Introduction

In the discharging process of green electrochemical devices like fuel cells or metal-air batteries, the output energy capacity is mainly determined by the cathodic oxygen reduction reaction (ORR). Great kinetic loss stemming from this sluggish reaction constitutes the main voltage drop as compared with the ones from the anode process, mass transport and ohmic resistance1. In principle, ideal catalysts for ORR should have the capability to efficiently break the O=O double bond (enthalpy change of 498 kJ mol−1; ref. 2). Platinum (Pt) based materials could meet this requirement. However, they are too scarce to realize widespread use, which thus results in another challenging issue that the consumption of Pt is required to be <0.2 g kW−1 above the cell voltage of 0.65 V (ref. 3). Therefore, various alternatives have been proposed, mainly including two categories of doped carbonaceous materials4,5 and late transition metals (that is, Mn, Fe, and Co; refs 6, 7, 8) in different chemical forms. In reality, fewer materials become available since both the activity and electrolyte corrosion at high potentials have to be considered under the extreme environments of the cathode. Especially, in acidic solutions, most of transition metals and their oxide derivatives are not thermodynamically stable, which is a major obstacle for catalysts developments9. The analogous situation also exists in the reverse process of oxygen evolution reaction (OER), which is significant for the applications of battery charging and hydrogen fuel production. This process is thermodynamically not favoured. The OER catalysts should be able to overcome both the activation energy barrier (Ea) and the standard free energy change (ΔG0=1.23 eV).

The catalytic activity in heterogeneous process is mainly governed by the adsorption enthalpy (Hads) of the substrate on catalysts as understood from the volcano plots10. The champion activity requires Hads to be controlled in a specific range, since higher Hads will block the followed elemental steps by forming stable intermediates; while lower Hads cannot effectively activate the substrate, which makes the adsorption as the rate limiting step. For transition metals, their Hads values essentially rely on the energetic states of d-electrons, as many works suggested that the catalytic activity varied with the electron energies tuned by the foreign orbitals with alloying11,12, coordination13,14 and oxidation15. According to the crystal field approach, d-orbitals in such processes are rearranged into t2g- and eg-orbitals. The eg-electrons are more energetic to be paired. The quantitative conclusion proposed by Gasteriger and co-workers16,17 well reflects this fact, which shows that eg-electrons≈1 is optimal for catalysis by studying a series of intrinsic perovskite materials. On the other side, the Faradaic processes involving O2 usually proceed with multi elemental steps as generally accepted. Previous density functional theory calculation showed that the oxygen containing intermediate (OCI) involved in ORR affected the global catalytic process18. Using a simplified dissociation/association mechanism, the higher activity of Pt over other metals can be explained since the energetic conditions of OCI on this metal surface are moderate for realizing multistep reactions, meeting the volcano relationship. In reality, the diversity of OCI makes ORR more complicated and the identification of these intermediates within their short lifetimes is a tough challenge using available analysis techniques. Up to now, both the global ORR and OER processes have not been fully understood yet. Despite, the volcano plots establish a brief relationship between activity and metal properties, they are still not powerful enough for guiding catalyst design as the detailed adsorption behaviours of OCI are not clarified.

For ORR, the ideal catalysts may be inspired from the biological systems in nature. The typical examples are cytochrome c oxidase and laccase, their active sites contain the similar metal clusters of assembled Cu2+ complexes19,20. Previous work showed that laccase in vitro is more desirable as the quasi-champion potential (onset at 1.2 V versus reversible hydrogen electrode (RHE), pH=4) is observed on its modified electrode21, suggesting that Cu2+ ion is supposed to be an ideal catalytic site as long as the energy level of the d-electrons is tuned to a reasonable state. However, directly mimicking the Cu2+ complexes as in enzyme might not be a reasonable strategy, by which the achieved activity is clearly not as good as the natural state22,23. Low overpotential and high reaction rate are hard to be simultaneously achieved22. The main reasons for this inefficient catalysis can be ascribed to the absence of appropriate mediators for sequential electron transfer by specific amino-acid chains in apo-enzyme components and the steric variation of the coordination structures after the complexes are assembled on the electrode. Thus, developing a proper strategy to utilize the catalytic ability of Cu2+ ion remains a challenge.

