Unleashing the power of zero-valent platinum single atoms for enhancing low-temperature oxygen activation

Abstract

Developing efficient catalysts for low-temperature O₂ activation is critical for energy-efficient heterogeneous catalysis, yet designing highly active and accessible active sites remains a formidable challenge. Here, we construct a new type of zero-valent platinum single atoms (Pt⁰ SAs) on two-dimensional Co₃O₄ through a feasible low-temperature reduction strategy. By eliminating oxygen coordination, we induce strong Pt−Co electronic interactions and optimize the Pt 5 d band center, promoting electron donation that strengthens O₂ activation. In contrast, conventional high-valent Pt⁴⁺ SAs, which rely on oxygen vacancy formation, exhibits weaker O₂ activation. Consequently, Pt⁰ SAs achieves a 9.3-fold higher turnover frequency in toluene oxidation than does the Pt⁴⁺ SAs at 140 °C (the temperature at which a toluene conversion efficiency of 90% is achieved). This work highlights electronic modulation in optimizing O₂ activation and affords a strategy for designing highly efficient zero-valent single atom catalysts.

Efficient low temperature oxygen activation is essential but hard to achieve with accessible metal sites. This work shows that zero-valent single Pt atoms on cobalt oxide donate electrons to oxygen, enabling low-temperature toluene oxidation.

Introduction

Thermal catalytic oxidation offers an efficient and economical strategy for eliminating toxic volatile organic compounds (VOCs)¹, major pollutants that are detrimental to atmospheric quality and human health^2–4. Enhancing catalyst performance at low temperatures hinges on improving the O₂ activation capacity, which broadens the operational window and conserves energy⁵. Effective O₂ activation typically requires electron-rich surface sites. While noble metal nanoparticles inherently possess zero-valent sites with good O₂ activation ability⁶, many remain encapsulated and underutilized. Conversely, noble metal single-atom catalysts (SACs), despite their high degree of atomic utilization^7,8, often feature coordinated saturation, forming electron-deficient high-valent metal centers^9,10 that impede O₂ activation. Therefore, the development of highly efficient and atomically economical active sites is imperative to unlock the full O₂ activation potential of noble metals.

The construction of zero-valent noble metal single atoms (SAs) is a promising solution to the above issue. Recent studies have highlighted the unique advantages of zero-valent SAs in activating small molecules. For example, Yao et al. observed dynamically detached Pt SAs from a nitrogen–carbon support that exhibited near-zero-valent electronic properties during the hydrogen evolution reaction, where the high 5d electron density of Pt facilitated H₂O activation¹¹. Similarly, Lu et al. stabilized zero-valent Pd SAs on boron-doped carbon nitride. These Pd SAs can enhance metal–support orbital hybridization on the substrate through synergistic interactions with adjacent boron atoms, inducing an upward shift in the Pd 4d orbitals, which promoted electron back-donation for efficient H–H bond cleavage¹². These findings demonstrate that the d-orbital electronic structure of zero-valent SAs favors reactant adsorption and activation, providing a compelling rationale for exploring the potential of zero-valent SAs in O₂ activation.

Reducing the degree of oxygen coordination around the noble metal center is feasible approach to construct zero-valent SAs. Ren et al. developed a rapid heat treatment (RTT) method in an inert atmosphere to obtain highly dispersed Pt SAs on Fe₂O₃. By increasing the RTT temperature from 500 to 600 °C, the Pt–O coordination number was reduced, and Pt–Fe coordination occurred, lowering the Pt valence to +0.56¹³. Furthermore, Chen et al. fabricated zero-valent Pt SAs on the CeO₂ (100) facet by coordinating Pt with two O atoms and one Ce atom, whereas positively charged Pt on the CeO₂ (110) facets coordinated to four oxygen atoms¹⁴. However, the aforementioned high-temperature reduction and crystal facet regulation approaches still face challenges in terms of cost-effectiveness and practical operability.

