermediate with significantly distorted geometry and electron transfer kinetics of 51 ns. On the other
hand, X-TAS experiments of the complete photocatalytic system in the presence of the electron donor showed the formation of a
square planar Co(I) intermediate species within a few nanoseconds followed by its decay in the microsecond timescales. The Co(I)
structural assignment is supported by calculations based on density functional theory (DFT). At longer reaction times, we observe
the formation of the initial Co(III) species concomitant to the decay of Co(I), thus closing the catalytic cycle. The experimental Xray absorption spectra of the molecular species formed along the catalytic cycle are modeled using a combination of molecular
orbital DFT calculations (DFT-MO) and Finite Difference Method (FDM). These findings allowed us to unequivocally assign the
full mechanistic pathway followed by the catalyst as well as to determine the rate limiting step of the process, which consists in the
protonation of the Co(I). This study provides a complete kinetics scheme for the hydrogen evolution reaction by a cobalt catalyst,
revealing unique information for the development of better catalysts for the reductive side of hydrogen fuel cells.

1-INTRODUCTION
Solar hydrogen fuel technology based on artificial photosynthesis will only be feasible if it can be performed efficiently and at low cost.1,2 From the first report of a functional water
splitting cell using UV light as energy source and metallic
platinum as a cathode for the hydrogen evolution catalysis,3
significant progress has been made towards the construction of
cells that use a wider range of the UV-visible-NIR spectrum
with increasing solar to hydrogen conversion efficiencies.4
They use a variety of configurations taking advantage of photovoltaics and photoelectrochemical technologies or both
combined. An important element for the optimum performance and competitive cost of these cells is the hydrogen
evolving catalyst (HEC) of the reductive half-cell, responsible
for the H—H bond formation. While metallic platinum is well
known to catalyze the water reduction very efficiently near the

thermodynamic potential, the high cost associated with this
precious metal has prompted the scientific community to
search for alternative catalysts based on earth-abundant materials. In this context, the use of molecular cobalt-based catalysts, stabilized by nitrogen donor ligands, has proven to be a
good alternative given the high turnover numbers and stability
achieved under catalytic conditions.4-9 The catalyst precursor
[LCoIIICl2]+ (L = macrocyclic ligand) in Scheme 1 belongs to
this family of HEC and has been used in electrochemical as
well as photochemical catalysis showing excellent results.10-14
In contrast to many other molecular HEC that are active only
in organic solvents, complex [LCoIIICl2]+ works in pure aqueous conditions showing remarkable stability over a period of
several hours. Although mechanistic studies based on electrochemical and spectroscopic methods such as electron paramagnetic resonance or UV-Vis spectroscopy have been re-

ported,12 a full understanding of the reaction pathway for the
hydrogen evolution reaction catalyzed by [LCoIIICl2]+ has not
been described. The main reason for this information gap is
the high reactivity of the intermediate species that are involved
in the catalysis, with lifetimes in the pico to microsecond
timescales. For this reason, time-resolved spectroscopic techniques have become a very powerful tool to track such intermediates. Indeed, pump-probe transient absorption spectroscopy with UV-Vis or IR probes have been successfully used to
study the electron transfer process from different photosensitizers such as [Ru(bpy)3]2+ to [LCoIIICl2]+ or other cobalt precatalysts.8,14-31 With these studies, it was possible to discriminate between oxidative and reductive quenching of the photosensitizers and to trap the Co(I) key intermediate species that
triggers the catalysis (Scheme 1). Early reports on mechanistic
studies were done with systems that work in organic solvents,
using organic acids as proton sources, a media that is unfavorable for a future commercialization of solar cells. More recently, the field has been directed towards the study of catalysts
working in pure aqueous conditions, which provide valuable
information for a realistic applied catalytic system.

Scheme 1. Left) Cobalt hydrogen evolution catalyst precursor [LCoIIICl2]+ studied in this work and its aquated
derivatives [LCoIIICl(OH2)]2+ and [LCoIII(OH2)2]3+.
Right)General scheme of plausible hydrogen evolution
pathways from cobalt (I) active species.

Scheme 2. Mechanistic pathway for the hydrogen evolution
reaction catalyzed by [LCoIIICl2]+.
In this work, we use X-ray transient absorption spectroscopy (X-TAS) at the Co K-edge to follow the dynamic processes
of a full catalytic cycle for the hydrogen evolution reaction
illustrated in Schemes 1 in acidic water. One of the advantages
of X-TAS is its element selectivity, allowing us to monitor
changes related to cobalt species with no contribution from
other components of the system containing ruthenium (from
the photosensitizer) or organic compounds. Furthermore,
valuable information about oxidation states as well as the
coordination sphere around the cobalt center can be extracted,
information that is not available using other techniques. XTAS has already been successfully used to study ultrafast
electron transfer of multicomponent systems or dyads all the
way from the picosecond to the microsecond timescale.22,32-42
By combining X-TAS in the ns-µs time regime with ab initio
theoretical calculations, we were able to characterize intermediate species involved in the catalysis, allowing us to build a
complete kinetics, electronic and structural scheme of the
reaction. The manuscript is divided according to the sequential
processes taking place in the reaction, as illustrated in Scheme
2. Thus, we start by reporting experimental and theoretical
XAS simulations of the bis(chloro) [LCoIIICl2]+ and bis(aqua)
[LCoIII(OH2)2]3+ complexes. We continue with the first electron transfer to convert the Co(III) initial compound to a reduced Co(II) intermediate and its interaction with the ascorbate electron donor. Finally, the second electron transfer to
form the Co(I) active species as well as the recovery of the
starting Co(III) derivative, closing the catalytic cycle, is discussed.
2-Results and Discussion
2·1-Static X-ray absorption spectroscopy analysis of
Co(III) pre-catalysts
and
bis(aqua)
The
bis(chlorido)
[LCoIIICl2]+
[LCoIII(OH2)2]3+ complexes shown in Scheme 1, both in solid
and solution form, were studied by X-ray absorption near edge
structure (XANES) and Extended X-ray absorption fine structure (EXAFS) spectroscopy (Fig. 1). The substitution of chlorido by aqua axial ligands in these complexes is characterized
by a 2.8 eV shift of the rising edge toward higher energy of the
XANES spectrum. The large value of the energy shift can be
accounted for by ligand to metal charge transfer shakedown
transitions. These transitions can affect the energy position43
and intensity44 of the shoulder on the rising edge of the
XANES spectrum upon chlorido to aqua ligand substitution as
shown in Kau et al44. This effect has been shown to be especially prominent for complexes with chlorido ligands as compared to those with aqua ligands44. The near edge region was
simulated with both DFT-MO and FDM methods (Fig. 1 A
and Fig. S1). Due to self consistency of the DFT-MO method,
it is more accurate in predicting chemical shift for different
oxidation states of Co and XANES in the first 40 eV above
absorption edge. This method is not appropriate for higher
energy region due to the limited size of the available basis set,
consisting of Slater type orbitals. In contrast FDM works well
for energies up to 200 eV and more, but this method is inaccurate within 1-2 eV in prediction of the absorption edge energy,
thus it is less suitable for the very near edge region of difference spectra discussed below.

