Single-atom catalysts (SACs) aim at bridging the gap between homogeneous and heterogeneous catalysis. The challenge is the development of materials with ligands enabling coordination of metal atoms in different valence states, and preventing leaching or nanoparticle formation. Graphene functionalized with nitrile groups (cyanographene) is herein employed for the robust coordination of Cu(II) ions, which are partially reduced to Cu(I) due to graphene-induced charge transfer. Inspired by nature's selection of Cu(I) in enzymes for oxygen activation, this 2D mixed-valence SAC performs flawlessly in two O2-mediated reactions: the oxidative coupling of amines and the oxidation of benzylic CH bonds toward high-value pharmaceutical synthons. High conversions (up to 98%), selectivities (up to 99%), and recyclability are attained with very low metal loadings in the reaction. The synergistic effect of Cu(II) and Cu(I) is the essential part in the reaction mechanism. The developed strategy opens the door to a broad portfolio of other SACs via their coordination to various functional groups of graphene, as demonstrated by successful entrapment of FeIII/FeII single atoms to carboxy-graphene.

1. Copyright WILEY-VCH Verlag GmbH & Co. KGaA, 69469 Weinheim, Germany, 2019.
Supporting Information
for Adv. Mater., DOI: 10.1002/adma.201900323
Mixed-Valence Single-Atom Catalyst Derived from
Functionalized Graphene
Aristides Bakandritsos, Ravishankar G. Kadam, Pawan
Kumar, Giorgio Zoppellaro, Miroslav Medved’, Jiří Tuček,
Tiziano Montini, Ondřej Tomanec, Pavlína Andrýsková,
Bohuslav Drahoš, Rajender S. Varma, Michal Otyepka, Manoj
B. Gawande,* Paolo Fornasiero,* and Radek Zbořil*

2. 0
Copyright WILEY-VCH Verlag GmbH & Co. KGaA, 69469 Weinheim, Germany, 2019.
Supporting Information
Mixed-Valence Single-Atom Catalyst Derived from Functionalized Graphene
Aristides Bakandritsos, Ravishankar G. Kadam, Pawan Kumar, Giorgio Zoppellaro, Miroslav
Medveď, Jiří Tuček, Tiziano Montini,Ondřej Tomanec, Pavlína Andrýsková, Bohuslav Drahoš,
Rajender S. Varma, Michal Otyepka, Manoj B. Gawande,* Paolo Fornasiero,* and Radek Zbořil*
Dr. A. Bakandritsos, Dr. R. G. Kadam, Dr. P. Kumar, Dr. G. Zoppellaro, Dr. M. Medveď, Dr. J.
Tuček, O. Tomanec, P. Andrýsková, Dr. R. S. Varma, Prof. M. Otyepka, Dr. M. B. Gawande, Prof.
R. Zbořil
Regional Centre of Advanced Technologies and Materials
Department of Physical Chemistry, Faculty of Science, Palacký University Olomouc
Šlechtitelů 27, 783 71 Olomouc, Czech Republic
E-mail: manoj.gawande@upol.cz; radek.zboril@upol.cz
Dr. B. Drahoš
Regional Centre of Advanced Technologies and Materials
Department of Inorganic Chemistry, Faculty of Science, Palacký University Olomouc
17. listopadu 12, 771 46 Olomouc, Czech Republic
Prof. P. Fornasiero, Prof. Tiziano Montini
Department of Chemical and Pharmaceutical Sciences, INSTM Trieste Research Unit and ICCOM-
CNR Trieste Research Unit, University of Trieste
via L. Giorgieri 1, I-34127 Trieste, Italy
E-mail: pfornasiero@units.it
Keywords: cyanographene, single-atom catalysis, cooperative catalysis, amine coupling, C-H
oxidation
This PDF file includes:
Materials and Methods
Figures M1–M5
Results - Supplementary Figures and Tables
Tables S1–S11
Figures S1–S15
Supplementary references S1-S29

3. 1
Materials and Methods
Reagents and materials. Graphite fluoride (>61 wt.% F, C1F1.1), NaCN (p.a. ≥97%), N,N-
Dimethylformamide (≥99.8%), CuCl2∙2H2O (99.99%), Na2SO4 anhydrous (≥99%), ethyl acetate
anhydrous (99.8%), 1,4-Dioxane anhydrous (99.8%), triethylamine (≥99%), Fe(NO3)3·9H2O (ACS
reagent, ≥98%), graphene oxide nanocolloids (product No 795534), all aromatic benzyl amines,
secondary cyclic amines and anilines were purchased from Sigma Aldrich and used as received
without further purification. All solvents were HPLC grade. All aqueous solutions were prepared
with ultrapure water (18 MΩ cm–1
).
Synthesis of G-CN, G-COOH, and metal immobilization. Cyanographene and graphene-acid
nanosheets were synthesized by following our previous method.30
Briefly, fluorinated graphite (120
mg, ∼4 mmol of C-F units) was added to 15 mL of DMF and stirred for 2 days. Then sonicated
(Bandelin Sonorex, DT 255H type, frequency 35 kHz, power 640 W, effective power 160 W) for 4
h under nitrogen atmosphere. Then 800 mg of NaCN (∼16 mmol) was added and the mixture was
heated at 403 K with a condenser under stirring (500 rpm). After 2 days, the mixture was left to
cool to room temperature. After washing and isolation of the pure product, the material was
suspended in distilled water. Hydrolysis of the nitrile groups on G-CN to carboxyl groups was
performed with 20% HNO3, according to the published procedure,30
in order to synthesize the
graphene acid derivative (G-COOH). Copper loading was performed by mixing an aqueous
suspension of G-CN (20 mL containing 120 mg of G-CN) with CuCl2∙2H2O (2 mL containing 60
mg of Cu ions). After 24 h of stirring, the mixture was separated with centrifugation. H2O was
added in the pellet, mixed and finally centrifuged in order to isolate the final copper-loaded G-CN.
The catalyst (the copper-loaded G-CN) was finally freeze-dried and stored for further use. The
determination of the copper content in the solid catalyst was performed with ICP-MS. The BET
surface area of G(CN)-Cu catalyst was found to be 153 m2
g–1
.
G-COOH loaded with Fe was performed following the same procedure, using
Fe(NO3)3·9H2O. ICP-MS indicated Fe content in the G(COOH)-Fe hybrid of 4.3 wt.%.
The reduced graphene oxide (rGO) copper loaded material was prepared starting with
commercial GO (PlasmaChem, Germany) and mixing with CuCl2∙2H2O following an identical
procedure as with G-CN. After isolation of the final solid, hydrazine was added to reduce GO and
solid was washed well with water and freeze-dried for further use.
Structural and physicochemical characterization techniques. The surface chemical properties of
materials were determined with X-ray photoelectron spectroscopy (XPS) performed on a PHI
VersaProbe II (Physical Electronics) spectrometer using an Al Kα source (15 kV, 50 W). The fine
morphological characteristics of catalyst were determined with transmission electron microscopy
(TEM) at 80 kV accelerating voltage on FEI Titan G2 60-300 transmission electron microscope
equipped with X-FEG electron gun, objective-lens image spherical aberration corrector and
ChemiSTEM EDS detector. While the ultrafine structure of catalysts was obtained on TEM JEOL
2010 with LaB6 type emission gun, operating at 160 kV with a resolution of 0.19 nm. For the
sample preparation, a very dilute dispersion of catalyst (~0.1 mg mL–1
) was prepared by sonication
and deposited on a carbon-coated copper grid and analyzed after drying for 24 h at room
temperature. STEM-elemental mapping for determining EDS pattern and distribution of elements
with the help of STEM-HAADF (high-angle annular dark-field imaging) analyses was recorded on
a FEI Titan HRTEM microscope operating at 80 kV. The nitrogen adsorption–desorption was
carried out using Micromeritics Flex 3 surface area and porosity analyzer. Before the measurements,
the sample was degassed under vacuum at 373 K for 24 hours in a degasser.
EPR spectra were recorded on JEOL JES-X-320 operating at X-band frequency (~9.1 GHz),
equipped with a variable temperature control ES 13060DVT5 apparatus. The cavity (Q) quality
factor was kept above 7000 in all measurements, microwave power was kept under non-saturating
conditions and highly pure quartz tubes were employed in the measurements (Suprasil, Wilmad,