Here, we introduce a facile low-temperature H₂ reduction strategy to construct zero-valent Pt SAs. This method circumvents the issue of Pt aggregation into nanoparticles that may occur during high-temperature H₂ reduction^15,16. The key to our approach is the use of two-dimensional (2D) metal oxide support, which can provide abundant atomic defects¹⁷ for anchoring highly dispersed SAs^18,19. Critically, the inherent unsaturated metal–oxygen coordination of this support renders lattice oxygen (O_lat) more labile, facilitating removal under mild reduction conditions to prevent atomic agglomeration²⁰. Building upon our previously developed defect-rich 2D Co₃O₄ support^21,22, we synthesized zero-valent Pt SAs using low-temperature (180 °C) H₂ reduction, substantially reducing the synthesis difficulty for zero-valent Pt SAs compared to current approaches. For comparative purposes, high-valent Pt SAs were prepared via air treatment. Comprehensive characterization and density functional theory (DFT) calculations elucidated the geometric and electronic structure of the zero-valent Pt sites. In situ techniques and DFT further unveiled the mechanism by which zero-valent Pt SAs enhances low-temperature O₂ activation and toluene (a model VOC) oxidation. These findings underscore the unique capability of the developed zero-valent Pt SAs to integrate the advantages of both Pt nanoparticles (efficient oxygen activation) and Pt SAs (high utilization efficiency).

Results

Construction and identification of zero-valent Pt SACs

The morphology of the Co₃O₄ support was characterized using electron microscopy. Scanning electron microscopy (SEM) (Supplementary Fig. 1a) and transmission electron microscopy (TEM) (Supplementary Fig. 1b) reveal a nanoflower-like structure composed of ultrathin nanosheets. The atomic force microscopy (AFM) image (Supplementary Fig. 1c) confirms the 2D nature of these nanosheets, with a thickness of ~2 nm. High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) reveals the presence of abundant surface defects (Supplementary Fig. 1d, e) and atomic vacancies (Supplementary Fig. 1f), providing anchoring sites for Pt SA dispersion.

The Pt SA precursor (Pt–P/Co₃O₄) was prepared via ethanol impregnation on 2D/Co₃O₄, followed by air oxidation to obtain Pt–A/Co₃O₄ and H₂ reduction to obtain Pt–H/Co₃O₄. The reduction process was investigated by constant-temperature H₂ temperature programmed reduction (H₂-TPR) (Supplementary Fig. 2a). The peak observed during heating process corresponds to Pt²⁺ reduction²³, while the peak emerging during constant-temperature process originates from the removal of ligands from the Pt(NH₃)₄(NO₃)₂ precursor, verified by a control experiment of replacing Pt precursor with Pt(NO₃)₂ (Supplementary Fig. 2b). The residual H₂ consumption is attributed to the facilitated reduction of Co₃O₄, driven by the hydrogen spillover effect from the newly reduced Pt species²⁴. For comparison, the reduction effect on the O_lat of Co₃O₄ is quite weak (Supplementary Fig. 2c), indicating that the reduced Pt can strongly interact with Co₃O₄. Moreover, the optimal reduction duration was evaluated. Supplementary Fig. 3 shows that only appropriate removal of O_lat can contribute to catalytic performance.

SEM (Supplementary Fig. 4) and TEM images (Supplementary Fig. 5) illustrate that Pt–A/Co₃O₄, Pt–H/Co₃O₄ and Co₃O₄-H retain the ultrathin nanosheet morphology of pristine Co₃O₄. X-ray diffraction (XRD) patterns (Supplementary Fig. 6) of all the samples match those of spinel Co₃O₄ (PDF# 42-1467), with minor peak broadening observed in Pt–H/Co₃O₄ due to O_lat depletion. In addition, no peak attributed to Pt nanoparticles is detected, which is due to the low Pt loading amount (approximately 0.6 wt%, as measured by inductively coupled plasma–mass spectrometry (ICP–MS)) or high Pt dispersibility. To confirm this, in situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) of CO adsorption was performed. Peaks at approximately 2090 cm⁻¹ for Pt–A/Co₃O₄ and approximately 2080 cm⁻¹ for Pt–H/Co₃O₄ are observed in Fig. 1a, which correspond to the CO that adsorbed linearly on isolated Pt atoms^25,26, suggesting that both samples are Pt SACs. In addition, the two different peak positions suggest different single-atom coordination²⁷. HAADF-STEM (Fig. 1b, c) and high-resolution energy dispersive spectroscopy (EDS) mapping (Fig. 1d, e) images verify the uniform distribution of Pt SAs on the surface lattices of both Pt–A/Co₃O₄ and Pt–H/Co₃O₄. Moreover, the Pt 4f  X-ray photoelectron spectroscopy (XPS) spectra (Fig. 1f) reveal Pt⁴⁺ dominance in Pt–A/Co₃O₄. Conversely, Pt–H/Co₃O₄ is dominated by Pt⁰, with its Pt 4f binding energy closely matching that of zero-valent Pt nanoparticles supported on Co₃O₄ (Supplementary Fig. 7). These results confirm the successful synthesis of high-valent and zero-valent Pt SAs (Pt⁴⁺ SA and Pt⁰ SA), respectively.

