DOI: 10.1002/cssc.

201600597 Communications

CO2 Reduction to CO in Water: Carbon Nanotube–Gold

Nanohybrid as a Selective and Efficient Electrocatalyst
Tran Ngoc Huan,[a] Praveen Prakash,[b] Philippe Simon,[a] Gwena[lle Rousse,[c] X. Xu,[d]
Vincent Artero,[e] Edmond Gravel,[b] Eric Doris,[b] and Marc Fontecave*[a]

A gold-based nanostructure has been demonstrated as promis- Furthermore, electrochemical reduction of CO2 in liquid phase
ing materials for the selective electroreduction of CO2 to CO in has been often studied in organic solvents, such as DMF or
aqueous conditions. In this work, we present a carbon nano- MeCN, and more recently in ionic liquids.[6] However, the reduc-
tube–gold nanohybrid as a selective and efficient electrocata- tion of CO2 in water is given priority as it may give rise to cost-
lyst for the reduction of CO2 in 0.5 m NaHCO3. The hybrid mate- effective and environmentally sound processes. Yet, CO2 solubili-
rial exhibits remarkable activity with a current density of ty is limited in aqueous solutions and proton reduction into H2
10 mA cm@2 at @0.55 V versus standard hydrogen electrode often competes with CO2 reduction. There is thus a challenge to
with a stable CO production rate (0.52 mmol s@1) after 4 h elec- be tackled that is to identify selective electrocatalysts that pref-
trolysis. Monodispersed gold nanoparticles anchored on erentially catalyze the CO2 reduction reaction [CO2-RR; Eq. (1)]
carbon nanotubes through a layer-by-layer method allows very over the H2 evolution reaction [HER; Eq. (2)]. In the last decade,
little Au loading and thus minimization of the cost of electrode various molecular and metal electrocatalysts have been studied
fabrication with a mass activity up to 100 A g@1 at @0.55 V but only few of them display high selectivity for CO2-RR in
versus reversible hydrogen electrode. It is 33 times higher than water.[5–8] Among them, gold-based nanostructures are promis-
a previous report for monodisperse Au nanoparticles (3 A g@1) ing electrocatalytic materials due to their high selectivity for CO
while ensuring selectivity (70 % faradaic yield of CO) at compa- production versus H2 production.[9–12] Their reactivity for the re-
rable reduction potential. duction of CO2 has been studied previously.[5, 10, 12]