4. 2
<0.5 OD). Experimental parameters in the spin trap experiments in Figure 4B of the main text were
as follows: mod. frequency 100 kHz, mod. amplitude of 0.35 mT, time const. of 30 ms, applied
microwave power of 0.6 mW, sweep time of 240 s, phase 0°, T = 248 K. Simulation of the EPR
envelope (Sim) using third-order perturbation theory (spin-Hamiltonian parameters: giso= 2.0053,
AN = 1.51 mT, AH (-proton) = 1.49 mT, tumbling effects included in the line-width function
(a+bm+cm2
) with coefficients a = 8.6, b = –0.1, c = –0.6 and Lorentzian/Gaussian ratio of 0.5.
Cu K-edge EXAFS and XANES data were measured on the SAMBA beamline at the
Synchrotron SOLEIL (Gif sur Yvette, France).[S1]
Spectra were collected by measuring the Kα
fluorescence line with a 36 pixels Germanium detector (Canberra), while transmission data was
simultaneously recorded with ionization chambers (IC-SPEC, FMB-Oxford) for the sample and a
Cu foil. Harmonic rejection has been accomplished by using two Pd coated mirrors. Data analysis
has been performed with the DEMETER software package.[S2]
ATHENA software has been used to
extract EXAFS signals, by background subtraction (Autobk algorithm) and normalization with the
edge height. Fourier transforms to R space of the k3
-weighted EXAFS data was performed in a k
range of 1–12 Å-1
using the Hanning window function, where ‘‘k’’ indicates a photoelectron wave
number and ‘‘R’’ represents the distance between the absorber atom and a scatterer atom. The
EXAFS signals were analyzed using Artemis software over a k range of 1–12 Å-1
and a R range of
0.8–3 Å. The maximum number of independent points can be estimated from the expression IP =
[(2kR/) + 1], where k is the extent of the data in k-space and R the R range to be modelled.
The number of variables used for the analysis (9 at most) remained well below the number of
independent points (IP = 16). The values of Fj(k), j(k) and j(k) for Cu-O and Cu-Cu were
generated by the Atoms 2.5 and FEFF 8.0 codes using the crystallographic data of Cu2O and Cu
metal. The fits were performed in the R-space. The passive electron reduction factor S0
2
was
determined experimentally on the Cu2O and Cu foil as standard materials. During the fit procedure
of the samples the coordination number (N), bond distance (R) and Debye–Waller factor (2
) were
allowed to adjust freely. The passive electron reduction factor (S0
2
) was held constant at the value
of reference sample.
The metal content of fresh and reused catalyst was determined with ICP-MS (Agilent 7700x,
Agilent, Japan). A weighted amount of sample from the catalyst (on a 0.01 mg read-out balance,
Kern ABT 220-5DNM) was digested with nitric acid in microwave digester followed by dilution
with water. The mixture was centrifuged to precipitate solid residues, and the upper half of the
supernatant was used for Cu determination. 1
H and 13
C NMR spectra of imines produced from the
catalytic reactions were recorded on 400 MHz NMR Varian spectrometer (Varian, Santa Clara, CA,
USA) and on an JNM-ECA600II NMR spectrometer (JEOL, Japan) at 298 K, using CDCl3/DMSO-
d6 as a solvent and TMS as an internal standard. The raw data were processed with Masternova®
6.1 software. The chemical shifts were expressed in parts per million relative to TMS. The reaction
products were analyzed by gas chromatography (Model: Agilent 6820) equipped with an Agilent
DB-5 capillary column (30 m × 0.32 mm, 0.5 m) under the operation parameters: inlet temperature
473 K, flame ionization detector (FID) temperature 523 K, oven temperature 523 K with a ramp
rate 10° min–1
from 373 K.
Transmission 57
Fe Mössbauer spectrum of the G(COOH)Fe sample was recorded employing
a Mössbauer spectrometer operating in a transmission geometry and constant acceleration mode and
equipped with a 50 mCi 57
Co(Rh) radioactive source of γ-rays radiation. For the Mössbauer
spectroscopy measurement at 5 K, the sample was placed inside a cryomagnetic system (Oxford
Instruments, U.K.), to which a Mössbauer spectrometer is mounted. The collected 57
Fe Mössbauer
spectrum was analyzed with mathematical fitting algorithms and routines in the MossWinn software
package;[S3]
prior to fitting, the signal-to-noise ratio was adjusted by the filtering algorithms built in
the MossWinn software program and by statistically-based approach developed by Prochazka et
al.[S4]
The values of the isomer shift are referred to metallic α-Fe at room temperature.
Computational details. The structural features, stabilities, spin and charge distributions, and other
characteristics associated with the binding of benzylic amines to copper atoms, which are bound to