Although the metal center in [LCoIIICl2]+
and
[LCoIII(OH2)2]3+ has the same formal oxidation state, the calculated Mulliken charge on the Co atom clearly manifest the
different crystal field exerted by these two ligands as expected
(see Table S2). The shift in energy position of the absorption
edge is also attributed to a slight geometrical changes in the
[LCoIIICl2]+ versus the [LCoIII(OH2)2]3+ complex (compare CoO and Co-Cl distances in Table S1, S2). We further monitored
the exchange of the labile chlorido ligands of the [LCoIIICl2]+
complex by aqua ligands in situ, by dissolving the [LCoIIICl2]+
complex in water and taking aliquots that were subsequently
frozen at successive time delays from 30 s to 4 h. A gradual
shift in the XANES spectrum confirms partial chlorido to aqua
ligand exchange and an eventual equilibrium after 4 h (Fig. S1
and Scheme 1). The final mixture is composed of an intermediate species with one aqua and one chlorido axial ligands
([LCoIIICl(OH2)]2+ together with the bis(aqua) complex
[LCoIII(OH2)2]3+ as shown in Scheme 1). The isolation of small
crystals of [LCoIIICl(OH2)]2+ suitable for X-ray single crystal
diffraction (XRD) analysis in an aqueous mixture of
[LCoIIICl2]+ further supports this point. (Table S3). These
results are of relevance for the water reduction catalytic experiments that are performed in aqueous conditions and can be
extrapolated to a wide range of reported catalysts of similar
structure bearing labile chlorido axial ligands, that are generally used as catalyst precursors.9

Figure 1. A) Top: Experimental normalized Co K-edge XANES
of the [LCoIIICl2]+ (black) and [LCoIII(OH2)2]3+ (red) complexes.
Bottom: Theoretical XANES simulation using FDM (solid) and
DFT-MO (dashed) methods. B) Experimental Fourier transforms
of k3-weighted Co EXAFS of the [LCoIIICl2]+ (black) and
[LCoIII(OH2)2]3+ (red) complexes. Inset: Back Fourier transforms,
experimental results (solid lines) and fitting (dashed lines) k3χ(k)
Experimental spectra were calculated for k values of 1.431 to
12.103 Å-1. C) Simulated EXAFS spectra. Atomic coordinates
were obtained from single crystal X-ray diffraction structure of
[LCoIIICl2]+ (dashed black line)10 and from DFT simulations
(solid black line for chloro and solid red line for aqua). Note that
there is no single crystal X-ray diffraction data for the
[LCoIII(OH2)2]3+ sample.

The EXAFS spectra of the bis(chlorido) and bis(aqua) complexes are shown in Fig. 1 B. Two prominent peaks are observed in the [LCoIIICl2]+ spectrum corresponding to the distinctive Co-N and Co-Cl bond distances while only one peak
is observed for the [LCoIII(OH2)2]3+ complex with contribution
of both the Co-N and Co-O bonds. EXAFS fits for the first
coordination sphere and the entire spectrum for both complexes are shown in Table S1 in the Supporting Information. Analysis of first peak in [LCoIIICl2]+ clearly resolves Co-N distanc-

es at 1.89 Å, and inclusion of two Co-Cl distances at 2.23 Å
improves the quality of the fit as shown by the decreased Rfactors and χ2 value (Table S1, Fit 2). Similarly, fitting of
[LCoIII(OH2)2]3+ illustrates improvement of the fit quality with
the addition of two Co-O distances at 1.93 Å in the first coordination sphere (Table S1, Fit 4). The fitted Co-N and Co-Cl
bond distances are in agreement with the reported XRD structure10 and relaxed structures from DFT geometry optimization.
Changes detected in experimental EXAFS for the [LCoIIICl2]+
and [LCoIII(OH2)2]3+ complexes correlated well with data from
XRD analysis and FEFF45 simulations of DFT optimized
coordinates (Fig. 1 C). These findings show that theoretical
methods including geometry optimization and ab-initio XAS
simulations used for [LCoIIICl2]+ and [LCoIII(OH2)2]3+ complexes reproduce the experimental data well, and therefore
these techniques can be reliably used for analysis of the intermediates species of the cobalt complex’s catalytic cycle.
2·2-Time-resolved X-ray absorption spectroscopy of binary mixtures Ru/Co: formation of Co(II)
Formation of the Co(II) intermediate from Co(III) (Eq. 1 in
Scheme 1) was probed in the ns-µs time range through timeresolved laser/X-ray pump/probe spectroscopy of a water
solution mixture consisting of the cobalt catalyst precursor
[LCoIIICl2]+ with the [Ru(bpy)3]2+ photosensitizer in 1:5 ratio,
that had been incubated for 24 hours allowing for partial chloro ligand substitution. XAS spectra of the binary mixture were
collected before (laser off) and after (laser on) the laser excitation. In particular, the laser-on XAS spectrum for an averaged
time delay of 918+/-612 ns between the laser pump excitation
and X-ray probing is shown in Fig. 2 A. Upon light excitation,
metal-to-ligand charge transfer at the [Ru(bpy)3]2+ triggers
electron transfer from the excited photosensitizer to the Co(III)
complex, forming a Co(II) species. By subtracting the laser-on
and laser-off spectrum, a transient signal is obtained for each
pump-probe time delay and provides information on the photoinduced dynamics. A transient signal with a maximum timedependent peak energy of 7720 eV is obtained which is typical
of the signature of Co(II) formation (Fig. 3, black trace). We
not only observe clear changes in the oxidation state of the Co
metal site, but also we see strong changes in the cobalt electronic configuration and coordination environment as indicated by the pre-edge region of the absorption spectrum at lower
photon energies around 7710 eV (red trace Fig. 2A, inset).
Presence of pre-edge features correspond to 1s to 3 d quadrupole transitions and dipole excitations of the core electrons
into the valence 3d states hybridized with ligand p orbitals4,5..
Non centrosymmetric geometries have been shown to have an
increased intensity in their pre-edge features due to an increase
in the metal 4 p mixing into the 3 d orbitals contributing towards the electric dipole 1s to 4 p character of this transition
as shown in Solomon et al46-48. Octahedral Co(III) complexes
have a d6 configuration showing a predominant pre-edge
peak46. However, local distortions of the octahedron around
Co result in additional splitting of the d levels (Figure S4,
Supporting Information). Upon laser excitation, the occupation
of the Co d shell changes to d7, thus enabling a multiplet preedge feature. The d level atomic rearrangements are accompanied with structure relaxation and hybridization of valence 3d
states with N/O ligand p-orbitals46,47. In order to elucidate the
structure of the Co(II) species, three possible geometries of the
transient Co(II) complex were tested by using DFT calculations. A Co(II) octahedral, square pyramidal and square planar
geometries were considered depending on the numbers of

water molecules coordinated to the cobalt For the calculations
presented in figure 3, both the spin polarization contribution
shown in Debeer et al.49 and the quadrupole transition operator
are taken into consideration.

Figure 2. A) Experimental transient X-ray absorption spectrum
for the [Ru(bpy)3]2+ (30mM)/[LCoIIICl2]+ (6 mM) binary mixture
at the Co K-edge, measured before (black) and an averaged time
delay of 918+/-612 ns after laser excitation. B) Fourier transforms of k3-weighted Co EXAFS of the laser off and reconstructed excited state, considering a 27 % excited state fraction. Inset:
Back Fourier transforms, experimental results (solid lines) and
fitting (dashed lines) k3χ(k) Fourier transforms were calculated for
k values of 1.431 to 10.358 Å-1.

.As shown in Fig. 3, there is poor agreement between experimental spectrum and theoretical simulation for a square planar structure (compare black and blue traces). On the other
hand, the simulations for the square pyramidal and distorted
octahedral geometries resolve most of the transitions observed
in the experimental spectrum including the two pre-edge features (red and pink traces, Fig. 3, inset). However, these preedge transitions as well as the two peaks in the near-edge
regions at 7717 and 7727 eV are most prominent in the square
pyramidal derivative. The difference in energy for the two preedge transitions was calculated by DFT-MO theory to be
around 2 eV for the square pyramidal derivative, comparable
to that observed in the experimental pre-edge XAS. Furthermore, analysis of the higher energy region of the spectrum
shows that the square pyramidal geometry best describes the
position of the maximum peak of the experimental transient
spectrum (see Fig. S3). Note that the energy position of calculated transient signal is better when a theoretical “laser off”
spectrum from a combination of [LCoIIICl(OH2)]2+ and
[LCoIII(OH2)2]3+ is considered as opposed to a pure
[LCoIIICl2]+ or a pure [LCoIII(OH2)2]3+complex (Fig. S3A and
S3B, respectively). This result is in agreement with an equilibrium between partially substituted [LCoIIICl(OH2)]+complex

and the bis(aqua) [LCoIII(OH2)2]3+ complex as illustrated in
Scheme 1.

Figure 3. Experimental difference spectrum (laser on-laser off)
corresponding to the Co(II) transient signal (black) and simulated
difference spectra for [LCoII]2+ square planar (blue),
[LCoII(OH2)]2+ square pyramidal (red) and [LCoII(OH2)2]2+ octahedral (pink) geometries calculated using DFT-MO method.
Molecular orbitals diagrams involved in the transition of the preedge region are illustrated in Figure S4. A 50/50 mixture of
[LCoIIICl(OH2)]2+ and [LCoIII(OH2)2]3+ was used as theoretical
“laser off” spectrum (Scheme 1).