Fig. 1. Construction of zero-valent Pt SACs.

a In situ CO DRIFTS of Pt–A/Co₃O₄ and Pt–H/Co₃O₄. HAADF-STEM images of Pt–A/Co₃O₄ (b) and Pt–H/Co₃O₄ (c). EDS mapping images of Pt–A/Co₃O₄ (d) and Pt–H/Co₃O₄ (e). f Pt 4f XPS spectra of Pt–A/Co₃O₄ and Pt–H/Co₃O₄. g Local coordination model of Pt⁴⁺ SA and Pt⁰ SA. The coordination bonds of Pt with O and Co are highlighted in yellow. h HAADF-STEM image of Pt⁰ SAs on the Co₃O₄ (110) facet in Pt–H/Co₃O₄. The inset shows the DFT-calculated location of the Pt⁰ SA. Pt⁰ SA is marked by an orange circle, and the surrounding Co atoms are indicated by blue hexagons. i Line profile of the Pt⁰ SA.

To elucidate the local coordination of Pt SAs, we initially attempted X-ray absorption fine structure (XAFS) spectroscopy at the Pt L₃-edge. However, the intense fluorescence background from the abundant Co atoms in the Co₃O₄ support overwhelmingly masked the weak signal from the low-loading Pt, making the Pt signal undetectable. Consequently, we employed DFT calculations and high-resolution HAADF-STEM to resolve the Pt coordination structure. Our previous studies demonstrated that the 2D Co₃O₄ surface possessed abundant cobalt vacancies, including octahedral and tetrahedral vacancies (VCo_Oh and VCo_Te) and oxygen vacancies with dicoordination (VO_II) and tricoordination (VO_III)²⁸, which served as potential anchoring sites for Pt SAs. The Co₃O₄ (110) facet was selected for analysis because of the distinguishable positions of octahedral Co (Co_Oh), tetrahedral Co (Co_Te) and oxygen atoms in the HAADF-STEM images (Supplementary Fig. 8). Furthermore, DFT calculations were performed on Co₃O₄ (110)-A and (110)-B facets, incorporating three distinct vacancy types in each (A-VCo_Oh, A-VCo_Te, A-VO_III, B-VCo_Oh, B-VO_II and B-VO_III, respectively; Supplementary Fig. 9). The calculations results indicate that Pt exhibits the lowest binding energy when it occupies a Co_Oh vacancy on the (110)-B facet (B-VCo_Oh-Pt, Supplementary Fig. 10), forming a surface O_II-Pt_Oh-O_III site. In this configuration, Pt coordinates with two surface O and two subsurface O atoms (Fig. 1g). In addition, Bader charge analysis reveals that the Pt atom in the O_II-Pt_Oh-O_III site also exhibits the highest charge, which is in accordance with the high Pt valence state of Pt–A/Co₃O₄.