5. 3
the G-CN surface, were analyzed. To this end, we selected two representative amines, namely
benzylamine (BAM) and 2-picolylamine (PAM), for our computational studies. The former exhibits
high conversion rates and selectivity in the studied G(CN)-Cu catalyzed cross-oxidative coupling
reactions, but the conversion and selectivity associated with the latter are noticeably lower. In
addition to the formation of G(CN)-Cu-BAM/PAM complexes, we have also explored the binding
of Cu(II)/Cu(I) cations to the G-CN surface. The reaction sites (copper atoms) are positively
charged. Therefore, we applied the finite cluster approach, where the size of the G(CN)-Cu model
systems was carefully chosen to obtain converged values of the relevant binding characteristics (see
below).
The ground state (GS) structures of all the investigated species were optimized via the B3LYP
method[S5,S6]
using the 6-31+G(d) basis set.[S7]
The spin-unrestricted formalism was applied for the
open-shell system. Furthermore, the suitability of this approach was verified by comparing the
results for selected structures with those obtained by the spin-restricted open-shell formalism. The
solvent effects were included by using the universal continuum solvation model based on solute
electron density (SMD).[S8]
The structures of ligands (H2O, BAM, PAM, O2, OH) were fully relaxed in the geometry
optimizations, to mimic the semi-local rigidity of graphene sheets. However, the G(CN)-Cu-
BAM/PAM structures were obtained via geometry optimizations of the local region containing the
copper atom(s) with ligands, cyano groups, and the two closest carbon atoms at the reaction site.
During these optimizations, the rest of the structure was frozen (see Figure M1a). The structure of
corronene-2CN (cor-2CN) was generated through a fully unconstrained optimization in a given
solvent. In contrast, the cyc14-2CN structure was pre-optimized in the gas phase and then re-
optimized in a given solvent with fixed positions of edge carbon atoms (see Figure M1b). All
calculations were performed with the Gaussian09 program.[S6,S9]
Figure M1. The structure of a) cor-2CN-Cu model. The red arrows indicate the carbon atoms that
are included (in addition to both cyano groups and the copper atom) in constrained optimizations in
our analysis, b) cyc14-2CN model, and c) cyc10-4CN-2Cu model.
Representative procedure for homo-oxidative coupling. In the typical experimental procedure, a
10 mL vial was charged with benzyl amines (1.83 mmol) and G(CN)-Cu catalyst (20 mg, 10.7 µmol
Cu, or 0.58% mol loading in reaction) and the resulting reaction mixture was sonicated for 5 min.
The ensuing reaction mixture was stirred at 358 K for 12 h under an air balloon, thereby yielding
the desired N-Benzyl-1-phenylmethanimine product. The progress of the reaction was monitored
via thin layer chromatography (TLC). After completion of the reaction, 5 mL of ethyl acetate was
added to the reaction mixture and the catalyst was removed by centrifugation. A crude product was
obtained by vacuum-drying the resulting supernatant. The conversion (94%) and selectivity (99%)
were determined via gas chromatography (GC) with flame-ionization detection (GC-FID).
Moreover, the imine product was further purified by column chromatography, thereby yielding 2a
in 91%. 1
H NMR (400 MHz, CDCl3): δ = 8.36 (t, J = 1.4 Hz, 1H), 7.78–7.76 (m, 2H), 7.41–7.37 (m,
3H), 7.33 (d, J = 4.4 Hz, 4H), 7.27–7.21 (m, 1H) 4.80 (d, J = 1.4 Hz 2H) ppm; 13
C NMR (100.5
MHz, CDCl3): δ = 162.05, 139.30, 136.14, 130.81, 128.64, 128.53, 128.31, 128.01, 127.03, 65.08
ppm. Further corresponding imines are synthesized following the above procedure. The compounds
are all reported in the literature.[S10–S17]

6. 4
Figure M2. 1
H NMR of N-Benzyl-1-phenylmethanimine in CDCl3 using TMS as internal standard.
Figure M3. 13
C NMR of N-Benzyl-1-phenylmethanimine in CDCl3 using TMS as internal standard.

7. 5
Representative procedure for oxidative cross-coupling. In the typical experimental procedure, a
10 mL vial was charged with benzyl amine (1.83 mmol) and aniline (5.49 mmol) in a 1:3 mol ratio,
and G(CN)-Cu catalyst (20 mg, 10.7 µmol Cu) was subsequently added to the vial. The resulting
reaction mixture was sonicated for 5 min and stirred at 358 K for 24-30 h under an air balloon, to
obtain the desired cross-coupled imine product. The progress of the reaction was monitored via
TLC. After completion of the reaction, 5 mL of ethyl acetate was added to the reaction mixture and
the catalyst was removed by centrifugation. The resulting supernatant was vacuum-dried, thereby
yielding (N-(4-Fluorophenyl)-1-(4-methoxyphenyl)methanimine (99% conversion and 98%
selectivity, by GC-FID). Further purification of the imine product was performed by column
chromatography that yielded 5a in 94%. 1
H NMR (400 MHz, CDCl3): δ = 8.35 (s, 1H), 7.84–7.81
(m, 2H), 7.18–7.13 (m, 2H), 7.08–7.03 (m, 2H), 6.99–6.95 (m, 2H), 3.85 (s, 3H) ppm; 13
C NMR
(100.5 MHz, CDCl3): δ = 162.24, 162.17, 159.50, 148.23, 130.46, 129.04, 122.17, 115.90, 114.18,
55.40 ppm. Other corresponding imines are synthesized following the above procedure. The
compounds are all reported in the literature.[S10–S17]
Figure M4. 1
H NMR of N-(4-Fluorophenyl)-1-(4-methoxyphenyl)methanimine in CDCl3 using
TMS as internal standard.

8. 6
Figure M5. 13
C NMR of N-(4-Fluorophenyl)-1-(4-methoxyphenyl)methanimine in CDCl3 using
TMS as internal standard.
Representative procedure for benzylic C-H oxidation. The oxidation reaction was carried out in
a 10 mL Schlenk tube. In a typical experimental procedure, tube was charged with ethyl benzene
(0.5 mmol), N-hydroxyphthalimide (NHPI) 15 mol%, 10 mg of G(CN)-Cu in 3 mL of acetonitrile
under O2 1 atm. This reaction mixture was heated at 60 °C for 24 h. Aliquots of the mixture were
extracted and checked by means of GC-FID (97% conversion and 99% selectivity). The
aforementioned synthesis procedure was followed for other benzylic C-H oxidation of hydrocarbon
derivatives.

9. 7
Results - Supplementary Figures and Tables
Table S1. Oxidative coupling of amines, a comparison of performance: mixed-valence G(CN)-Cu
catalyst vs. state-of-the-art catalysts.
Entry
Catalyst
(Qty)
Substrates
Time
(h)
Temp
. (K)
Amine
(mmol
)
Conv./
Yield
(%)
Select.
(%)
Conv./Select.
after
recycling
1
CuO
nano-
flakes
(20 mg)
12 373 1.83 98 98
~42/- after
five cycles
ref. 34
Comments: Poor reusability, relatively high temperature, oxygen balloon, high
metal loadings.
2
Cu(0)
(0.05
mol)
20 363 9.3 88 100
N/A
ref. [S18]
Comments: Non-reusable (due to oxidative corrosion), relatively high temperature,
poor conversions/selectivity, relatively long reaction time, poor conversion for
cross-coupling products.
3
Graphite
oxide
(50
wt.%)
4 373 5 99 98
92/- after five
cycles
ref. 39
4 373 5 - -
Comments: High catalyst loading, high O2 pressure (5 atm), high temperature, poor
conversion and selectivity for aliphatic amines, scalable production of GO is tedious
and requires hazardous oxidizing agents and chemicals.
4
(ba-GO)
(5 wt.%)
12 363 9.3 98 - 93/- after six
cycles
ref. 37
Comments: Scalable production of GO is tedious and requires hazardous oxidizing
agents and chemicals, non-active for alkylamines, applicability of catalyst to cross
coupling of amines is unknown.

