Tandem supported, high metal-loading, non-PGM electrocatalysts for oxygen reduction reaction

Oxygen reduction reaction (ORR) is a critical process for a diverse set of applications ranging from fuel cells to metal–air batteries and wastewater treatment.^1–6 In particular, due to the rapidly expanding hydrogen economy, fuel cells have attracted significant attention in recent years. Among the existing technologies, alkaline electrolyte-based fuel cells, especially alkaline anion exchange membrane fuel cells (AAEMFCs), are highly promising due to their superior efficiency and high stability compared to acid electrolyte-based counterparts.⁷ Traditional ORR catalysts are based on platinum group metal (PGM), among which Pt/C has been considered the benchmark.^4,8–10 However, these catalysts are limited by the scarcity and high cost of Pt, as well as carbon corrosion effects during operation, especially in start-up and shut-down processes.^9,11,12 To reduce the cost, significant efforts in the research community have been devoted to synthesizing and improving non-PGM catalysts, achieving ORR activities close to those of Pt/C.^13–18 These catalysts require the dispersion of isolated metal atoms on a substrate support, which is typically based on a carbon framework with nitrogen or sulfur dopants to anchor the metal atoms. Despite their decent ORR activities, such structures still suffer from carbon corrosion effects.¹⁹ In addition, their synthesis usually requires high-temperature pyrolysis above 700 °C, which tends to produce nanoparticle aggregates and limits the loading of single metal atoms.

To improve both the stability and metal-loading capacity of non-PGM catalysts, an alternative support, g-C₃N₄, has gained interest in recent years.^13,20,21 As a polymeric material that can be produced from the pyrolysis of nitrogen rich molecules, g-C₃N₄ is known for its excellent physicochemical stability, remaining stable up to ∼600 °C in air or inert gas.^22,23 Computational studies have shown that the N-rich pores of g-C₃N₄ can bind a range of transition metal atoms with a higher binding energy than the aggregated metal clusters, enabling the effective stabilization of metal single atom sites.^24,25 Due to these favorable properties, g-C₃N₄ based non-PGM catalysts have been widely used for thermal and photocatalysis, and metal loading up to >20 wt. % has been reported.^26–30 Hindering its use in electrochemistry, however, is its high intrinsic bandgap of 2.7 eV and its tendency to form thick layers, both of which limit the electrical conductance.³¹ Efforts have been made on coating metal-doped g-C₃N₄ thin layers on carbon black to form hybrid-supported catalysts, but the metal loading has been limited to <2 wt. %.³² 

In this work, we use one-pot pyrolysis to produce Cu–g-C₃N₄/C catalysts, where the carbon black and g-C₃N₄ serve as core/shell tandem support for the Cu atoms (Fig. 1). Carbon black is chosen to enhance the overall electrical conductivity, g-C₃N₄ to maximize loading capacity and stability, and Cu to enable ORR activity while ensuring stability. By controlling the pyrolysis temperature, we modulate the extent of g-C₃N₄ polymerization as well as Cu loading, thus tuning and optimizing the eventual alkaline ORR activity and stability.

The synthetic method is summarized in Fig. 1. The Cu–g-C₃N₄/C catalysts [CuDCDVu-(x)T (x = temperature in °C)] were prepared via the following steps. We first stirred dicyandiamide (DCDA) with CuCl₂ in methanol for 1 hour at room temperature. Cu (II) salts have previously been demonstrated to complex with DCDA and catalyze the insertion of alcohol groups, resulting in the precipitation of a Cu DCDA complex.³³ The resulting slurry was sonicated with Vulcan XC 72R carbon black (Cabot Corp.) and dried under vacuum. The resulting powder was then pyrolyzed at a temperature between 550 and 650 °C under nitrogen for 2 h. Following pyrolysis, the materials were immersed in 1M HCl for ∼16 h and then filtered and rinsed with ultrapure water to remove weakly adsorbed Cu and/or CuCl₂ species.