The percentage of the Co(II) reduced species obtained after
laser excitation was determined to be around 27% by integrating the pre-edge region of the excited state and comparing to a
pure Co(II) sample that was prepared chemically (Fig. S5).
Analysis of the reconstructed EXAFS spectrum for the Co(II)
photo-induced species reveals two prominent peaks in the first
coordination sphere corresponding to the Co-N and the elongated Co-O distances, with values of 1.96 and 2.19 Å respectively (Fig. 2B, Fig S5 D and Table S1). These distances are in
agreement with the theoretical relaxed coordinates (see Table
S2). In addition, calculated EXAFS profile agrees well with
experimental when using these coordinates (Fig. S5 E and
Table S2). Note that the Co-O distance is much longer than in
the initial Co(III) complex as expected due to the reduction of
the metal site. The presence of an additional peak corresponding to a Co-O distance of 2.19 Å rules out a Co(II) square
planar structure in agreement with the simulations discussed
Importantly, the [LCoIIICl2]+/[Ru(bpy)3]2+ system is very
stable during time-resolved X-ray absorption measurements
over long exposure to X-ray and laser beams at room tempera-

ture, as no sample degradation was observed between different
scans. This illustrates the robustness of the [LCoIIICl2]+ macrocyclic complex as compared to the more commonly used
cobaloxime type complexes.50 For instance, we carried out
time-resolved XAS of a water soluble cobaloxime/[Ru(bpy)3]2+ system with the same photon flux and we
observed noticeable spectral changes in the XANES features
due to the gradual degradation of the initial Co(III) complex
(Fig. S6). In addition, a Co(II) transient signal analogous to
that observed in the [LCoIIICl2]+/[Ru(bpy)3]2+ is obtained with
other photosensitizers such as water soluble CdTe quantum
dots,14 confirming the versatility of this catalyst (Fig. S7).
It was possible to measure the kinetics of the formation and
decay of the Co(II) formed in situ by fixing the X-ray photon
energy at 7720 eV and varying the time delay between the
laser and X-ray pulses. The Co(II) species shows a rise time of
51 ns, consistent with fast electron transfer from the excited
photosensitizer to the Co(III). On the other hand, the decay
time of the back electron transfer reaction, responsible for the
decay of the Co(II) species was determined to be 5.3 µs (Fig.
4). These kinetics determine the lifetime of the Co(II), which
should be long lived enough to form the Co(I) active species
that enters the catalytic cycle (Eq. 2 and 3, Scheme 1).

Figure 4. Pump-probe time delay scans recorded at 7720 eV of a
binary mixture [LCoIIICl2]+/[Ru(bpy)3]2+, corresponding to the
formation and decay of the Co(II) photo-induced species. Inset:
kinetics fittings of the Co(II) formation (left) and decay (right).

2·3-Interaction of Co(II) species with ascorbate electron
donor
One of the key elements of a homogeneous photochemical
HEC system is the sacrificial electron donor that provides the
necessary reducing equivalents to produce hydrogen from
water. In this study we used ascorbic acid/sodium ascorbate
mixtures (H2Asc/NaHAsc, 1:1), that have shown excellent
results
when
used
in
combination
with
the
[LCoIIICl2]+/[Ru(bpy)3]2+ hybrid photocatalyst used here.12,31 In
this system a spontaneous “dark” reduction from Co(III) to
Co(II) precedes the light-induced hydrogen evolution events
due to the reduction potential of the ascorbate agent, that is
sufficiently strong to reduce the pre-catalyst [LCoIIICl2]+ into a
Co(II) derivative.12,14 This process is evident from XAS analysis of a mixture of [LCoIIICl2]+ and H2Asc/NaHAsc, showing
the expected energy shift in the XANES region when compared with the parent Co(III) complex. Interestingly, we find
that the resulting spectrum for the [LCoIIICl2]+/ascorbate mixture is different to that obtained for a Co(II) derivative that had

been prepared in the absence of ascorbate,51 suggesting that
the ascorbate ion is coordinating the Co(II) metal site. For
instance, comparison of the XANES spectra reveals a 0.4 eV
energy shift and a change in the spectral features between the
two Co(II) species (Fig. 5 A). On the other hand, EXAFS
analysis of the two Co(II) species illustrates a slight decrease
of intensity in the first peak that corresponds to the Co-N bond
distance as well as a decrease in the back-scattering amplitude
corresponding to the Co-O bond distance (Fig. 5 B), suggesting a change in the first coordination sphere around the cobalt
site. The UV-Visible absorption spectrum is also sensitive to
the ascorbate coordination to Co(II) by showing gradual
changes in the absorption bands at 350, 460 and 670 nm (Fig.
S9 A).

Figure 5. A) Normalized Co K-edge XANES of the Co(II) species (6 mM) prepared chemically in the absence of ascorbate
(black) frozen within 5 min, and Co(II) prepared from the reaction
of [LCoIIICl2]+ (6 mM) with ascorbic acid/ascorbate (200 mM)
(red) frozen within 4 h B) Fourier transforms of k3-weighted Co
EXAFS of the Co(II) prepared chemically in the absence of
ascorbate (black) frozen within 30 s, and Co(II) prepared from the
reaction of [LCoIIICl2]+ with ascorbic acid/ascorbate (red) and
frozen within 4 hrs. Inset: Back Fourier transforms, experimental
results (solid lines) and fitting (dashed lines) k3χ(k) Experimental
spectra were calculated for k values of 1.431 to 10.358 Å-1.

In order to understand the Co(II)/ascorbate interaction, we
performed simulation for different structures with ascorbate
ion coordination, namely a square pyramidal structure with
one ascorbate coordinated ([LCoII(Asc)]+) and two octahedral
complexes with two ascorbate ions ([LCoII(Asc)2)]) or one
ascorbate and one water coordinated in the axial positions
([LCoII(Asc)(OH2)]+) (Fig. S2 and Table S2). These calculations reveal that a Co(II) center with two coordinated ascorbate ions is unstable. On the other hand, XANES simulations
show that [LCoII(Asc)(OH2)]+ and [LCoII(Asc)]+ species both
show a shift of the rising edge towards higher energy in
agreement with experimental data (Fig. S9 C). In addition, the
octahedral derivative [LCoII(Asc)(OH2)]+ features a characteristic shoulder at around 7722 eV, resembling the XANES

shape obtained for the [LCoIIICl2]+/ascorbate mixture (compare Fig. S9 B and C). These results suggest that an equilibrium between different Co(II) species with distinct ascorbate
coordination exist in the mixture as illustrated in Scheme 2.
2·4-Time-resolved X-ray absorption spectroscopy of ternary mixtures Ru/Co/Asc: formation of Co(I) and recovery
of Co(III) species
Time-resolved X-ray transient absorption spectroscopy of
catalytic mixtures containing the pre-catalyst [LCoIIICl2]+ (6
mM), the photosensitizer [Ru(bpy)3]2+ (30 mM) and the sacrificial electron donor H2Asc/NaHAsc, (1:1, 214 mM) in water
shows the formation and decay of at least two different cobalt
species in the nanosecond to microsecond time scale (Fig. 6).
First of all, a reduced species is formed within a few nanoseconds with a maximum peak at 7714 eV in the difference spectrum, suggesting the formation of a Co(I) derivative. This
reduced species has a lifetime of 1.75 µs as determined from
the kinetics trace obtained at different time delays (Fig. 6 C).
In order to ascertain the structure of this reduced species,
XANES spectra of several Co(I) models were simulated and
used to calculate the difference spectrum by subtraction with
the Co(II) square pyramidal “laser-off” spectrum
([LCo(OH2)]2+). Additional comparison with Co(II) species
with ascorbate coordinated are shown in Fig. S10 in the Supporting Information. Different geometries with different coordinated ligands for a Co(I) species were considered, namely a
square planar, a square pyramidal with an attached water molecule and a square pyramidal with an ascorbate ion in the axial
position (Fig. S2 and Table S2). Both low and high spin configurations were calculated for all the Co(I) models, as it has
been suggested that both configurations have similar energies,
however no significant differences were found in the respective XANES spectra.52,53 As previously mentioned, DFT-MO
gives a better description for the pre-edge region, the low
energy region below the rising edge, while FDM is preferred
to describe the shape of the spectrum in a higher energy range
above the edge. Indeed, we observe that the pre-edge peak
characteristic of the Co(I) species at 7714 eV is well reproduced by the DFT-MO simulation of a square planar geometry
(red line in Fig. 7 A). The FDM simulated spectra for this
geometry also matches well with the high energy range of the
experimental spectrum (>7720 eV) (Fig. S10 A) and therefore
we conclude that the best agreeable structure for the reduced
derivative formed after laser excitation is a Co(I) square planar
complex, a geometry that is strongly favored by a d8 electronic
configuration. These findings are supported by reported X-ray
diffraction structures and calculations of similar intermediates
for the CO2 and proton reduction reactions.52,53 Since the
Co(II) species slowly interacts with the ascorbate ion with
time, the difference spectra considering a “laser-off” Co(II)
square pyramidal with coordinated ascorbate ([LCo(Asc)]+)
are also shown in Figure S10.