10. 8
Entry
Catalyst
(Qty)
Substrates
Time
(h)
Temp
. (K)
Amine
(mmol
)
Conv./
Yield
(%)
Select.
(%)
Conv./Select.
after
recycling
Ref.
5
P-doped
graphene
(10
wt.%)
12 373 0.4 82 -
73/- after 6
cycles
ref. [S19]
12 373 0.4 <10 -
Comments: High reaction temperature, poor selectivity, poor conversion for
aliphatic amines, inapplicable to cross couplings, catalyst synthesis requires a
template, high temperature, and acid washing.
6
Au/graph
ite
(5 mol%
Au)
17 383 0.4 69 -
80/- after 10
cycles
ref. [S20]
Comments: Long reaction time, high temperature, high amount of noble metal
required resulting in expensive protocol, very low conversion, solvent was used.
Only with O2 the performance could be improved.
7
NHPI/Fe
(BTC)
MOFs
(75 mg)
24 373
4.57
(0.5
mL)
98 90 96/91 after 2
cycles
ref. [S21]
10 373
4.57
(0.5
mL)
- -
Comments: Increased reaction time, high temperature, relatively large amount of
catalyst for small amount of amine, O2 balloon, tedious synthesis, narrow range of
substrates, relatively low selectivity, catalyst ineffective for coupling of alkyl
amines, and cross-coupling reactions.
8
MOF-
253
(0.15
mmol)
6 373 10 >99 -
98/- after 6
cycles
ref. [S14]
Comments: High reaction temperature, O2 balloon, tedious synthetic procedure for
MOF by using various expensive organic ligands.
9
CoTPP(C
F3)4
(5 × 10−4
mmol)
3 403 5 87 -
N/A
ref. 36
Comments: High reaction temperature, homogeneous catalyst, low conversion,
high pressure of O2 (6 atm), catalyst explored for self-coupling only.

11. 9
Entry
Catalyst
(Qty)
Substrates
Time
(h)
Temp
. (K)
Amine
(mmol
)
Conv./
Yield
(%)
Select.
(%)
Conv./Select.
after
recycling
Ref.
10
PdCu
NPs
(10 mg)
3 383 1.83 87 87.5
85.2/86.3
after 4 cycles
ref. [S22]
Comments: High reaction temperature, low conversion, precious metal and O2
balloon.
11
meso
Cs/MnOx
(25 mg)
3 383 0.5 99 93 2.2 (TON)
after four
recycling
ref. 35
+
(1:3)
3 383
0.5:1.5
(1:3)
15 70
Comments: High reaction temperature, relatively high catalyst loading, toluene
solvent.
12
Au-
Pd@CN
T
(0.10 g)
(Au:Pd,
1:1 mol
ratio)
4 393 1.0 95 98
N/A
ref. [S23]
Comments: High reaction temperature, use of noble metal, scalable synthesis of
CNTs is difficult, O2 balloon, p-xylene solvent.
13
MnOx/Ce
O2
(100 mg)
5 393 3.5 98.4 95.4
82.3/97.2
after 3 cycles
ref. 33
Comments: Increased reaction temperature, relatively large amount of catalyst,
poor conversion and selectivity, was explored for self-coupling only, O2 balloon.
14
G(CN)-
Cu
(20 mg,
0.58%
molCu)
12 358 1.83 94 >99
89/98 after 5
cycles
this work
12 358 1.83 30 -
+
(1:3)
12 358
1.8:5.5
(1:3)
99 92
Comments: Low reaction temperature and time, high conversion and selectivity for
self- and cross-coupling reactions, low catalyst loading, improved reusability, and
catalytic performance after reuse; scalable production of catalyst, no need for an
oxygen-rich atmosphere.

12. 10
Figure S1. X-ray diffraction diagrams from i) pristine fluorographene (FG, the precursor of G-CN),
ii) cyanographene (G-CN) and iii) the copper-loaded cyanographene (G(CN)-Cu). X-Ray radiation
used: Co K-alpha (1.789 Å). Comments: The catalyst powder (as well as the rest of the solids)
lacks completely reflections from any inorganic nanoparticles. Only the very broad reflection at
30.4 degrees is present, typical of non-restacked graphene powders and present in carbon materials.
The extent of the parallel stacking of graphenes (thickness of the crystallites, Lc) can be estimated
from the Scherrer formula.[S24, S25]
In the case of the G(CN)-Cu is 1.8 nm, which confirms that the
catalyst is composed of few layer graphene stacks, which are in a disordered state between each
other.
Figure S2. High-angle annular dark field scanning transmission electron microscopy images
showing heavier (and thus brighter) single metal atoms embedded in the G˗CN support.

13. 11
Figure S3. a) First derivative of the normalized X-ray absorption edge structure (XANES)
spectra of a G(CN)-Cu sample and relevant standard materials. Inset: the associated linear
combination analysis of the G(CN)-Cu spectrum, based on the spectra of Cu(H2O)6
2+
and Cu2O
standards (i.e. Cu(I) and Cu(II) species), confirming the copresence of Cu(I) and Cu(II). The
quality of the fit was compromized becasue aquaeous solutions of CuCl undergo spontaneous
oxidation to Cu(II), making it impossible to prepare a suitable atomically diapsersed Cu(I)
standrad. b) Fourier-Transformed k3
-weighted extended X-ray absorption fine structure
(EXAFS) spectra of G(CN)-Cu sample and relevant standard materials. Comments: The
XANES spectrum of G(CN)-Cu (panel a) resembles that of Cu(H2O)6
2+
more closely than those
of Cu2O and CuO, indicating that the Cu species in G(CN)-Cu are atomically dispersed. The
EXAFS spectra in panel b further confirm the presence of atomically dispersed Cu species in
the catalyst: peaks at the Cu-Cu distance corresponding to the 2nd coordination shell are visible
in the Fourier-transformed EXAFS signals of Cu2O and CuO, but not in those of G(CN)-Cu and
Cu(H2O)6.

14. 12
Figure S4. HR-XPS analysis of a) the Cu 2p3/2 core-level spectra of the two copper/graphene
systems (G-CN and reduced GO after interaction with CuCl2) b) the C 1s core level spectra of the
same graphene/copper systems. Reduced GO/Cu was prepared using the same procedure as G(CN)-
Cu, but with reduced graphene oxide (rGO) instead of G-CN. HR-XPS analysis revealed that Cu in
rGO occurs mainly in the Cu(II) valence state, as indicated by the intense satellite peaks, which are
characteristic of divalent copper ions. This experiment also demonstrates that (i) the cyano group of
the G(CN)-Cu catalyst is crucial for the charge transfer and stabilization of the mixed valence
Cu(I)/Cu(II) system, and (ii) the presence of an electron-rich 2D graphene skeleton alone is
insufficient for Cu(II) reduction (the very high content of sp2
carbon centers in the rGO is
demonstrated by the strong peak at 284.7 eV,shown in Figure S4b). This control sample (rGO-Cu)
confirms that the presence of Cu(I) ions should be attributed to specific interactions with the G-CN
support rather than reduction by the XPS electron beam.[S26]
Figure S5. EPR spectra of the G(CN)-Cu catalyst dispersed in hexane before (blue line) and after
(red line) adding hydrogen peroxide (30 %, 30 μL) directly into the EPR tube. An increase in the
Cu(II)-induced spectral signal was observed. We thus conclude that Cu(I) cations were indeed
present in the as-prepard catalyst, and they became EPR detectable after oxidation. Therefore, their