To simulate reduction treatment, O atoms near the Pt atom in the B-VCo_Oh-Pt model were successively removed. The energies of oxygen vacancy formation (EOVs) upon removal of one O atom from the O_II-Pt_Oh-O_III site (yielding the B-VCo_Oh-Pt-RO_II and B-VCo_Oh-Pt-RO_III models, respectively; Supplementary Fig. 11a, b) are lower than those for the O_II-Co_Oh-O_III site in the Co₃O₄ model (Supplementary Fig. 11c, d), indicating enhanced O_lat reactivity, which is consistent with the constant-temperature H₂-TPR results. The Bader charge of Pt_Oh in the B-VCo_Oh-Pt-RO_II and B-VCo_Oh-Pt-RO_III models are +0.81|e| and +0.80|e|, respectively, which are higher than those of the zero-valence Pt state of Pt–H/Co₃O₄. Further removal of two oxygen atoms (B-VCo_Oh-Pt-R2O, Supplementary Fig. 12a) still results in a positive Pt Bader charge (+0.40|e|). Moreover, removing one subsurface O atom connected to Pt_Oh in the B-VCo_Oh-Pt-R2O model can obtain the B-VCo_Oh-Pt-R3O_U model (Supplementary Fig. 12b) and B-VCo_Oh-Pt-R3O_D model (Supplementary Fig. 12c) with Pt_Oh Bader charge of +0.10|e| and 0|e|, respectively, among which the latter matches that of the Pt⁰ SA in Pt–H/Co₃O₄. In this configuration, Pt coordinates with one subsurface O, one subsurface Co_Oh, and one tertiary surface Co_Oh atom (Fig. 1g). The DFT-predicted Pt location and its surrounding Co atoms correspond well with the respective atoms in the high-resolution HAADF-STEM image (Fig. 1h), where the bright Pt SA signal aligns precisely with an octahedral site, as illustrated by the contrast profile (Fig. 1i). Further validation comes from the vibrational frequencies of adsorbed CO. The peaks obtained by theoretical calculation (Supplementary Fig. 13) at 2079 cm⁻¹ (Pt⁴⁺ SA) and 2070 cm⁻¹ (Pt⁰ SA) exhibit the same qualitative trend in position as observed experimentally (Fig. 1a). This consistency strongly supports the DFT-predicted coordination environment of Pt SAs. Moreover, DFT calculation (Supplementary Fig. 14) reveal that the binding energy of Pt⁰ SA is comparable to the Pt cohesive energy (−5.5 eV), indicating its sufficient thermodynamic stability^29,30.

Catalytic performance of toluene oxidation

As shown in Fig. 2a, Pt–H/Co₃O₄ exhibits superior catalytic performance, with T₁₀, T₅₀, T₉₀, and T₁₀₀ values (corresponding to the evaluation temperatures for the conversion of toluene into CO₂ (η) with efficiencies of 10%, 50%, 90%, and 100%, respectively) of 130, 136, 140 and 150  °C, respectively. In contrast, Pt–A/Co₃O₄ requires higher temperatures (T₁₀ = 142 °C, T₅₀ = 153 °C, T₉₀ = 160 °C, and T₁₀₀ = 170 °C, respectively), whereas the Co₃O₄ support alone demonstrates markedly inferior performance (T₁₀ = 190 °C, T₅₀ = 207 °C, T₉₀ = 220 °C, and T₁₀₀ = 240 °C). These results indicate that Pt⁰ SA sites significantly enhance the catalytic performance. In addition, Co₃O₄-H prepared by H₂ reduction shows only a marginal improvement over pristine Co₃O₄ (Supplementary Fig. 15), indicating that reduction treatment alone does not substantially enhance the catalytic performance.

Fig. 2. Catalytic performance of toluene oxidation.

a Toluene conversion efficiency into CO₂ (η). b r_s and TOF of Pt–A/Co₃O₄ and Pt–H/Co₃O₄ at 125 °C. c Long-term catalytic stability of Pt–H/Co₃O₄ at T₉₀. d Toluene conversion efficiency into CO₂ of Pt–H/Co₃O₄ at T₁₀₀ and under different relative humidities.

Furthermore, the apparent activation energy (E_a) values are calculated (Supplementary Fig. 16) and follows the order of Co₃O₄ (101.8 kJ/mol) > Pt–A/Co₃O₄ (69.3 kJ/mol) > Pt–H/Co₃O₄ (54.3 kJ/mol), which is consistent with their catalytic performance at low temperature. Additionally, the specific surface area normalized reaction rate (r_s) at 125 °C (At this temperature, η < 10%), derived from characterization of N₂ adsorption and desorption (Supplementary Fig. 17 and Supplementary Table 1), demonstrates that Pt–H/Co₃O₄ exhibits a 41.8-fold higher r_s than does Pt–A/Co₃O₄ (Fig. 2b). Moreover, the turnover frequency (TOF) of Pt–H/Co₃O₄ is 9.3-fold higher than that of Pt–A/Co₃O₄ at 125 °C, underscoring the superior catalytic efficiency of Pt⁰ SA sites over Pt⁴⁺ SA sites.