CO2ðsolÞ þ 2 Hþ ðsolÞ þ 2 e@ ! COðgasÞ þ H2 O ð1Þ

Recycling and valorization of CO2 by converting it to energy-
2 Hþ sol þ 2 e@ ! H2ðgasÞ ð2Þ
dense liquid or gaseous fuels is a fascinating scenario.[1] Howev-
er, the development of this technology has been hampered by
the cost, low efficiency, and little selectivity of the catalysts. Elec- Nanostructuration of metallic surfaces has indeed been
trochemical reduction of CO2 is a promising approach as it shown to improve activity (high current densities) and selectivity
allows an easy control of the production of a variety of useful (high faradaic yields for a given product) relative to planar met-
products, mainly HCOOH, CO, CH3OH, CH4, and higher hydrocar- allic electrodes.[10–13] Hence, an ideal catalyst should combine an
bons.[2–5] However, highly negative electrode potentials are re- active metallic nanomaterial with a highly porous scaffold to
quested in most cases, resulting in poor energetic efficiencies. meet both high activity and selectivity. In this study, monodis-
persed gold nanoparticles (AuNPs) compactly and uniformly
[a] Dr. T. N. Huan, Dr. P. Simon, Prof. M. Fontecave
anchored onto carbon nanotubes (CNTs) through a layer-by-
Laboratoire de Chimie des Processus Biologiques layer method are shown for the first time to behave as efficient
CollHge de France catalysts for CO2-RR in 0.5 m NaHCO3 aqueous solution.
11 Place Marcelin Berthelot, 75005 Paris (France) Carbon nanotube–gold nanohybrids (AuCNTs; Figure 1 a)
E-mail: mfontecave@cea.fr
were synthesized as previously described.[14, 15] Amphiphilic nitri-
[b] P. Prakash, Dr. E. Gravel, Dr. E. Doris
iBiTecS, Service de Chimie Bioorganique et de Marquage lotriacetic–diyne lipid (DANTA) was first self-assembled on multi-
CEA Saclay, 91191 Gif-sur-Yvette (France) walled carbon nanotubes by hydrophobic interactions. Then,
[c] Dr. G. Rousse the diyne motif of DANTA was photopolymerized by irradiation
FRE 3677 “Chimie du Solide et Energie” at 254 nm. Cationic poly(diallyldimethylammonium chloride)
CollHge de France
(PDADMAC) was next deposited as a second layer, which inter-
11 Place Marcelin Berthelot, 75005 Paris (France)
acted with the primary anionic layer by electrostatic interactions.
[d] Dr. X. Xu
Laboratoire de Physique et d’Etude des Mat8riaux Finally, the doubly-coated DANTA–PDADMAC nanotubes were
ESPCI-ParisTech, PSL Research University recovered by centrifugation and treated with a solution of fresh-
75005 Paris (France) ly prepared colloidal AuNPs. The multilayer polymer assembly
[e] Dr. V. Artero served for interfacing AuNPs and the nanotubes optimally. A so-
Laboratoire de Chimie et Biologie des M8taux
lution of AuCNTs was then prepared by suspending the gold
Universit8 Grenoble Alpes, CNRS UMR 5249, CEA
17 rue des Martyrs, F-38000 Grenoble (France) nanohybrid in water. Au concentration ([Au] = 3.4 : 0.05 mm)
Supporting Information for this article can be found under http:// was measured by inductively coupled plasma–mass spectrome-
dx.doi.org/10.1002/cssc.201600597. try (ICP–MS). Transmission electron microscopy (TEM) images

ChemSusChem 2016, 9, 2317 – 2320 2317 T 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
 Communications

Figure 2. LSV of AuCNT/GDL electrodes (Au loading: 90 mg cm@2) in 0.5 m

NaHCO3 saturated with N2 (red) and CO2 (black).