Figure 6. Co K-edge X-ray transient absorption spectra of an
aqueous mixture containing [LCoIIICl2]+ (6 mM), [Ru(bpy)3]2+ (30
mM) and H2Asc/NaHAsc (1:1, 214 mM). A) Stacked spectra
corresponding to a series of time delays between laser and X-ray
pulses. These measurements were carried out for a range of averaged time delays from 0 to 25 µs. Each transient signal represents
an average of a set of delays over 459 ns. B) Selected transient
spectra showing the Co(I) decay and Co(III) formation in the 1.15
µs (magenta) and 5.28 µs (blue) time delays. In black, the inverse
of the transient spectrum obtained for binary mixtures
[LCoIIICl2]+/[Ru(bpy)3]2+ corresponding to the Co(III) to Co(II)
transformation(See black trace in Fig. 3) C) X-ray kinetics taken
at 7713.9 eV of a ternary mixture [LCoIIICl2]+/[Ru(bpy)3]2+/Asc,
corresponding to the decay of the Co(I) photo-induced species
(red dots) with the corresponding kinetics fitting (black solid
line). D) X-ray kinetics taken at 7732.9 eV of a ternary mixture
[LCoIIICl2]+/[Ru(bpy)3]2+/Asc, corresponding to the formation and
decay of the Co(III) photo-induced species (red dots) with the
corresponding kinetics fitting (black solid line). The decay of the
Co(I) species have been fitted using an exponential function. The
rise and decay of the Co(III) species have been fitted by using a
system of ordinary differential equation with three parameters; the
rise rate, the decay rate, and the population of the excited transient
Co(I) species.

As the Co(I) intermediate decays in the measured transient
spectra, a second oxidized species is formed, as evidenced by
a broad band appearing in the energy range of 7720-7750 eV
(Figure 6). The formation of this oxidized species shows a rise
time of 1.63 µs that correlates well within error bars with the
observed Co(I) decay of 1.75 µs showing that that both processes are linked to the same reaction step (Fig. 6 C and D).
According to the different hydrogen evolution pathways considered in Scheme 1, a Co(III) hydride forms after protonation
of the Co(I) intermediate and can generate hydrogen via heterolytic or homolytic pathways by reacting with a proton or a
second Co(III) hydride molecule, respectively. Alternatively,
it can be reduced to form a Co(II) hydride species that can
analogously react through homolytic or heterolytic mechanisms. Although there is a consensus that the formation of a
Co(III) hydride species is formed after protonation of the
Co(I) intermediate species, little experimental evidence is
found in the literature about the steps that follow the formation
of this Co(I) derivative, particularly with systems that work in
pure aqueous conditions.30,32-34,54-59 Therefore it was of para-

mount interest to elucidate the nature of the oxidized species
derived from the Co(I) intermediate generated in our X-TAS
experiment. Simulation of XANES spectra of all plausible
candidates in Scheme 1, including both Co(II) and Co(III)
hydride species, were performed and thoroughly analyzed
(Fig. S2 and Table S2). Figure 7 B compares the experimental
transient spectrum obtained 5.28 µs after laser excitation with
the DFT-MO simulated spectra for three different candidates,
namely a Co(III) octahedral geometry with two water molecules in both axial positions, a Co(III) hydride square pyramidal complex as well as Co(III) hydride octahedral derivative
with an axial water molecule. Additional spectral comparisons
of Co(II) hydride species and FDM calculations are found in
Fig. S11 in the Supporting Information. The best agreement
between experimental and simulated spectra is found for the
Co(III) octahedral geometry with two aqua axial ligands
[LCoIII(OH2)2]3+, suggesting that this is the compound formed
at longer experimental times (compare red and black traces in
Fig. 7 B and Fig. S11). This hypothesis is confirmed by the
good matching of the transient spectrum with the difference
spectrum of the isolated compounds [LCoIII(OH2)2]3+ and
[LCoII(OH2)]+ or inverting the transient signal shown in Fig. 3
for the binary mixture Ru/Co that represents the reverse transition from a Co(III) to Co(II) species (compare black and blue
lines in Fig. 6 B).
These findings allowed us to unambiguously assign the heterolytic pathway through a Co(III) hydride species as the
preferred mechanistic route for the hydrogen evolution reaction catalyzed by a derivative of complex [LCoIIICl2]+
(Scheme 2). The rate limiting step of this process is the formation of the intermediate hydride that is so reactive that
rapidly combines with a proton in the medium to generate
hydrogen gas and recovers the starting complex
[LCoIII(OH2)2]3+ within a time scale of less than 2 microseconds. Taking into account the 100 ps pulse duration of the Xrays, we can further affirm that the lifetime of the hydride
species is lower than 100 picoseconds. The relatively slow rate
(microseconds) associated with the formation of the Co(III)
hydride is a consequence of the low reactivity of the d8 Co(I)
square planar intermediate complex that is highly stabilized by
the macrocyclic nature of the ligand surrounding the cobalt
coordination sphere. Our results suggest that future efforts to
enhance the catalytic rate of similar hydrogen evolution catalysts should focus on facilitating the protonation of the Co(I)
intermediate, that is, increasing the basicity of the metal center
by either introducing higher electron density to the cobalt or
by modifying the stable macrocyclic structure of the ligand.
Our findings not only prove the mechanistic pathway followed
by [LCoIIICl2]+ with spectroscopically and kinetically characterized intermediates all along the catalytic cycle, but also
confirm the molecular nature of the catalyst, ruling out the
formation of zero-valent metallic colloids upon catalytic turnover.

reaction in water and suggest designs of new catalysts that
perform beyond the microsecond time regime. They also provide promise for further investigations of light harvester/Co
supramolecular assemblies in aqueous medium, where the
electron transfer is more efficient through the addition of an
electron relay. The electron transfer by the incorporation of a
relay is consequently significantly faster in the femtosecond/sub-picosecond timescales. Although sub-ps ultrafast
time-resolved studies are not possible at synchrotrons, with
temporal resolutions in the 100’s picoseconds/nanosecond
timescales, development of such types of complexes will
provide novel avenues for future studies at free electron lasers
where femtosecond time resolution is possible.