The catalytic stability of Pt–H/Co₃O₄ with the optimal performance was further investigated. As shown in Fig. 2c, its toluene conversion efficiency remains at 90% after 48 h of reaction at 140 °C. Subsequently, post-reaction XPS (Supplementary Fig. 18) and HAADF-STEM (Supplementary Fig. 19) analyses also confirm that Pt⁰ SAs retain their zero-valent state and atomic dispersion. In addition, the T₉₀ and T₁₀₀ of Pt–H/Co₃O₄ remain stable during five consecutive test cycles, respectively (Supplementary Fig. 20). Moreover, Pt–H/Co₃O₄ also possesses remarkable moisture resistance at T₁₀₀, maintaining near-complete toluene conversion into CO₂ (100%, 97%, and 95%) at different relative humidity (RH) levels of 20%, 50%, and 80%, respectively (Fig. 2d). Impressively, Pt–H/Co₃O₄ outperforms previously reported SACs and noble metal nano particles supported on cobalt-based catalysts for toluene oxidation (Supplementary Table 2 and Table 3), demonstrating its exceptional catalytic performance at low temperature.

Electronic metal–support interaction (EMSI) and local atomic structure

Investigating the EMSI and local atomic structure of active sites is critical for understanding the electronic and geometric configurations that underpin the enhanced catalytic performance upon Pt SA loading. X-ray absorption near-edge structure (XANES) spectra of the Co K-edge were obtained to probe the modulation of the electronic Co structure caused by loading Pt SAs. Figure 3a reveals that the near-edge absorption thresholds adhere to the trend of Co₃O₄ > Pt–A/Co₃O₄ > Co₃O₄-H > Pt–H/Co₃O₄. A lower absorption threshold implies a reduced valence state³¹, indicating that both Pt SA loading and H₂ reduction can decrease the Co valence state. This finding is corroborated by XPS result (all data have been corrected by C 1s spectra, Supplementary Fig. 21), where deconvolution of Co 2p XPS spectra (Fig. 3b) into the Co³⁺ and Co²⁺ components revealed a Co³⁺/Co²⁺ ratio consistent with the trend in the XANES data (Supplementary Table 4). For Co₃O₄-H, the reduction in Co valence derives from the H₂ reduction of surface Co³⁺. For Pt SACs, the decrease in Co valence likely arises from the EMSI between the Pt SAs and Co₃O₄. To verify this inference, the B-VO_III-Pt, B-VCo_Oh-Pt-R3O_D, and B-VO_III models are applied to represent Pt–A/Co₃O₄, Pt–H/Co₃O₄, and Co₃O₄, respectively. The charge density difference (Fig. 3c) suggests that both types of Pt SAs cause local electron rearrangement of the coordinated atoms. For Pt⁰ SA, local electrons are enriched between the Pt_Oh atom and the coordinated Co_Oh atoms in subsurface and tertiary layers, whereas Pt⁴⁺ SA exhibit electron depletion near Pt and electron enrichment around the coordinated oxygen.

Fig. 3. EMSI and local atomic structure.

Co K-edge XANES spectra (a) and Co 2p XPS spectra (b) of Pt–A/Co₃O₄, Pt–H/Co₃O₄, Co₃O₄, and Co₃O₄-H. c Charge density difference of Pt–A/Co₃O₄ and Pt–H/Co₃O₄, green represents electron depletion and yellow represents electron enrichment. d Average Bader charge of the surface Co atoms of Pt–A/Co₃O₄, Pt–H/Co₃O₄, and Co₃O₄. e PDOS spectra of the Pt 5d-orbitals of Pt–A/Co₃O₄ and Pt–H/Co₃O₄. f Co K-edge FT k³χ spectra of Pt–A/Co₃O₄, Pt–H/Co₃O₄, Co₃O₄, and Co₃O₄-H. A typical value of 0.3–0.4 Å requires to be added to the radial distance for phase correction.