Figure 1. (a) Schematic representation of the AuCNTs assembly, (b) TEM in 0.5 m NaHCO3 in the presence and absence of CO2
image of the AuCNT nanohybrid, and (c) SEM image of AuCNT nanohybrid
(Figure 2). The current density observed in CO2-free solution
deposited on GDL.
was entirely attributed to the HER while the current density
measured in the CO2-saturated solution was significantly larger
showed densely decorated CNTs with monodisperse spherical after the onset potential (@0.15 V vs. standard hydrogen elec-
AuNPs (Figure 1 b) with a narrow size distribution centered at trode (SHE)), indicating additional CO2-RR activity of AuCNT/
around 3 nm. Gold valence state was established by X-ray pho- GDL electrodes.
toelectron spectroscopy (XPS) analysis, which showed only the The results obtained for controlled potential electrolysis
characteristic Au 4f contributions of the Au0 species. X-ray using the AuCNT/GDL electrode under CO2 saturation at differ-
powder patterns of two samples with different AuNPs loading, ent potentials are presented in Figure S4. The current density
shown in Figure S1 in the Supporting Information, display two in all cases was stable during electrolysis and increased as
peaks at 2 q = 38.28 and 64.78 that can be attributed to Au’s a function of increased cathodic potential to reach a remark-
(111) and (220) reflections, respectively.[12] This is also confirmed able value of 15 mA cm@2 at @0.65 V versus SHE. Only CO and
by the higher intensity observed for these two peaks for the H2 were detected as reaction products with no evidence for
sample of higher Au loading on the electrode surface. formate, methanol, methane, and formaldehyde formation.
Variable amounts of the AuCNTs solution were sprayed onto Figure 3 a shows that the rates of H2 and CO production
the surface of a gas diffusion layer (GDL) electrode using the varied extensively between @0.15 and @0.65 V versus reversible
air spray coating method, after which the GDL electrode was hydrogen electrode (RHE). Between @0.15 and @0.45 V versus
dried under air conditions. The scanning electron microscopy RHE, H2 formation was slightly preponderant. Similar production
(SEM) image of the AuCNT/GDL electrode is presented in Fig- rates of H2 and CO were observed at @0.45 V versus RHE. The
ure 1 c, showing a nanoporous structure of the deposited two production rates increased drastically for potentials more
AuCNT film on the electrode. The nanotubes were shown to negative than @0.45 V versus RHE up to @0.65 V versus RHE,
be pure enough by Raman spectroscopy (Figure S2) and X-ray but this acceleration was much more pronounced for the CO2-
photoelectron spectroscopy, which indicated an O/C ratio of RR. The faradaic yields (FY) for the production of CO and H2 pre-
roughly 5 % but no nitrogen. sented in Figure 3 b are consistent with these variations in rates,
Electrochemical cyclic voltammetry (CV) using a AuCNT/GDL with the highest FY value for CO production (70 : 2 %) obtained
electrode in 0.5 m H2SO4 was first recorded for the quantifica- at @0.55 V versus RHE (current density of 10 mA cm@2) with a FY
tion of electroactive surface sites. The CVs in Figure S3 a show of 25 : 2 % for H2 production. A long-term electrolysis of CO2 re-
characteristic anodic and cathodic peaks at approximately 1.3 duction at @0.55 V versus RHE has been also carried out (Fig-
and 0.9 V versus Ag/AgCl, corresponding to the oxidation of ure S5). The current density was very stable during 4 h electroly-
Au and reduction of Au oxide, respectively.[16, 17] Estimation of sis with a constant rate of CO production (0.52 mmol s@1). As
the electrochemical surface area (ECS) was carried out by nor- a control experiment and confirming that CO was only derived
malizing the charge of the oxide reduction peak with a value from CO2 reduction, bulk electrolysis under saturated N2 at all
of 400 mC cm@2.[18, 19] The data in Figure S3 b show a linear in- potentials from @0.15 to @0.65 V versus RHE generated H2 as
crease of ECS as a function of the Au surface density of the the only reaction product with FY = 100 %. This excluded any
electrode. The Au loading of 90 mg cm@2 was selected based on possibility of CO deriving from the decomposition or transfor-
previous publications for the ensuing experiment of CO2 elec- mation of the carbon structure. This has been confirmed by
troreduction. electrolysis in the presence of 13CO2. Mass spectrometric analysis
The electrocatalytic performance of the AuCNT/GDL elec- of the gaseous product unambiguously established the forma-
trode was then evaluated by linear sweep voltammetry (LSV) tion of 13CO.

ChemSusChem 2016, 9, 2317 – 2320 www.chemsuschem.org 2318 T 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
 Communications
of this material is that it allows very little Au loading and thus
minimization of the cost of electrode fabrication. For example,
their mass activity, up to 100 A g@1 at @0.55 V versus RHE, is 33
times higher than that reported for monodisperse Au nanopar-
ticles (3 A g@1) at comparable reduction potential.[10, 11]
In conclusion, we have shown that the layer-by-layer assembly
of small gold nanoparticles (AuNPs) on carbon nanotubes (CNTs)
provided access to a promising material for selective electrore-
duction of CO2 to CO in aqueous conditions. The high density of
AuNPs decorating the CNTs promotes increased production rates
of CO thanks to the large active surface area provided by the
nanoporous structure of the CNT film. Furthermore, this assem-
bly allows a remarkably high mass activity, resulting in high cur-
rent density (10 mA cm@2 at @0.55 V vs. SHE) in water with very
little Au. This remarkable result makes these hybrid nanomateri-
als exciting systems for the further development of electrocata-
lytic CO2 reduction reaction under aqueous conditions.

