N-heterocyclic carbene-functionalized magic-number gold nanoclusters

Article abstract of DOI:10.1038/s41557-019-0246-5

Magic-number gold nanoclusters are atomically precise nanomaterials that have enabled unprecedented insight into structure–property relationships in nanoscience. Thiolates are the most common ligand, binding to the cluster via a staple motif in which only

Full text of DOI:10.1038/s41557-019-0246-5

Articles
https://doi.org/10.1038/s41557-019-0246-5
N-heterocyclic carbene-functionalized
magic-number gold nanoclusters
Mina R. Narouz1, Kimberly M. Osten2, Phillip J. Unsworthꢀ ꢀ1, Renee W. Y. Man2, Kirsi Salorinne2,
Shinjiro Takanoꢀ ꢀ3, Ryohei Tomihara3, Sami Kaappaꢀ ꢀ4, Sami Malola4, Cao-Thang Dinhꢀ ꢀ5,
J. Daniel Padmos1, Kennedy Ayoo1, Patrick J. Garrett1, Masakazu Nambo2, J. Hugh Horton1,
Edward H. Sargentꢀ ꢀ5, Hannu Häkkinenꢀ ꢀ4*, Tatsuya Tsukudaꢀ ꢀ3,6* and Cathleen M. Cruddenꢀ ꢀ1,2*
Magic-number gold nanoclusters are atomically precise nanomaterials that have enabled unprecedented insight into struc-
ture–property relationships in nanoscience. Thiolates are the most common ligand, binding to the cluster via a staple motif
in which only central gold atoms are in the metallic state. The lack of other strongly bound ligands for nanoclusters with dif-
ferent bonding modes has been a significant limitation in the field. Here, we report a previously unknown ligand for gold(0)
nanoclusters—N-heterocyclic carbenes (NHCs)—which feature a robust metal–carbon single bond and impart high stability to
the corresponding gold cluster. The addition of a single NHC to gold nanoclusters results in significantly improved stability and
catalytic properties in the electrocatalytic reduction of CO2. By varying the conditions, nature and number of equivalents of the
NHC, predominantly or exclusively monosubstituted NHC-functionalized clusters result. Clusters can also be obtained with up
to five NHCs, as a mixture of species.
agic-number gold nanoclusters stabilized by organic NHCs have, however been described as stabilizing ligands in small,
ligands are intriguing species that bridge the gap between mixed-valency clusters of three to four atoms in seminal works by
M
molecules and materials1–3. Although clearly nanomateri- the Corrigan34,35, Sadighi36 and Bertrand37 groups.
als, with size-dependent properties, they can be described with a
Here, we report the first example of NHC-stabilized Au(0) nano-
single molecular formula, have discrete electronic transitions due clusters. Au11 nanoclusters functionalized by up to five NHC ligands
to their well-defined molecular orbitals and can be characterized have been prepared and characterized. Remarkably, the addition of
by techniques usually employed in molecular science. Kornberg even one NHC to the Au11 cluster significantly increases the thermal
and colleagues’ determination of the precise structure of the clus- stability of this important nanocluster and improves its activity in
ter Au102(SR)44 (SR=p-mercaptobenzoic acid) using single-crystal the electrochemical reduction of CO2 to CO, which is a critically
X-ray diffraction (XRD) was a watershed moment in understanding important reaction in the valorization of CO2.
these nanomaterials4. The crystallographically determined struc-
ture partially confirmed earlier computational predictions on the Results and discussion
bonding at the gold–thiolate interface and provided strong support Development of cluster synthesis. Our studies began with pre-
for the ‘superatom’ theory of closed electronic shells, first described formed phosphine-stabilized undecagold clusters Au11(PPh3)7Cl3
by Häkkinen and colleagues to explain why certain cluster sizes are and [Au11(PPh3)8Cl2]Cl (ref. 38). Phosphine-stabilized clusters were
over-represented among known species, and hence ‘magic’2,5.
chosen because NHC and phosphine ligands are both neutral and
Since the seminal paper by Kornberg, research into magic- thus require no change in oxidation state or cluster charge to accom-
number nanoclusters featuring different thiolates, metals, core con- pany an exchange reaction. Furthermore, phosphine-stabilized
figurations and applications has abounded6–10. However, with few clusters do not have a layer of Au(i) atoms on the exterior, as is the
exceptions, thiols and phosphines remain virtually the only organic case for thiolate-stabilized clusters. Treatment of [Au11(PPh3)8Cl2]Cl
ligands used to stabilize these structures (other than some work (2) with di-isopropyl benzimidazolium hydrogen carbon-
on alkynes)11–17
.
ate 1a·H2CO3 (Fig. 1a) in THF at 66°C gave predominantly a single
N-heterocyclic carbenes (NHCs) are phosphine analogues that NHC-containing cluster (3a, Table 1, entry 1). This cluster results
have attracted considerable attention in organometallic chemistry from substitution of one phosphine ligand for the NHC (Fig. 1b).
and recently in surface science18–31. In organometallics, these ligands The less stable cluster Au11(PPh3)7Cl3 was not effective as a substitu-
are known for their ability to form strong, substitutionally inert tion partner.
bonds to various metals32,33. Recently they have been shown to form
Conclusive information about the cluster identity was obtained
highly robust self-assembled monolayers (SAMs) on Au(111) sur- by mass spectrometry, which demonstrated that mono-substi-
faces and nanoparticles20,30; however, their use as stabilizing ligands tuted cluster 3a is the predominant product, accompanied by
in magic-number nanoclusters has not been described previously. small amounts of di-substituted cluster 4a and starting material 2
1Department of Chemistry, Queen’s University, Kingston, Ontario, Canada. 2Institute of Transformative Bio-Molecules (WPI-ITbM), Nagoya University,
Nagoya, Japan. 3Department of Chemistry, School of Science, The University of Tokyo, Bunkyo-ku, Tokyo, Japan. 4Departments of Chemistry and Physics,
Nanoscience Center, University of Jyväskylä, Jyväskylä, Finland. 5Department of Electrical and Computer Engineering, The University of Toronto, Toronto,
Ontario, Canada. 6Elements Strategy Initiative for Catalysts and Batteries (ESICB), Kyoto University, Katsura, Kyoto, Japan.