Figure 7. A) Comparison of the experimental difference spectrum
taken at an averaged delay of 689+/-230 ns after laser excitation
(black) and DFT-MO simulated difference spectra for [LCoI]+
square planar (red), [LCoI(OH2)]+ square pyramidal (pink) and
[LCoI(Asc)]+ square pyramidal (green) geometries. B) Comparison of experimental spectrum taken at an averaged delay 5.28+/0.230 µs after laser excitation (black) and DFT-MO simulated
difference spectra for [LCoIII(OH2)2]+ octahedral (red),
[LCoIIIH)]2+ square pyramidal (cyan) and [LCoIIIH(OH2)]2+ octahedral (pink) geometries. The simulated spectrum of a square
pyramidal Co(II) complex with an aqua ligand in the axial position has been used as “laser off” spectrum to calculate the difference spectra in A and B.(See also Fig. S10 and S11 in the supporting information for further comparisons and FDM simulations)

3-Summary and Conclusions
In summary, we report the application of transient X-ray absorption spectroscopy in combination with DFT calculations
in monitoring the electronic and structural dynamics of water
soluble [Ru(bpy)3]2+/[LCoIIICl2]+ hybrid systems. Upon excitation of the chromophore, the cobalt(III) octahedral complex is
reduced to a Co(II) species and undergoes a Co-O bond elongation of 0.24 Å. We additionally measured the rise time kinetics for the formation of the Co(II) intermediate to be 51 ns
followed by its decay to form the starting Co(III) complex in
the order of 5.3 µs. The study of the complete multimolecular
system in the presence of sodium ascorbate/ascorbic acid
electron donor ([Ru(bpy)3]2+/Co/Asc) shows unambiguous
experimental proof for the formation of a Co(I) square planar
species followed by the formation of the octahedral Co(III)
starting complex with two axial aqua ligands. The formation
of Co(I) was found to be in the hundreds of nanoseconds time
scale while its decay lasts for a few microseconds. The formation of a Co(III) derivative accompanies the decay of the
Co(I) decay, with comparable kinetics in the microsecond
timescale. These structural and kinetics results allow us to
conclusively determine that the catalyst follows an heterolytic
mechanistic pathway where the rate limiting step is the formation of the Co(III) hydride derivative, which has a lifetime
lower than a hundred picoseconds as illustrated in Scheme 2.
These findings are crucial to understand the charge separation
dynamics in photocatalytic systems for the hydrogen evolution

4-Experimental Section
Samples Preparation. Throughout this study, the
[LCoIIICl2]+, [LCoIII(OH2)2]3+ and [LCoII(OH2)]2+ complexes
were synthesized as described in the literature.11,12,53 Purity of
the compounds was verified by 1H-NMR, cyclic voltammetry
and/or
UV-Vis
spectroscopy.
Small
crystals
of
[LCoIIICl(OH2)](ClO4)2 suitable for single crystal XRD analysis were obtained from slow evaporation of an aqueous reaction mixture during the synthesis of [LCoIII(OH2)2](ClO4)3
from [LCoIIICl2]Cl. The crystals of [LCoIIICl(OH2)](ClO4)2
were too small for XAS collection and XAS analysis. Ultrapure (Type 1) water (resistivity 18.2 M .cm at 25°C, TOC
4 µg/L) sourced from a Q-POD unit of Milli-Q integral water
purification system (Millipore) was used to prepare the catalytic solutions.
UV-Vis Spectroscopy. A Cary 300 Bio UV-Vis Spectrophotometer (Varian Inc.) was used to monitor UV-visible
spectra changes of a solution of [LCoIIICl2]+ upon reduction
with a mixture of sodium ascorbate and ascorbic acid. UVvisible spectral changes were also monitored for the chloro
ligands substitution by aqua for solutions of [LCoIIICl2]+ in
water.
Static XANES and EXAFS Measurements. X-ray absorption spectra were collected at the Advanced Photon
Source (APS) at Argonne National Laboratory on bending
magnet beamline 20 at electron energy 7.7 keV and average
current 100 mA. The radiation was monochromatized by a
Si(110) crystal monochromator. The intensity of the X-ray
was monitored by three ion chambers (I0, I1 and I2 ) filled with
70% nitrogen and 30% helium and placed before the sample
(I0) and after the sample (I1 and I2 ). A Co metal foil was
placed between the I1 and I2 and its absorption recorded with
each scan for energy calibration. Plastic (Lexan) EXAFS
sample holders (inner dimensions of 12 mm x 3 mm x 3 mm)
filled with frozen solutions samples were inserted into a precooled (20 K) cryostat. The samples were kept at 20 K in a He
atmosphere at ambient pressure. Data was recorded as fluorescence excitation spectra using a 13-element energy-resolving
detector. Solid samples were diluted with BN powder, pressed
between polypropylene and mylar tape, and measured in the
cryostat in transmission mode. In order to reduce the risk of
sample damage by X-ray radiation, 80% flux was used in the
defocused mode (beam size 1 x 10 mm) and no damage was
observed to any samples scan after scan. The samples were
also protected from the X-ray beam during spectrometer
movements by a shutter synchronized with the scan program.
No more than 5 scans were taken at each sample position in
any condition. The Co XAS energy was calibrated by the first

maximum of the second derivative of the cobalt metal XANES
spectrum.
Time-resolved XAS measurements. Time-resolved X-ray
absorption spectra were collected at 7 ID-D60 and 11 ID-D37
beamlines at the APS using undulator radiation at electron
energy 7.7 keV. The experiments were carried out using the 24
bunch timing mode of APS (in top up mode with a constant
102 mA ring current) which consists of a train of xrays separated by 153 ns. This mode easily allows for gateable detectors
that selects X-ray pulses. This timing mode was suitable for
this type of experiments in which a fresh sample was required
for every incoming X-ray pulse as well as for the time resolution of the Avalanche Photodiode (APD)/ fast scintillator
detectors.
Experiments carried out at 7 ID-D used an amplified laser
system (Duetto from Time-Bandwidth) to generate 10 ps pulses at its second harmonic wavelength of 532 nm. The X-ray
beam was focused via Kirkpatrick-Baez optics to around 8 µm
spot size and overlapped with the laser spatially and temporally. The sample was flown into a free-flowing flat liquid jet
sheet with a thickness of 100 µm oriented at 45˚ to both X-ray
and laser beams and continuously degassed with N2 gas. Timeresolved XAS measurements were carried out in fluorescence
mode with two fast scintillators positioned at 90˚ to the X-ray
beam. Laser excitation conditions such as the spot size, repetition rate and power were optimized in order to provide maximum excitation in each probed volume while allowing a fresh
sample to be excited from shot to shot. In this case, a laser
spot size of 15 µm (V) x 11 µm (H), repetition rate of 210 kHz
and laser power of 138 mW were used. The [LCoIIICl2]+
complex is very robust, and no degradation was found in the
XAS spectra during X-ray and laser irradiation for several
hours. In order to monitor the formation of intermediate species in the microsecond time scale, a field programmable gate
array logic system was implemented that allowed the collection of data for each of the X-ray pulses in the train between
laser pump pulses. With a liquid flow speed of around 6 m/s
and laser spot size of around 15 µm, the pumped laser volume
moved out of the gaussian full-width-half-maximum (FWHM)
probed region in about 1.1 µs, limiting the temporal range that
could be explored. Reducing the flow speed would require a
reduction in the pump/probe repetition rate, which is undesirable as more time would be needed to collect the necessary
statistics. Instead, a pump-flow-probe scheme was incorporated where the pump beam was spatially separated from the
probe position and the flow speed of the sample dictated the
pump/probe delay, a variation on the scheme from Smolentev
et al50. Such a setup was useful for monitoring the formation
of Co(II) in a binary Ru/Co mixture.
Studies of the binary Ru/Co mixture and the complete photocatalytic system were additionally repeated at beamline 11
ID-D. The sample was pumped at 527 nm wavelength using a
regenerative amplified laser with 1.6 kHz repetition rate 5psFWHM pulse length and laser power of 630 mW. The sample
was flown into a free-flowing 550 µm cylindrical jet inside an
airtight aluminium chamber and continuously degassed with
nitrogen. The X-ray and laser beam was spatially overlapped
with an X-ray spot size of 100 µm(V) x 450 µm(H) and laser
spot size of 170 µm(V) x 550 µm. With a liquid flow speed of
3 m/s, the pumped laser volume was calculated to move out of
the gaussian FWHM region in around 24 µs. This temporal
range ensured that the excited state volume was probed more