Transmission electron microscopy (TEM) was used to image the structure of the catalyst particles as well as control samples (Figs. 2 and S1). Similar aggregates of spheres with diameters ∼40 to 50 nm were present in the carbon black, DCDVu-600T (made from the same pyrolysis procedure as CuDCDVu-600T except that no Cu precursors were used), and CuDCDVu-600T [Figs. S1(a)–S1(c)]. This size is consistent with that of Vulcan XC 72R as described by the manufacturer, suggesting that the pyrolysis process did not induce large changes in the particle diameter. No metal nanoparticles were visible in any of the TEM images, although this cannot rule out the possibility of trace amounts of small particles/clusters present in the overall sample. High resolution images revealed the layered structures of the catalysts [Fig. 2(a), right, and Fig. S1(d)]. The interlayer distance varies between 0.3 and 0.4 nm in most regions and can reach up to ∼0.6 nm near the surface [Fig. S1(c)]. Considering that pristine graphite and g-C₃N₄ have similar interlayer distances (∼0.34 vs ∼0.33 nm),³⁴ it is hard to distinguish these compositions directly from the TEM images. The larger spacings we observed are likely due to the Cu incorporation into g-C₃N₄ and/or strain effects of the curved layers. To further evaluate the chemical distribution of these catalysts, we carried out energy dispersive spectroscopy scanning transmission electron microscopy (EDS-STEM) measurements. As shown in Fig. 2(b), we observed a homogeneous distribution of Cu and N on the carbon black particles, indicating an effective coverage of Cu–g-C₃N₄ on the carbon support as well as a uniform dispersion of Cu in the catalyst layer.

Powder x-ray diffraction (XRD) (Bruker D8, Cu kα₁) patterns were taken of the catalysts and are shown in Fig. 3(a). Peaks were present at 24.7° and 43.63° for all samples, consistent with the (100) and (111) reflections of graphite domains in amorphous carbon³⁵ and in good agreement with the unmodified Vulcan carbon black. Note that for the CuDCDVu-625T and CuDCDVu-650T samples, the 43.63° peak, though still present, is beneath the two sharper Cu and Cu₂O peaks, as discussed later. For CuDCDVu-550T, an extra-sharp peak at 27.51° was observed, which is absent in the bare carbon black sample. This new peak corresponds to the (002) reflection of g-C₃N₄,³⁶ further confirming the presence of g-C₃N₄ in this pyrolyzed sample. At increasing temperatures, the 27.51° peak broadened and was undetectable for the sample pyrolyzed at 650 °C, implying a loss of crystalline g-C₃N₄ layers.³⁷ In fact, the TEM images of CuDCDVu-600T (Figs. 2 and S1) already revealed disorder effects in the interlayer packing. Above 600 °C, peaks indicating the presence of Cu nanoparticles and Cu₂O were present,³⁸ suggesting a breakdown of Cu–N and aggregation of Cu at higher temperatures. At 600 °C or lower, Cu and Cu₂O particles are either absent or present only in trace amounts. As 600 °C has been reported to be a stability limit for g-C₃N₄,³⁹ the lack of Cu nanostructures in samples below this temperature is consistent with preferential incorporation of isolated Cu atoms into carbon nitride rather than agglomeration into nanoparticles. Note that, even when small amounts of Cu and/or Cu₂O particles are present, they are not expected to affect the ORR performance as discussed later, due to the weak or negligible ORR activities of Cu and Cu₂O as reported before.^40,41

We further carried out confocal Raman measurements^42,43 and observed pronounced peaks at 1347 and 1594 cm⁻¹ for carbon black and all the pyrolyzed samples, corresponding to the D and G modes of graphite, respectively (Fig. S2). The nearly constant peak positions and intensity ratio confirm that the carbon black structure remained unchanged during all the pyrolysis processes.