*e-mail: hannu.j.hakkinen@jyu.fi; tsukuda@chem.s.u-tokyo.ac.jp; cruddenc@chem.queensu.ca
NATURE CHEMiSTRY | www.nature.com/naturechemistry

NaTure CHeMisTry
Articles
a
b
penta-substituted clusters 6a and 7a. Thus, the use of the bicarbon-
ate salt 1a·H2CO3 at ~1equiv. is important to produce clusters that
are predominantly mono-substituted.
[Au11(PPh3)7(NHC)Cl2]+
1x
–
3x
HCO3
[Au11(PPh3)8Cl2]Cl
THF
reflux
2 h
[Au11(PPh3)6(NHC)2Cl2]+
With conditions in hand for the preparation of NHC-stabilized
clusters 3a and 4a, we then examined the effect of the NHC struc-
ture (Table 1). NHC precursors 1b–d·H2CO3 were used, where
the nitrogen atoms of the NHC are substituted with Et, Me and
Bn groups. Each of these relatively minor changes in NHC struc-
ture had an influence on the reaction. NHC precursor 1b·H2CO3
(R=Et) consistently gave a mixture of mono- and di-substituted
clusters, regardless of the conditions or stoichiometry, suggest-
ing that the initial substitution activates the cluster for a second
displacement reaction (Fig. 1d and Table 1, entries 5—8). Again,
increasing the amount of the NHC gave greater substitution
(Supplementary Section 2).
2
N
N
R
R
4x
H
Pure mono-substituted
(3) or mixed mono- and
di-substituted (4) NHC clusters
1a·H2CO3, R = iPr
1b·H2CO3, R = Et
1c·H2CO3, R = Me
1d·H2CO3, R = Bn
c
3a
3a
3a
Obs.
Calcd.
NHC precursor 1c·H2CO3 (R=Me) behaved more like 1a·H2CO3
and allowed for the isolation of clean, mono-substituted cluster
3c with no evidence for formation of the di-substituted cluster 4c
under optimized conditions. On average, lower isolated yields were
observed in comparison to reactions with 1a·H2CO3 and 1b·H2CO3,
and the reaction was best run at lower temperatures (Table 1,
entries 9 and 10). The addition of larger quantities of this small
NHC gave cluster degradation rather than increased substitution
(Supplementary Sections 2 and 3).
Mass number (m/z)
4a
2
3,600
3,800
4,000
4,200
4,400
4,600
4,800
5,000
Mass number (m/z)
d
3b
4b
Obs.
3b
4b
4b
Calcd.
Finally, benzylated NHC precursor 1d·H2CO3 behaved similarly
to 1b·H2CO3, giving a mixture of mono-, di- and tri-substituted
clusters 3d, 4d and 5d even at 1.2equiv. of the precursor. At higher
loadings (9.0equiv.), a mixture of clusters including mono- (3d), di-
(4d), tri- (5d) and tetra- (6d) NHC-containing clusters was observed
(Table 1, entries 11 and 12 and Supplementary Figs. 45–48).
Mass number (m/z)
3,600
3,800
4,000
4,200
4,400
4,600
4,800
5,000
Mass number (m/z)
Fig. 1 | Synthesis and characterization of NHC-modified Au11 nanoclusters.
a, Structure of NHC precursors employed in this study featuring a variety
of organic substituents on the nitrogen atoms flanking the key carbon.
b, Reaction of undecagold cluster [Au11(PPh3)8Cl2]Cl (2) with NHC
precursors from a, resulting in formation of new NHC-containing clusters.
c, Mass spectrometric characterization of the reaction mixture leading to
[Au11(PPh3)7(NHCi-Pr)Cl2]Cl 3a, along with the structure obtained by X-ray
crystallography. Inset, observed versus calculated isotope pattern. The
high level of purity is notable from the mass spectrum, which shows 3a as
the major product, along with small quantities of starting cluster 2 and di-
substituted cluster [Au11(PPh3)6(NHCi-Pr)2Cl2]Cl 4a. Colours in the atomistic
structures: yellow, Au; purple, Cl; brown, P; blue, N, cyan, C; white, H.
d, Mass spectrometric characterization of the reaction of 1b with 2,
showing a mixture of 3b and 4b along with DFT-predicted structures. Note
that the structure of 3a shown in c is an actual crystal structure while 3b
and 4b are DFT-simulated.
Computational studies and structure prediction. Density func-
tional theory (DFT, implemented as described in ref. 39; for techni-
cal details see Supplementary Information) was used to investigate
the energetics of substitution of the various phosphines in cluster
2 by NHCs (Fig. 2a). Introduction of NHC (1a) into cluster 2 was
found to be thermodynamically favourable at most sites, with the
exception of phosphine 8 (P8), which gave an unfavourable reac-
tion energy change of +0.05eV, probably due to steric constraints
at this site (Fig. 2b). Introduction of the NHC by displacement of
phosphines adjacent to the Au–Cl sites was uniformly preferred,
with the single most favourable exchange predicted to occur at P2
for all NHCs examined. Differences between exchange at P2 and the
next favourable phosphine energetically were 0.18eV, 0.07eV and
0.06eV for 3a, 3b and 3c, respectively (Fig. 2b and Supplementary
Table 7). The calculated highest occupied molecular orbital–lowest
unoccupied molecular orbital gaps for the mono-substituted clus-
ters were largest for substitution at P2 for each NHC type, showing
(Fig. 1c). Quantification of the cluster purity was accomplished by that the trends in electronic stability followed exchange energetics
1
detailed H NMR spectroscopic analysis. For precise details and (Supplementary Table 8).