Experimental Section
Chemicals
All chemicals including sodium bicarbonate (99 %), H2SO4 (99.8 %),
cationic poly(diallyldimethylammonium chloride) (PDADMAC), tet-
rahydroxymethylphosphonium chloride (THPC), and HAuCl4 were
obtained from Sigma–Aldrich. Multi-walled carbon nanotubes
(MWCNTs) were obtained from Prof. Haiyan Li (Xiamen University,
China) and prepared by catalytic decomposition of methane ac-
cording to the previously reported method.[20] Amphiphilic nitrilo-
triacetic–diyne lipid (DANTA) was synthesized according to our pro-
Figure 3. (a) Production rate and (b) faradaic yield of CO (black square) and
cedure reported elsewhere.[15]
H2 (red circle) during CO2 reduction in 0.5 m NaHCO3 solution using a AuNP/
GDL (90 mg cm@2 AuNP loading) electrode at different potentials.

The high selectivity and activity of the AuCNT/GDL electrode Synthesis and characterization of carbon nanotubes
for CO2 reduction into CO were in marked contrast to that of Multi-walled carbon nanotubes (MWCNTs) were prepared by cata-
control electrodes, namely CNT/GDL, DANTA–PDADMAC–CNT/ lytic decomposition of methane on a Ni–Mg–O catalyst at the Uni-
GDL and AuNP/GDL, within the investigated potential range versity of Xiamen.[20] Raw MWCNTs were purified through treat-
(@0.15 to @0.65 V vs. RHE). Under CO2 saturation conditions, ment with 8 n HNO3 for 12 h under refluxing conditions. After
the CNT/GDL and DANTA–PDADMAC–CNT/GDL electrodes washing several times with deionized water and drying at 373 K in
a vacuum oven, purified MWCNTs were obtained with open ends.
were exclusively active for HER, with no observed formation of
The nanotubes were highly uniform in diameter along the tube
CO and formate. The AuNP/GDL electrode displayed a good length with outer diameters between 10–50 nm and lengths in the
selectivity for CO2-RR with FY reaching up to 75 : 2 % for CO 10–100 mm range.
production at @0.55 V versus RHE, however with much lower Carbon nanotubes were the only carbonaceous species in the puri-
current density (2 mA cm@2) at the same AuNP loading fied sample. However, trace amounts of the catalyst were detected
(90 mg cm@2), thus indicating lower catalytic activity. In addition, by inductively coupled plasma–atomic emission spectroscopy (ICP–
a blank experiment conducted with CNTs in the absence of AES). The samples were calcined at 750 8C for 6 h before they were
supported gold showed no intrinsic catalytic activity. dissolved in concentrated HNO3, diluted with deionized water, and
analyzed using ICP–AES. The nickel and manganese content were
Some of the performances of the AuCNT/GDL electrode, no-
respectively 0.24 and 0.05 mass percentage.
tably the high mass activity and the low overpotentials, com-
pare favorably with previously reported Au-based systems op-
erating in water. For example, Au nanoparticles derived from Assembly of the CNT–gold nanohybrid
reduction of thick Au oxide films showed FY of > 98 % for CO,
larger than those reported here but with less current stability Step 1: Amphiphilic DANTA (20 mg) was dissolved in aqueous Tris-
buffer (2 mL of 25 mm, pH 8) before multi-walled carbon nano-
(the current density dropped from 10 to 6 mA cm@2 in 8 h).[12]
tubes (50 mg) were added. The dispersion was sonicated, and the
Monodisperse 8 nm Au nanoparticles were also shown to dis- stable suspension transferred into two tubes. The tubes were cen-
play very high FY for CO (> 90 %) but at quite high cathodic trifuged at 5000 g, and the supernatants were collected. The latter
potentials, and the onset potential was significantly more neg- were centrifuged at 15 000 g for 45 min. The supernatant was dis-
ative than those reported here.[10] The most interesting feature carded and the pellets taken in buffer and centrifuged again at