15. 13
presence was not a product of interaction with the XPS beam. Experimental parameters: Frequency
9.168 GHz, mod. Freq. 100 KHz, mod. amplitude of 0.8 mT, time const. of 30 ms, applied
microwave power of 0.6 mW, sweep time of 240 s, phase 0 deg, T = 143 K.
Comments on the analysis of the HR-XPS N1s envelope. Two different batches of the
starting G˗CN and of the catalyst (G(CN)-Cu) were evaluated, with identical results (Figure S6a).
Deconvolution clearly showed that after the binding of Cu ions, the ratio of the areas of the two
main N1/N2 components of G-CN changed significantly, from 1.3 to 1.7. That is, the lower binding
energy N1 component was enriched after coordination of the Cu(II) ions. This was surprising
because the Cu dication should attract electron density from its ligands. For instance, both
experimental and computational studies have shown that nitrogen atoms in N-doped graphene shift
to higher binding energies upon complexation with Co2+
. [S27]
The increased area of the lower
binding energy component after coordination was thus a clear indication of charge-transfer toward
the -C≡N groups in this case. The shift in the binding energy also confirms the interactions (i.e. the
formation of coordinative bonds) between the nitrile groups and Cu cations.
The two main components in the N1s XPS spectra can be ascribed to the attachment of
nitrile groups in different local environments (i.e. environments with and without nearby defects in
the graphene skeleton). This is also clearly reflected in the IR band of the nitrile groups, which is
asymmetric and was fitted with two major C≡N components (Figure S6b). The observation of
nitrile groups in multiple local environments is also consistent with the fact that not all copper
atoms are reduced upon interaction with cyanographene.
Figure S6. a) Deconvoluted HR-XPS N1s spectra for two different batches of G-CN and G(CN)˗Cu
samples. Circles represent experimental data and solid lines the fitting results. b) IR spectrum of
G˗CN showing the asymmetric nitrile band at 2200 cm-1
. The inset shows this asymmetric band
after deconvolution into three components, supporting the hypothesis that the material contains
nitrile groups in distinct local environments.

16. 14
Table S2. The binding of Cu(II) cations to G-CN in various solvents. The structures, selected bond
lengths (Å), binding energies (kcal mol–1
), spin density plots (contour value: 0.001), atomic spin,
and natural charge densities (a.u.) on the copper atom of model R-CN-Cu(II) systems were
computed at the U-B3LYP/6-31+G(d)/SMD level of theory. The structures were obtained via
constrained geometry optimizations at the same level of theory (see text for details on the
constraints).
Model ACN-Cu(II) Corronene-2CN-Cu(II) Cyc14-2CN-Cu(II)
Structure
RC-N/RN-Cu
Water 1.157/2.108 1.159/1.867 1.159/1.863
PAM 1.156/2.090 1.157/1.871 1.158/1.872
BAM 1.157/2.071 1.157/1.873 1.158/1.875
Binding energy
Water –8.1 –29.7 (–28.7)a
–28.0
PAM –17.7 –56.5 (–55.7) –59.5
BAM –28.7 –87.6 (–86.8) –92.8
Spin density
Spin density on Cu
Water 1.00 0.01 0.00
PAM 0.72 0.00 0.00
BAM 0.68 0.00 0.00
Charge density on Cu
Water 1.96 (1.68)b
1.04 (0.74) 1.03 (0.77)
PAM 1.67 (1.31) 1.02 (0.72) 1.01 (0.78)
BAM 1.63 (1.28) 1.01 (0.71) 1.00 (0.74)
a
The values in parentheses were obtained from the RO-B3LYP calculations.
b
The values in parentheses were determined via Mulliken population analysis.
Comments:
1. The similarity of the main structural characteristics, binding energies as well as spin and charge
densities of the two model G(CN)-Cu(II) systems justifies the use of a smaller model
(corronene-2CN-Cu(II)) for further analysis.
2. The spin unrestricted (U-B3LYP) and restricted-open shell (RO-B3LYP) formalisms yielded
similar binding energies for corronene-CN-Cu(II). This similarity and the low spin
contamination of the U-B3LYP ground state density (S2
< 0.751) confirmed that the U-B3LYP
approach is suitable for studying the open-shell states of G(CN)-Cu species.
3. Owing to the possibility of charge transfer, the binding of Cu(II) ions to the CN group in
G(CN)-Cu(II) systems differed considerably from and was significantly stronger than the
binding in ACN-Cu(II). This significantly stronger binding was accompanied by a significant

17. 15
decrease in the charge and spin densities on the copper atom. The charge transfer from
cyanographene to Cu(II) can thus facilitate the reduction of the copper oxidation state.
4. The solvent polarity had a significant effect on the stability of the G(CN)-Cu(II) systems.
Namely, less polar solvents were more effective in stabilizing the complex compared with more
polar solvents, where the solvation energy of Cu(II) ions is larger.
Similar C≡N bond lengths occurred in all R-CN-Cu(II) species and the bond lengths were
practically independent of the solvent polarity. The RCN of R-CN-Cu(II) species was only slightly
smaller than the C≡N bond in ACN (1.165 Å) indicating that the triple bond character was
preserved in R-CN-Cu(II). We collected the FT-IR spectra of the G(CN)-Cu catalyst focusing on
the very characteristic band of the nitrile (cyano) groups (occurring at 2200 cm–1
, as previously
reported30
), and no shift of the band frequency was observed. This concurred with the DFT
calculations, where no changes in the C≡N bond length were observed, since the charge transfer
originated from the aromatic graphene lattice (rather than from the electrons of the nitrile bond).
Table S3. Optimization studies and the effect of various reaction parameters on amine to imine
conversion using the G(CN)-Cu catalyst.a
Entry Catalyst Time (h) Temp (K) Conv. (%)b
Select. (%) TOF (h–1
)
1
2
3
4
5
6
7
8
9
10
11
-
G(CN)-Cu
G(CN)-Cu
G(CN)-Cu
G(CN)-Cu
G(CN)-Cu
G(CN)-Cu
G(CN)-Cu
CuO
rGO-Cu
GO
24
12
12
12
12
12
12
18
18
12
18
358
358
358
358
RT
348
373
358
358
358
358
-
96c
94
-d
-
78
99
97
64h
48h
13
-
99
>99
-
-
99
98
>99
97
94
92
-
14
14
-/-
-/-
11
14
9
6
6
1
a
Reaction conditions: 20 mg catalyst (10.7 µmol Cu), benzylamine 0.2 mL (1.83 mmol), air balloon (1 atm);
b
determined via GC; c
under O2 atmosphere (1 atm); d
nitrogen balloon-20 mg; RT: room temperature.
Comments:
1. No reaction occurred in the absence of the catalyst (entry 1). However, a 96% conversion (12 h
at 358 K under O2, >99% selectivity) was achieved for the imine (entry 2) when the G(CN)-Cu
catalyst was included. Approximately the same conversion (i.e., 94%) was attained, when the
reaction was conducted under air (entry 3, highlighted). This indicated the effectiveness of the
catalyst under low-oxygen conditions and its value for increased safety and cost-effectiveness
of the process. No reaction product was observed under nitrogen, thereby confirming that the
reaction requires O2 as an oxidant for the oxidative dehydrogenation of the amines (entry 4).
2. The robustness of the G(CN)-Cu catalyst was verified by examining various catalysts under the
same conditions. Pristine G˗CN and CuO yielded 25% and 64% conversions, respectively

