Introduction

The increasing demand for renewable energy has been motivating researchers around the world to pursue the development of new technologies. Various methods and devices to generate sustainable forms of energy that are efficient and affordable have been reported [1–4]. In addition to clean energy, the high amount of CO2 emitted in the atmosphere must also be mitigated [5, 6]. The use of CO2 as a feedstock to generate fuels is intriguing and inspired by photosynthesis [7]. The consumption of CO2 from the atmosphere via its conversion to fuel could provide new sources of energy. In this view, electrochemical reduction of CO2 appears promising since it can lead to formation of methane, ethylene, methanol, formate, carbon monoxide, formic acid and other compounds that can be potentially used as fuels [8–10]. The applicability of the CO2 reduction process is currently limited by the high overpotentials even when using noble metal catalysts as active electrodes [11].

Different materials have been investigated as catalysts for CO2 reduction. Some candidates such as Hg, In and Cd have shown high current efficiency associated with high hydrogen overvoltage but remain poor choice for CO2 reduction catalysts because of their low selectivity in producing the desired products [12]. On the other hand, materials such as Pt, Ni, Fe and Ti favor hydrogen evolution reaction (HER) in aqueous media and also possess high affinity to CO, leading to rapid reduction of catalytic activity [13]. Copper is the most promising catalyst for CO2 reduction reaction because it enables the production of hydrocarbons at significant current densities, exhibits good selectivity, and has low surface affinity to CO [14]. Theoretical studies have shown that CO acts as an important precursor for hydrocarbon synthesis, which progresses in successive hydrogenation/reduction reactions [15] leading to the 16 products identified during CO2 reduction on copper [9]. Despite its rich electrochemistry, Cu suffers from high overpotentials compared to potentials of standard catalysts for hydrocarbon production (0.17 V vs. RHE) [9, 16]. Challenges associated with high overpotential and low electrochemical stability must be overcome if copper is to be seriously considered as a practical catalyst for CO2 reduction reaction [16]. It has been demonstrated that the oxide passivation layer on a Cu film can reduce the overpotential for the CO2 reduction reaction as well enhance the stability [17]. Although passivated Cu films are more robust catalysts, it leads to a reduction of the current density due to the copper oxide layer. Cu nanoparticles (NPs) could be beneficial in facilitating the CO2 reduction reaction but their instability caused by combined effects of aggregation and rapid deactivation of the catalytic surface via oxidation has limited their consideration. One strategy recently proposed to overcome the stability issues was to use Au2Cu NPs that are less prone to oxidation instead Cu NPs for CO2 reduction [18]. However, Au is expensive and also known to be a good catalyst for HER that can disrupt the production of hydrocarbons [19]. In the present work, we demonstrate that reduced graphene oxide decorated with copper NPs (CuNPs on rGO) is a promising catalyst system for CO2 reduction that exhibits remarkable stability in a medium with low proton availability. CuNPs on rGO electrodes exhibit lower overpotential for CO2 reduction reaction and show a strong increase in the current density. To assess the stability of the catalysts, we have tested CuNPs on rGO for several hours and found limited decrease in the current density and negligible change in the overpotential.

Graphene oxide (GO) has been investigated for a broad range of applications in electronics [20], biology [21], energy storage [22, 23], and more recently in catalysis [24–26]. GO consists of atomically thin sheets with lateral dimensions of a few micrometers and thickness of less than 1 nm. GO is composed of sp2 and sp3 carbon bonds as well as oxygen functional groups such as hydroxyls and carbonyls [27, 28]. Its chemically heterogeneous properties can be tuned to some extent by varying the oxygen content through reduction and offer a remarkable platform for performing novel chemistry [29, 30]. The partial removal of oxygen functional groups leads to an enhancement in the π conjugation so that the semi-metallicity of graphene is partially restored and its conductivity increases by several orders of magnitude compared to as-prepared GO [31, 32]. Recently, it has been reported that reduced graphene oxide (rGO) works synergistically with catalyst particles to enhance their stability and performance [33–36].

Experimental

The synthesis was performed in two steps as shown in Scheme 1. First, an aqueous solution of GO was prepared at a concentration of 1.5 mg/mL and this solution was mixed with 9 mg of CuSO4 (copper II sulfate). NaOH was then added and the mixture was vigorously stirred at 70 °C for 5 min. At this step of the synthesis, the color of the solution changed to dark grey. Then, 625 μL of hydrazine was quickly added to the mixture and stirred for additional 40 min. Upon the addition of hydrazine, the color turned dark violet. The mixture was further washed several times with EtOH by centrifugation (10 min at 30 k rpm).