replicate runs, see Supplementary Section 2. NHC nanoclusters
For the reaction of NHC precursor 1b·H2CO3 with 2 to form
were analysed before and after purification by chromatography on di-substituted clusters, the four lowest-energy isomers were found
silica gel, which indicated that there was no change in the cluster within a 0.1eV window. Three of these included introduction of the
distribution after chromatography and highlighted the stability of NHC at P2 (Supplementary Table 9). NHC precursors 1a·H2CO3
the cluster to chromatography (Supplementary Fig. 33). The main and 1c·H2CO3 were also predicted to have favourable energetics for
effect of purification was to remove molecular by-products such as the production of di-substituted clusters. For example, substitution
[Au(NHC)2]+ and OPPh3 (Supplementary Fig. 12).
of 1a and 1c at P2 and P7 led to energy changes of −1.39eV and
Increasing the amount of 1a·H2CO3 to 5equiv. resulted in larger −1.43eV, respectively, suggesting that kinetic control rather than
amounts of the di-substituted cluster 4a, with the addition of water thermodynamics must play a role in the observation of predomi-
attenuating reactivity and giving a more selective process (Table 1, nantly or exclusively mono-substitution for these NHCs.
entries 2 and 3). The use of the free NHC in place of the bicarbonate
In the later stages of this work we were able to determine the
salt gave similar overall yields, but a mixture of clusters was obtained atomistic structure of 3a from a single-crystal X-ray analysis (vide
(entry 4). The mono-substituted cluster 3a still predominated, but infra). The crystal structure confirmed our DFT predictions that it
di-substituted cluster 4a was observed along with tri-substituted is energetically optimal to replace phosphine P2 with an NHC in
cluster 5a. ESI-MS (electrospray ionization mass spectrometry) cluster 3a (see the comparison of the predicted and observed struc-
revealed the presence of trace amounts of the tetra- and tures in Supplementary Fig. 52). Three computational structure
NATURE CHEMiSTRY | www.nature.com/naturechemistry

NaTure CHeMisTry
Articles
a
Table 1 | Preparation of NHC-containing clusters by reaction of
phosphine-containing cluster 2 with benzimidazolium hydrogen
carbonates 1a–d
N
N
1x
[Au11(PPh3)8Cl2]+
[Au11(PPh3)8-n(NHC)nCl2]+
1a
–PPh3
Conditions
Entry 1x·H2CO3
Additive
Cluster(s) produceda Yieldb
(%)
Ligand exchange
(equiv.)
1
1a (1.2)
1a (5.0)
1a (5.0)
1a (1.2)e
1b (1.2)
1b (1.2)
1b (5.0)
1b (5.0)
1c (1.2)
1c (1.2)
1d (1.2)
1d (9.0)
None
H2Od
None
None
H2O
3a,4a (9:1)
3a,4a (3:1)
3a,4a (1:1)
3a,4a,5a (2.6:1:1)
3b,4b (2:1)
3b,4b (2:1)
3b,4b (1:1)
3b,4b (1:1)
3c
22c
26c
38
24
41
[Au11(PPh3)7(NHCi-Pr)Cl2]+
[Au11(PPh3)8Cl2]+
2
3
b
c
3a (Exp.)
4
5
3a (DFT CS opt)
3a (DFT CS)
3a (DFT pred)
a
Phosphine
Energy (eV)
6
7
None
H2O
20
64
43
6c
P1
P2
P3
P4
P5
P6
P7
P8
–0.64
–0.93
–0.73
–0.55
–0.41
–0.40
–0.75
+0.05
8
9
10
11
12
None
H2O
H2Of
None
None
3c
10c
3d,4d,5dg
3d,4d,5d,6dg
18
aChanges in potential energy for
reaction of 2 with 3a by DFT.
38
aConditions: THF solvent, 66ꢀ°C, 2ꢀh. 3x clusters contain 1 NHC (nꢀ=ꢀ1), 4x (nꢀ=ꢀ2), 5x (nꢀ=ꢀ3),
6x (nꢀ=ꢀ4), ratios determined by 1H NMR spectroscopy, traces of starting cluster 2 present in
some cases. bIsolated yields. cAverage of 3–5 runs. dWater added at 500ꢀppm unless otherwise
noted. eFree carbene employed. f40ꢀ°C, 1,000ꢀppm H2O. gExact ratios difficult to elucidate by NMR.
Electrospray ionization mass spectrometry (ESI-MS) analysis shows decreasing amounts of more
substituted clusters.
300
400
500
Wavelength (nm)
Fig. 2 | DFT and UV–vis spectroscopic prediction of structure for NHC-
modified nanoclusters. a, Labelling of phosphines in [Au11(PPh3)8Cl2]Cl
(2) and exchange with 1a. b, DFT-predicted changes in potential energy for
reaction shown in a at each phosphine. c, Comparison of calculated optical
spectra (red, blue, green curves) of cluster 3a to experimental data (black).
Three computed spectra are calculated for the initial DFT-predicted
structure obtained by replacing phosphine P2 in cluster 2 by NHC 1a
(red), for the experimentally determined crystal structure (CS) (blue) and
for DFT-optimization of the experimentally determined crystal structure
(green). The computed spectra are blueshifted by 0.4ꢀeV for better visual
comparison to the experimental data.
models for 3a were now in our hands: one based upon the structure
predicted prior to obtaining an experimental crystal structure (red),
one employing exact coordinates from the experimentally obtained
crystal structure (blue) and one applying DFT optimization to the
experimentally derived crystal structure (green).
The computed optical absorption spectra from these models are
compared to the experimental data measured for 3a in Fig. 2c. We
notice that all computed spectra have spectral features that match
the experimental data very well; however, the computed spectra
systematically underestimate all transition energies by ~0.4eV.
This is understandable based on the properties of the DFT func-
tional used in the calculation. The shift is corrected in Fig. 2c to the NHC-functionalized cluster starts to show decoalescence at
enable a better visual comparison to experiment. The nature of the −10°C, while the all-phosphine cluster 2 does not begin to deco-
dominant two absorption features in the computed spectra (close to alesce until close to −70°C.