18. 16
(entry 9). Furthermore, the importance of the nitrile groups of G-CN and the mixed valence
character was revealed by testing rGO-Cu that contained Cu(II) ions mainly (rGO-Cu used
previously for XPS comparisons with G(CN)-Cu). The results revealed a conversion of only
48% for rGO-Cu (entry 10). Pure GO was subsequently tested with conversion of 13% (entry
11).
3. TON = moles of desired product (Nz)/(moles of catalyst (Nc) or moles of active sites)
Conversion (C%) = moles of all product (Np × 100)/moles of reactant (Nr), or Np = (C%) × Nr/100
Selectivity (S%) = [moles of desired product (Nz) × 100/moles of all product (Np)]
or Nz = (S%) × Np/100, or Nz = (S%) × [(C%) × Nr/100]/100 = (S%) × (C%) × Nr/104
Turnover number (TON) = Nz/Nc = [(S%) × (C%) × Nr/104
]/Nc
Turnover frequency (TOF) = rate of product formation over catalyst = TON/time
Example: Table S3, entry 11:
TON = [92 × 13 × 1.83 (mmol)/104
]/10.7 × 10–3
(mmol) = 20.4
TOF = 20.4518 = 1.1 h–1
Table S4. Optimization of benzylamine and aniline molar ratio for the generation of asymmetrical
imine products using the G(CN)-Cu catalyst.a
Entry Benzylamine:Aniline
(mole ratio)
Conv. (%)b
Product 1a Select.
(%)
Product 2a Select. (%)
1
2
3
4
5
3:1
2:1
1:1
1:2
1:3
97
>99
>99
>99
>99
83
72
44
12
8
17
28
56
88
92
a
Reaction conditions: Catalyst 10.7 µmol Cu, benzylamine 1.83 mmol, and aniline 5.49 mmol (for 1:3 ratio), air balloon
(1 atm); 24–30 h; b
determined via GC.
Table S5. Elemental chemical composition (at.%) as determined from HR-XPS of the fresh and the
recycled G(CN)-Cu catalyst.
Times used C 1s N 1s O 1s Cu 2p
Fresh
G(CN)-Cu
77.7 14.1 7.6 0.7
After 5th
recycle
76.8 15.5 6.8 0.9a
a
The higher content of Cu in the fifth cycle is due to the measurement uncesrtainty of XPS.

19. 17
Figure S7. HR-XPS analysis. G(CN)-Cu catalyst after the 1st (left panels) and the 5th recycling
(right panels) steps. Deconvolution in the a),b) Cu 2p region, c,d) C 1s region, and e),f) N 1s region.

20. 18
Table S6. The stability of G(CN)-Cu2+
aqua complexes in various solvents. The structures, selected
bond lengths (Å), and complex formation energies (kcal mol–1
) of model G(CN)-Cu2+
aqua
complexes in selected solvents were computed at the B3LYP/6-31+G(d)/SMD level of theory. The
structures were obtained via constrained geometry optimizations at the same level of theory (see
text for details on the constraints).
Model cor-2CN-Cu
2+
/1w cor-2CN-Cu
2+
/2w-Y cor-2CN-Cu
2+
/3w-Tg
Structure
Description
One water molecule bound
at the axial position
Two water molecules in a Y
position
Three water molecules
at the Tg coordination
RN-Cu/RCu-O
Water 1.851/1.961 unstable 2.008/2.002 (2.004, 2.004)
PAM 1.864/1.956 unstable unstable
BAM 1.866/1.965 1.860/2.004 (2.358) unstable
Binding energy
Water -15.1 NA -27.1
PAM -13.4 NA NA
BAM -17.5 -23.3 NA
Model cor-2CN-Cu
2+
/3w-Th cor-2CN-Cu
2+
/4w-tbpy cor-2CN-Cu
2+
/5w-Oh
Structure
Description
Three water molecules
at the Th coordination
Four water molecules at the
tbpy coordination
Five water molecules
at the Oh coordination (non-
equivalent equatorial bonds)
RN-Cu/RCu-O
Water unstable unstable 2.045/2.040 (2.04, 2.50)
PAM
1.873/2.191 (2.197,
2.179)
unstable 1.965/2.027 (2.11, 2.34)
BAM
1.873/2.218 (2.204,
2.185)
unstable unstable
Binding energy
water NA NA -27.6
PAM -19.4 NA -26.8
BAM -28.6 NA NA
Comments:
1. The relative stability of G(CN)-Cu2+
aqua complexes with a given coordination number largely
depends on the solvent. In a water environment, the mono- and three-aqua (Tg) have been
identified as the stable structures. The Oh structure exhibits two types of equatorial bonds (in
trans position), which correspond possibly to the Tg structure.

21. 19
2. In BAM and PAM environments, the three-aqua (Th) complex occurs as a stable structure (in
addition to the 1w-axial structure).
Table S7. The binding of benzylic amines to G(CN)-Cu(II) with the amine also acting as a solvent.
The structures, selected bond lengths (Å), complex formation energies (kcal mol–1
), spin density
plots (contour value: 0.001), atomic spin, and charge densities (a.u.) on the copper atom of model
G(CN)-Cu(II)/amine complexes in the corresponding amine solvent were computed at the
B3LYP/6-31+G(d)/SMD level of theory. The structures were obtained via constrained geometry
optimizations at the same level of theory (see text for details on the constraints).
Model
cor-2CN-Cu(II)/
BAM/bare-axial
cor-2CN-Cu(II)/
PAM/bare-Y
cor-2CN-Cu(II)/
PAM/bare-axial
Structure
Description
One BAM molecule
bound
at the axial position
One PAM molecule
bound
at the Y position
One PAM molecule bound
at the axial position
RN-Cu/RCu-N(am)
BAM/PAM 1.861/1.949 1.860/2.032 (2.084) 1.848/1.932
RC-
N(Am)/RN(am)-H
BAM/PAM 1.504/1.024 1.482/1.021 1.494/1.022
Binding energy
BAM/PAM –36.7 (–36.9)a
3.1 11.6
Spin density
plot
Spin density on the copper atom
BAM/PAM 0.00 0.22 0.01
Charge density on the copper atom
BAM/PAM 0.39 0.38 0.36
a
The value obtained when the cyc14-2CN-Cu(II) model is applied.
Comments:
1. Although a PAM molecule can form two bonds with the copper atom (in the Y-like
configuration), its binding is considerably weaker than that of BAM.

22. 20
2. The binding of BAM on G(CN)-Cu(II) leads to the prolongation (i.e., weakening) of C—N and
N—H bonds compared to non-bound BAM (RC-N(am) = 1.475 Å and RN(am)-H = 1.020 Å). This
prolongation increases the bond susceptibility to other reactive agents.
3. The spin and charge density analysis results concur with the findings obtained for the G(CN)-
Cu(II) systems, i.e., the charge of the copper atom decreases significantly and an unpaired
electron is delocalized over the G-CN surface.