Bader charge analysis of the surface and subsurface Co atoms of the B-VO_III-Pt, B-VCo_Oh-Pt-R3O_D, and B-VO_III models (Supplementary Fig. 22) indicates that Pt loading and H₂ reduction result mainly in a decrease in the average valence of the subsurface Co atoms, whereas the average valences of Co_Oh atoms on the Pt–H/Co₃O₄ and Pt–A/Co₃O₄ surfaces increase (Fig. 3d). In general, an increase in the valence of surface Co_Oh is conducive to the chemical adsorption of toluene³², implicating that loading Pt SAs can facilitate this process. Projected density of electronic states (DOS) analysis further reveals the EMSI effects. Figure 3e reveals that the d band center of the 5d orbital of Pt⁰ SA is closer to the Fermi level than that of Pt⁴⁺ SA, which is consistent with the lower Pt valence state in Pt–H/Co₃O₄ relative to Pt–A/Co₃O₄³³. In addition, both Pt–H/Co₃O₄ and Pt–A/Co₃O₄ have obvious d-orbital occupation near the Fermi level, signifying a strong EMSI between the Pt SAs and Co₃O₄¹¹. Furthermore, the 3d orbital of the two coordinated Co atoms strongly hybridizes with the 5d orbital of the Pt⁰ SA, but the overlap with the 2p orbital of the coordinated O atom is weaker (Supplementary Fig. 23a). In addition, the d band center of these two coordinated Co atoms is reduced in contrast with that of Co₃O₄ (Supplementary Fig. 23b), proving that the abundant electrons around Pt are derived from the Pt–Co interaction (formation of Pt–Co bond) after the removal of partially coordinated oxygen atoms³⁴.

Extended X-ray absorption fine structure (EXAFS) was utilized to identify the variation in Co coordination. The peaks located at 1.47, 2.45, and 3.00 Å in the Fourier transform (FT) Co K-edge k³χ spectra (Fig. 3f) are attributed to Co–O coordination, octahedral Co³⁺–Co³⁺ coordination (Co_Oh–Co_Oh) and tetrahedral Co²⁺–Co coordination (Co_Te–Co, including Co_Te–Co_Te and Co_Te–Co_Oh), respectively³⁵. The fitting results (Supplementary Fig. 24 and Supplementary Table 5) indicate that the Co–O, Co_Oh–Co_Oh, and Co_Te–Co coordination of Pt–A/Co₃O₄ decreased slightly compared with that of Co₃O₄, suggesting that the Pt⁴⁺ SA leads to a small increase in lattice disorder and oxygen vacancy, which is consistent with the reduced E_OV of the O_II-Pt_Oh-O_III site relative to that of the O_II-Co_Oh-O_III site. In Co₃O₄-H, the pronounced decrease in Co–O coordination is due to the removal of coordinated oxygen atoms via H₂ reduction. Moreover, the Co−O distance increases and the attenuation of Co_Oh−Co_Oh coordination is greater than that of Co_Te−Co, implying preferential oxygen removal from Co_Oh−O sites. Notably, Pt–H/Co₃O₄ has the lowest Co–O, Co_Oh−Co_Oh and Co_Te−Co coordination, which is consistent with highly active O_lat that facilitates extensive oxygen vacancy formation and induces pronounced surface lattice disorder during H₂ reduction³⁶. This substantial reduction in Co coordination in Pt–H/Co₃O₄ provides a structural foundation for the formation of Pt–Co bonds. Furthermore, the HAADF-STEM images (Supplementary Fig. 25) directly visualize the lattice-disordered structures surrounding the Pt SAs in both Pt–A/Co₃O₄ and Pt–H/Co₃O₄. Wavelet transform analysis of the Co K-edge (Supplementary Fig. 26) also provides atomic-scale insight, revealing a systematic decrease in Co-Co and Co–O coordination signal intensity in the order of Co₃O₄ > Pt–A/Co₃O₄ > Co₃O₄-H > Pt–H/Co₃O₄, which is consistent with the FT-EXAFS results. Collectively, these findings may indirectly suggest that Pt⁰ SA in Pt–H/Co₃O₄ possesses less oxygen coordination.

Oxygen activation mechanism

Both the characterization and DFT calculations reveals that the Pt⁰ SA site in Pt–H/Co₃O₄ exhibits optimal O₂ activation potential. To verify this, the O 1s XPS spectra was acquired at room temperature, and the peaks could be deconvolved into two components (Supplementary Fig. 27), assigned to O_lat and surface-absorbed reactive oxygen species (O_ads)³⁷. The O_ads/O_lat value of Pt–H/Co₃O₄ follows the trend of Co₃O₄ < Pt–A/Co₃O₄ < Co₃O₄-H < Pt–H/Co₃O₄ (Supplementary Table 4). Co₃O₄-H benefits from H₂-induced oxygen vacancies to promote O₂ activation, whereas Pt–A/Co₃O₄ with Pt⁴⁺ SAs and marginally increased the surface oxygen vacancies exhibits weaker enhancement of O₂ activation. In contrast, Pt–H/Co₃O₄, featuring a Pt⁰ SA demonstrates the highest level of reactive oxygen species (ROS) generation. Further insights were obtained via H₂-TPR (Supplementary Fig. 28). For Co₃O₄, the reduction peaks at 220–320 and 320–500 °C correspond to Co³⁺−O and Co²⁺−O reduction, respectively. Upon Pt SA loading, these peaks shift to lower temperatures, indicating enhanced O_lat reactivity. In addition, a new reduction peak (100–250 °C) emerges only for Pt–H/Co₃O₄ and is attributed to surface ROS reduction^38,39. Similarly, O₂ temperature-programmed desorption (O₂-TPD) spectra (Supplementary Fig. 29) reveals the O₂ adsorption capabilities of Pt–A/Co₃O₄ and Pt–H/Co₃O₄. The peak in the range of 120–200 °C corresponds to the desorption of surface O_ads, while the peak around 200–450 °C is assigned to the desorption of surface O_lat⁴⁰. Notably, the intensity of the surface O_ads peak for Pt–H/Co₃O₄ is significantly stronger, suggesting its superior O₂ activation capability. Electron paramagnetic resonance (EPR) analysis using 5,5-dimethyl-1-pyrroline N-oxide (DMPO) trapping under O₂-saturated conditions (Fig. 4a) further confirms that the strongest superoxide/peroxide signal intensity for Pt–H/Co₃O₄. While DMPO-EPR cannot distinguish superoxide from peroxide⁴¹, subsequent DFT calculations would address this issue.

Fig. 4. Oxygen activation mechanism.

a EPR signals of DMPO-captured superoxide/peroxide. Quasi-in situ XPS spectra of O₂ absorption under heating over Pt–A/Co₃O₄ (b) and Pt–H/Co₃O₄ (c). Charge density difference diagrams of O₂ adsorption and contours of 2D ELF slices through O₂ on Pt_Oh–OV (d), Pt⁰ SA (e), and Co_Oh-OV sites (f). In the charge density difference diagram, green represents electron depletion and yellow represents electron enrichment. In the ELF contours, a high ELF value indicates high electron localization. g Proposed effect of electronic modulation of the Pt⁰ SA on O₂ activation. h PDOS spectra of the Pt 5d-orbitals of Pt–A/Co₃O₄ and Pt–H/Co₃O₄ and the O 2p orbital of the adsorbed O₂.