300nm and 450nm in the shifted spectra) are further analysed in
Information on the bonding and electronic properties of the clus-
Supplementary Fig. 53. The dominant transitions were found to be ters was also obtained through Au L3-edge X-ray absorption spectros-
either from Au s-type (450nm) or Au d-type (300nm) states to the copy (XAS). X-ray absorption near edge structure (XANES) spectra
empty states of the phenyl rings of the phosphines.
show a decrease in intensity when the NHC is introduced into the
clusters (Supplementary Fig. 55), particularly for the Me derivative
Spectroscopic characterization. The clusters were also extensively 3c, which may be due to the electron-donating ability of NHCs33,40,41
.
1
characterized by NMR spectroscopy, examining H, 31P and 13C
Extended X-ray absorption fine structure (EXAFS) R-space spec-
nuclei (Supplementary Section 3). The NHC carbon bound to Au tra of the clusters revealed information on Au–C, Au–P and Au–Cl
could not be observed by 13C NMR spectroscopy at natural abun- bonding, and Au–Au bonding within the cluster (Supplementary
dance, but when cluster 3a was prepared with 13C enrichment at C2 Fig. 54 and Supplementary Table 10). Analysis of Au–C bond lengths
of the NHC, the Au–C bond could be observed at 209ppm. This revealed that the Au–C(NHC) bond length in 3c is 11.4pm shorter
signal appeared as an octet, indicating equivalent coupling to each than the corresponding Au–C bond in 3a, consistent with steric dif-
of the seven phosphines (Fig. 3a). Because there are at least two ferences between the NHC types. Most importantly, the introduc-
unique environments for phosphines in any NHC-containing clus- tion of NHC 1a resulted in a 3pm increase in the average Au–Au
ter, exchange processes must be happening on the timescale of the bond length for gold atoms on the exterior of the cluster relative to
NMR experiment. 31P NMR spectroscopy supports the presence of the all-phosphine cluster 2, which suggests that NHC ligation leads
exchange phenomena at room temperature because a clean singlet to expansion of the outermost Au atoms in the cluster. A similar
was observed (Fig. 3b). To address this further, low-temperature cluster expansion was not observed on ligation of the smaller NHC
31P NMR spectra were obtained for both 2 and 3a (Supplementary in cluster 3c, which is interesting considering that this cluster has
Figs. 50 and 51). These spectra confirm the presence of an exchange higher stability than 3a (vide infra), suggesting an interplay between
phenomenon (or phenomena) on the NMR time scale. Interestingly, the experimental observations of cluster structure and stability.
NATURE CHEMiSTRY | www.nature.com/naturechemistry

NaTure CHeMisTry
Articles
a
b
a
[Au11(PPh3)8Cl2]+ (2)
+
AuP2
AuP+
b
c
–AuPCl–P
–AuPCl
–P
+
AuC2
AuCP+
[Au11(PPh3)7(NHCi-Pr)Cl2]+ (3a)
90 80 70 60 50 40 30 20
(ppm)
(ppm)
–AuPCl–P
–AuCCl–P
–AuPCl
–AuCCl
–P
d
AuC+
c
0
1,000
2,000
3,000
4,000
5,000
Mass number (m/z)
Fig. 4 | Assessment of stability and ligand dissociation by CiD mass
spectrometry. a, In-source CID mass spectra of cluster 2 recorded at
the CID voltage of 0ꢀV. Notations C and P represent NHCi-Pr and PPh3,
respectively. b, In-source CID mass spectra of cluster 2 recorded at 200ꢀV.
c, In-source CID mass spectra of cluster 3a recorded at the CID voltage of
0ꢀV. d, In-source CID mass spectra of cluster 3a recorded at 200ꢀV.
The structure of the cluster obtained from this study was found to be
the same as that predicted by DFT, resulting from displacement of P2
(Fig. 3c). Although static disorder in the region of the NHC ligand
complicates a detailed analysis of the Au–C bond length, the value
obtained (2.09(2)Å) is in the same range as that predicted by EXAFS
analysis (2.161(6)Å, Supplementary Table 10). Taken in aggregate, the
Au–Au bonds in the cluster were seen to contract upon introduction
of the NHC ligand; however, localized bonds experience contractions
and expansions that may be significant for catalysis (Supplementary
Figs. 56 and 57 and Supplementary Tables 11 and 12).
Fig. 3 | Spectroscopic and XRD characterization of NHC-substituted
cluster 3a. a, 13C {1H} NMR spectrum of 3a bearing 13C-labelled NHC,
which enables identification of the Au–C resonance. The signal appears
as an octet from equal coupling to all seven phosphines. b, 31P {1H} NMR
spectrum showing a single resonance at room temperature consistent with
rapid exchange. c, Single-crystal XRD structure for NHC cluster 3a showing
NHC incorporation at P2 as proposed from DFT studies.
Stability and catalytic activity of nanoclusters. Stability and ligand
dissociation were also assessed using collision-induced dissociation
mass spectrometry (CID-MS). In Fig. 4 we compare the in-source
CID mass spectra of 2 and 3a, recorded under identical conditions.
Cluster 2 undergoes loss of Au(PPh3)Cl and PPh3 as major CID paths
(Fig. 4a). In contrast, loss of the Au(NHCi-Pr)Cl unit was the most
dominant CID pathway for 3a (Fig. 4b), even though the ratio of
NHCi-Pr to PPh3 in the cluster is 1:7. In addition, direct dissociation of
The longer Au–C bond and its effect on the underlying struc- the NHCi-Pr ligand by Au–C bond cleavage was not observed under
ture has precedent in prior DFT studies, which have proposed that any conditions, in sharp contrast to 2, where the loss of PPh3 was
binding of an NHC results in restructuring of the surface Au atoms observed. These results suggest that NHCi-Pr has significantly higher
in Au(111) systems20,42,43; however, this effect has been difficult to binding affinity to Au than PPh3 does. These conclusions were sup-
observe experimentally. The EXAFS results reported herein repre- ported by thermogravimetric analysis–mass spectrometry (TGA–
sent a direct measurement of the NHC influence on the underlying MS) data, which also showed that phosphines are the first ligands
structure in metallic Au materials.
to be lost (Supplementary Table 13 and Supplementary Figs. 61–64).