at the center and less at the edges where the excitation fraction
would be less, due to movement of the sample. Beamline 11
ID-D has an automated data digitization system which allows
for all x-ray pulses after laser excitation to be collected. Such a
system, together with the larger X-ray beam spot size, was
very useful for our experiments, as multiple X-ray pulses after
laser excitation were averaged to monitor the dynamics for the
formation and decay of Co(II), Co(I) and Co(III) transient
species in the ns-µs time regime. In addition, by averaging
multiple pulses, we obtained a better resolution of the main
features in the pre-edge and edge regions of the transient signals.
The delay between the laser and X-ray pulses was adjusted
by a programmable delay line (PDL-100A-20NS, Colby Instruments) and the X-ray fluorescence signals were collected
with two APDs positioned at 90˚ on both sides of the liquid
jet. Moreover, a combination of Z-1 filters and soller slits with
conical geometry were used to reduce the background from
elastically scattered x-rays. Note that the kinetics for the formation and decay of the Co(II) species was measured by varying the delay between the laser and X-ray pulses from 0-25µs
with the programmable delay line. Because of the lower transient signals for the Co(I) and Co(III) transient species, their
rise and decay kinetics were determined by recording the
intensity of the laser (on-off) signal at the peak energy of the
transient signal.
At both beamlines 7 ID-D and 11 ID-D, a Co metal foil was
placed between two ionization chambers downstream to the Xray beam, and its transmission recorded with each scan for
energy calibration.
EXAFS data analysis. Athena software61 was used for data
processing. The energy scale for each scan is normalized using
the cobalt metal standard and scans made for the same samples
were added. Data in energy space are pre-edge corrected,
normalized, deglitched (if necessary), and background corrected. The processed data are next converted to the photoelectron
wave vector (k) space and weighted by k3. The electron wave
number is defined as k = [2m( E − E0 ) / h 2 ] 12 , E0 is the energy
origin or the threshold energy. k-space data were truncated
near the zero crossings (k = 1.403 to 10.358/12.103 Å-1) in Co
EXAFS before Fourier transformation. The k-space data are
transferred into the Artemis Software for curve fitting. In
order to fit the data, the Fourier peaks are isolated separately,
grouped together, or the entire (unfiltered) spectrum was used.
The individual Fourier peaks were isolated by applying a
Hanning window to the first and last 15% of the chosen range,
leaving the middle 70% untouched. Curve fitting is performed
using ab initio-calculated phases and amplitudes from the
FEFF845 program and ab initio-calculated phases and amplitudes are used in the EXAFS equation62
χ ( k ) = S02 ∑

Nj
kR

2
j

−2 R j

f eff j (π , k , R j )e

− 2σ 2 k 2
j

e

λ j (k )

sin( 2kR j + φij ( k ))

(1)
where Nj is the number of atoms in the jth shell; Rj the mean
distance between the absorbing atom and the atoms in the jth
j

shell;

f eff j

(π,k, Rj ) is the ab initio amplitude function for

shell j, and the Debye-Waller term e −2σ k accounts for damping
due to static and thermal disorder in absorber-backscatterer
2
j

2

−2 R j

distances. The mean free path term e λ ( k ) reflects losses due to
inelastic scattering, where λj(k), is the electron mean free path.
j

The oscillations in the EXAFS spectrum are reflected in the
sinusoidal term sin(2kR j + φij (k )) , where φij (k ) is the ab initio
phase function for shell j. This sinusoidal term shows the
direct relation between the frequency of the EXAFS oscillations in k-space and the absorber-back scatterer distance. S02 is
an amplitude reduction factor.
The EXAFS equation (Eq. 1) is used to fit the experimental
Fourier isolated data ( in q-space) as well as unfiltered data (in
k-space) and Fourier transformed data (in R-space) using N,
S02, E0, R, and σ2 as variable parameters. N refers to the number of coordination atoms surrounding Co for each shell. The
quality of fit is evaluated by R-factor and the reduced Chi2
value. The deviation in E0 was required to be less than or equal
to 10 eV. An R-factor less than 2% denotes that the fit is good
enough whereas an R-factor between 2 and 5% denotes that
the fit is correct within a consistently broad model62. The
reduced Chi2 value is used to compare fits as more absorberbackscatter shells are included to fit the data. A smaller reduced Chi2 value indicates a better fit. Similar results were
obtained from fits done in k, q, and R-spaces.
DFT-MO and FDM Calculations. The ground state electronic structures and atomic geometries of the different Co
complexes were calculated using DFT B3LYP level of theory
using the ADF-2015 program package 63. There is ongoing
discussion in the literature about the best exchange correlation
functional 64 and in our work we tested the popular local
GGA-PBE65 and hybrid B3LYP 66 functionals for all cobalt
complexes. Both functionals slightly overestimate Co-N and
Co-ligand bond distances during geometry optimization as
compared to the crystallography data. The covalency of metalligand or even metal-metal bonds is strongly affected by the
amount of Hartree-Fock exchange in the functional and GGA
functionals may overestimate the covalency of bonds while
hybrid functionals will give too ionic bonds. The bond length
is also affected by the solvent effects which were simulated by
means of the COSMO model with a dielectric medium around
the molecule. 67 B3LYP was found to better describe unoccupied molecular orbitals which are subsequently used for
XANES calculations, thus in the main text we show results
based on this functional. The QZ4P basis set was used in all
calculations. Thus each atomic orbital was represented as a
combination of four Slater type orbitals (STO) with different
exponential powers and 4 polarization functions. STOs are
better from a theoretical point of view since they have the
correct behavior near the nucleus and the correct asymptotic
long-range behavior. Typically, fewer STOs than gaussian
GTOs are needed to reach a certain level of accuracy, as discussed in Güell et al. 68 The X-ray absorption cross section
obeys the Fermi golden rule, and the matrix element for the
electron transition can be written as
σ (E) ~ ψ

2

ε ⋅ r ⋅ e i k ⋅r ψ 1 s

⋅ δ ( E − E f − E1 s )

(2)
where ψ1s is a core−electron wave function, ψf is a wave
function for the unoccupied state, ε is a photon polarization, k
is a photon wavevector. Both dipole and quadrupole terms
were taken into account:
f

σ ( E ) ~ ψ f ε ⋅ r ⋅ (1 + ik ⋅ r ) ψ 1s

2

⋅ δ ( E − E f − E1s )

(3)
The averaging was performed for all directions of photon
polarization and wave vector with a restriction of perpendicu-

larity of these two vectors. Calculation of the wave function
for unoccupied states is performed using two independent
approaches. The first is based on a basis set method within the
molecular orbital DFT approach already described above
(DFT-MO in the text). Molecular orbitals of the 1s core level
and unoccupied levels are used to calculate dipole matrix
elements and Lorentzian convolution is subsequently applied
for comparison with experiment69. In order to represent a final
state after the photon absorption process, a core hole is introduced on a 1s cobalt level in a self-consistent way. In a second
approach shown in supplemetary we use a finite difference
method (FDM in the text) realized in FDMNES software.70 In
this method, the Schrodinger equation is solved on a 3d grid of
points and unknowns are the values of the wave function on
this grid. The FDM is preferable to describe the shape of the
spectrum in a wide energy range but it is less accurate in prediction of absorption edge energy shifts than quantum chemistry DFT-MO approach. Molecular orbitals obtained in a quantum chemistry basis set approach have more accurate energies
but fail to describe energy region above 30-40 eV above the
absorption edge due to basis set limitations. However in the
main text we are interested mainly in the pre-edge region and
near edge XAS leaving only DFT-MO in the text for clarity.
For the DFT-MO calculations shown in text, we first obtain
eigenvalues and corresponding wavefunctions for spin polarizations for a given Co complex in the presence of a core hole.
The matrix elements are then evaluated for transitions between
1s core level and unoccupied MOs using dipole and quadrupole transition operators. In order to compare with experimental results, a convolution of calculated matrix elements
was performed with a Lorentzian profile using energy dependent linewidth. In the pre-edge region, the width of the Lorentzian profile corresponds to a core hole lifetime broadening
for Co. This value is then increased in higher energy interval
with a smooth arctangent function. The parameters of the
matrix elements calculations (grid step, size number of unoccupied MOs) and energy convolution are fixed once for all
complexes. The DFT-MO calculated spectra were subsequently aligned according to the energy value of the Co 1s orbital,
thus reproducing the chemical shift for different Co oxidation
states. A rigid shift with the identical value of 1 eV was applied for all spectra in order to align the energy scale between
experimental data and theoretical calculations.
As mentioned above, it is important to note that metal to
ligand charge transfer shakedown transitions can cause a shift
in the energy position43 and intensity44 of the shoulder on the
rising edge of the XANES spectrum. However, the proper
theoretical description of such state requires the use of multireference computational methods which are at present capable
of dealing only with the first several transitions71.
In our simulations, we intended to reproduce both pre-edge
transitions and main edge features which require calculation of
energies and wavefunctions of more than 500 electronic states
that were possible only within DFT level of theory. Thus, we
expect the discrepancies between energy position of the
shakedown transition observed in the experiment and calculated spectra to be in the order of 1eV. However, the effect of the
intensity of the peak associated with shakedown transitions is
much suppressed for complexes with oxygen bonds 44.