In this work, we propose an effective approach to tune the d-electron density of Cu2+ active site and simultaneously increase electron transfer rate via the synergistic effect of electronic connection between Cu2+ and N, and between Cu2+ and metallic Cu (Cu0) in graphene formed by pyrolysis. Impressive ORR performance on the resultant catalyst is achieved in acidic and alkaline media. Especially, in the latter case, superior activity than commercial Pt/C catalyst is observed with an onset potential of 0.978 V versus RHE. Interestingly, the resultant catalyst also shows the excellent activity towards OER with the onset potential closely approaching the standard potential. The catalytic process has been probed by using electrochemical methods and in situ fluorescence spectroelectrochemistry, and the results imply the importance of overcoming the same energy barrier for HO· generation during ORR and OER.

Results

Structural analysis

The catalysts were prepared from the pyrolysis of the mixture of graphene oxide (GO) and Cu2+-1,10-phenathroline (Cu(phen)2) at different temperature, and labelled as CPG-t (t denotes the pyrolysis temperature). After pyrolysis, GO as the support of the catalytic sites is converted into graphene, which greatly increases the conductivity of the resultant materials. The typical morphology of CPG-900 was characterized by transmission electron microscopy (TEM, Fig. 1a). It can be identified that both the basal plains and edges of graphene are coated with heavier nanoparticles as revealed by the darker and discrete regions. The energy dispersive X-ray spectroscopy of the imaged region shows the presence of Cu and N elements (Supplementary Fig. 1). The high resolution TEM image identifies the lattice spacing in one of the nanoparticles as ~0.202 nm (Fig. 1b). The fringes revealed by the X-ray diffraction (XRD) pattern are assigned to the Cu (111) crystal face of 2θ=43.3°. Combining with the coexisted diffractions at 50.5° and 74°, the existence of Cu0 is confirmed in CPG-900 (ref. 24), which infers that the Cu2+ ions can be reduced via the dehydrogenation of 1,10-phenanthroline (Fig. 1c). The composition of CPG-900 was further determined by X-ray photoelectron spectroscopy (XPS), where the total atomic percentage of Cu is measured to be ~3.6% (Supplementary Fig. 2 and Supplementary Table 1, that is, mass percentage of Cu approaching 15%). In the Cu 2p survey, both the 2p3/2 and 2p1/2 bands are broadened compared with that of the Cu(phen)2 precursor, indicating continuum of states for Cu species in CPG-900 (Fig. 1d). According to the results from the XRD patterns, the shoulder signals at binding energies of 932.7 and 952.5 eV are attributed to Cu0. The main signals at higher binding energies of 935.4 and 955.2 eV are originated from the desired Cu2+ ions in Cu2+–N bond because, (i) the presence of Cu 2p satellites (944.3 and 963.4 eV) demonstrates the unfilled electron state of Cu 3d9 orbitals25; (ii) the possible composition of copper oxide is excluded from the TEM and XRD results. The result of the broadening characters of Cu 2p are speculated as the coexistence of Cu0 species and Cu2+ ions connected to the N atoms with different configurations in graphene. The deoxygenation process of GO allows N doping by replacing some active oxygen atoms. In the N 1 s spectra, the signal band for Cu(phen)2 precursor split into three after pyrolysis, reflecting the different N configurations (for example, pyridine- and pyrrole-like species). Meanwhile, the Cu2+–N–C connection is partially preserved during the N doping process as the binding energies of N 1s at 398.6 and 400.9 eV are coincident with the values reported previously26. (Supplementary Fig. 3) The Raman scattering spectra of CPG-900 show several bands as in the case of Cu(phen)2 precursor at low wavenumbers. The Cu2+-N vibration shifts to lower wavenumber of 277 cm−1 for CPG-900 as compared with the one (299 cm−1) for Cu(phen)2, demonstrating the change of the ligand’s environment (Supplementary Fig. 4)27.