23. 21
Table S8. Structures, selected bond lengths (Å), and complex formation energies (kcal mol–1
) of
model G(CN)-Cu(I)/amine complexes in the corresponding amine solvent were computed at the
B3LYP/6-31+G(d)/SMD level of theory. The structures were obtained via constrained geometry
optimizations at the same level of theory (see text for details on the constraints).
Model
cor-2CN-Cu(I)/
BAM/bare-axial
cor-2CN-Cu(I)/
PAM/bare-Y
Structure
Description
One BAM molecule
bound
at the axial position
One PAM molecule
bound
at the Y position
RN-Cu/RCu-N(am)
BAM/PAM 1.858/1.949 1.860/2.055 (2.086)
Binding energy
BAM/PAM –35.8 –2.3
Comments:
1. As in the case of G(CN)-Cu(II), a benzylamine molecule attacks the axial position of the
copper atom bound to the G˗CN surface in a bare fashion, i.e., the simultaneous binding of
water molecules from a side is unfavorable. The binding parameters (relevant bond lengths and
the complexation energy) are almost identical to those of the G(CN)-Cu(II)/BAM complex.
This indicates that the Cu…BAM binding is mainly determined by the positive charge on the
copper atom, which is similar in both cases.
2. The binding of a PAM molecule is considerably weaker than that of BAM.
Catalytic mechanism governing the oxidative dehydrogenation of amines
A theoretical study of the catalytic mechanism governing the oxidative dehydrogenation of amines
was performed at the B3LYP/6-31+G(d)/SMD level using two model systems representing single-
center (Figure M1a) and double-center (Figure M1c) sites on the G(CN)-Cu catalyst. The former
was used for modeling reaction steps involving a single Cu(I)/Cu(II) center (e.g., oxygen activation
and formation of hydroperoxyl radical and release of water, see step 1-I, 2I-2II, in Figure S8).
However, the latter was applied for processes involving two neighboring Cu centers (e.g., formation
of a cyclic intermediate and subsequent hydrogen abstraction, see step 3˗I in Figure S8).
Benzylamine (BAM) was selected as a representative amine, participating in the reaction, as well as
the solvent. The structures playing key roles in the mechanism elaborated in Figures S8–S11 are
displayed in Figure S12.

24. 22
Figure S8. Initial steps of the catalytic mechanism. In step 1-I oxygen binding and activation takes
place. Calculations suggest that this configuration requires a small amount of energy (ca. 20 kcal
mol–1
), in close agreement to the findings on oxygen binding in copper enzymes.17
Calculations also
indicated partial charge-transfer from the G-CN-Cu(I)-amine system (ca. 0.34 e, Figure S12,
structure 1-I) and increase of the nucleophilic character of O-O, thus promoting the first homolytic
hydrogen abstraction from the co-coordinated amine. In step 2-I, the formed hydroperoxo-copper
complex can mediate a second homolytic hydrogen atom abstraction, leading to formation and
release of a water molecule (step 2-II), of the imine and a copper-oxyl complex (step 3-I). The oxyl-
intermediate withdraws one electron from the Cu(I) center, and then facilitates a proton abstraction,
which leaves behind two electrons, one of them reducing the Cu(II) center. The reaction energies
are given in kcal mol–1
; the estimated activation energies are reported in brackets.

25. 23
Figure S9. Oxidation steps involving the cyclic intermediate (3-II) that interacts with a hydrogen
atom on the BAM molecule, which is coordinated to the neighboring Cu center. After abstraction of
the proton, the reaction can proceed either through a single-step or a two-step release of a water
molecule. The former appears to be more favorable (than the latter) due to lower reaction energy
(kcal mol–1
); the estimated activation energies are reported in brackets.

26. 24
Figure S10. Exchange of ligands (amine and imine) on the a) Cu(II) and b) Cu(I) centers. The
reaction energies are given in kcal mol–1
; no activation energy is required for the processes. The low
magnitude of the binding and releasing energies (independent of copper oxidation state) suggests
easy ligand exchange. Nevertheless, the most stable structure (5-III) enables formation of the final
product (see Figure S11).

27. 25
Figure S11. Final steps of the mechanism. a) Reaction of an amine with imine in a concerted
fashion is thermodynamically favorable; although the activation barrier can be rather high (the
reported value represents an upper limit). b) Alternatively, an intermediate state from step (4) can
act as an active site for the reaction leading to a product isomer that transforms to an energetically
more stable product (E = –18.6 kcal mol–1
). In the last step, ammonia can be readily formed via
abstraction of hydrogen from the environment (not shown).

28. 26
Figure S12. The key structures occurring in the catalytic mechanism, as described in Figure S8–
S11. The NBO charges are displayed in selected structures.

29. 27
Table S9. Calculations for the estimation of the copper ions surface density.
Number of atoms in G(CN)-Cua
Cin graphene
b
Cin CN Nin CN
c
Cu
76.3 11.5 11.5 0.7
100 atoms in total
Mass of atoms (g)
Cin graphene Cin CN Nin CN Cu
1.52×10–21
2.29×10–22
2.68×10–22
7.1×10–23d
Total mass of atoms in G(CN)-Cu: 2.09×10–21
g
Surface area calculations
1 g of G(CN)-Cu occupies 153 m2
(the experimental value)
2.09×10–21
g of G(CN)-Cu occupies 3.19×10–19
m2
or 32 Å2
0.7 Cu atoms occupies 32 Å2
or 1 Cu atom occupies ~48 Å2
a
The ratio of the number of atoms was based on the present XPS results, which are in full agreement with the previously
reported findings in ref 28. 100 atoms were considered in total.
b
According to these XPS results, 15% of the carbon atoms in graphene lattice are functionalized with CN groups (i.e.,
15% functionalization degree).
c
The N-content used here is 11.5 (lower than the nominal N-content in the XPS results in Table S5), because, as we
have analyzed in detail in our previous work on the synthesis of G-CN,30
there is a small amount of background N,
coming from the solvent (DMF) during the synthesis of G-CN.
d
This amount equals 3.4 wt.% of Cu atoms in the catalyst, according to the ICP results.
Table S10. Control experiments on the oxidative cross-coupling of benzylamine with aniline.a
Entry Copper–based Catalysts Conv. (%)b
Product 1a
Select. (%) (homo)
Product 2a
Select. (%) (hetero)
1
2
3
4
5
CuCl
CuCl2
CuCl + CuCl2
GCN-Cu (0.5wt%)c
GCN-Cu (3.4wt%)
47
60
56
93
>99
93
100
92
62
8
7
0
8
38
92
a
Reaction conditions: Catalyst (10.7 µmol Cu), 1.83 mmol benzylamine, and 5.49 mmol aniline (giving a 1:3 ratio), air
balloon (1 atm); 24–30 h; b
determined via GC. c
The amount of catalyst used was adjusted to ensure that the total copper
content of the reaction mixture remained constant (3.4 wt%).
6.9Å
6.9Å

30. 28
Figure S13. Two-center amine coupling reaction mechanism: a comparison of benzylamine and
aniline focused on step 2 of the simplified mechanism presented in Fig. 4a. The numbers beneath
the arrows illustrate the energetics of cooperative proton abstraction from an amine by the
neighboring G-(CN)-Cu(II)-O–
complex. a) Proton abstraction from the α˗carbon of benzylamine
(BAM); b) Proton abstraction from the amino group of BAM; and c) Proton abstraction from the
amino group of aniline (AN). In the case of BAM, the C-H bond cleavage is energetically
preferable. For aniline, which lacks α-hydrogens, proton abstraction from N-H is also
thermodynamically accesible through the cyclic intermediate; this can be attributed to the
comparative weaker N-H bond in AN resulting from the mesomeric effect and the stabilization of
resulting radical by the neighboring aromatic ring. The energies (reported in kcal mol-1
) were
obtained at the B3LYP/6-31+G(d) level of theory using the SMD implicit solvent model (solvent =
benzylamine) with the ovalene platform kept frozen during geometry optimizations.
Comments on the importance of Cu(I)/Cu(II) cooperativity for high selectivity toward
crosscoupling. As reported previously (Table S1), BAM undergoes self˗coupling without requiring
cooperativity because the intermediate imine can be formed at a single Cu center (Figure S8, steps
1-I to 2-II). Imine formation is required for the coupling reaction with a second BAM molecule
(Figure S11, step 6-IV). However, AN cannot undergo hydrogen abstraction and activation as BAM
does. Although the first H-abstraction would be possible (Figure S8, step 1˗I), forming a more
nucleophilic N-centered radical anion (Figure S8 step 2-I), this step is energetically demanding
(+21.9 kcal mol-1
) and thus improbable for both BAM and AN. The following step is energetically
very favorable (-32.2 kcal mol-1
), rendering the overall process viable. However, AN, completely
lacking α˗hydrogens, cannot undergo this second step, making its activation improbable.
Consequently, the only possible pathway in a single metal-center reaction is the nucleophilic attack
of AN on the BAM-derived imine, as widely accepted.[S14,S28,S29]
This process gives low yields and
poor selectivity for the cross-coupled product[S21,35]
becasue of AN’s low nucleophilicity.
Nevertheless, the activation of AN through the cyclic two-center mechanism is energetically much
more favorable (Figure S13c). Selectivities comparable to those achieved with G(CN)-Cu have