ASSOCIATED CONTENT

Field Code Changed

Supporting Information. Additional supporting information is
available on EXAFS analysis, XRD data, DFT-MO, and FDM
calculations . This material is available free of charge via the
internet at www.//pubs.acs.org

AUTHOR INFORMATION
Corresponding Authors
*dmoonshi@gmail.com
*cgimbert@iciq.es
*allobet@iciq.es

Author Contributions
‡These authors contributed equally.

ACKNOWLEDGMENT
We acknowledge support from the U.S. Department of Energy,
Office of Science, Basic Energy Sciences, Chemical Sciences,
Geosciences, and Biosciences Division under contract no. DEAC02-06CH11357. This research uses resources of the Advanced
Photon Source(APS), a U.S Department of Energy (DOE) Office
of Science User Facility operated for the DOE Office of Science
by Argonne National Laboratory under Contract No. DE-AC0206CH11357. We thank Dr. Chengjun Sun and Dr. Qingyu Kong
for help with experiments at beamlines 20 BM and 11 ID-D, APS.
We also acknowledge financial support from MINECO and the
“Fondo Europeo de Desarrollo Regional” (FEDER) through
grants CTQ-2013–49075-R, SEV-2013-0319, and CTQ-201452974-REDC. Financial support from the EU COST actions
CM1202 and CM1205 are also acknowledged. C.G.S. is grateful
to AGAUR and Generalitat de Catalunya for a “Beatriu de Pinós”
postdoctoral grant. A.G. and A.S. would like to thank the Russian
Ministry of Education and Science for the support (agreement №
14.587.21.0002, unique identifier RFMEFI58714X0002).

REFERENCES
(1) Walter, M. G.; Warren, E. L.; McKone, J. R.; Boettcher, S. W.;
Mi, Q.; Santori, E. A.; Lewis, N. S. Chem.Rev. 2010, 110, 6446.
(2) McKone, J. R.; Lewis, N. S.; Gray, H. B. Chem.Mat. 2014, 26,
407.
(3) Fujishima, A.; Honda, K. Nature 1972, 238, 37.
(4) Queyriaux, N.; Jane, R. T.; Massin, J.; Artero, V.; ChavarotKerlidou, M. Coord. Chem. Rev. 2015, 304–305, 3.
(5) Eckenhoff, W. T.; Eisenberg, R. Dalton. Trans. 2012, 41, 13004.
(6) Thoi, V. S.; Sun, Y.; Long, J. R.; Chang, C. J. Chem. Soc. Rev.
2013, 42, 2388.
(7) Berardi, S.; Drouet, S.; Francas, L.; Gimbert-Surinach, C.;
Guttentag, M.; Richmond, C.; Stoll, T.; Llobet, A. Chem. Soc. Rev.
2014, 43, 7501.
(8) Khnayzer, R. S.; Thoi, V. S.; Nippe, M.; King, A. E.; Jurss, J. W.;
El Roz, K. A.; Long, J. R.; Chang, C. J.; Castellano, F. N. Energy
Environ.Sci. 2014, 7, 1477.
(9) Tong, L.; Kopecky, A.; Zong, R.; Gagnon, K. J.; Ahlquist, M. S.
G.; Thummel, R. P. Inorg.Chem. 2015, 54, 7873.
(10) Leung, C.-F.; Chen, Y.-Z.; Yu, H.-Q.; Yiu, S.-M.; Ko, C.-C.;
Lau, T.-C. Int. J.-Hydrogen Energy 2011, 36, 11640.
(11) McCrory, C. C. L.; Uyeda, C.; Peters, J. C. J.Am.Chem.Soc.
2012, 134, 3164.
(12) Varma, S.; Castillo, C. E.; Stoll, T.; Fortage, J.; Blackman, A. G.;
Molton, F.; Deronzier, A.; Collomb, M.-N. Phys.Chem.Chem.Phys.
2013, 15, 17544.
(13) Roy, S.; Bacchi, M.; Berggren, G.; Artero, V. ChemSusChem
2015, 8, 3632 – 3638
(14) Gimbert-Suriñach, C.; Albero, J.; Stoll, T.; Fortage, J.; Collomb,
M.-N.; Deronzier, A.; Palomares, E.; Llobet, A. J.Am.Chem.Soc.
2014, 136, 7655.
(15) Natali, M.; Orlandi, M.; Chiorboli, C.; Iengo, E.; Bertolasi, V.;
Scandola, F. Photochem.Photobiol.Sci. 2013, 12, 1749.

(16) Dong, J.; Wang, M.; Zhang, P.; Yang, S.; Liu, J.; Li, X.; Sun, L.
J. Phys. Chem C 2011, 115, 15089.
(17) Shan, B.; Baine, T.; Ma, X. A. N.; Zhao, X.; Schmehl, R. H.
Inorg. Chem. 2013, 52, 4853.
(18) Lewandowska-Andralojc, A.; Baine, T.; Zhao, X.; Muckerman,
J. T.; Fujita, E.; Polyansky, D. E. Inorg. Chem. 2015, 54, 4310.
(19) Rodenberg, A.; Orazietti, M.; Probst, B.; Bachmann, C.; Alberto,
R.; Baldridge, K. K.; Hamm, P. Inorg. Chem. 2015, 54, 646.
(20) Cropek, D. M.; Metz, A.; Muller, A. M.; Gray, H. B.; Horne, T.;
Horton, D. C.; Poluektov, O.; Tiede, D. M.; Weber, R. T.; Jarrett, W.
L.; Phillips, J. D.; Holder, A. A. Dalton Trans. 2012, 41, 13060.
(21) Veldkamp, B. S.; Han, W.-S.; Dyar, S. M.; Eaton, S. W.; Ratner,
M. A.; Wasielewski, M. R. Energy Environ.Sci. 2013, 6, 1917.
(22) Mukherjee, A.; Kokhan, O.; Huang, J.; Niklas, J.; Chen, L. X.;
Tiede, D. M.; Mulfort, K. L. Phys.Chem.Chem.Phys.2013, 15, 21070.
(23) Huang, J.; Mulfort, K. L.; Du, P.; Chen, L. X. J.Am.Chem.Soc.
2012, 134, 16472.
(24) Guttentag, M.; Rodenberg, A.; Kopelent, R.; Probst, B.;
Buchwalder, C.; Brandstätter, M.; Hamm, P.; Alberto, R. Eur. J.
Inorg. Chem. 2012, 2012, 1,59
(25) Reynal, A.; Lakadamyali, F.; Gross, M. A.; Reisner, E.; Durrant,
J. R. Energy Environ.Sci. 2013, 6, 3291.
(26) Lakadamyali, F.; Reynal, A.; Kato, M.; Durrant, J. R.; Reisner,
E. Chem. Eur. J. 2012, 18, 15464.
(27) Reynal, A.; Willkomm, J.; Muresan, N. M.; Lakadamyali, F.;
Planells, M.; Reisner, E.; Durrant, J. R. Chemical Communications
2014, 50, 12768.
(28) Soltau, S. R.; Niklas, J.; Dahlberg, P. D.; Poluektov, O. G.;
Tiede, D. M.; Mulfort, K. L.; Utschig, L. M. Chem. Commun. 2015,
51, 10628.
(29) Deponti, E.; Luisa, A.; Natali, M.; Iengo, E.; Scandola, F. Dalton
Trans. 2014, 43, 16345.
(30) Dempsey, J. L.; Winkler, J. R.; Gray, H. B. J.Am.Chem.Soc.
2010, 132, 16774.
(31) Gueret, R.; Castillo, C. E.; Rebarz, M.; Thomas, F.; Hargrove,
A.-A.; Pécaut, J.; Sliwa, M.; Fortage, J.; Collomb, M.-N. J J. Photochem. Photobiol. B, Biol. 2015, 152, Part A, 82.
(32) Smolentsev, G.; Guda, A.; Zhang, X.; Haldrup, K.; Andreiadis,
E. S.; Chavarot-Kerlidou, M.; Canton, S. E.; Nachtegaal, M.; Artero,
V.; Sundstrom, V. J.Phys.Chem.C 2013, 117, 17367.
(33) Smolentsev, G.; Guda, A. A.; Janousch, M.; Frieh, C.; Jud, G.;
Zamponi, F.; Chavarot-Kerlidou, M.; Artero, V.; van Bokhoven, J.
A.; Nachtegaal, M. Farad. Discuss. 2014, 171, 259.
(34) Smolentsev, G.; Cecconi, B.; Guda, A.; Chavarot-Kerlidou, M.;
van Bokhoven, J. A.; Nachtegaal, M.; Artero, V. Chem. Eur. J. 2015,
21, 15158.
(35) Chen, L. X.; Zhang, X.; Lockard, J. V.; Stickrath, A. B.;
Attenkofer, K.; Jennings, G.; Liu, D.- -J. Acta Crystallogr. Sect. A
2010, 66, 240.
(36) Smolentsev, G.; Canton, S. E.; Lockard, J. V.; Sundstrom, V.;
Chen, L. X. J. Electron. Spectrosc. Relat. Phenom. 2011, 184, 125.
(37) Chen, L. X.; Zhang, X. J.Phys.Chem.Lett. 2013, 4, 4000.
(38) Canton, S. E.; Zhang, X.; Zhang, J.; van Driel, T. B.; Kjaer, K.
S.; Haldrup, K.; Chabera, P.; Harlang, T.; Suarez-Alcantara, K.; Liu,
Y.; Pérez, J.; Bordage, A.; Pápai, M.; Vankó, G.; Jennings, G.; Kurtz,
C. A.; Rovezzi, M.; Glatzel, P.; Smolentsev, G.; Uhlig, J.; Dohn, A.
O.; Christensen, M.; Galler, A.; Gawelda, W.; Bressler, C.; Lemke, H.
T.; Møller, K. B.; Nielsen, M. M.; Lomoth, R.; Wärnmark, K.;
Sundström, V. J.Phys.Chem.Lett. 2013, 4, 1972.
(39) Chen, L. X.; Zhang, X.; Shelby, M. L. Chem.Sci. 2014, 5, 4136.
(40) Canton, S. E.; Kjær, K. S.; Vankó, G.; van Driel, T. B.; Adachi,
S.-i.; Bordage, A.; Bressler, C.; Chabera, P.; Christensen, M.; Dohn,
A. O.; Galler, A.; Gawelda, W.; Gosztola, D.; Haldrup, K.; Harlang,
T.; Liu, Y.; Møller, K. B.; Németh, Z.; Nozawa, S.; Pápai, M.; Sato,
T.; Sato, T.; Suarez-Alcantara, K.; Togashi, T.; Tono, K.; Uhlig, J.;
Vithanage, D. A.; Wärnmark, K.; Yabashi, M.; Zhang, J.; Sundström,
V.; Nielsen, M. M. Nat. Commun 2015, 6,1
(41) Smolentsev, G.; Sundström, V. Coord.Chem.Rev. 2015, 304–
305, 117.
(42) Mulfort, K. L.; Mukherjee, A.; Kokhan, O.; Du, P.; Tiede, D. M.
Chem.Soc.Rev. 2013, 42, 2215.