The thermal stability of the clusters was assessed by heating clus-
DFT calculations supported the trends observed experimentally
by EXAFS. When the bond lengths were averaged over each of the ters in various solvents and monitoring the molecular transitions
optimized clusters, DFT studies predicted a 1.9pm shorter Au–C by UV–vis spectroscopy. The all-phosphine cluster 2 was used as a
bond in 3c compared to 3a. Similarly, the averaged calculated Au– benchmark, which underwent complete decomposition after heating
Au bond length for the exterior gold atoms increased by 1.0pm for to 70°C in pentanol (12h) and toluene at 70°C (24h). By contrast,
cluster 3a compared to 2, while cluster 3c showed no core expansion NHC-stabilized clusters showed dramatically improved stability,
(within error).
with 3c displaying virtually no change over 24h at 70°C in toluene
(Supplementary Fig. 58). Stability was reduced in the polar solvent
Single-crystal XRD study of nanocluster 3a. X-ray quality single pentanol, which allowed us to differentiate between the different
crystals of 3a were grown by layering of n-hexane onto a solution of 3a NHC-functionalized clusters (Fig. 5a,b and Supplementary Fig. 59).
in dichloromethane at room temperature. After several days, reddish Among the clusters examined, cluster stability tracked with sterics,
prisms of 3a could be observed. Because dichloromethane co-crys- with 3c being the most stable, showing virtually no decomposition
tallizes with 3a, degradation of the crystal was observed on drying. after 12h at 70°C in pentanol (Fig. 5a,b). Clusters 2, 3a and 3b
NATURE CHEMiSTRY | www.nature.com/naturechemistry

NaTure CHeMisTry
Articles
a
[Au11(PPh3)8Cl2]Cl
[Au11(PPh3)7(NHCMe)Cl2]Cl
0 h
1 h
2 h
3 h
4 h
5 h
6 h
7 h
8 h
9 h
10 h
11 h
12 h
2
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0.0
0 h
1 h
2 h
3 h
4 h
5 h
6 h
7 h
8 h
9 h
10 h
11 h
12 h
3c
0.8
0.6
0.4
0.2
0.0
SPR
300
400
500
600
700
300
400
500
600
700
Wavelength (nm)
Wavelength (nm)
b
3b
3c
3c
2
3a
2
3a
3b
Pentanol
70 °C
12 h
c
3c
2
3c
3c
10
8
100
2
2
30
20
10
0
3a
3a
3a
80
60
40
20
0
6
4
2
0
–0.6
–0.8
–1.0
–0.6
–0.8
–1.0
–0.6
–0.8
–1.0
E versus RHE (V)
E versus RHE (V)
E versus RHE (V)
Fig. 5 | Nanocluster stability and activity in the electrocatalytic reduction of CO2 to CO. a, UV–vis spectra of clusters 2 and 3c in pentanol at 70ꢀ°C over
12ꢀh. Cluster 2 decomposes to nanoparticles, evident by the loss of molecular signals at 405 and 310ꢀnm and growth of the signal at 550ꢀnm, while 3c is
completely stable. b, Nanoclusters 2, 3a, 3b and 3c before and after thermal treatment showing the decomposition of 2, 3a and 3b while nanocluster 3c
remains intact. c, Catalytic activity and selectivity of nanoclusters 2, 3a and 3c in the CO2 to CO reduction showing that nanocluster 3c has the highest
Faradaic efficiency for CO, the highest current density and the highest mass activity among the nanoclusters tested. Although Faradaic efficiency seems
independent of applied current, catalytic activity (as assessed through current density and mass activity) is highest at −1.0ꢀV (versus reversible hydrogen
electrode, RHE).
change colour from yellow (clusters) to purple (nanoparticles) after to 180°C for 2h. This temperature was coincident with changes
heating in pentanol (Fig. 5b). The relative stability of the clusters in the TGA-MS of the nanoclusters that are attributed to loss of a
is thus 2<3a<3b<3c. TGA–MS studies also showed that 3c was phosphine ligand (Supplementary Table 13 and Supplementary
the most stable of all clusters, with onset of ligand loss occurring Figs. 61–64). Lower temperatures resulted in significantly lower
15–20°C higher than all other clusters (Supplementary Figs. 60–64). activity, and longer treatment (6h) at 180°C was also detrimental
With a detailed understanding of structure and stability in hand, (Supplementary Fig. 66).
we examined these clusters as catalysts for the electrocatalytic
Electroreduction of CO2 was performed using a three-electrode
reduction of CO2 to CO (Fig. 5c)44–49. This is an important reaction system, with Ag/AgCl and Pt mesh used as reference and coun-
because CO is the key component in many high-volume carbon– ter electrodes, which were submerged in a CO2-saturated, 0.1M
carbon bond-forming reactions such as the oxo reaction and the KHCO3 electrolyte solution. During the reaction, CO2 was intro-
Cativa process.
duced by bubbling the gas through the electrolyte at 15 standard
Nanoclusters 2, 3a and 3c were dissolved in a mixture of isopro- cubic centimetres per minute, and the product gases were analysed
panol and toluene (1:2vol/vol) and deposited on carbon-paper elec- by gas chromatography.
trodes with a catalyst loading of 0.25mgcm−2. The electrodes were
As shown in Fig. 5c, nanocluster 3c outperformed all-phosphine
then dried at room temperature before activating in air by heating cluster 2 and cluster 3a bearing an isopropylated NHC. Nanocluster
NATURE CHEMiSTRY | www.nature.com/naturechemistry

NaTure CHeMisTry
Articles
The solvent was removed in vacuo to give an orange solid, which was dissolved
in dichloromethane (DCM) and passed through a Celite plug to remove any
insoluble orange particles. The solvent was evaporated under an air stream
or by rotary evaporator and the resulting orange solid was triturated with
Et2O (2 ×5 ml) to remove triphenylphosphine oxide. The resulting clusters were
dissolved in a minimum amount of DCM to give a red solution, which was
loaded onto a silica gel column packed with DCM/MeOH (25:1 vol/vol).
