In Situ Synthesis of Co₃O₄ Nanoparticles on N-Doped Biochar as High-Performance Oxygen Reduction Reaction Electrocatalysts

Abstract

The development of sustainable and high-performance oxygen reduction reaction (ORR) electrocatalysts is fundamental to fuel cell implementation. Non-precious transition metal oxides present interesting electrocatalytic behavior, and their incorporation into N-doped carbon supports leads to excellent ORR performance. Herein, we prepared a shrimp shell-derived biochar (CC), which was doped with nitrogen via a ball milling approach (N-CC), and then used as support for Co₃O₄ nanoparticles growth (N-CC@Co₃O₄). Co₃O₄ loading was optimized using three different amounts of cobalt precursor: 1.56, 2.33 and 3.11 mmol in N-CC@Co₃O₄_1, N-CC@Co₃O₄_2 and N-CC@Co₃O₄_3, respectively. Interestingly, all prepared electrocatalysts, including the initial biochar CC, presented electrocatalytic activity towards ORR. Both N-doping and the introduction of Co₃O₄ NPs had a significant positive effect on ORR performance. Meanwhile, the three composites showed distinct ORR behavior, demonstrating that it is possible to tune their electrocatalytic performance by changing the Co₃O₄ loading. Overall, N-CC@Co₃O₄_2 achieved the most promising ORR results, displaying an E_onset of 0.84 V vs. RHE, j_L of −3.45 mA cm⁻² and excellent selectivity for the 4-electron reduction (n = 3.50), besides good long-term stability. These results were explained by a combination of high content of pyridinic-N and graphitic-N, high ratio of pyridinic-N/graphitic-N, and optimized Co₃O₄ loading interacting synergistically with the porous N-CC support.

1. Introduction

Fuel cells (FCs) are electrochemical devices that directly convert chemical energy into electrical energy [1]. FCs constitute a promising alternative to internal combustion engines, as they present higher energy efficiency and lower environmental impact compared to conventional energy conversion devices [2,3]. Nevertheless, most of the environmental impact of and economic barriers to FCs results from the conventional precious metal electrocatalysts required by several electrochemical reactions behind FC operation [2].

The oxygen reduction reaction occurs at the FC’ cathode, where O₂ molecules are reduced to water. Two ORR mechanisms are possible, involving the direct transfer of four electrons in a single step or via two steps, each involving two-electron transfer [4]. The 4-electron reduction is the most economically viable for FC applications [5]. ORR is a challenging reaction, presenting large overpotentials and sluggish kinetics, which must be overcome by an efficient electrocatalyst [6,7].

While Pt-based materials are currently the state-of-the-art electrocatalysts for ORR, their industrial application is economically and environmentally problematic [8]. Alternatively, non-precious transition metals and their oxides display interesting electrocatalytic properties arising from the variable oxidation state, besides lower cost and higher availability than platinum [9,10]. In particular, spinel Co₃O₄ has attracted attention due to its intrinsic electrocatalytic activity towards ORR, with Co²⁺ sites favoring O₂ adsorption and electron transfer [10,11]. However, these oxides have low electrical conductivity, low stability and a tendency to agglomerate, reducing the exposure of active sites to reactants [12,13]. These limitations can be mitigated by employing a conductive support, such as carbon materials [14]. Moreover, heteroatom doping of carbon materials confers additional active sites for ORR [15,16].

Reportedly, N-doping enhances ORR performance of carbon materials by changing its charge distribution, as nitrogen atoms have higher electronegativity than carbon [17]. However, this enhancement is not always related to nitrogen content present in the doped material, but to the position in which nitrogen atoms are introduced into the graphitic structure [18,19]. The role played by each type of nitrogen in ORR is still not fully disclosed [20,21], but pyridinic and graphitic nitrogen are often considered responsible for the improvement of ORR activity by reducing the O₂ adsorption energy in adjacent carbon atoms [22,23,24]. Rocha et al. reported that the interaction between these two species displays an important effect on ORR performance of N-doped MWCNTs, with higher pyridinic/graphitic ratios leading to better electrocatalytic activity [18].

Therefore, the combination of N-doped carbon materials and Co₃O₄ particles is a popular strategy to obtain high performance ORR electrocatalysts [25,26,27,28,29]. While graphene and MWCNTs are regularly used, it should be considered that they are obtained from petrochemical sources, through energy-intensive procedures. To lower the production cost and improve the sustainability of the final electrocatalyst, without compromising its effectiveness, biomass-derived carbons (biochars) may be used [30,31]. Additionally, the numerous oxygen-containing functional groups usually present on biochar’s surface may help to anchor the metal oxide particles, facilitating the growth/incorporation of these particles under mild conditions [30].

The composition of shrimp shells, rich in nitrogen and containing significant amounts of sulfur and phosphorus, makes them a promising biomass source for biochar production [32,33]. Moreover, they also contain CaCO₃, which promotes porosity formation without requiring any activation procedure by acting as a self-template, thus eliminating time-consuming and energy-demanding steps and the use of expensive (e.g., gaseous CO₂) or corrosive (e.g., NaOH, FeCl₃, etc.) activation agents [34,35]. Crustacean shells are a common waste, available globally, with 6 to 8 million tons being produced annually [36]. As the shells are currently returned to oceans, incinerated or discarded in landfills [37], their utilization in biochar production also contributes to mitigating environmental problems generated by this disposal.

Herein, we prepared composites containing N-doped biochar and Co₃O₄ NPs, with different Co₃O₄ loadings (N-CC@Co₃O₄_1, N-CC@Co₃O₄_2 and N-CC@Co₃O₄_3), through simple procedures: the initial biochar (CC), obtained from pyrolysis of shrimp shells, was doped via a ball milling approach, followed by in situ co-precipitation of Co₃O₄. Both N-doping and Co₃O₄ incorporation led to significant improvements in ORR performance. While all prepared composites showed good ORR activity, the Co₃O₄ loading had a considerable effect on their results, with N-CC@Co₃O₄_2, containing an intermediary loading, achieving the best performance. The synergistic effect between Co₃O₄ NPs and our N-doped biochar, containing high amounts of pyridinic and graphitic nitrogen, was responsible for N-CC@Co₃O₄_2’s excellent ORR performance.

2. Results and Discussion

2.1. Electrocatalysts Characterization

The morphology of the prepared electrocatalysts was assessed by SEM. Figure 1 shows the micrographs of CC, N-CC, bare Co₃O₄ and the three prepared N-CC@Co₃O₄ composites. The CC micrograph shows several large pores in thin carbonaceous sheets. After doping, the biochar developed visibly smaller and interconnected pores, and an overall rougher surface. The Co₃O₄ micrograph revealed a large network of small needle-like particles. The original morphology of both N-CC and the metal oxide particles were maintained on the composites. The elemental distribution maps (Figures S1–S3) revealed that the oxide particles are uniformly dispersed into the biochar support. EDS analyses were also performed for all prepared catalysts (Figure S4), showing the presence of C, O, P, S, Cl, Ca and a small amount of K in CC and N-CC, all remaining from shrimp shell composition (the N signal was significantly overlapped with the high intensity C peak). The same elements were found in the composite’s spectra, with additional Co peaks appearing. It should be noticed that Pd and Au signals in all EDS spectra should be attributed to the thin AuPd film coating the samples, necessary to ensure enough conductivity to acquire good quality SEM images.

The content of C, H, N and S was ascertained by elemental analysis. CC contained mostly carbon (71.00%), but also a reasonable amount of nitrogen (3.12%) and sulfur (1.13%), besides a residual content of hydrogen (0.85%). As expected, the content of nitrogen increased significantly to 6.64% after doping, while the amount of sulfur decreased to 0.11% and the contents of carbon and hydrogen remained similar to the original biochar (69.54% and 0.84%, respectively).

Co contents in the prepared composites were determined by ICP-OES. Co content increased from 3.35 wt% in N-CC@Co₃O₄_1 to 18.35 wt% in N-CC@Co₃O₄_3. N-CC@Co₃O₄_2 obtained an intermediary value of 14.02 wt%.

The success of the doping approach and of the incorporation of metal oxide particles into N-CC was evaluated by Raman spectroscopy. Figure 2 displays the Raman spectra of the CC, N-CC and the prepared composites, as well as bare Co₃O₄ NPs for comparison. CC and N-CC spectra exhibit the characteristic profile of carbon materials [38,39], presenting the D and G bands in the range of 1365–1368 cm⁻¹ and 1580–1590 cm⁻¹, respectively (

Table 1. Raman shift of D and G bands, and I_D/I_G values for CC, N-CC and the prepared composites.

Material	Raman Shift (cm−1)		ID/IG
	D Band	G Band
CC	1362.15	1582.88	1.25
N-CC	1354.52	1577.33	1.26
N-CC@Co3O4_1	1354.29	1583.06	1.16
N-CC@Co3O4_2	1358.14	1587.14	1.12
N-CC@Co3O4_3	1355.07	1581.20	1.21

). Both bands are also detected in the spectra of all composites, around the same Raman shifts. As the D band occurs due to phonon scattering at defective sites and the G band is related to the conjugated structure of sp² carbon, the ratio between the intensity of these two bands (I_D/I_G) gives a measure of the disorder level of the material [38,40]. I_D/I_G values can also be consulted in Table 1. CC presented a disordered carbon structure, as suggested by the high I_D/I_G ratio (1.25). This structure is mainly preserved during N-doping and the growth of metal oxide NPs, with I_D/I_G values of N-CC and the three composites varying from 1.12 to 1.26. The Raman spectrum of Co₃O₄ NPs presents the characteristic bands of inverted spinel Co₃O₄ structure at 469.89, 515.21, 611.67 and 675.48 cm⁻¹ [41]. The most intense band, around 675 cm⁻¹, is also observed on N-CC@Co₃O₄ spectra, indicating the successful preparation of the three composites.

The crystallinity of the prepared materials was evaluated by XRD (Figure 3). CC and N-CC diffractograms display a broad peak around 25°, corresponding to plane (002) of graphite [42,43], besides several peaks over 30–50° attributed to mineral phases remaining from the shrimp shells’ original composition [44,45,46]. All these diffraction peaks are observed in the composites’ diffractograms, in addition to extra peaks around 31.6, 36.8, 39.4, 44.8, 59.4 and 65.3°, associated to planes (220), (311), (222), (400), (511) and (440) of Co₃O₄ [47,48]. The crystallite size (D) of bare Co₃O₄ NPs, as well as Co₃O₄ NPs on the three composites, was estimated by the Debye–Scherrer equation (Equation (1)).

$$D={\displaystyle \frac{\mathrm{K}\lambda }{\beta \cos \theta }}$$

(1)

where K stands for the Scherrer constant, λ is the X-ray wavelength, β is the FWHM of the selected peak and θ is the diffraction angle, applied to the most intense plane (311) [49,50]. Bare Co₃O₄ NPs showed a crystallite size of 7.9 nm, while the particles present on the N-CC@Co₃O₄_1, N-CC@Co₃O₄_2 and N-CC@Co₃O₄_3 displayed sizes of 12.1, 9.2 and 5.0 nm, respectively.

The surface composition of the prepared materials was assessed by XPS (

Table 2. Surface composition of the prepared materials.

Material	At %
	C 1s	N 1s	O 1s	S 2p	P 2p	Ca 2p	Co 2p
CC	90.55	2.99	4.95	1.01	0.50	---	---
N-CC	87.24	6.87	4.59	0.15	0.77	0.38	---
N-CC@Co3O4_1	69.45	5.72	17.90	---	1.12	0.48	5.33
N-CC@Co3O4_2	66.77	5.85	19.56	---	0.93	0.54	6.36
N-CC@Co3O4_3	47.17	3.48	34.56	---	---	---	14.79

). CC and N-CC contain a large amount of carbon (90.55 and 87.24 at %, respectively), which progressively decreased in the composites to 69.45 (N-CC@Co₃O₄_1), 66.77 (N-CC@Co₃O₄_2) and 47.17 at % (N-CC@Co₃O₄_3). As expected, and in accordance with CHNS results, the N content increased significantly after doping, from 2.99 at% in CC to 6.87 at % in N-CC. In contrast, N-CC@Co₃O₄_1, N-CC@Co₃O₄_2 and N-CC@Co₃O₄_3 presented N contents of 5.72, 5.58 and 3.48 at %, respectively. The amount of O remained constant during the biochar doping (4.95 at % in CC vs. 4.59 at % in N-CC) but increased progressively in the composites according to the loading of Co₃O₄: N-CC@Co₃O₄_1, N-CC@Co₃O₄_2 and N-CC@Co₃O₄_3 obtained values of 17.90, 19.56 and 34.56 at %, respectively. Co content in the composites increased, according to the amount of Co precursor used, to 5.33, 6.36 and 14.79 at % in N-CC@Co₃O₄_1, N-CC@Co₃O₄_2 and N-CC@Co₃O₄_3, respectively. Residual amounts of S 2p were detected only on CC and N-CC (1.01 and 0.15 at %, respectively), while P 2p was identified on all materials, except N-CC@Co3O4_3, in amounts of 0.50–1.12 at %. Ca 2p was found in small amounts (0.38–0.54 at %) on N-CC, N-CC@Co₃O₄_1 and N-CC@Co₃O₄_2, but not on CC, suggesting that the doping procedure, especially the ball milling approach, may expose residual amounts of calcium remaining in the bulk of CC after acid leaching.

C 1s spectra of all prepared materials present an intense peak at 284.60 eV, corresponding to sp² C domains, as well as five less intense peaks related to sp³ C, C–N, C–O–C, C=O and O–C=O groups at 285.20, 285.90, 286.90, 288.20 and 289.30 eV, respectively, and a satellite peak attributed to π–π* transitions around 290.54–291.17 eV (Figure S5) [29,51]. Additionally, contributions from C–O–P at 284–285 eV should be taken into account in all spectra except N-CC@Co₃O₄_3 [52], and C–S/C=S may contribute to CC and N-CC’ spectra around 286–287 eV [53].

N 1s spectra of CC, N-CC and the composites were deconvoluted in four peaks: pyridinic-N, pyrrolic-N, graphitic-N and oxidized-N, around 398.07–398.44, 399.66–400.28, 400.99–401.87 and 403.18–404.81 eV, respectively (Figure 4) [19,54]. Our doping strategy led to a significant increase in pyridinic-N content and decrease in pyrrolic-N compared to the undoped biochar (52.33% vs. 33.11% and 22.38% vs. 32.44%, respectively), while the contents of graphitic-N and oxidized-N did not change considerably (Table S1). Nitrogen in the composites was found majorly under pyridinic-N (46.87, 43.17 and 37.36% in N-CC@Co₃O₄_1, N-CC@Co₃O₄_2 and N-CC@Co₃O₄_3, respectively), followed by pyrrolic-N (33.96–36.36%), graphitic-N (10.84–22.70%) and oxidized-N (5.46–7.17%) (Table S1).

The deconvolution of O 1s spectra in all prepared materials is represented in Figure S6. Both CC and N-CC displayed three peaks at 530.88–530.90, 532.13–532.33 and 533.60–533.61 eV, corresponding to C=O, C–O and O–C=O groups, respectively [55,56]. As expected, the co-precipitation of Co₃O₄ NPs drastically affected the O 1s region, with the composite’s spectra presenting peaks related to O²⁻ in Co₃O₄ lattice (529.94–530.05 eV), adsorbed hydroxyl groups into Co₃O₄ structure (531.13–531.28 eV), C–O and C=O groups from N-CC (531.97–532.32 eV), hydroxyl groups from MIPA (532.97–533.48 eV) and adsorbed water molecules (533.97–534.95 eV) [27,29].

The composite’s XPS spectra of the region Co 2p showed two main peaks around 781 and 797 eV corresponding to Co 2p_3/2 and Co 2p_1/2 spin orbital splitting (Figure 5). All spectra presented a doublet at 779.88–779.95 and 795.02–795.11 eV corresponding to Co(III) 2p_3/2 and Co(III) 2p_1/2 contributions, and a Co(II) doublet at 781.36–781.43 and 796.03–797.93 eV. Two shake-up satellite peaks were fitted on the Co 2p_3/2 main peak, at 782.86–783.41 and 788.35–788.70 eV, while a single satellite of Co 2p_1/2 was located at 803.90–804.31 eV [27,29]. The same peaks were observed in the bare Co₃O₄ spectrum, at similar binding energies (Figure 5).

S 2p was detected only in CC and N-CC (Figure S7). While CC presented a well-resolved spectrum, showing S 2p_3/2 and S 2p_1/2 contributions of C–S–C (163.81 and 164.94 eV, respectively), C–SH (165.39 and 166.52 eV, respectively), C–SO_x–C (167.92 and 169.05 eV, respectively) and a single contribution for C–SO₃ groups (170.11 eV) [55], the low amount of S at the N-CC surface only allowed to deconvolute its spectrum in S 2p_3/2 and S 2p_1/2 at 164.48 and 165.66 eV, respectively. Ca 2p was found in N-CC, N-CC@Co₃O₄_1 and N-CC@Co₃O₄_2 (Figure S8), showing the Ca 2p_3/2 and Ca 2p_1/2 contributions of Ca(II) in CaCO₃, around 346.93–347.33 and 350.43–350.82 eV, respectively [57,58]. P 2p spectra of CC, N-CC, N-CC@Co₃O₄_1 and N-CC@Co₃O₄_2 are represented in Figure S9. The CC spectrum was fitted into three doublets corresponding to P–C (131.35 and 132.22 eV), [P₂O₇]⁴⁻ (133.04 and 133.91 eV) and [PO₃]⁻ (135.37 and 136.24 eV) [59]; the remaining materials showed only the last two doublets at similar binding energies.

2.2. ORR Performance

The electrochemical behavior of the prepared catalysts was initially evaluated by cyclic voltammetry, in N₂ and O₂-saturated electrolyte. As displayed in Figure S10, CC, N-CC and the three composites showed an irreversible cathodic peak around 0.75–0.80 V vs. RHE in presence of O₂ that is not visible in N₂-saturated electrolyte, corresponding to oxygen reduction; this peak was also observed in bare Co₃O₄ around 0.50 V vs. RHE. The CVs of the composites and Co₃O₄ also displayed additional cathodic and anodic peaks between 1.1 and 1.5 V vs. RHE, both in N₂ and O₂-saturated electrolyte, associated with redox processes occurring in the cobalt oxide [60,61].

ORR polarization curves of the prepared materials at 1600 rpm are shown in Figure 6a. Through this technique, onset potential (E_onset) and diffusion-limited current density (j_L) values were calculated, and the results can be found in

Table 3. ORR results obtained in O₂-saturated 0.1 mol dm⁻³ KOH electrolyte, including the onset potentials (E_onset), diffusion-limited current densities (j_L, determined at 1600 rpm and 0.26 V), numbers of electrons transferred by O₂ molecule (n) and Tafel slopes (TS).

Catalyst	Eonset/V vs. RHE	jL/mA cm−2	n	TS/mV dec−1
CC	0.78	−3.11	2.83	74
N-CC	0.85	−3.31	3.30	125
N-CC@Co3O4_1	0.85	−2.92	3.73	64
N-CC@Co3O4_2	0.84	−3.45	3.50	35
N-CC@Co3O4_3	0.82	−3.16	3.42	44
Co3O4	0.64	−1.69	---	115
Pt/C	0.93	−4.62	3.58	91

. Some catalysts, such as CC, N-CC and Co₃O₄, did not reach a plateau region at the evaluated potential window. In these cases, j_L was estimated as the current density obtained at the last screened potential (E = 0.26 V vs. RHE), although these values may not correspond to diffusion-limited current density by definition, as kinetics may still be controlling the ORR process along with diffusion. CC displayed decent j_L and E_onset of −3.11 mA cm⁻² and 0.79 V vs. RHE, respectively, which were further enhanced after the N-doping process: N-CC obtained a significantly more positive E_onset (0.88 V vs. RHE) and higher j_L (−3.31 mA cm⁻²). This improvement is explained by increasing contents of pyridinic-N and decreasing amounts of pyrrolic-N after doping, while the graphitic-N was mostly maintained. Indeed, N-CC presented higher pyridinic-N/pyrrolic-N and pyridinic-N/graphitic-N ratios (Table S2), which are reportedly related to higher ORR activity [18,62].

While Co₃O₄ NPs showed poor electrocatalytic behavior, with E_onset of 0.63 V vs. RHE and j_L of only −1.69 mA cm⁻², their incorporation into N-CC resulted in a considerable improvement of the composites’ catalytic performance. The composites achieved E_onset between 0.83 and 0.86 V vs. RHE, with decreasing loading of Co₃O₄ leading to more positive values (Table 3). Although N-CC@Co₃O₄_3 and N-CC@Co₃O₄_1 presented lower j_L in absolute value than N-CC (−3.16 and −2.92 mA cm⁻², respectively), N-CC@Co₃O₄_2 exhibited the highest value of the prepared materials (−3.45 mA cm⁻²), evidencing the importance of loading optimization. Therefore, the composite with an intermediary amount of Co₃O₄ showed the most promising balance between E_onset and j_L. Nevertheless, it is still slightly outperformed by Pt/C in terms of these parameters (E_onset = 0.93 V vs. RHE and j_L = −4.62 mA cm⁻²). On the other hand, the obtained values are similar to the ones reported in the literature for composites containing carbon materials and Co₃O₄, as can be confirmed in Table 4. For instance, our composites presented more positive E_onset and higher j_L than Co₃O₄@GF_O₃ and Co₃O₄@GF_KMnO₄, prepared using oxidized graphene flakes [27], and very close values to the ones reported to Co/N₃S₃-GF, containing N,S-doped graphene [29].

To further evaluate the electrocatalytic performance of the prepared materials, LSV was also performed at different rotation speeds (400, 800, 1200, 1600, 2000 and 3000 rpm, Figure S11), allowing the analysis of K–L plots (Figure S12) and the determination of the number of electrons transferred per O₂ molecule (n); as the LSV of Co₃O₄ did not present a well-defined diffusion-controlled region, it was not possible to apply the K–L equation and, consequently, determine its n value. All K–L plots showed good linearity, indicating that the prepared materials (except Co₃O₄) followed first order kinetics regarding the dissolved O₂ concentration [63]. It is important to mention that there is some controversy about the correct way to estimate n values, with recent reports indicating that the K–L method may not apply fully to porous carbon materials with high surface area (S_BET > 550 m² g⁻¹) [64]. However, as our catalysts showed low surface areas (see N₂ adsorption–desorption results below), the K–L method could be successfully applied herein.

The variation of n to the applied potential was plotted in Figure 6b for all the catalysts plus Pt/C, added for comparison purposes. N-CC and the composites showed practically constant n values in the potential window of 0.27–0.42 V vs. RHE, while CC displayed slightly decreasing n values at higher potentials. Therefore, all materials but CC followed a mechanism independent of applied potential. CC achieved an average n of 2.83, suggesting a mixed process between the 2-electron reduction and the 4-electron pathway. N-doping led to an n closer to 4, with N-CC obtaining a value of 3.30: high pyridinic-N/graphitic-N ratios on carbon materials are correlated with higher n [18,62]. The incorporation of Co₃O₄ NPs into N-CC further improved the selectivity for the direct reduction. Interestingly, the composite prepared with a lower amount of cobalt precursor obtained a value of n closer to 4 (n = 3.72), following a predominantly 4-electron reduction. In contrast, N-CC@Co₃O₄_3, with the highest loading of Co₃O₄, obtained n of 3.42, while N-CC@Co₃O₄_2 achieved an intermediate value of 3.50. Therefore, while the presence of Co₃O₄ was important for improving the selectivity for the 4-electron pathway, its loading should be carefully optimized. Once again, the obtained n values are in accordance with the ones reported in the literature for heteroatom doped carbon/Co₃O₄ composites (Table 4).

Tafel plots were constructed to evaluate the kinetics of all prepared materials (Figure 6c). CC and N-CC obtained the highest Tafel slopes (TS), with values of 74 and 125 mV dec⁻¹. On the other hand, the composites displayed considerably lower values, with N-CC@Co₃O₄_2 showing the lowest TS (35 mV dec⁻¹), followed by N-CC@Co₃O₄_3 (44 mV dec⁻¹) and N-CC@Co₃O₄_1 (64 mV dec⁻¹). Pt/C obtained a TS of 91 mV dec⁻¹. The significantly lower TS values presented by the composites compared to Pt/C indicates that O₂ is easily adsorbed at the catalyst surface [63]. Comparing all these values, it is possible to infer that the rate-limiting step was different for the evaluated catalysts: CC, N-CC and Pt/C kinetics were likely limited by Equation (2), while the mechanism rate of the three composites was dominated by Equations (3) and (4), where M is an empty site on the catalyst surface [65]. Therefore, the introduction of Co₃O₄ led to changes in the rate-limiting step.

$${\mathrm{MO}}_2+{\mathrm{e}}^{-}\rightleftharpoons {\mathrm{MO}}_2^{-}$$

(2)

$${\mathrm{MO}}_2^{-}+{\mathrm{H}}_2\mathrm{O}\rightleftharpoons {\mathrm{MO}}_2\mathrm{H}+{\mathrm{OH}}^{-}$$

(3)

$${\mathrm{MO}}_2\mathrm{H}+{\mathrm{e}}^{-}\rightleftharpoons \mathrm{MO}+{\mathrm{OH}}^{-}$$

(4)

Overall, N-doping led to enhanced electrocatalytic activity, which was further improved by incorporation of Co₃O₄ NPs into the biochar support. Herein, N-CC@Co₃O₄_2 achieved the most promising ORR performance, showing the best balance between E_onset, j_L, n and TS. To help clarify these results, the textural properties of N-CC@Co₃O₄_2 (and CC and N-CC for comparison purposes) were further investigated by N₂ adsorption–desorption isotherms at 77 K. CC, N-CC and N-CC@Co₃O₄_2 displayed similar specific surface areas (S_BET of 265.76, 286.24 and 277.11 m² g⁻¹, respectively), indicating that this parameter was not determining their ORR performance. So, we also investigated the pore distribution on these catalysts (Figure S13). CC, N-CC and N-CC@Co₃O₄_2 showed broad peak width distribution along 0 to 50 nm (in the first two catalysts) or up to 60 nm (in the composite). CC presented predominantly pores with width of 2.73 and 4.75 nm, containing therefore mostly mesopores, according to IUPAC classification. Similarly, pores of 2.41 and 4.72 nm are predominant in N-CC, indicating that the doping procedure did not cause significant changes in the biochar’s porosity. In contrast, most of N-CC@Co₃O₄_2 pores presented width equal or inferior to 1.75 nm, indicating the presence of abundant micropores along with a smaller amount of mesopores. Considering these results, the composite’s microporosity could be attributed to the incorporation of Co₃O₄ into the N-CC support. Although mesopores promote fast transport of reactants and products towards or from the catalytic sites during the ORR process [66], some works also associated them with high selectivity for H₂O₂ production [67,68]. Micropores are reported to provide active sites for ORR [69,70]. Therefore, the combination of microporosity and mesoporosity may help explain the enhanced ORR performance of N-CC@Co₃O₄_2 relative to CC and N-CC [71].

Moreover, while the amount of nitrogen is not directly related to a better ORR performance [18,19], our doping procedure allowed us to increase significantly the content of pyridinic-N, while also obtaining a considerable amount of graphitic-N. Although the exact nature of active sites in N-doped carbon materials is still controversial [21], most studies revealed that pyridinic-N and graphitic-N play an important role in the electrocatalytic process [72,73,74,75]. On the other hand, the introduction of transition metal oxides, including Co₃O₄, into N-doped carbons is a well-known strategy to improve its ORR performance, as it synergistically combines Co₃O₄ electrocatalytic activity and the support’s conductivity and active sites [12,13]. Nevertheless, it should be noted that our support presented a defective structure with a low graphitization degree, which could negatively affect the composite’s performance.

The long-term stability of our most promising material was assessed by chronoamperometry at E = 0.47 V vs. RHE for 15 h (Figure 6d). N-CC@Co₃O₄_2 displayed good stability, with current retention of 80.80%, a similar value to the one obtained with Pt/C (84.08%).

Table 4. ORR performance of carbon/Co₃O₄-based composites reported in the literature, in 0.1 mol dm⁻³ KOH electrolyte.

Catalyst	Eonset/V vs. RHE	jL/mA cm−2	n	Reference
N-CC@Co3O4_1	0.85	−2.92	3.73	This work
N-CC@Co3O4_2	0.84	−3.45	3.50	This work
N-CC@Co3O4_3	0.82	−3.16	3.42	This work
Co3O4@bio-C	0.86	−5.20	3.86	[25]
Co/Co3O4@NC	0.94	−4.78	3.82	[13]
Co3O4/OPAC	0.76	−2.49	2.46	[76]
Co3O4/NCMT-800	0.91	−4.50	3.95	[26]
Co3O4/NHPC	0.96	−6.0	3.91	[12]
Co/N3S3-GF	0.87	−3.98	~ 4.0	[29]
Co3O4@GF_O3	0.82	−2.70	3.5	[27]
Co3O4@GF_KMnO4	0.79	−2.50	3.1	[27]
N-rGO/CNT/Co3O4	0.87	~−4.0	~4.0	[28]
Co3O4/CW	0.89	−4.79	~4	[77]
Co3O4/N-rGO	---	~−2.5	2.9–3.4	[78]

Table 4. ORR performance of carbon/Co₃O₄-based composites reported in the literature, in 0.1 mol dm⁻³ KOH electrolyte.

Catalyst	Eonset/V vs. RHE	jL/mA cm−2	n	Reference
N-CC@Co3O4_1	0.85	−2.92	3.73	This work
N-CC@Co3O4_2	0.84	−3.45	3.50	This work
N-CC@Co3O4_3	0.82	−3.16	3.42	This work
Co3O4@bio-C	0.86	−5.20	3.86	[25]
Co/Co3O4@NC	0.94	−4.78	3.82	[13]
Co3O4/OPAC	0.76	−2.49	2.46	[76]
Co3O4/NCMT-800	0.91	−4.50	3.95	[26]
Co3O4/NHPC	0.96	−6.0	3.91	[12]
Co/N3S3-GF	0.87	−3.98	~ 4.0	[29]
Co3O4@GF_O3	0.82	−2.70	3.5	[27]
Co3O4@GF_KMnO4	0.79	−2.50	3.1	[27]
N-rGO/CNT/Co3O4	0.87	~−4.0	~4.0	[28]
Co3O4/CW	0.89	−4.79	~4	[77]
Co3O4/N-rGO	---	~−2.5	2.9–3.4	[78]

3. Experimental

3.1. Materials and Chemicals

Shrimp shells were obtained from Mar Cabo—Produtos Congelados Lda and ground to a fine powder before utilization. Melamine (99%, Alfa Aesar, Kandel, Germany), hydrochloric acid (37%, Fluka, Porto Salvo, Portugal), 1-amine-2-propanol (MIPA, 93%, Sigma-Aldrich, Algés, Portugal), cobalt(II) chloride hexahydrate (98–102%, Thermo Scientific, Porto Salvo, Portugal), N,N-dimethylformamide (99.5%, Sigma-Aldrich, Algés, Portugal), Nafion 117 (5 wt% in lower aliphatic alcohols, Aldrich, Algés, Portugal) and Pt/C 20 wt% HiSPEC^® 3000 (Alfa Aesar, Kandel, Germany) were used as received. Ultrapure water (resistivity 18.2 MΩ cm at 25 °C, Interlab, Portugal) was used in all electrochemical analysis and material preparations, except for the initial biochar preparation in which deionized water was used.

3.2. Electrocatalysts Preparation

3.2.1. Preparation of the Biochar (CC)

To obtain the initial biochar, denoted as CC, powdered shrimp shells were carbonized at 900 °C for 2 h, with a heating rate of 10 °C min⁻¹, under a constant N₂ flow (100 mL min⁻¹). Then, the resulting material was treated with a 2.0 mol dm⁻³ HCl solution under reflux at 80 °C for 4 h, collected by filtration and washed with deionized water until reaching neutral pH. The material was dried at 80 °C under vacuum overnight.

3.2.2. N-Doping of CC (N-CC)

N-CC was obtained by doping CC with nitrogen, following a procedure described elsewhere [54]. Briefly, 0.60 g of CC and 0.39 g of melamine were milled together in a Retsch ball mill for 5 h at a frequency of 15 s⁻¹, using a zirconium oxide ball with diameter of 1.5 cm. The obtained material was pyrolyzed at 600 °C for 1 h, under N₂ flow (100 mL min⁻¹), with a heating rate of 10 °C min⁻¹.

3.2.3. Preparation of N-CC@Co₃O₄ Composites

N-CC@Co₃O₄ composites were prepared through in situ co-precipitation of Co₃O₄ particles according to the procedure described by Fernandes et al. [29]. A quantity of 250 mg of N-CC was dispersed in 50 mL ultrapure water by ultrasonication during 10 min. A certain amount of CoCl₂∙6H₂O was added to this dispersion under magnetic stirring. A 3.0 mol dm⁻³ aqueous solution of MIPA was added drop-by-drop (addition rate of 50 mL h⁻¹), under vigorous magnetic stirring, until pH 10 was reached; the stirring was maintained during 24 h at room temperature. The precipitated solid was collected by filtration, washed with ultrapure water until neutral pH was obtained and then with ethanol, and dried at 60 °C in a glass vacuum oven. Finally, the material was calcinated under a constant flow of air (100 mL min⁻¹) at 250 °C for 3 h with a heating rate of 5 °C min⁻¹. Composites with tunable loading of Co₃O₄ were prepared by changing the amount of CoCl₂∙6H₂O added: 3.11, 2.33 and 1.56 mmol of Co(II) precursor were used in the synthesis of N-CC@Co₃O₄_3, N-CC@Co₃O₄_2 and N-CC@Co₃O₄_1, respectively. For comparison, Co₃O₄ nanoparticles were also prepared through the procedure described without the addition of the biochar support.

3.3. Characterization Techniques

All the prepared materials were extensively characterized by several techniques, namely micro-Raman, powder XRD, XPS and SEM. Additionally, CC and N-CC were analyzed by elemental analysis (CHNS mode), while the cobalt content of N-CC@Co₃O₄_1, N-CC@Co₃O₄_2 and N-CC@Co₃O₄_3 was determined by ICP-OES. The textural properties of CC, N-CC and N-CC@Co₃O₄_2 were evaluated by N₂ adsorption-desorption isotherms. More details of the characterization techniques employed can be found in the Supplementary Materials.

3.4. Electrochemical Measurements

The electrochemical tests were carried out in a conventional three-electrode compartment cell, in 0.1 M KOH electrolyte, using an Ag/AgCl (3 mol dm⁻³ KCl, Metrohm, Lisboa, Portugal) reference electrode, a glassy carbon rod (diameter 2.0 mm, Metrohm, Lisboa, Portugal) auxiliary electrode and a rotating disk electrode (RDE) of glassy carbon (diameter of 3.0 mm, Metrohm, Lisboa, Portugal) as working electrode. The measurements were obtained on an AutoLab PGSTAT 302N potentiostat/galvanostat controlled by Nova 2.1 software. The RDE was extensively polished, before being modified by drop-cast of 5.0 μL of the catalyst ink (containing 1.0 mg of catalyst, dispersed in 250 μL DMF and 20 μL Nafion 117) and dried under an air flow. The ORR performance of the prepared materials was assessed in 0.1 mol dm⁻³ KOH solution via cyclic voltammetry (CV), linear sweep voltammetry (LSV) and chronoamperometry [29,56,79,80]. More detailed information can be found in the Supplementary Materials.

4. Conclusions

Composites containing Co₃O₄ NPs and N-doped biochars were prepared through the carbonization of shrimp shells, followed by doping via a solvent-free approach and in situ co-precipitation of Co₃O₄ under mild conditions. The loading of Co₃O₄ was optimized by preparing three composites using different amounts of cobalt precursor.

All the prepared materials were extensively characterized by several techniques. SEM micrographs revealed that CC and N-CC presented porous structures, which were modified by Co₃O₄ growth. Elemental analysis and XPS confirmed the successful N-doping of CC, with N content increasing from ~3 to ~7 wt%. Raman, XRD and XPS verified the correct preparation of spinel Co₃O₄ into N-CC structure.

N-doping and Co₃O₄ incorporation improve CC’s electrocatalytic performance. N-CC@Co₃O₄_2 achieved the most promising activity, obtaining E_onset and j_L of 0.84 V vs. RHE and −3.45 mA cm⁻², respectively, and n_O2 and stability comparable to Pt/C. The high content of pyridinic-N and graphitic-N, besides an optimized Co₃O₄ loading that allowed an enhanced synergistic effect between the NPs and N-CC support, contributed to N-CC@Co₃O₄_2’s excellent performance. Moreover, our catalyst, derived from shrimp shell waste, displayed comparable performance to other composites reported in the literature containing conventional carbon nanomaterials obtained from petrochemical resources.