Scheme 1
scheme 1

Schematic representation of the synthesis method consisting of mixing GO with copper(II) and NaOH first. Hydrazine was then added and the mixture was stirred for 40 min. The final material corresponds to rGO nanosheets decorated with Cu NPs

Full size image

Graphene oxide synthesis

Graphene oxide (GO) was prepared chemically via modified Hummer’s method [37]. The graphite powder (Branwell Graphite Inc.) with size of >425 μm was chemically oxidized, exfoliated, and purified by repeated centrifugation.

Copper nanoparticles passivated with polyallylamine

Cu nanoparticles are prepared following the procedure proposed by Wang and Asefa [38], using polyallylamine as passivating agent. In the present case, 220 μL polyallylamine (Sigma-Aldrich) was completely dissolved in distilled water under vigorous stirring at 70 °C. Then, 9 mg of copper sulfate (Sigma-Aldrich) was added followed by 700 μL of NaOH 0.5 M. After 5 min, ~60 μL of hydrazine was poured in the mixture. The solution color changed from an initial blue color to brown and 40 min later the solution color became red.

Copper thin film

A thin film of copper was formed by electrochemical deposition using a solution with 19 mg of Cu(I) in 10 mL acetonitrile. The copper was deposited on a glassy carbon electrode with 3 mm diameter. The deposition was carried out using a constant current of 10 μA for 10 min giving approximately 4 μg of deposited copper. Considering the density of bulk copper and the diameter of the glassy carbon, the thickness of the copper film is estimated to be around 70 nm.

Characterization

X-ray photoelectron spectroscopy (XPS) measurements were performed with a Thermo Scientific K-Alpha spectrometer. All spectra were taken using an Al Kα microfocused monochromatized source (1,486.6 eV) with a resolution of 0.6 eV. Raman spectra were obtained using a Renishaw 1,000 system operating at 514 nm. The UV–vis absorption spectra were measured with a Lambda 950 spectrophotometer (PerkinElmer). Transmission electron microscope (TEM) images of the samples were obtained with a Topcon 002B TEM that was working at an acceleration voltage of 200 kV.

Electrochemical measurements

The measurements (VersaStat 3 potentiostat from Princeton Applied Research) consist of linear sweep voltammetry (LSV) and were performed in two distinct environments (Argon and CO2) using acetonitrile (10 mL) and tetra-n-butylammonium hexafluorophospate (NBu4PF6) (387 mg, Alfa Aesar) as electrolyte. The electrolyte solution was first degassed using argon for 30 min. Under this condition, no reduction reaction is expected. The same procedure was repeated under CO2 to saturate the solution prior the measurements.

All materials were deposited by drop casting on glassy carbon electrode (CH Instruments, Inc.) and the current of the voltammograms is normalized by the geometric area (~7 × 10−2 cm2) of the glassy carbon. An Ag/Ag+ (0.1 m) electrode in acetonitrile and a platinum wire were used as reference and counter electrode, respectively. The reference electrode was calibrated using the well-known reduction oxidation signals of ferrocene (FCN) (Figure S1). Potentials in the voltammograms are corrected to NHE.

Results and discussion

RGO decorated with CuNPs was synthesized by co-reduction of copper(II) and GO using hydrazine solution as reducing agent. Briefly, copper(II) sulfate (9 mg) and GO (3 mg) were dispersed in 10 mL of 0.5 M NaOH at 70 °C under vigorous stirring. In basic solution, copper(II) cations remain in close contact with the negatively charged oxygen functional groups of GO. 625 μL of 80 % hydrazine solution was then poured into the mixture to reduce copper(II) and the GO nanosheets (see Methods for details). The addition of hydrazine results in a significant change in the color of the solution, becoming intensely reddish due to surface plasmon resonance (SPR) effect of metallic copper nanoparticles. This effect causes strong absorption around 580–600 nm [39]. As a result of the co-reduction process, conductive rGO is obtained along with metallic copper nanoparticles formed via reduction of Cu(II) as shown in Scheme 1. It is worth noting that no additional passivating agents, such organic molecules or polymers, were used to stabilize the as-prepared copper particles. TEM images of the composite indicate that the CuNPs are uniformly distributed on the basal plane of the reduced graphene oxide and it was also observed that some nanoparticles were partially wrapped by the rGO sheets (Fig. 1a). TEM observations suggest that rGO acts as a stabilizing support for the 20–40 nm Cu nanoparticles by immobilizing and preventing their aggregation into larger particles (inset Fig. 1a).