ChemSusChem 2016, 9, 2317 – 2320 www.chemsuschem.org 2319 T 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
 Communications
15 000 g for 45 min. The pellets were finally resuspended in 1.5 mL Rietveld refinements were performed with the FullProf suite of pro-
buffer and submitted to UV irradiation at 254 nm for 6 h. grams.
Step 2: After polymerization, the buffer volume was adjusted to Analysis of 13CO from 13CO2 reduction: The mass spectrometer
1.5 mL. The suspension was stirred in the presence of the cationic (Aspec Ltd) was directly connected to the electrochemical cell
polymer PDADMAC (700 mL of a 20 % water solution) for 1 h. The during standard bulk electrolysis under 13CO2 saturation. The gas
ensuing centrifugation at 15 000 g for 30 min permitted removal of reference was Argon (molecular weight MW = 40). Gaseous prod-
the excess polymer. The pellets were taken in 2 mL of buffer. This ucts were then analyzed by mass spectrometry every third minute.
operation was repeated twice using the buffer solution and two Control experiments were also run: (i) electrolysis using an Ar-satu-
more times using pure water. rated solution; (ii) electrolysis saturation of non-labelled CO2.
Step 3: The final pellets were resuspended in water (1 mL). 50 mL
of the latter suspension was transferred to Eppendorf tubes (V 20).
To each tube a colloid suspension of the gold nanoparticles (1 mL,
1 mm) was added, and the mixture was vortex-stirred at room tem- Acknowledgements
perature for 1 min every 30 min (during 4 h).[20] The suspension
was then centrifuged at 3000 g for 5 min. The supernatant was dis- P.P. thanks the Enhanced Eurotalents program for support. The
carded, and 1 mL of a fresh gold colloid suspension was added. “Service de Chimie Bioorganique et de Marquage” belongs to the
The same process was repeated two more times. The pellets were Laboratory of Excellence in Research on Medication and Innova-
washed three times by centrifugation/redispersion in water. The 20 tive Therapeutics (ANR-10-LABX-0033-LERMIT). We acknowledge
pellets were combined and 4 mL of water was finally added. support from the French National Research Agency (ANR, Carbi-
ored ANR-12-BS07-0024-03; Labex program ARCANE, ANR-11-
Methods LABX-0003-01 and DYNAMO, ANR-11-LABX-0011) and from Fon-
dation de l’Orangerie for individual Philanthropy and its donors.
Electrochemical measurements were performed in a three-elec-
trode two-compartment cell using a Bio-logic SP300 potentiostat.
Ag/AgCl/3 m KCl (hereafter abbreviated as Ag/AgCl) was used as Keywords: carbon nanotubes · co2 reduction ·
the reference electrode and placed in the same compartment as electrocatalysis · gold · nanoparticles
the working electrode. A platinum counter electrode was placed in
a separate compartment connected by a glass frit and filled with
the electrolytic solution. The change of potential scale versus RHE [1] M. Aresta, A. Dibenedetto, A. Angelini, Chem. Rev. 2014, 114, 1709 –
was done according to Equation (3): 1742.
[2] Y. Hori, A. Murata, R. Takahashi, J. Chem. Soc. Faraday Trans. 1 1989, 85,
EðV vs: RHEÞ ¼ EðV vs: Ref:Þ þ EðV of Ref: vs: NHEÞ þ 0:059 > pH 2309 – 2326.
[3] J. Qiao, Y. Liu, F. Hong, J. Zhang, Chem. Soc. Rev. 2014, 43, 631 – 675.
ð3Þ [4] C. Costentin, Chem. Soc. Rev. 2013, 42, 2423 – 2436.
[5] K. P. Kuhl, T. Hatsukade, E. R. Cave, D. N. Abram, J. Kibsgaard, T. F. Jara-
where E(V of Ref. vs. NHE) = 0.205 V and the pH of 0.5 m NaHCO3 millo, J. Am. Chem. Soc. 2014, 136, 14107 – 14113.