For the pyrolyzed samples, CHN elemental analysis was performed to quantify the carbon, nitrogen, and hydrogen contents, and inductively coupled plasma optical emission spectroscopy (ICP-OES) was used to measure the copper content [Table S1 and Fig. 3(b)]. With increasing pyrolysis temperatures, the weight percentage of nitrogen decreased, carbon increased, and copper first increased, then decreased. The maximum Cu loading was observed to be ∼26 wt. %, occurring at a pyrolysis temperature of 600 °C. This metal loading is among the best of all the reported non-PGM catalysts.^44–48 In addition, a trace amount of hydrogen, between 0.2 and 0.5 wt. %, was detected for each sample (Table S1).

To quantify the local chemical coordination of the pyrolyzed samples, we carried out x-ray photoelectron spectroscopy (XPS). The survey spectra, shown in Fig. S3, reveal the presence of Cu, N, C, and O. The presence of oxygen is typical for single-atom catalysts, resulting from either adventitious oxygen in the air or from the oxygen present in the solvated methanol complexes used as precursors.^33,49 The Cl from the CuCl₂ precursor was either absent or only present in trace amounts after the synthesis, likely due to evaporation during pyrolysis or removal in the following water-rinsing step.

Figure 4 shows the high-resolution spectra of N 1s and Cu 2p, as well as their quantitative analysis. In the N 1s spectra [Fig. 4(a)], for all samples, three peaks were observed at 398.6, 399.5, and 401.0 eV, which can be assigned to pyridinic-N (nitrogen present within triazine and heptazine rings of g-C₃N₄), amine-N (terminal and linking amine groups), and the quaternary nitrogen present in the center of triazine rings, respectively.³² As the pyrolysis temperature increased, the proportion of pyridinic-N steadily declined [Fig. 4(c)]. Amine-N proportion was largely unaffected by the pyrolysis temperature and averaged 39.3%. The quaternary-N proportion was maximized at 17.9% for CuDCDVu-650T and averaged 14.1% for all samples. An extra peak emerged at ∼397.4 eV in samples pyrolyzed at 600 °C or higher, and it was assigned to Cu–N species based on previous reports.⁵⁰ The presence of C–N bonds is further confirmed by the C 1s spectra (Fig. S4).

For the Cu 2p spectra [Fig. 4(b)], the 2p_1/2 and 2p_3/2 peaks have nearly identical shapes, each containing two components. Here we focus on the analysis of the 2p_3/2 peak. The peak component at ∼934.6 eV is attributed to Cu²⁺, while the component at 932.7 eV is between the values commonly assigned to Cu⁺ and Cu⁰, suggesting an oxidation state less than +2.^51,52 We attribute the 932.7 eV component to mainly Cu⁺, considering that crystalline Cu⁰ was either not observed or existed only in small amounts in XRD [Fig. 3(a)]. The presence of broad Cu²⁺ satellites at ∼944 eV further verifies the presence of Cu²⁺.⁵³ The ratio of Cu⁺ to Cu²⁺ was maximized for CuDCDVu-575T at 1.52 but was near 1.0 for all other pyrolysis temperatures [Fig. 4(d)]. As the source of Cu was CuCl₂, Cu(I) must have been produced during the pyrolysis, presumably from Cu(II) reduction by either ammonia or generated organic species.⁵⁴ 

Considering the intact graphitic structure of the carbon black [Figs. 3(a) and S2], the presence of crystalline layered g-C₃N₄ backbone structure [Fig. 3(a)], and the existence and roughly uniform distribution of C, N, and Cu [Figs. 2(b) and 4] in the pyrolyzed catalysts, we believe that the core/shell configuration shown in Fig. 1 is a reasonable estimate of the Cu–g-C₃N₄/C catalyst structure.

CuDCDVu(x)T were used as catalysts for ORR in an alkaline electrolyte (0.1M KOH aqueous solution). Catalyst ink was drop-cast onto a 5 mm diameter glassy carbon rotating disk electrode (RDE) to a mass loading of 609.2 μg/cm². Scanning electron microscopy (SEM) and energy dispersive spectroscopy (EDS) images reveal the nearly complete coverage of the glassy carbon surface by the catalysts (Fig. S5). To quantify the ORR activity, we performed linear sweep voltammetry (LSV) measurements with disk rotation in O₂-saturated electrolytes (Fig. 5).⁵⁵ Control measurements in nitrogen-saturated electrolytes, shown in Fig. S6, confirm that the cathodic currents obtained in Fig. 5 are largely due to ORR. The following discussions on ORR are based on results measured under O₂-saturation.