The product was eluted with DCM/MeOH (9:1 v/v) to give the desired cluster as
a red solid after solvent evaporation in vacuo. For details of specific experiments
see Supplementary Information.
3c affected the electrocatalytic CO2 reduction with the high-
est Faradaic efficiency (selectivity for CO production versus H2).
Similarly, cluster 3c had the highest current density (mAcm−2) and
mass activity (Ag−1) of all clusters tested, at all voltages employed
(Fig. 5c), with the greatest differences being observed at −1.0V
(versus RHE). Higher loadings of this nanocluster on the surface led
to decreased performance, suggesting that more highly dispersed
nanoclusters are more active than any agglomerated species that
may form (Supplementary Fig. 67).
Compared with previously reported related NHC-functionalized
nanoparticles, nanocluster 3c displays similar levels of Faradaic effi-
Data availability
Spectral and purity data are available for all new compounds, along with original
NMR, MS, XPS, UV–vis, DFT, TGA–MS and electrochemical data. Single-crystal
X-ray crystallographic data are included for cluster 3a, while crystallographic data
for cluster 3a have been deposited at the Cambridge Crystallographic Data Centre
under deposition no. CCDC 1878623. Copies of the data can be obtained free
of charge via https://www.ccdc.cam.ac.uk/structures/. All other data supporting
the findings of this study are available within the Article and its Supplementary
Information, or from the corresponding author upon reasonable request.
ciency and current density, although at more negative voltages43,44
.
The best nanostructured catalysts in the literature again give similar
results, but at lower overpotentials, as a consequence of high catalyst
loadings43,47. With the high mass activity already observed for 3c,
performance improvements would be expected through the use of
higher-surface-area supports such as carbon nanoparticles43.
To ensure that CO2 is the source of the observed CO, the reac-
tion was carried out with 13C-labelled CO2. Analysis of this reac-
tion by gas chromatography–mass spectrometry (GC–MS) analysis
(Supplementary Fig. 65) shows clear incorporation of the isotope into
carbon monoxide (13CO, m/z=29), confirming that the CO is pro-
duced from CO2 by comparison with CO obtained from 12CO2 where
GC-MS analysis gave an m/z of 28 mass units for 12CO. Only trace
amounts of formate were observed after prolonged exposure (1%),
illustrating the high selectivity of the reaction (Supplementary Fig. 68).
Interestingly, cluster 3c is the only cluster of those examined that
retained its structure and did not decompose to nanoparticles fol-
lowing extended thermal treatment. This observation leads to the
intriguing possibility that having an intact nanocluster is impor-
tant for catalytic activity, and illustrates the advantage of employing
molecular species such as clusters as catalysts. It should be noted
that NHC substitution does not uniformly lead to improved cata-
lysts because cluster 3a was outperformed by all-phosphine cluster
2. Although the precise reasons for this difference are not presently
clear, it may be that comparisons across cluster types are not as valid
as comparisons within the NHC series 3a, 3b and 3c. However, it is
clear that the most stable cluster examined (3c) also gives the high-
est activity.
Received: 6 April 2018; Accepted: 28 February 2019;
Published: xx xx xxxx
References
1. Bürgi, T. Properties of the gold–sulphur interface: from self-assembled
monolayers to clusters. Nanoscale 7, 15553–15567 (2015).
2. Häkkinen, H. Te gold–sulfur interface at the nanoscale. Nat. Chem. 4,
443–455 (2012).
3. Yamazoe, S., Koyasu, K. & Tsukuda, T. Nonscalable oxidation catalysis of gold
clusters. Acc. Chem. Res. 47, 816–824 (2014).
4. Jadzinsky, P. D., Calero, G., Ackerson, C. J., Bushnell, D. A. & Kornberg, R. D.
Structure of a thiol monolayer-protected gold nanoparticle at 1.1Å resolution.
Science 318, 430–433 (2007).
5. Häkkinen, H., Walter, M. & Grönbeck, H. Divide and protect: capping gold
nanoclusters with molecular gold–thiolate rings. J. Phys. Chem. B 110,
9927–9931 (2006).
6. Qian, H., Zhu, M., Wu, Z. & Jin, R. Quantum sized gold nanoclusters with
atomic precision. Acc. Chem. Res. 45, 1470–1479 (2012).
7. Hesari, M., Workentin, M. S. & Ding, Z. NIR electrochemiluminescence from
–
Au25 nanoclusters facilitated by highly oxidizing and reducing co-reactant
radicals. Chem. Sci. 5, 3814–3822 (2014).
8. Jensen, K. M. Ø. et al. Polymorphism in magic-sized Au144(SR)60 clusters.
Nat. Commun. 7, 11859 (2016).
9. Chevrier, D. M., Yang, R., Chatt, A. & Zhang, P. Bonding properties of
thiolate-protected gold nanoclusters and structural analogs from X-ray
absorption spectroscopy. Nanotechnol. Rev. 4, 193–206 (2015).
10. Chevrier, D. M., Zeng, C. J., Jin, R. C., Chatt, A. & Zhang, P. Role of Au4
units on the electronic and bonding properties of Au28(SR)20 nanoclusters
from X-ray spectroscopy. J. Phys. Chem. C 119, 1217–1223 (2015).
11. Jin, R. Quantum sized, thiolate-protected gold nanoclusters. Nanoscale 2,
343–362 (2010).
Conclusion
We have reported an example of the use of NHCs to stabilize gold(0)
nanoclusters. The introduction of the NHC can be accomplished
by a simple displacement reaction employing benzimidazolium
hydrogen carbonate salts. The number of NHCs introduced is con-
trollable by NHC structure, equivalents and reaction conditions.