Figure 1: Characterizations of CPG-900.
figure 1

Low (a) and high (b) resolution TEM images of CPG-900. The XRD patterns (c) and Cu 2p XPS survey (d) of pristine graphene, Cu(phen)2 and CPG-900. Scale bars, 10 nm (a) and 2 nm (b).

Full size image

ORR activity and catalytic processes

By rotating ring-disk electrode (RRDE) measurements, the current densities of ORR and apparent electron transfer number (n) are summarized in Fig. 2a–d. It is found that the pyrolysis temperature plays an important role in determining ORR activity. Among the CPG-t catalysts prepared at pyrolysis temperature ranging from 600–950 °C, the CPG-900 sample shows the optimum result. The Raman scattering spectra show that the intensity ratio of D-band to G-band increases with the pyrolysis temperature, possibly suggesting the gradual incorporation of Cu2+-N structure into the graphene forming active sites (Supplementary Fig. 5). In 0.5 M H2SO4 solution saturated with oxygen, the onset potential of ORR appears at 0.855 V versus RHE on CPG-900 and the Tafel slope at low overpotentials is fitted as 71 mV per decade, closely approaching the commercial Pt/C catalyst (66 mV per decade, Fig. 2e). The current density is, however, not high under the present conditions. This could indicate that the Cu(phen)2 precursor changes the thermal expansion processes of GO during pyrolysis (that is, generating CO or CO2)28, resulting in the low specific surface area of the resultant CPG-900. Nevertheless, the CPG-900 in 0.1 M KOH exhibits the superior activity towards ORR over the commercial Pt/C catalyst as evidenced by both the higher current density and onset potential (0.978 V versus RHE). The Tafel plots show the smaller slope of 49 mV per decade (Fig. 2f). Combining with the higher reduction potential, the larger exchange current density on CPG-900 can thus be estimated in the early stage of ORR (ref. 29). As expected, the CPG-900 shows little cross effect from methanol oxidation and good CO tolerance as compared with the Pt/C catalyst (Supplementary Fig. 6). In addition, the long-term performance test reveals that the CPG-900 exhibits much better stability than the Pt/C catalyst (Supplementary Fig. 7). The N2 adsorption/desorption isotherms demonstrate the slight variation of BET surface area (from 79.6–76.2 m2 g−1) and inner pore distribution of the CPG-900 after the stability test and the XPS characterizations confirm the remaining Cu2+ and Cu0 components in the catalyst except a minor shift in the binding energy for Cu 2p (that is, 0.3 eV for Cu 2p3/2, Supplementary Fig. 8).

Figure 2: ORR catalysis on CPG-900.
figure 2

ORR polarizations (a,c) and corresponding n (b,d) of the CPG-t catalysts prepared with the pyrolysis temperature of 600 °C (1), 800 °C (2), 900 °C (3) and 950 °C (4) and the corresponding results for the 20% Pt/C catalyst (5) are displayed for comparison. The electrolytes were O2 saturated 0.5 M H2SO4 (a,b) and 0.1 M KOH (c,d). Rotating speed: 1,000 r.p.m., scan rate: 10 mV s−1. (e,f) The Tafel plots are derived respectively from the data in Fig. 1a,c.