31. 29
previously only been attained by combining high reaction temperatures with pure O2 atmosphere
and high catalyst reaction loadings.[S14,39]
Figure S14. Electron microscopy on G(COOH)-Fe. a) representative HR˗TEM of a graphene flake
from the G(COOH)-Fe verifying the absence of any nanoparticles. b) HR-TEM image of
G(COOH)-Fe showing several high-contrast isolated spots originating from single Fe atoms
entrapped on the G-COOH flake. c),d) EDS chemical maps on G(COOH)˗Fe for c) Fe and d)
combined C and Fe.

32. 30
Figure S15. 57
Fe Mössbauer spectrum of the G(COOH)-Fe sample, recorded at a temperature of 5
K in absence of external magnetic field. The asymmetric line-shape of the spectrum denotes the
presence of two different Fe components. Deconvolution of the spectrum suggested the presence of
Fe ions in different valence states (Fe(III) and Fe(II)), according to the isomer shift of the
components (see Table S10), which was ascribed to a partial charge transfer from graphene to Fe
centers, as in the case of the G(CN)˗Cu catalyst. In addition, the quadrupole splitting of the two Fe
species underlined the change of the their coordination environment in the G(COOH)-Fe sample, in
comparison to the starting Fe source (Fe(NO3)3·9H2O).

33. 31
Table S11. Values of the Mössbauer hyperfine parameters, derived from the least-square
fitting of the 57
Fe Mössbauer spectrum of the G(COOH)-Fe sample, collected at a temperature
of 5 K, where δ is the isomer shift, ΔEQ is the quadrupole splitting, and RA is the spectral area
of individual spectral components, identified upon spectrum fitting.
Component δ ± 0.01
(mm s–1
)
ΔEQ ± 0.01
(mm s–1
)
RA ± 1
(%)
Assignment
Doublet –0.02 0.43 12 Fe(II)
Doublet 0.56 0.77 88 Fe(III)
Supplementary References
[S1] V Briois, E. Fonda, S. Belin, L. Barthe, C. La Fontaine, F. Langlois, M. Ribbens, F.
Villain, Applications et Développements Récents; EDP Sciences: Les Ulis, France,
2011; 41–47.
[S2] B. Ravel, M. Newville, J. Synchrotron Radiat. 2005, 12, 537–541.
[S3] Z. Klencsár, E. Kuzmann, A. Vértes, J. Radioanal. Nucl. Chem. 1996, 210, 105.
[S4] R. Prochazka, P. Tucek, J. Tucek, J. Marek, M. Mashlan, J. Pechousek, Meas. Sci.
Technol. 2010, 21, 025107.
[S5] P. J. Stephens, F. J. Devlin, C. F. Chabalowski, M. J. Frisch, J. Phys. Chem. 1994, 98,
11623–11627.
[S6] A. D. Becke, J. Chem. Phys. 1993, 98, 5648–5652.
[S7] R. Ditchfield, W. J. Hehre, J. A. Pople, J. Chem. Phys. 1971, 54, 724–728.
[S8] A. V. Marenich, C. J. Cramer, D. G. Truhlar, J. Phys. Chem. B. 2009, 113, 6378–6396.
[S9] M. Frisch et al., Gaussian, Inc., Wallingford, CT, 2009.
[S10] R. D. Patil, S. Adimurthy, Adv. Synth. Catal. 2011, 353, 1695–1700.
[S11] E. Zhang, H. Tian, S. Xu, X. Yu, Q. Xu, Org. Lett. 2013, 15, 2704–2707.
[S12] A. E. Wendlandt, S. S. Stahl, Org. Lett. 2012, 14, 2850–2853.
[S13] J. L. García Ruano, J. Alemán, I. Alonso, A. Parra, V. Marcos, J. Aguirre, Chem. –
Eur. J. 2007, 13, 6179–6195.
[S14] X. Qiu, C. Len, R. Luque, Y. Li, ChemSusChem 2014, 7, 1684–1688.
[S15] Z. Hu, F. M. Kerton, Org. Biomol. Chem. 2012, 10, 1618–1624.
[S16] L. Liu, S. Zhang, X. Fu, C.-H. Yan, Chem. Commun. 2011, 47, 10148–10150.
[S17] F. Raza, J. H. Park, H.-R. Lee, H.-I. Kim, S.-J. Jeon, J.-H. Kim, ACS Catal. 2016, 6,
2754–2759.
[S18] R. D. Patil, S. Adimurthy, RSC Adv. 2012, 2, 5119–5122.
[S19] F. Yang, X. Fan, C. Wang, W. Yang, L. Hou, X. Xu, A. Feng, S. Dong, K. Chen, Y.
Wang, Y. Li, Carbon 2017, 121, 443–451.
[S20] M.-H. So, Y. Liu, C.-M. Ho, C.-M. Che, Chem. – Asian J. 2009, 4, 1551–1561.
[S21] A. Dhakshinamoorthy, M. Alvaro, H. Garcia, ChemCatChem 2010, 2, 1438–1443.
[S22] V. S. Marakatti, S. C. Sarma, B. Joseph, D. Banerjee, S. C. Peter, ACS Appl. Mater.
Interfaces 2017, 9, 3602.
[S23] W. Deng, J. Chen, J. Kang, Q. Zhang, Y. Wang, Chem. Commun. 2016, 52, 6805–
6808.
[S24] T. Kyotani, N. Sonobe, A. Tomita, Nature 1988, 331, 331.
[S25] F. Sun, X. Liu, H. B. Wu, L. Wang, J. Gao, H. Li, Y. Lu, Nano Lett. 2018, 18, 3368.
[S26] A. M. Salvi, F. Langerame, A. Macchia, M. P. Sammartino, M. L. Tabasso, Chem.
Cent. J. 2012, 6 Suppl 2, S10.

34. 32
[S27] K. Artyushkova, B. Kiefer, B. Halevi, A. Knop-Gericke, R. Schlogl, P. Atanassov,
Chem. Commun. 2013, 49, 2539.
[S28] A. Monopoli, P. Cotugno, F. Iannone, F. Ciminale, M. M. Dell’Anna, P. Mastrorilli, A.
Nacci, Eur. J. Org. Chem 2014, 2014, 5925.
[S29] B. Chen, L. Wang, S. Gao, ACS Catal. 2015, 5, 5851.