The structures of the NHC-containing clusters were predicted
by DFT and confirmed by a combination of mass spectrometry,
NMR, EXAFS, XANES, UV–vis spectroscopy and single-crystal
XRD. The stability of the NHC-containing clusters was assessed
by treatment at high temperature in a variety of solvents and by
CID-MS and TGA–MS. NHC-containing clusters were found to be
more stable than the all-phosphine clusters, with absolute stabil-
ity dictated by the nature of the NHC. The performance of various
clusters in electrocatalytic CO2 reduction was found to correlate
with cluster stability, with the most stable cluster having the high-
est Faradaic efficiency, catalytic activity and current density. These
observations suggest that these novel NHC-stabilized gold clusters
can have quantifiable benefits for electrocatalytic performance,
which can be enhanced even further due to an atomically precise
understanding of their structure.
12. Tsukuda, T. & Häkkinen, H. Protected Metal Clusters: From Fundamentals to
Applications (Elsevier, Amsterdam, 2015).
13. Wang, Y. et al. Atomically precise alkynyl-protected metal nanoclusters as a
model catalyst: observation of promoting efect of surface ligands on catalysis
by metal nanoparticles. J. Am. Chem. Soc. 138, 3278–3281 (2016).
14. Wang, Y. et al. Site preference in multimetallic nanoclusters: incorporation
of alkali metal ions or copper atoms into the alkynyl-protected body-
centered cubic cluster [Au7Ag8(C≡ CtBu)12]+. Angew. Chem. Int. Ed. 55,
15152–15156 (2016).
15. Wang, Y. et al. An intermetallic Au24Ag20 superatom nanocluster stabilized by
labile ligands. J. Am. Chem. Soc. 137, 4324–4327 (2015).
16. Wan, X. K., Tang, Q., Yuan, S. F., Jiang, D. E. & Wang, Q. M. Au19 nanocluster
featuring a V-shaped alkynyl-gold motif. J. Am. Chem. Soc. 137, 652–655 (2015).
17. Maity, P., Tsunoyama, H., Yamauchi, M., Xie, S. H. & Tsukuda, T.
Organogold clusters protected by phenylacetylene. J. Am. Chem. Soc. 133,
20123–20125 (2011).
18. Weidner, T. et al. NHC-based self-assembled monolayers on solid gold
substrates. Aust. J. Chem. 64, 1177–1179 (2011).
19. Zhukhovitskiy, A. V., Mavros, M. G., Voorhis, T. V. & Johnson, J. A.
Addressable carbene anchors for gold surfaces. J. Am. Chem. Soc. 135,
7418–7421 (2013).
20. Crudden, C. M. et al. Ultra stable self-assembled monolayers of
N-heterocyclic carbenes on gold. Nat. Chem. 6, 409–414 (2014).
21. Crudden, C. M. et al. Simple direct formation of self-assembled
N-heterocyclic carbene monolayers on gold and their application in
biosensing. Nat. Commun. 7, 12654 (2016).
Methods
In a two-neck flask equipped with a condenser and an argon balloon, a mixture
of [Au11(PPh3)8Cl2]Cl (2) and the corresponding benzimidazolium hydrogen
carbonate (1x) were dissolved in THF (1 ml of THF per 1 mg of 2). The resulting
mixture was heated at 70 °C for 2 h before being cooled to room temperature.
NATURE CHEMiSTRY | www.nature.com/naturechemistry

NaTure CHeMisTry
Articles
22. Zhukhovitskiy, A. V., MacLeod, M. J. & Johnson, J. A. Carbene ligands in
surface chemistry: from stabilization of discrete elemental allotropes to
modifcation of nanoscale and bulk substrates. Chem. Rev. 115,
11503–11532 (2015).
44. Tang, Q. et al. Lattice-hydride mechanism in electrocatalytic CO2 reduction
by structurally precise copper-hydride nanoclusters. J. Am. Chem. Soc. 139,
9728–9736 (2017).
45. Cao, Z. et al. A molecular surface functionalization approach to tuning
nanoparticle electrocatalysts for carbon dioxide reduction. J. Am. Chem. Soc.
138, 8120–8125 (2016).
23. Larrea, C. R. et al. N-heterocyclic carbene self-assembled monolayers on
copper and gold: dramatic efect of wingtip groups on binding, orientation
and assembly. ChemPhysChem 18, 3536–3539 (2017).
46. Cao, Z. et al. Chelating N-heterocyclic carbene ligands enable tuning of
electrocatalytic CO2 reduction to formate and carbon monoxide: surface
organometallic chemistry. Angew. Chem. Int. Ed. 57, 4981–4985 (2018).
47. Gray, H. B. Powering the planet with solar fuel. Nat. Chem. 1, 7 (2009).
48. Liu, M. et al. Enhanced electrocatalytic CO2 reduction via feld-induced
reagent concentration. Nature 537, 382–386 (2016).
49. Zhang, L., Zhao, Z.-J. & Gong, J. Nanostructured materials for heterogeneous
electrocatalytic CO2 reduction and their related reaction mechanisms.
Angew. Chem. Int. Ed. 56, 11326–11353 (2017).
24. MacLeod, M. J. & Johnson, J. A. Pegylated N-heterocyclic carbene anchors
designed to stabilize gold nanoparticles in biologically relevant media.
J. Am. Chem. Soc. 137, 7974–7977 (2015).
25. Kim, H. K. et al. Reduction of the work function of gold by N-heterocyclic
carbenes. Chem. Mater. 29, 3403–3411 (2017).
26. Wang, G. et al. Ballbot-type motion of N-heterocyclic carbenes on gold
surfaces. Nat. Chem. 9, 152–156 (2016).
27. Engel, S., Fritz, E. C. & Ravoo, B. J. New trends in the functionalization of
metallic gold: from organosulfur ligands to N-heterocyclic carbenes.
Chem. Soc. Rev. 46, 2057–2075 (2017).