Full size image

A recent report30 suggested that ORR could be mediated by the redox of metal active sites existing in the catalysts. This mechanism is not operative in the present study since there is no direct correlation between the redox potential of Cu2+ ions and ORR potential as indicated by the cyclic voltammogram (Supplementary Fig. 9) and differential pulse voltammogram (Supplementary Fig. 10). To explore the catalytic mechanism of CPG-900, Cu species was leached by HNO3 treatment. This process increases the BET surface area (from 79.6–128.1 m2 g−1, Supplementary Fig. 11) and produces more inner micropores, accompanied with the disappearance of Cu signals in TEM images, XRD patterns and Raman scattering spectra (Supplementary Fig. 12). The leached Cu accounts for ~13.3% (mass ratio) of CPG-900 sample as measured by atomic absorption spectroscopy, which is in agreement with the XPS result. The N atomic content of HNO3-treated CPG-900 slightly increases due to the physically attached HNO3 on graphene (Supplementary Table 2). However, the binding energies of these N species shift to lower values, indicating the positively charged Cu2+ ions detached from the N ligand (Supplementary Fig. 13d). As expected, HNO3 treatment brings significant degradation in the catalytic activity of GPG-900 towards ORR. As shown in Fig. 3a–d, the half-wave potential (E1/2), reduction current and n greatly decrease in both 0.5 M H2SO4 and 0.1 M KOH. Thus, it can be reasonably inferred that, (i) the Cu2+ ions in CPG-900 act as the catalytic sites since HNO3-treated CPG-900 sample becomes less active for ORR and Cu0 is known to be inert towards ORR. The observed catalysis of HNO3-treated CPG-900 is from the remaining Cu2+ ions (Supplementary Fig. 13), and the catalytic current can be enhanced by re-coordinating more Cu2+ ions into the residual N ligands doped in graphene (Supplementary Fig. 14); (ii) Cu2+ ions are both connected with the N ligand and surface Cu0 of Cu nanoparticles (Fig. 3e). The coordination of N to Cu2+ ions should not be as perfect as that in the case of the Cu(phen)2 precursor due to the steric rigidity of graphene structure and nanoparticle. Such imperfect coordination environments undoubtedly changes the spatial position of O2 adsorption on Cu2+ ions, which might be more close to mimic the ones of the Cu2+ centre in laccase; (iii) after pyrolysis, the electron donating ability of N ligand becomes weaker as the electron pairs are conjugated by the π-electrons of graphene30 as well as the imperfect coordination configurations. In addition, the Cu0 can donate electrons to the connected Cu2+ ions as indicated by the slight shift of the binding energy of Cu 2p (Fig. 1d). X-ray absorption spectroscopy (Fig. 3f) measurements show a sharp band of the 2p3/2 to 3d transition of Cu2+ ions in CPG-900 at 930.1 eV, which is lower than Cu(phen)2, but higher than Cu(Ac)2. This result indicates that the electron donating ability of Cu0 seems to be relatively moderate, but is still stronger than the 1,10-phenanthroline ligand. (iv) The extended X-ray absorption fine structure measurements of the Cu K-edge in CPG-900 exhibit the typical Cu nanocrystalline features. In the derived oscillation spectra, the amplitude of CPG-900 is whereas smaller than that of Cu foil, indicating the structural defects in Cu nanoparticle due to the size effect (Supplementary Fig. 15). The surface Cu0 atoms should provide the free electrons to fill the d-orbitals of Cu2+ ions via dangling bonds. A similar phenomenon has been reported that the Au1+/3+ ions, which are stable on the Au0/CeO2 nanoparticle system, function as the active sites for catalysis31. Thereby, the synergistic electronic effect from Cu0 and N ligands should make a preferable bonding between O2 and Cu2+ ions. This configuration is distinctly different from the cases for the neighbouring transition metals (that is, Fe/Co2+ coordination structure)8,30,32. The present results reflect that the catalytic activity of transition metals is governed by both the electron arrangement of metal centres and its coordination surroundings.

Figure 3: Active sites in CPG-900.
figure 3

ORR polarizations (a,c) and corresponding n (b,d) obtained on CPG-900 before and after treated with HNO3 in O2 saturated 0.5 M H2SO4 (a,b) and 0.1 M KOH (c,d). Rotating speed: 1,000 r.p.m., scan rate: 10 mV s−1. (e) Schematic illustration shows the proposed active sites in CPG-900. (f) XAS characterizations of the Cu L3-edge in Cu(Ac)2, Cu(phen)2, CPG-900 and Cu foil.

Full size image

It is clear that the n of ORR strongly depends on the overpotential and electrolyte pH. In the RRDE experiments, the n is governed by the ratio of the H2O2 oxidation reaction current on ring to that of ORR current on disk. Although the reactions of H2O2 on the disk electrode are usually neglected, they surely affect both the ring and disk currents (Fig. 4a). As generally accepted, the reaction rates of H2O2 are determined by its adsorption/desorption processes33,34. At open circuit potential, the influence of H2O2 decomposition on the RRDE results is estimated to be rather limited (Supplementary Fig. 16). While within the potentials for ORR, the redox behaviours of H2O2 are involved on CPG-900, which closely correlate to the n of ORR. As shown in Fig. 4b, in the early stage of ORR (0.5 M H2SO4, A zone), H2O2 is reoxidized to O2 on the disk electrode, which reflects moderate adsorption energy of H2O2 on CPG-900. Thus, a more favoured H2O2 decomposition is also expected on CPG-900 comparing to that at open circuit potential. These processes result in much less H2O2 detected on the ring electrode and higher n (4-3.8). In the following stage (B zone), the redox of H2O2 appears to be relatively slow, showing that the desorption of H2O2 from CPG-900 becomes dominant, thus, more H2O2 is detected on the ring electrode and the n gradually reaches the minimum of 3.7. At more negative potentials (C zone), the n arises due to the contribution from the H2O2 reduction reaction (HRR). On the other hand, the n of ORR continuously decreases with increasing the overpotential in 0.1 M KOH (Fig. 4c). In this case, only two zones (A and B) in the relationship between n and overpotential appear due to the absence of fast reduction of H2O2 process. In A zone, H2O2 is not detected at the ring electrode since all the formed H2O2 during ORR is reoxidized to O2 or/and decomposed, and thus, the n keeps as 4. This result benefits for the terminal fuel cell assembly because the transient state of ORR is used to assess the half-cell voltage, where the catalytic current and overpotential reach balanced3. With the increase of overpotential, the n decreases continuously from 4 to 3.7, reflecting that the H2O2 reduction and/or decomposition gradually become trivial comparing to the desorption of H2O2 from CPG-900.

Figure 4: Influence of H2O2 adsorption/desorption on n of ORR.
figure 4

(a) Schematic illustration of the different reaction pathways of H2O2 involved in ORR (the subscripts of ‘sol’ and ‘ads’ stand for the reaction species dissolved in bulk solution and adsorbed on CPG-900, respectively). The polarization curves of 5 mM H2O2 on CPG-900 in 0.5 M H2SO4 (b) and 0.1 M KOH (c) was characterized using cyclic voltammetry at a rotating rate of 1,000 r.p.m., 10 mV s−1 in N2 atmosphere. ORR current, H2O2 oxidation reaction (HOR) current and n in O2 saturated electrolyte are all summarized in the same figure, aiming to explain the relationships of n versus overpotential.

Full size image

Stepwise electron transfer of ORR

It is assumed that OCI is involved in the ORR. However, verifying such assumption is rather difficult as the lifetime of OCI is quite short in aqueous media, especially within the interface layer with applied electrical field35,36. To solve this issue, we applied in situ fluorescent spectroscopy combining with electrochemical techniques to probe the formation of OCI using 2′,7′-dichlorodihydrofluorescein (DCDHF) as an indicator37. Upon reacting with the OCI generated in ORR, DCDHF will be converted to fluorescent dichlorofluorescein (DCF, Fig. 5a). As shown in Fig. 5b, during ORR at 0.848 V versus RHE (the cathodic current is about −20 μA), an emission at 522 nm continuously grows with electrolyzing time in 0.1 M KOH containing 2.5 μM DCDHF. To verify the detected OCI is a transient state rather than a byproduct, ORR on CPG-900 was performed at 0.849 V versus RHE in a solution of 0.01 M KOH in 1:1 (v/v) water/ethanol solvent saturated with air. A distinct decrease in the reduction current occurs upon addition of DCDHF (Fig. 5c), demonstrating that the formation of OCI is followed by an electrochemical reaction that it can further receive electron from electrode to accomplish the ORR and a parallel chemical reaction with DCDHF to form fluorescent DCF. The appearance of the fluorescence emission from H2O2 can be ruled out as no significant fluorescence emission is observed when 8 μM H2O2 is solely added into 0.1 M KOH without applying potential (Supplementary Fig. 17). Further experiments identified the observed OCI as HO· radical by using a specific signal molecule, coumarin. As shown in Fig. 5d, the fluorescence emission at 500 nm increases due to the hydroxylation of coumarin seventh position during the ORR. This result agrees with the observation acquired on a Pt electrode38.

Figure 5: Probing the multi elemental steps of ORR.
figure 5

(a) Schematic illustration of the conversion from non-fluorescent DCDHF to fluorescent DCF by the OCI formed in ORR on CPG-900. (b) In situ fluorescence spectroelectrochemistry performed at a constant potential in O2 saturated 0.1 M KOH containing 2.5 μM DCDHF for 10 min (from bottom to top). (c) I–T curve of ORR on CPG-900 at 0.849 V versus RHE. The electrolyte is 0.01 M KOH dissolved in 1:1 (v/v) water/ethanol solution with saturated air. The red arrow denotes the addition of 12.5 μM DCDHF into the electrolyte, showing an abrupt decrease in reduction current due to the competitive chemical reaction of OCI with DCDHF to the electrochemical reaction by receiving electrons from CPG-900. (d) In situ fluorescence spectroelectrochemistry performed at 0.799 V versus RHE with 60 μM coumarin to specifically verify the presence of HO· radicals. To avoid the hydrolysis of ester ring in coumarin, the electrolyte was diluted to be 0.01 M KOH saturated with O2 (refs 39, 40). The electrolysis lasted for ~50 min (the black and red lines stand for the initial and final signals, respectively).

Full size image

Extended application for OER

The catalytic process of OER on CPG-900 is also analyzed by RRDE method to avoid the influence from produced bubbles (Fig. 6a). The obtained onset potential closely approaches the standard potential, as low as 1.34 V (in 0.1 M KOH) and 1.23 V (in 0.5 M H2SO4) versus RHE, which are superior to many OER catalysts reported recently (Supplementary Table 3). Being analogous to ORR, the selectivity of OER is studied by recording the ring current at a potential with respect to the H2O2 oxidation. In 0.5 M H2SO4, a much smaller anodic current at the ring electrode (1/5,000~1/2,500 of the disk current) occurs, suggesting that the global OER is almost a 4e-transfer but is still involved with trace amount of the H2O2 byproduct. Interestingly, in 0.1 M KOH, a relatively small and cathodic ring current is observed although the ring electrode potential is kept more positive for H2O2 oxidation (Fig. 6b). At 1.348 V versus RHE, the reduction behaviours of O2, H2O2 or H2O on the Pt ring electrode are rather impossible. The observed cathodic ring current is more likely originated from the reduction of OCI. Thereby, the OER is probed by the in situ fluorescence spectroelectrochemistry. Using DCDHF indicator, the fluorescence emission at 522 nm keeps increasing at an anodic current of 3–4 μA, demonstrating that the OCI is exactly involved in OER (Fig. 6c and Supplementary Fig. 18).

Figure 6: A brief correlation between ORR and OER.
figure 6

OER polarizations (a) and corresponding ring currents (b) obtained on CPG-900 in 0.5 M H2SO4 (olive line) and 0.1 M KOH (orange line). Rotating speed: 1,000 r.p.m., scan rate: 10 mV s−1. (c) In situ fluorescence spectroelectrochemistry performed at constant potential in 0.1 M KOH containing 2.5 μM DCDHF. (d) Schematic illustration of the electrochemical conversion between O2 and HO−, which experiences a common transient state of HO· radical.

Full size image

Discussion

According to the above results, the most possible OCI in OER could be O2− and HO· radicals. But injecting electrons to O2− consumes too much energy. In contrast, the elemental reaction (HO· to HO−, E0=1.9 V versus standard hydrogen electrode38) is more energetically favourable, which can reasonably explain the origin of the reduction event on the Pt ring in 0.1 M KOH. Using coumarin indicator, the presence of HO· radicals is confirmed by the increased fluorescence emission at 500 nm (Supplementary Fig. 19). Accordingly, we establish a brief relationship between ORR and OER. With respect to the standard potential of HO· radical, the overpotential for both ORR and OER should be at least 1.25 V (O2+2H2O+ne−→(4−n)HO·+nHO− E0=−0.024 V versus standard hydrogen electrode38). The actual energetic state of this specie on the CPG-900 surface is altered, which would be a major reason to decrease the overpotentials of OER and ORR in alkaline solutions (Fig. 6d). In addition, OER is normally not thermodynamically favoured. But in the present study, the observed overpotential of OER is lower than ORR, suggesting that the initial adsorption of HO− and O2 on CPG-900 is different. The more favourable bonding configuration might be the adsorption of polarized HO− on Cu2+ ions.