was 7.2. [6] C. Costentin, M. Robert, J. M. Sav8ant, A. Tatin, Proc. Natl. Acad. Sci. USA
2015, 112, 6882 – 6886.
To check the stability of the catalyst, a long-term electrolysis (4 h)
[7] A. Tatin, C. Comminges, B. Kokoh, C. Costentin, M. Robert, J. M. Sav8ant,
of CO2 reduction was performed at @0.55 V versus RHE and CO2 Proc. Natl. Acad. Sci. USA 2016, 113, 5526 – 5529.
was continuously bubbled during electrolysis. [8] M. Alvarez-Guerra, J. Albo, E. Alvarez-Guerra, A. Irabien, Energy Environ.
The production rate of CO was calculated before and after long- Sci. 2015, 8, 2574 – 2599.
term electrolysis. The solution was continuously bubbled with CO2 [9] Y. J. Jang, J. W. Jang, J. Lee, J. H. Kim, H. Kumagai, J. Lee, T. Minegishi, J.
until the head space of the electrochemical cell was filled with Kubota, K. Domen, J. S. Lee, Energy Environ. Sci. 2015, 8, 3597 – 3604.
CO2. After that, 5 min electrolysis was carried out in a closed elec- [10] W. Zhu, R. Michalsky, O. Metin, H. Lv, S. Guo, C. J. Wright, X. Sun, A. A.
trochemical cell. At the end of electrolysis, CO was quantified and Peterson, S. Sun, J. Am. Chem. Soc. 2013, 135, 16833 – 16836.
the result was used to calculate the production rate. [11] A. S. Hall, Y. Yoon, A. Wuttig, Y. Surendranath, J. Am. Chem. Soc. 2015,
137, 14834 – 14837.
H2 measurements were performed by gas chromatography on
[12] Y. Chen, C. W. Li, M. W. Kanan, J. Am. Chem. Soc. 2012, 134, 19969 –
a Shimadzu GC-2014 equipped with a Quadrex column, a thermal 19972.
conductivity detector, and using N2 as a carrier gas. CO, methane, [13] C. W. Li, M. W. Kanan, J. Am. Chem. Soc. 2012, 134, 7231 – 7234.
and other volatile hydrocarbons from the gas phase were analyzed [14] A. Morozan, S. Donck, V. Artero, E. Gravel, E. Doris, Nanoscale 2015, 7,
using a gas chromatograph (Shimadzu GC-2010) equipped with 17274 – 17277.
a methanizer, a flame ionization detector (FID), and a shincarbon [15] J. John, E. Gravel, A. Hagege, H. Y. Li, T. Gacoin, E. Doris, Angew. Chem.
ST (Restek) column. Methanol was assayed by gas chromatography Int. Ed. 2011, 50, 7533; Angew. Chem. 2011, 123, 7675.
(Shimadzu GC 2010) using an Rtx-1 column (Restek) and a FID. [16] N. Alexeyeva, K. Tammeveski, Anal. Chim. Acta 2008, 618, 140.
Formate, oxalate, and glyoxylate concentrations were determined [17] H. Erikson, G. Juermann, A. Sarapuu, R. J. Potter, K. Tammeveski, Electro-
chim. Acta 2009, 54, 7483.
by ionic exchange chromatography (883 Basic IC, Metrohm).
[18] S. Trasatti, O. A. Petrii, Pure Appl. Chem. 1991, 63, 711.
SEM images were acquired using a Hitachi S-4800 scanning elec- [19] H. Angerstein-Kozlowska, B. E. Conway, A. Hamelin, L. Stoicoviciu, J.
tron microscope. TEM and high-resolution (HR)TEM images were Electroanal. Chem. 1987, 228, 429.
obtained on a JEM-2100F transmission electron microscope (JEOL, [20] P. Chen, H. B. Zhang, G. D. Lin, Q. Hong, K. R. Tsai, Carbon 1997, 35,
Japan) using an accelerating voltage of 200 kV. 1495 – 1501.
X-ray powder diffraction (XRD) patterns were recorded using an
X’Pert Pro Panalytical diffractometer equipped with either a CuKa Received: May 4, 2016
radiation source (lKa1 = 1.541 a, lKa2 = 1.54 a) or a CoKa radiation Revised: June 14, 2016
source (lKa1 = 1.789 a, lKa2 = 1.793 a) using an X’Celerator detector. Published online on August 5, 2016

ChemSusChem 2016, 9, 2317 – 2320 www.chemsuschem.org 2320 T 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim