Antibiotics persist in aquatic environments, posing threats to ecosystems and human health due to hazards such as bioaccumulation and induction of antimicrobial resistance [1,2]. In recent years, persulfate-based advanced oxidation processes (PS-AOPs) have demonstrated remarkable efficiency in degrading organic pollutants [3–5]. As common reactive oxygen species (ROS) in SR-AOP systems, sulfate radical (SO₄^•−) and hydroxyl radical (^•OH) achieve high mineralization rates for contaminants owing to their high redox potential and relatively long half-life [6]. However, in complex aqueous matrices, these radicals are readily scavenged, compromising oxidation performance. In contrast, the non-radical pathway dominated by singlet oxygen (¹O₂) offers superior selectivity, anti-interference capacity and mild oxidation [7–9]. How to achieve the directional generation of ¹O₂ in PS-AOPs through the structural regulation of materials is a current research hotspot.

Cobalt-based oxides, exemplified by Co₃O₄, demonstrate significant catalytic potential in water treatment. Oxygen vacancies (O_v), formed by the departure of lattice oxygen atoms from metal oxides, are recognized as effective active sites for activating peroxymonosulfate (PMS) to generate ¹O₂ [10–12]. During catalysis, O_v typically functions as electron acceptors, adsorbing PMS to initiate intermediate species (e.g., superoxide radicals, O₂^•−) that subsequently facilitate ¹O₂ production via chain reactions [13,14]. Conventional strategies for creating O_v in metal oxides—including ion doping, chemical reduction, and hydrothermal treatment—although effective, still face challenges such as high synthesis temperature requirements and impure products [15–17]. In contrast, plasma etching emerges as a distinct approach, enabling rapid surface defect formation at ambient temperature through interactions with reductive active species or electrons. This technique offers advantages of low energy consumption and minimal byproduct generation [18,19]. Therefore, plasma treatment provides a promising way to construct surface O_v on Co₃O₄, which may facilitate PMS to generate ¹O₂.

In this study, Co₃O₄ was treated with N₂ plasma for different durations at room temperature to synthesize O_v-rich Co₃O₄-x (x denotes the reaction time). Comprehensive characterizations via X-ray photoelectron spectroscopy (XPS), Raman spectroscopy, and solid-state electron spin resonance (ESR) confirmed the successful introduction of O_v. Co₃O₄-x/PMS systems exhibited significantly enhanced catalytic oxidation performance compared to the pristine-Co₃O₄/PMS system. Probe experiments confirmed that this enhancement was primarily attributed to the facilitated generation of ¹O₂, whose steady-state concentration exhibited a significant positive correlation with the improved TC degradation rates. As expected, the incorporation of O_v is shown to be effective in promoting the generation of ¹O₂. Additionally, Co₃O₄-x exhibited excellent stability, reusability, and promising applicability in complex water matrices.

The synthesis procedure is illustrated in Fig. 1a. Pristine Co₃O₄ was prepared via the sol-gel method, followed by thermal treatment in a muffle furnace at 400 ℃ for 2 h. Subsequently, Co₃O₄ were subjected to plasma treatment under a N₂ atmosphere to obtain Co₃O₄-x (x = 15, 30, 45 min). The phase purity and crystallinity of all samples were confirmed by X-ray diffraction (XRD), as their patterns (Fig. 1b) exhibited diffraction peaks matching well with the standard pattern of Co₃O₄ (PDF #42–1467) [20]. Furthermore, energy-dispersive X-ray spectroscopy (EDS) mapping images (Fig. S1 in Supporting information) clearly demonstrated the homogeneous distribution of Co and O elements across the material surface. Scanning electron microscopy (SEM) images (Fig. S2 in Supporting information) revealed that all samples exhibit irregular bulk morphologies. Notably, increased surface roughness and porosity of Co₃O₄-x were observed, attributable to high-energy bombardment induced by plasma. This morphological change was further substantiated by Brunauer-Emmett-Teller (BET) surface area analysis (Fig. 1c), which showed a gradual increase in the specific surface area of Co₃O₄-x with prolonged plasma treatment time. At the microscopic scale, however, comparison of the Transmission electron microscope (TEM) images between pristine Co₃O₄ and Co₃O₄–30 revealed comparable morphologies (Figs. 1d and e). High resolution transmission electron microscope (HRTEM) imaging of Co₃O₄–30 (Fig. 1f) further resolved its crystalline nature, displaying distinct lattice fringes with spacings of 0.286 nm and 0.244 nm, corresponding to the (220) and (311) planes of the Co₃O₄ spinel structure, respectively [21]. Additionally, the selected-area electron diffraction (SAED) pattern (Fig. 1g) confirmed the good match between the Co₃O₄-x and standard Co₃O₄ [22], indicating that the plasma treatment did not alter the crystal structure of the material.

Solid ESR was employed to detect O_v in pristine Co₃O₄ and Co₃O₄–30 (Fig. S3 in Supporting information). Compared to pristine Co₃O₄, Co₃O₄–30 exhibited a more pronounced ESR signal at a g-factor of 2.003, confirming the successful introduction of O_v [23]. The result was further substantiated by Raman scattering. Each sample exhibited four Raman active modes (F_2g, E_g, F_2g², A_1g), corresponding to the crystal cube Co₃O₄ [24,25]. The Raman peak intensity of Co₃O₄-x weakened and shifted towards lower frequencies (Fig. 2a). Among them, the A_1g peak (685 cm⁻¹) associated with the Co-O bond stretching of the octahedral locus showed a relatively significant redshift change, indicating that plasma treatment altered the local electron density and lattice vibration frequency of the Co₃O₄-x, revealing the introduction of O_v [26]. Furthermore, XPS was performed to monitor the valence state distribution and electronic structure of the samples. The O 1s spectra included three distinct peaks (Fig. 2b): Lattice oxygen (O_latt, ~530.02 eV), adsorbed oxygen species (O_ads, ~531.38 eV), and surface hydroxyl groups/water (O_H2O, ~533.38 eV) [27–29]. O_ads is commonly used as a relative indicator for O_v concentration. The relative content of O_ads in Co₃O₄–30 (48.91%) was significantly higher than that in Co₃O₄ (28.98%) (Fig. 2c), confirming the increased concentration of O_v on the surface of Co₃O₄–30. The Co 2p spectra (Fig. 2d) could be deconvoluted into characteristic peaks corresponding to Co²⁺ (binding energies at 781.89 eV and 796.66 eV) and Co³⁺ (binding energies at 780.06 eV and 795.03 eV) [25]. Compared to the pristine Co₃O₄, the Co²⁺/Co³⁺ ratio in Co₃O₄–30 increased from 0.48 to 0.68. The decrease in the average valence state of cobalt is attributed to the transfer of electrons from the O atoms to the Co d-bond center during the formation of O_v [30].

The performance of different catalysts for PMS activation was further evaluated. As shown in Fig. 3a, direct PMS oxidation contributed negligibly to TC removal (< 1%). Although the physical adsorption performance improved slightly with the increase of plasma treatment duration, the adsorption capacity was still very limited (Fig. S4 in Supporting information). In the oxidative degradation stage, the Co₃O₄-x/PMS systems exhibited significantly higher TC degradation efficiency than the pristine Co₃O₄/PMS system (Fig. S5 in Supporting information). Specifically, the Co₃O₄/PMS system only achieved 58.9% TC removal within 30 min, with a first-order reaction kinetic constant (k) of 0.023 min⁻¹. In contrast, the Co₃O₄–15/PMS, Co₃O₄–30/PMS, and Co₃O₄–45/PMS systems demonstrated markedly improved performance, attaining removal efficiencies of 84.7% (k = 0.055 min⁻¹), 93.2% (k = 0.078 min⁻¹), and 92.7% (k = 0.077 min⁻¹), respectively. The catalytic activity of Co₃O₄-x was significantly enhanced with longer plasma processing time. Given the nearly identical performance of Co₃O₄–30 and Co₃O₄–45, it can be inferred that a 30 min-plasma treatment is sufficient to achieve complete formation of O_v on the surface of Co₃O₄. Besides, key parameters of reaction (PMS dosage, catalyst dosage, pH and TC dosage) were optimized to establish the best degradation conditions (Fig. S6 in Supporting information). Moreover, Comparison of the performance with advanced catalysts in recent researches demonstrated the superior activity of Co₃O₄–30 for TC degradation via PMS activation (Fig. S7 and Table S1 in Supporting information).

To identify the dominant ROS, quenching experiments and ESR measurements were performed. As shown in Fig. 3b, the impact of tert-butanol (TBA, k_{TBA, •OH} = 6.0 × 10⁸ L mol⁻¹ s⁻¹) and methanol (MeOH, k_MeOH, SO_4•- = 1.1 × 10⁷ L mol⁻¹ s⁻¹, k_{MeOH, ^•OH} = 9.7 × 10⁸ L mol⁻¹ s⁻¹) on TC degradation was relatively weak [31,32]. Meanwhile, no significant signal peaks corresponding to DMPO-SO₄^•− or DMPO-^•OH were observed (Fig. S8a in Supporting information), suggesting that the production of SO₄^•− and ^•OH radicals was limited in the Co₃O₄–30/PMS system. In contrast, distinct TEMP-¹O₂ and DMPO—O₂^•− signals were also detected (Fig. 3c and Fig. S8b in Supporting information), confirming the presence of ¹O₂ and O₂^•−[33,34]. The addition of p-benzoquinone (p-BQ, k_p-BQ, O₂^•− = 8.3 × 10⁸ L mol⁻¹ s⁻¹) and furfuryl alcohol (FFA, k_{FFA, ¹O2} = 1.2 × 10⁸ L mol⁻¹ s⁻¹) to the Co₃O₄–30/PMS system resulted in significant inhibition of TC degradation, reducing the removal rate by 32% and 60%, respectively [35,36]. The highest quenching efficacy observed with FFA suggests that the dominant degradation mechanism in the Co₃O₄-x/PMS system is a non-radical pathway involving ¹O₂. Besides, considering that the redox potential of O₂^•− is as low as −0.33 V (vs. NHE), which is theoretically difficult to effectively degrade TC [37]. The introduction of trace amounts of p-BQ exerted a significant inhibitory effect on the degradation process, suggesting that O₂^•− may serve as a key precursor to ¹O₂ [38].

Moreover, quenching experiments conducted on the pristine Co₃O₄/PMS system revealed the similar trend, indicating the same types of ROS produced (Fig. S9 in Supporting information). However, the inhibitory effect of FFA on the pristine Co₃O₄/PMS system was less pronounced, suggesting a lower contribution of ¹O₂ compared to Co₃O₄-x/PMS system. Therefore, FFA was further employed as a probe to quantitatively determine the steady-state concentrations of ¹O₂ ([¹O₂]_ss) generated in each system [39]. As shown in Fig. 3d and Fig. S10 (Supporting information), the Co₃O₄-x/PMS system generated significantly higher [¹O₂]_ss than Co₃O₄/PMS system, and the amount of [¹O₂]_ss positively correlated with the TC degradation performance in each system, which confirmed ¹O₂ originated from plasma-induced O_v and drived the degradation (Fig. 3e). In addition, the influence of O₂ and N₂ on the degradation reaction can be ignored, thus the path of converting dissolved oxygen to ¹O₂ was excluded (Fig. S11 in Supporting information) [40]. Based on the above experimental results, it can be inferred that the O_v on the Co₃O₄-x surface serve as crucial adsorption sites and electron donors, selectively generating ¹O₂ via PMS activation rather than SO₄^•− or ^•OH radicals. Herein, it is speculated that O_v received electrons from PMS and generated O₂^•−, which functioned dually as intermediates for ¹O₂ formation and as effective ROS directly attacking TC (Fig. 3f, Eqs. 1–4) [41].

¹O₂ typically exhibits high selectivity towards organic compounds containing electron-rich moieties [42]. Therefore, six contaminants with varying electron-donating capacities (methyl orange (MO), sulfamethoxazole (SMX), bisphenol A (BPA), TC, ciprofloxacin (CIP), and benzoic acid (BA)) were selected for investigation [43,44]. As shown in Fig. 4a and Fig. S12 (Supporting information), the Co₃O₄–30/PMS system exhibits significantly higher degradation efficiency for electron-rich organics (MO, SMX, BPA, TC) compared to electron-deficient organics (CIP, BA). The distinct selectivity is attributed to the electrophilic property. Another distinguishing feature of ¹O₂ is its strong resistance to Inorganic anions. Then, high concentrations of interfering anions (H₂PO₄⁻, Cl⁻, HCO₃⁻, NO₃⁻, SO₄²⁻) and humic acid (HA) were introduced (Fig. 4b and Fig. S13 in Supporting information). In the presence of anions, the degradation efficiency of TC remained above 80%. Even with the addition of HA, the removal rate still reached 75%. Among these interfering substances, SO₄^2- as a reaction product, may inhibit the activation of PMS, thereby slightly attenuating TC degradation. In contrast, HA, as a competing pollutant, likely consumed a portion of ¹O₂, which also contributed to the decrease in TC removal. Overall, the Co₃O₄–30 material demonstrated favorable environmental tolerance. Furthermore, the adaptability of Co₃O₄–30/PMS system was evaluated under realistic water matrices, including tap water and lake water. As depicted in Fig. 4c, the Co₃O₄–30/PMS system maintained high oxidation efficiency in these complex environments, achieving over 90% TC removal within 30 min in both cases. Catalyst recyclability was confirmed over four consecutive cycles, where TC degradation efficiency consistently retained > 70%, demonstrating strong durability for pollutant abatement (Fig. 4d). The slight decrease in the removal rate of TC after four cycles may be attributed to the decline of the active sites on the catalyst surface and the catalytic activity of Co₃O₄–30 can be restored through simple plasma treatment. Additionally, the XRD pattern reveals that the characteristic peaks of the catalyst remained consistent before and after the reaction (Fig. S14 in Supporting information), confirming the structural stability of the Co₃O₄–30. Although O 1s XPS analysis shows a slight decrease in the proportion of O_ads on cycled Co₃O₄–30, indicating a reduction of O_v, the O_ads content remains higher than that of the pristine Co₃O₄ (Fig. S15 in Supporting information). Thus, Co₃O₄–30 maintains relatively high catalytic activity within a certain number of cycles, which can be attributed to its surface-rich O_v. Furthermore, these results provide additional evidence that O_v indeed serves as the key active site for catalyzing PMS to generate ¹O₂.

In summary, we successfully introduced O_v to the surface of Co₃O₄-x via a facile and efficient N₂ plasma treatment strategy. The O_v rich Co₃O₄–30/PMS system achieved 93.2% TC degradation within 30 min (k = 0.078 min⁻¹), significantly outperforming the untreated catalyst. The O_v served as key electron acceptors, facilitating the adsorption of PMS to generate the intermediate O₂^•−, which subsequently which subsequently led to the production of ¹O₂. Dominated by the ¹O₂, the Co₃O₄–30/PMS system exhibited strong oxidative degradation capability toward electron-rich pollutants, demonstrating high selectivity in wastewater purification. The system also showed excellent tolerance to high concentrations of interfering ions and organic matter. As the primary technique employed to engineer catalyst defects, plasma treatment offers a simple and controllable approach for enabling non-radical AOPs for selective pollutant degradation, holding significant potential for practical water purification applications.