In summary, a high performance Cu nanocomposite (CPG-900) is synthesized by the pyrolysis of the mixture of GO and Cu(phen)2 as inspired by the catalytic sites in natural enzyme. The electron density of Cu2+ active sites in CPG-900 is tuned by the electron donation effect from both the neighbouring Cu0 of Cu nanoparticle and the N ligand incorporated in the rigid graphene. This imperfect coordination configuration provides an optimal environment for electronic bonding of O2 to Cu2+ ions. The ORR occurs on the CPG-900 with overpotential of only ~0.25 V in 0.1 M KOH, referencing to the standard potential. Since the reactions of H2O2 involve in the ORR process, the n of ORR varies with the overpotential and electrolyte pH. In addition, the CPG-900 shows excellent catalytic activity towards OER with the onset potential approaching the standard potential. Results from the in situ fluorescence spectroelectrochemistry measurements establish a brief correlation between ORR and OER in alkaline solutions, showing that both the processes are intermediated by the HO· radical. This work offers a new insight into the OER/ORR mechanism and possible strategies to design high performance catalysts.

Methods

Reagents

Graphite powder (99.9995%, 100 mesh) was purchased from Alfa Aesar Company (UK). Both Cu(Ac)2 and 1,10-phenathtoline (phen) were purchased from Nanjing Chemical Reagent Co. Ltd (China). Other reagents were of analytical-reagent grade. All aqueous solutions were prepared with Millipore water (resistivity of 18.2 MΩ cm).

Synthesis of Cu(phen)2-graphene nanocomposite

The Cu2+-1,10-phenanthroline (Cu(phen)2, Supplementary Fig. 20) precursor was synthesized by heating the mixed solution of Cu(Ac)2 (0.5 mmol dissolved in 30 ml dimethyl formamide) and 1,10-phenanthroline (1 mmol dissolved in 70 ml CH2Cl2) at 35 °C for 8 h. The solvents of the solution were then removed by rotary evaporation. The remaining solid was successively washed by acetone and CH2Cl2 until turning into emerald. The GO precursor was synthesized via chemical exfoliation of graphite. Briefly, the mixture of graphite (6 g), K2S2O8 (3 g) and P2O5 (3 g) were heated at 80 °C for 8 h in concentrated H2SO4. The as-obtained graphite (2 g) was oxidized by KMnO4 (6 g) in concentrated H2SO4 at 35 °C for 2 h. The mixture of Cu(phen)2 and GO was pyrolyzed with a mass ratio of 2:1 in Ar atmosphere, resulting in Cu(phen)2-graphene nanocomposites (CPG-t, t denotes the pyrolysis temperature ranging from 600–950 °C). To exclude any unstable phases in acidic media, the pyrolyzed product was treated in H2SO4 (0.5 M) at 80 °C for 2 h before each characterization. In a controlled experiment, the CPG-900 was further treated with HNO3 (1.3 M) for the leaching of Cu. This process was performed at 80 °C for 2 h in air atmosphere. Afterwards, the HNO3 solution was removed by centrifugation and the resultant solid product was washed with ethanol several times and finally dried under pressure at room temperature.