Fig. 1
figure 1

TEM image and FTIR of CuNPs on rGO prepared by hydrazine reduction: a TEM image showing the dispersion of nanoparticles on rGO. The size distribution histogram indicates an average particle size of about 30 nm. b FTIR spectra of bare rGO and CuNPs on rGO. The oxygen peaks are virtually absent in bare rGO while they are present in Cu NP decorated rGO and are presumably helping the stabilization of the Cu NPs

Full size image

Figure 1b presents FTIR spectra of the CuNPs on rGO and bare rGO films reduced by hydrazine. Clear differences between the two spectra can be observed. First, the rGO spectrum reveals the absence of oxygen functionalized groups, suggesting a strong reduction [40]. In contrast, the intense bands between 800 and 1,330 cm−1 in the CuNPs/rGO spectrum correspond to stretching vibrations (mainly from epoxy and ketone groups), which suggests that decoration with CuNPs inhibits a more severe reduction of GO. However, contributions in the 1,500–1,600 cm−1 range are typically not related to oxygen groups, and can be assigned to the asymmetric stretch of sp2-hybridized C=C within the rGO basal plane [41]. The presence of residual hydroxyl has been demonstrated to be beneficial in catalysis through enhancement of electrochemical stability and catalytic activity [41, 42]. In addition, hydroxyl groups that remain on rGO nanosheets decrease the interface energy between rGO and copper, leading to stabilization of CuNPs.

To further investigate the structure of copper-decorated rGO, X-ray photoelectron spectroscopy (XPS) was performed. From the XPS survey spectrum of CuNPs/rGO hybrid material shown in Fig. 2, it is possible to identify signals from oxygen, carbon and copper. The upper left inset shows the high-resolution signal of the C1s region from rGO with C–C/C=C, C–OH/C–O, C=O and C=O–OH peaks at 284.4, 286, 287.3, 289 and the satellite peak at 290.6 eV. After hydrazine treatment, the carbon/oxygen ratio in the CuNPs on rGO sample increased from 2.1 up to 7.3. This value is in good agreement with those previously reported for mildly reduced GO [43, 44]. On the other hand, peaks from Cu 2p3/2 and 2p1/2 can be identified at 932.5 and 952.4 eV, respectively. The fitting of the Cu peaks reveals a main signal at 932.5 eV that can be attributed to the zero-valence states of Cu and smaller peaks at 934.4 eV that are due to the presence of Cu(II) [17]. The latter indicates partial oxidation of the Cu nanoparticle surface in the form of CuO. We believe that the CuO shell is formed during exposure to air after the synthesis since the presence of hydrazine during nanoparticle synthesis is likely to prevent copper from being oxidized. It is also possible to verify the presence of two Cu satellites bands at 940 and 943.9 eV. Moreover, the presence of copper in zero-valence state is also confirmed by our UV–Vis spectroscopy results (see Supplementary Information, Figure S2) that show an intense absorption band at around 600 nm attributed to the surface plasmon resonance effect (SPR) in metallic copper NPs as previously reported [45–47]. This peak is not observed in the UV–Vis spectrum of the bare rGO, Figure S2. The SPR band was also observed at 560 nm for CuNPs prepared using polyallylamine as capping agent. The shift in position of copper SPR band between CuNPs prepared with polyallylamine and CuNPs on rGO can be attributed to differences in the size of the nanoparticles. That is, our TEM image analyses show that the Cu NPs prepared using polyallylamine are significantly smaller in size than the CuNPs on rGO (Figure S3).

Fig. 2
figure 2

XPS survey spectrum of CuNPs on rGO. High-resolution spectra from the C1s and Cu2p regions can be deconvoluted with different components from various chemical bonds, as indicated by the fits in the upper portion of the figure

Full size image

CO2 reduction reaction

To gain insight into the catalytic activity of copper nanoparticles on rGO, cyclic voltammetry was performed in argon (Ar)- and CO2-saturated electrolyte solutions (Fig. 3). It has been recently reported that the selectivity toward the chemical species is highly dependent on the morphology of polycrystalline copper [48]. To limit the number of products from the CO2 reduction, all the measurements reported here have been performed in the absence of water. Specifically, the experiments were performed in acetonitrile using Tetra-n-butylammonium hexafluorophosphate (NBu4PF6, 0.1 M) electrolyte. Acetonitrile was chosen because it can dissolve approximately eightfold more CO2 than water [49], and also because it is a low hydrogen availability medium so that the contribution from HER on the cathodic current can be largely disregarded. In addition, the mechanism for CO2 reduction in low hydrogen availability medium is well known. Only three species can be obtained: carbon monoxide (CO), carbonate (CO3 2−) or oxalate (C2O4 2−) involve 2 electrons each (Equation I and II). Oxalate formation requires C–C coupling which is known to occur at higher overpotentials (>0.9 V vs. RHE) [9, 10]. Experiments in DMF with low proton content similar to acetonitrile have demonstrated that oxalate and CO are the two main products of the reaction and only traces of HCO2 − have been detected [50].

Fig. 3
figure 3

Linear sweep voltammetry (LSV) of CuNPs/rGO, Cu thin film and polymer passivated CuNPs under Argon- and CO2-saturated solutions. The Cu NPs on rGO electrodes exhibit lower overpotential and higher current density compared to other forms of copper tested in this study

Full size image
$$2{\text{CO}}_{2} + 2e^{ - } \to {\text{C}}_{2} {\text{O}}_{4}^{2 - }$$
(1)
$$2{\text{CO}}_{2} + 2e^{ - } \to {\text{CO}} + {\text{CO}}_{3}^{2 - }$$
(2)

The activity of the electrodes can be easily tested by comparing the cathodic current in solution saturated with CO2 with the cathodic current in solution saturated with Ar. It is worth noting that the potentials to reduce Cu (II) to Cu (I) and Cu(I) to Cu metal in organic medium are approximately −0.3 and −0.6 V vs. NHE, respectively [51]. Thus, at the range of negative potentials used here for CO2 reduction, the reduction of the copper oxide shell is thermodynamically favored and the nanoparticles are most likely being reduced to copper metal [48]. From Fig. 3, the CO2 reduction is clearly demonstrated by the high cathodic currents obtained from CuNPs on rGO electrodes immersed in CO2-saturated solution compared to Ar-saturated solution, whereas rGO alone shows a limited catalytic activity (Figure S4). Furthermore, catalytic activity of CuNPs on rGO electrodes is significantly higher for ~70 nm-thick Cu film or CuNPs stabilized by polyallylamine (see Methods for details). Interestingly, a significant cathodic current was not observed using an electrode with passivated CuNPs with polyallylamine. Therefore, it can be stated that the presence of the polymer on the surface of CuNPs inhibits the copper activity. These results demonstrated the clear superior activity of the CuNPs supported on rGO nanosheets. Comparing the electrocatalytic activity of CuNPs on rGO and Cu thin film, it can be observed that higher current density can be achieved with lower overpotentials with rGO decorated with CuNPs. For instance, at −1.2 V vs. NHE the current density observed in the CuNPs/rGO electrode is −0.24 mA cm−2, while the same current density is reached at −1.54 V for Cu thin film. The current density obtained in the CuNPs/rGO at −1.54 V vs. NHE is −0.97 mA cm−2, fourfold better than the Cu thin film.

Stability is an additional parameter in catalysis. To demonstrate the potential use of CuNPs on rGO for CO2 reduction reaction, chronoamperometric measurements were carried out to investigate the stability (Fig. 4a). A constant overpotential of 1.4 V vs. NHE was applied on the CuNPs/rGO electrode for 3 h. Linear sweep voltammetry (LSV) was also performed with intervals of 1 h during the chronoamperometric measurements to check any overpotential shift. CuNPs on rGO exhibit stable current during the CO2 reduction reaction (inset Fig. 4a). The improvement of the overpotential in the first hour under negative potential is likely due to the partial reduction of the oxide passivation layer on the copper nanoparticles as demonstrated by XPS and supported by Li et al. [17]. To gain insight into the reaction kinetics of the reduction process, we have calculated the turnover frequency (TOF) of the CuNP on rGO and Cu thin-film catalysts. The results of which are shown in Fig. 4b. It can be seen that the TOF obtained considering two-electron process (CO, CO3 2− and C2O4 2− products) is larger for CuNPs on rGO than the Cu thin films, suggesting that the CO2 reduction is more favorable for CuNPs on rGO. It can also be seen from Fig. 4b that the number of turnovers increases linearly, indicating no substantial degradation of the activity of the Cu nanoparticles when deposited on rGO.

Fig. 4
figure 4

a Linear sweep voltammetry (LSV) curves for CuNPs on rGO as a function of time (0–3 h). The electrode was held at −1.4 V vs. NHE during testing. The corresponding chronoamperometric curve for CuNPs/rGO is shown in inset. b Turnover frequency (TOF) as a function of the overpotential calculated for CuNPs/rGO (black) and Cu thin film (red). Inset is the evolution of the turnover from CuNPs on rGO measured over 24 h with an overpotential of 1.4 V. The black arrow indicates the change of the electrolyte solution. Calculations are based considering the total amount of copper and a 2-electron process (CO, CO3 2− and C2O4 2− products)

Full size image

Conclusions

Our results suggest that rGO is a promising support for nanosized catalysts. The stabilization of Cu NPs on rGO not only improves the overall electrochemical stability but also enhances the catalytic activity for CO2 reduction, in contrast with Cu NPs with polymer capping layer. Specifically, Cu nanoparticles on rGO have demonstrated lower overpotential and higher current density in comparison to pure Cu thin films and polymer capped Cu NPs. The enhanced stability may be attributed to partial covering of the NPs by a shell of rGO as observed in the TEM. Recent work proposes rough copper surface like nanoparticles as a way to enhance the selectivity towards hydrocarbons through uncoordinated copper sites [48]. These results suggest that rGO decorated with CuNPs may be a promising support for reducing CO2.