As the pyrolysis temperature increases from 550 to 650 °C, the half-wave potential (E_1/2) and onset potential (E_onset) both become more positive before shifting negative, peaking at E_1/2 = 0.81 V and E_onset = 0.90 V vs RHE with T = 600 °C, comparable to the ORR activity of the commercially standard 20% Pt/C catalysts reported at E_1/2 = 0.84 ± 0.03 V and E_onset = 0.96 V.^56–58 A similar trend is observed for the limiting current density (j_l), although catalysts pyrolyzed at more than 600 °C tend to show large variations. To achieve a more comprehensive quantification of the ORR activity trend, we performed three separate LSV measurements for each pyrolysis temperature (Fig. S7) and summarized the average E_1/2, E_onset, and j_l and their 95% confidence intervals in Table S2. These results reveal the consistent peaking of ORR activity at 600 °C. This activity trend is consistent with the elemental analysis results [Fig. 3(b) and Table S1]. Specifically, we find that the Cu loading is strongly correlated with the ORR activity, as a higher Cu concentration always leads to more positive E_1/2 and E_onset (Fig. S8). In contrast, the C and N contents exhibit no correlation with the ORR activity (Fig. S8). The observed structure-activity relationship reveals the importance of Cu loading in boosting ORR activity for non-PGM catalysts.

To examine the origin of the catalytic activities of the tandem supported catalysts, a series of control samples, including carbon black itself and other pyrolyzed structures with fewer precursor components, were prepared, measured, and compared to CuDCDVu-600T [Fig. 5(b)]. We observed much lower ORR activity in all these control samples, revealing that the carbon support, g-C₃N₄ framework, and Cu doping are all essential for promoting ORR.

We further quantified the ORR kinetics by extracting the electron transfer number (n) from Koutecký–Levich plots (Fig. S9). For CuDCDVu-600T, n was found to be 3.84 ± 0.10, implying a high selectivity of 92% ± 5% for four-electron reduction. Other catalysts measured also showed n values around 3.8. Tafel analysis, derived from chronopotentiometry measurements, is shown in Fig. S10. We find that CuDCDVu-600T has a Tafel slope of 53.3 mV dec⁻¹, lower than that of the reported Pt catalysts for 0.1M KOH (∼60 mV dec⁻¹),⁵⁹ indicating a facile reaction kinetics.

To assess the stability of catalysts, accelerated stability tests (AST) were run to measure the loss of current and half-wave potential, and the results are displayed in Fig. 6. To measure the stability of the catalysts under oxygen reduction conditions, the potential was scanned from 0.6 to 1.0 V (vs RHE) and back in a triangle waveform over 10 000 cycles at a rate of 100 mV s⁻¹, adopted from similar test protocols from the literature.^60–62 Polarization curves taken before and after potential cycling are displayed in Fig. 6(a). CuDCDVu-550T showed practically no change in E_1/2 after AST. CuDCDVu-600T experienced only a small drop in E_1/2 of ∼16.5 mV. The E_1/2 of CuDCDVu-575T and CuDCDVu-625T decreased by ∼9 mV each, while CuDCDVu-650T dropped by ∼50 mV. Note that the 575, 600, and 650 T samples exhibited a larger limiting current after AST, which is partially responsible for the decrease in E_1/2 and likely due to further catalyst activation effects.