Apparatus

The ultraviolet–visible spectra were obtained on a Nanodrop-2000C spectrophotometer (Thermo Fisher Scientific Company, USA). The electron paramagnetic resonance characterizations were performed on an EMX-10/12 spectrometer (Bruker Company, Germany). The morphologies of each sample were characterized by TEM (JEM-2100, JEOL Company, Japan). The TEM samples were prepared by drying a droplet of sample suspension on Ni-grids with carbon film. The XRD and Raman spectra were carried out on a X'TRA (ARL Company, Switzerland) and a FT-Raman spectrometer (Bruker company), respectively. The XPS were obtained on a PHI 5000 VersaProbe spectrometer (UlVAC-PHI Company, Japan). The curve fitting of elemental fine spectra was performed using 20% Gaussian–Lorentzian peak shape. The atomic absorption spectroscopy was performed for measuring the leached Cu (180-80 type, Hitachi Company, Japan).

The hard X-ray absorption experiments at Cu K-edge were performed at room temperature in the transmission mode using a ion chamber detector at beam line BL14W1 of the Shanghai Synchrotron Radiation Facility (SSRF), China. The photon energy was calibrated with the first inflection point of Cu K-edge in Cu foil. The soft X-ray absorption experiments at Cu L-edge were performed at beam line BL08U. During the measurements, the synchrotron was operated at an energy level of 3.5 GeV and a current between 150–210 mA.

Electrochemical measurements

A glassy carbon (GC, 5.61 mm in diameter) disk and a Pt (0.84 mm in width) ring were applied as the working electrodes fixed on the rotating apparatus (PINE Company, USA). Prior to electrochemical measurements, the GC disk and Pt ring electrodes were polished with 0.05 μm alumina slurries. Then, the electrodes were successively sonicated in ethanol and deionized water. The Pt ring electrode was further electrochemically polished in 0.5 M H2SO4. A 6 μl suspension (10 mg ml−l CPG-t sample dispersed in ethanol) was casted on the GC disk and was later covered with a layer of Nafion film while without contacting the Pt ring electrode. The ORR and OER electrochemical measurements were performed on a CHI 900D potentiostat (CH Instrument Company, USA) with Ag/AgCl electrode and Pt wire as reference and counter electrodes, respectively. The onset potential is determined by the value where current starts to show above (OER)/ below (ORR) zero. The collection efficiency (N) of the working electrodes modified with catalysts was measured by using 5 mM [Ru(NH3)6]3+/0.1 M KCl solution in N2 atmosphere. The apparent electron transfer number (n) of ORR was calculated by the following equation:

where, id stands for the current of ORR on GC disk electrode and ir stands for the current of H2O2 oxidation reaction on Pt ring electrode.

The potentials of ORR and OER were converted to the values referring to the RHE. In 0.5 M H2SO4, the open circuit potential remains at 0.2249 V (E(RHE)=E (Ag/AgCl)+0.2249 V), rarely showing fluctuation; in 0.1 M KOH, the open circuit potential remains at 0.9482 V (E(RHE)=E (Ag/AgCl)+0.9482 V), showing slight variation of 0.0001 V (Supplementary Fig. 21).

In situ fluorescence spectroelectrochemistry

The experiment was designed by combining an electrolysis unit (CHI 660D, CH Instrument Company) with fluorescence spectrophotometer (RF-5301PC, Shimadzu, Japan). Typically, CPG-900 suspension (1 μl) was dropped on GC electrode (3 mm in diameter) in a three-electrode system. The DCDHF (2.5 μM) signal probe was added into 0.1 M KOH before ORR/OER. The stock solution of 1.25 mM DCDHF was prepared using 0.01 M KOH in water/ethanol (1:1, v/v) solvent. During electrolyzing, the OCI involved in ORR/OER instantaneously react with DCDHF, generating fluorescent DCF. Verification of the presence of HO· radicals was performed at room temperature (14 °C estimated by an infrared thermometer) using specific coumarin signal probe. To avoid the hydrolysis of contained ester ring, the concentration of KOH electrolyte was diluted to 0.01 M. With the generation of HO· species, the hydroxylation of coumarin will occur at the seventh position, greatly increasing the fluorescent emission upon excitation at 360 nm.

Additional information

How to cite this article: Wang, J. et al. Bioinspired copper catalyst effective for both reduction and evolution of oxygen. Nat. Commun. 5:5285 doi: 10.1038/ncomms6285 (2014).