Acknowledgements
28. Moller, N. et al. Stabilization of high oxidation state upconversion
nanoparticles by N-heterocyclic carbenes. Angew. Chem. Int. Ed. 56,
4356–4360 (2017).
C.M.C., J.H.H. and E.H.S. acknowledge support from the Natural Sciences and
Engineering Research Council of Canada (NSERC) and the Canada Foundation for
Innovation (CFI) and the Ministry of Research Innovation (MRI) in terms of discovery
grants and infrastructure grants, respectively. K.A. and P.J.G. thank NSERC for support
through the awarding of USRA fellowships. M.R.N. thanks the Ontario Graduate
Scholarship programme and Queen’s University for fellowship support. This work was
supported by KAKENHI from the Japan Society for the Promotion of Science (JSPS;
17H03030 and 16K13962 to C.M.C. and 17H01182 to T.T.), Nanotechnology Platform
(project no. 12024046) and the Elements Strategy Initiative for Catalysts & Batteries
(ESICB). J.S.P.S. and N.U. acknowledge funding of this research through The World
Premier International Research Center Initiative (WPI) programme. The computational
work was supported by the Academy of Finland through the Academy Professorship of
H.H. All computations were carried out at the Finnish CSC computer centre. S.K. thanks
the Väisälä Foundation for a personal PhD study grant. K. Itami is thanked for assistance
with the preparation of this manuscript.
29. Salorinne, K. et al. Water-soluble N-heterocyclic carbene-protected gold
nanoparticles: size-controlled synthesis, stability and optical properties.
Angew. Chem. Int. Ed. 56, 6198–6202 (2017).
30. Man, R. W. Y. et al. Ultrastable gold nanoparticles modifed by bidentate
N-heterocyclic carbene ligands. J. Am. Chem. Soc. 140, 1576–1579 (2018).
31. Rühling, A. et al. Modular bidentate hybrid NHC–thioether ligands for the
stabilization of palladium nanoparticles in various solvents. Angew. Chem. Int.
Ed. 55, 5856–5860 (2016).
32. Crudden, C. M. & Allen, D. A. Stability and reactivity of N-heterocyclic
carbene complexes. Coord. Chem. Rev. 248, 2247–2273 (2004).
33. Hopkinson, M. N., Richter, C., Schedler, M. & Glorius, F. An overview of
N-heterocyclic carbenes. Nature 510, 485–496 (2014).
34. Polgar, A. M., Weigend, F., Zhang, A., Stillman, M. J. & Corrigan, J. F. A
N-heterocyclic carbene-stabilized coinage metal–chalcogenide framework
with tunable optical properties. J. Am. Chem. Soc. 139, 14045–14048 (2017).
35. Azizpoor Fard, M., Levchenko, T. I., Cadogan, C., Humenny, W. J. & Corrigan,
J. F. Stable -ESiMe3 complexes of CuI and AgI (E=S, Se) with NHCs: synthons
in ternary nanocluster assembly. Chem. Eur. J 22, 4543–4550 (2016).
36. Robilotto, T. J., Bacsa, J., Gray, T. G. & Sadighi, J. P. Synthesis of a trigold
monocation: an isolobal analogue of [H3]+. Angew. Chem. Int. Ed. 51,
12077–12080 (2012).
37. Jin, L. et al. Trinuclear gold clusters supported by cyclic (alkyl)(amino)
carbene ligands: mimics for gold heterogeneous catalysts. Angew. Chem. Int.
Ed. 53, 9059–9063 (2014).
38. McKenzie, L. C., Zaikova, T. O. & Hutchison, J. E. Structurally similar
triphenylphosphine-stabilized undecagolds, Au11(PPh3)7Cl3 and
[Au11(PPh3)8Cl2]Cl, exhibit distinct ligand exchange pathways with
glutathione. J. Am. Chem. Soc. 136, 13426–13435 (2014).
39. Enkovaara, J. et al. Electronic structure calculations with GPAW: a real-space
implementation of the projector augmented-wave method. J. Phys. Condens.
Matter 22, 253202 (2010).
40. Zhang, P. & Sham, T. K. X-ray studies of the structure and electronic
behavior of alkanethiolate-capped gold nanoparticles: the interplay of size
and surface efects. Phys. Rev. Lett. 90, 245502 (2003).
Author contributions
C.M.C., P.J.U., M.R.N. and K.S. designed and carried out the synthesis of the nanoclusters,
assisted by K.A., P.J.G., M.N., K.M.O. and R.W.Y.M. K.M.O. and R.W.Y.M. optimized
the synthetic procedures and purifications and acquired TGA–MS data. MS analysis was
performed and interpreted by S.T., R.T. and T.T., including CID MS. Crystallization of
3a was carried out by M.N. and S.T. on a sample prepared and purified by K.M.O. DFT
studies, including prediction of structure and optical spectra, were carried out by S.K.,
S.M. and H.H. EXAFS and XANES studies were carried out and interpreted by J.H.H. and
J.D.P. Electrocatalytic studies were performed and interpreted by C.-T.D. and E.H.S.
The manuscript was written by C.M.C. with assistance from co-authors.
Competing interests
The authors declare no competing interests.
Additional information
Supplementary information is available for this paper at https://doi.org/10.1038/
s41557-019-0246-5.
Reprints and permissions information is available at www.nature.com/reprints.
41. Lopez-Cartes, C. et al. Gold nanoparticles with diferent capping systems: an
electronic and structural XAS analysis. J. Phys. Chem. B 109, 8761–8766 (2005).
42. Tang, Q. & Jiang, D.-E. Comprehensive view of the ligand–gold interface
from frst principles. Chem. Mater. 29, 6908–6915 (2017).
43. Rodríguez-Castillo, M. et al. Reactivity of gold nanoparticles towards
N-heterocyclic carbenes. Dalton Trans. 43, 5978–5982 (2014).
Correspondence and requests for materials should be addressed to H.H., T.T. or
C.M.C.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.
© The Author(s), under exclusive licence to Springer Nature Limited 2019
NATURE CHEMiSTRY | www.nature.com/naturechemistry