During the start-up and shutdown processes of fuel cells, the cathode may experience transient voltage spikes up to 1.4 V vs RHE.⁶³ To evaluate the stability of our catalysts under such extreme positive potentials, we further performed a series of anodic AST by scanning the potential back and forth between 1.0 and 1.5 V (vs RHE) at a rate of 500 mV s⁻¹ for 5000 cycles. This protocol was adopted from previous reports on the carbon corrosion effects of Pt/C catalysts.^61,64 Polarization curves before and after the cycling are shown in Fig. 6(b). CuDCDVu-550T again exhibited the greatest stability, with a nearly negligible change in the current density. For the 575, 600, 625, and 650 T samples, E_1/2 dropped by 16.6, 33.1, 57.4, and 37.8 mV, respectively. Considering the harsh test conditions, these catalysts have demonstrated high stability. In comparison, Fe–N–C and Pt/C catalysts have been known in some cases to lose practically all ORR activity from extensive carbon corrosion.^65,66 Further improvements in our catalyst stability will likely be achieved by optimizing the Cu–g-C₃N₄ coverage on the carbon black via the modulation of precursor composition and ratio, pyrolysis time, and heating rate.

In addition to AST in pure electrolytes, we further performed stability tests in the presence of methanol, both as model impurities in hydrogen and for evaluating potential direct methanol fuel cell applications.⁶⁷ We observed superior stability of CuDCDVu-600T compared to commercial Pt/C catalysts under 1M methanol in O₂-saturated 0.1M KOH (Fig. S11).

We have compared the ORR activity and stability of our CuDCDVu-600T catalysts with existing single-atom catalysts synthesized at low pyrolysis temperature (≤800 °C) and tested under alkaline conditions, as summarized in Table S3. The ORR activity of our catalysts is among the best on this list, while their stability cannot be directly compared due to a lack of extensive cycling tests in the existing literature.

In summary, Cu-based ORR electrocatalysts were prepared by pyrolysis of CuCl₂ with dicyandiamide and carbon black at 550–650 °C. We observed the successful incorporation of Cu atoms into the C/g-C₃N₄ tandem support and pronounced structural evolution as a function of pyrolysis temperature. At 600 °C, we produced a Cu loading of ∼26 wt. %, resulting in high alkaline ORR activity of E_1/2 = 0.81 V and E_onset = 0.90 V vs RHE, as well as negligible change in the overall ORR current after 10 000 cycles. These structures offer a promising platform for the optimization of ORR catalysts and AAEMFC applications.

Experimental methods, CHN and ICP-OES elemental analysis data, a list of ORR activity parameters for CuDCDVu-T, a list of reported low-temperature processed single-atom catalysts for ORR in alkaline electrolytes, additional TEM results, Raman spectra of carbon black and CuDCDVu-T catalysts, XPS survey spectra of carbon black and CuDCDVu-T catalysts, C 1s XPS spectra, SEM image, and EDS elemental mapping of CuDCDVu-600T catalyst layer on glassy carbon, comparison of the LSV results measured under N2 and O2 saturation, summary of all the LSV results of the pyrolyzed catalysts, effect of pyrolysis temperature on the C, Cu, and N contents, and alkaline ORR activity of CuDCDVu-T catalysts, Koutecký–Levich analysis, steady state Tafel analysis of CuDCDVu-600T catalyst, and methanol susceptibility tests of CuDCDVu-600T and Pt/C.

This material is based on work supported by the Air Force Office of Scientific Research under Award No. FA9550-22-1-0014. Materials characterization was carried out in the Materials Research Laboratory Central Research Facilities and the School of Chemical Sciences Microanalysis Laboratory at the University of Illinois at Urbana-Champaign.

The authors have no conflicts to disclose.

Jonathan Matsuura: Conceptualization (equal); Data curation (lead); Formal analysis (lead); Investigation (lead); Methodology (lead); Resources (equal); Validation (equal); Visualization (equal); Writing – original draft (lead); Writing – review & editing (equal). Anjaiah Sheelam: Formal analysis (supporting); Writing – review & editing (supporting). Yingjie Zhang: Conceptualization (equal); Funding acquisition (lead); Project administration (lead); Resources (equal); Supervision (lead); Validation (equal); Visualization (equal); Writing – review & editing (equal).

The data that support the findings of this study are available within the article and its supplementary material.