(43) Bair, R. A.; Goddard, W. A. Phys.Rev.B. 1980, 22, 2767.
(44) Kau, L. S.; Spira-Solomon, D. J.; Penner-Hahn, J. E.; Hodgson,
K. O.; Solomon, E. I. J. Am. Chem. Soc. 1987, 109, 6433.
(45) Rehr, J. J.; Albers, R. C. Rev. Mod. Phys. 2000, 72, 621.
(46) Westre, T. E.; Kennepohl, P.; DeWitt, J. G.; Hedman, B.;
Hodgson, K. O.; Solomon, E. I. J. Am. Chem. Soc. 1997, 119, 6297.
(47) Loeb, K. E.; Westre, T. E.; Kappock, T. J.; Mitic, N.; Glasfeld,
E.; Caradonna, J. P.; Hedman, B.; Hodgson, K. O.; Solomon, E. I. J.
Am. Chem. Soc. 1997, 119, 1901.
(48) Hahn, J. E.; Scott, R. A.; Hodgson, K. O.; Doniach, S.;
Desjardins, S. R.; Solomon, E. I. Chem. Phys. Lett. 1982, 88, 595.
(49) Chandrasekaran, P.; Stieber, S. C. E.; Collins, T. J.; Que, J. L.;
Neese, F.; DeBeer, S. Dalton Trans. 2011, 40, 11070.
(50) Smolentsev, G.; Guda, A.; Zhang, X.; Haldrup, K.; Andreiadis,
E. S.; Chavorot-Kerlidou, M.; Canton, S. E.; Nachtegaal, M.; Artero,
V.; Sundstrom, V. J.Phys. Chem.C 2013, 117, 17367.
(51) Long, K. M.; Busch, D. H. Inorg. Chem.1970, 9, 505.
(52) Lacy, D. C.; McCrory, C. C. L.; Peters, J. C. Inorg. Chem. 2014,
53, 4980.
(53) Zhang, M.; El-Roz, M.; Frei, H.; Mendoza-Cortes, J. L.; HeadGordon, M.; Lacy, D. C.; Peters, J. C. J. Phys. Chem. C 2015, 119,
4645.
(54) Hu, X.; Brunschwig, B. S.; Peters, J. C. J.Am.Chem.Soc. 2007,
129, 8988.
(55) Muckerman, J. T.; Fujita, E. Chem.Commun. 2011, 47, 12456.
(56) Solis, B. H.; Hammes-Schiffer, S. J.Am.Chem.Soc. 2011, 133,
19036.
(57) Bhattacharjee, A.; Weiss, A. K. H.; Artero, V.; Field, M. J.;
Hofer, T. S. J.Phys.Chem.B. 2014, 118, 5551.
(58) Niklas, J.; Mardis, K. L.; Rakhimov, R. R.; Mulfort, K. L.; Tiede,
D. M.; Poluektov, O. G. J.Phys.Chem.B. 2012, 116, 2943.
(59) Estes, D. P.; Grills, D. C.; Norton, J. R. J.Am.Chem.Soc. 2014,
136, 17362.
(60) March, A. M.; Stickrath, A.; Doumy, G.; Kanter, E. P.; Krassig,
B.; Southworth, S. H.; Attenkofer, K.; Kurtz, C. A.; Chen, L. X.;
Young, L. Rev.Sci.Instrum. 2011, 82, 073110.
(61) Ravel, B.; Newville, M. J. Synchrotron Radiat. 2005, 12, 537.
(62) Koningsberger, D. C.; Prins, R. X Ray Absorption: Principles,
Applications, Techniques of EXAFS, SEXAFS and XANES; John
Wiley & Sons, 1988.
(63) Guerra, C. F.; Snijders, J. G.; te Velde, G.; Baerends, E. J. Theor.
Chem. Acc. 1998, 99, 391.
(64) Cramer, C. J.; Truhlar, D. G. Phys.Chem.Chem.Phys 2009, 11,
10757.
(65) Perdew, J. P.; Burke, K.; Ernzerhof, M. Phy.Rev.Lett. 1996, 77,
3865.
(66) Reiher, M.; Salomon, O.; Artur Hess, B. Theor.Chem.Acc., 107,
48.
(67) Klamt, A.; Schuurmann, G. J. Chem. Soc., Perkin Trans. 2,
1993, 799.
(68) Güell, M.; Luis, J. M.; Solà, M.; Swart, M. J.Phys.Chem.A. 2008,
112, 6384.
(69) Alperovich, I.; Smolentsev, G.; Moonshiram, D.; Jurss, J. W.;
Concepcion, J. J.; Meyer, T.; Soldatov, A.; Pushkar, Y. J .Am. Chem.
Soc . 2001, 133 , 15786
(70) Guda, S. A.; Guda, A. A.; Soldatov, M. A.; Lomachenko, K. A.;
Bugaev, A. L.; Lamberti, C.; Gawelda, W.; Bressler, C.; Smolentsev,
G.; Soldatov, A. V.; Joly, Y. J. Chem. Theory Comput. 2015, 11,
4512.
(71) Tomson, N. C.; Williams, K. D.; Dai, X.; Sproules, S.; DeBeer,
S.; Warren, T. H.; Wieghardt, K. Chem. Sci.2015, 6, 2474.