To elucidate the effect of low-temperature elevation (spanning the T₁₀, T₅₀ and T₉₀ of Pt SACs) on O₂ activation, quasi-in situ XPS characterization was performed in an O₂ atmosphere. Both Pt–A/Co₃O₄ (Fig. 4b) and Pt–H/Co₃O₄ (Fig. 4c) display increasing O_ads/O_lat ratios with elevating temperature, with the ratio for Pt–H/Co₃O₄ being consistently higher at each temperature point (Supplementary Fig. 30), confirming its stronger O₂ activation capability. Notably, the O_lat peak shifts to a higher binding energy for both catalysts, indicating electron depletion around the O_lat and the weakened electronic interaction between Co and the O atom, which demonstrates the O_lat activation during the heating process⁴². Based on the preceding analysis, the O_lat activation in Pt–H/Co₃O₄, stems from its highest degree of lattice disorder (Fig. 3f), whereas in Pt–A/Co₃O₄, it arises from the enhanced reactivity of the O_II-Pt_Oh-O_III site compared to the O_II-Co_Oh-O_III site (Supplementary Fig. 11). Given that the weak O₂ activation ability of isolated Pt⁴⁺ SA site alone cannot explain the markedly enhanced low-temperature activity of Pt–A/Co₃O₄ over Co₃O₄, it is proposed that the actual O₂ activation site in Pt–A/Co₃O₄ is oxygen vacancies (Pt_Oh–OV) associated with Pt⁴⁺ SAs, which formed by the removal of activated O_lat from O_II-Pt_Oh-O_III site and exhibit stronger O₂ activation capacity than those oxygen vacancies near Co atoms (Co_Oh-OV). This hypothesis is further supported by subsequent DFT calculations.

O₂ adsorption was modeled on B-VCo_Oh-Pt, B-VCo_Oh-Pt-RO_III, B-VCo_Oh-Pt-R3O_D, and B-OV_III models, simulating O₂ adsorption on Pt⁴⁺ SA, Pt_Oh–OV, Pt⁰ SA, and Co_Oh-OV sites, respectively. Charge density difference analysis (Fig. 4d–f and Supplementary Fig. 31) reveals electron transfer from all surfaces to the adsorbed O₂, confirming its activation. The calculated O₂ adsorption energy (E_{ads, O2}) values follow the order of Pt⁴⁺ SA (−0.9 eV) < Co_Oh-OV (−1.2 eV) < Pt_Oh–OV (−1.5 eV) < Pt⁰ SA (−3.1 eV), with corresponding O=O bond lengths (L_O=O) of 1.33, 1.37, 1.39 and 1.45 Å, respectively. Thus, O₂ is activated as a superoxide on the oxygen vacancy and Pt⁴⁺ SA and as a peroxide on the Pt⁰ SA⁴³. Pt⁰ SA exhibits the strongest O₂ activation capability, as evidenced by its largest E_{ads, O2} and longest O=O bond length. Based on the above analysis, the Co_Oh-OV, Pt_Oh–OV, and Pt⁰ SA serve as the O₂ activation sites for Co₃O₄, Pt–A/Co₃O₄, and Pt–H/Co₃O₄, respectively.

The electron localization function (ELF) is applied to resolve the differences in the geometric and electronic properties of activated O₂ at different active sites, where the electron localizations also follow the order of Pt⁴⁺ SA < Co_Oh-OV < Pt_Oh–OV < Pt⁰ SA, indicating that the Pt⁰ SA possesses the highest covalent character upon its interaction with absorbed O₂. As shown in Fig. 4g, Pt–H/Co₃O₄ has an upshifted d band center relative to that of Pt–A/Co₃O₄, which facilitates easier delocalization of Pt electrons into antibonding orbitals during O₂ adsorption, strengthening the interaction of the O 2p π* band of O₂ for activation⁴⁴. Consequently, the O₂ chemisorbed on the Pt⁰ SA exhibits strengthened overlap between the 2p orbital of O₂ and the 5d orbital of Pt than does Pt_Oh–OV. This overlap shifts closer to the Fermi level for Pt⁰ SA (Fig. 4h), further evidencing its enhanced O₂ activation^45,46. Additionally, analysis of the Pt 5d orbitals before and after O₂ adsorption reveals significant electron loss in the d_xy, d_yz, and d_xz orbitals of the Pt⁰ SA (Supplementary Fig. 32), which spatially align with the 2p π* orbital of O₂. Subsequent DFT calculations (Supplementary Fig. 33) revealed that the dissociation process for O₂ adsorbed on Pt⁰ SA site is thermodynamically more favorable, exhibiting an energy barrier 0.4 eV lower than that at the Pt_Oh–OV SA site. This further demonstrates the superior O₂ activation capability of Pt⁰ SA sites.

Catalytic mechanism of toluene oxidation

On the basis of the above analysis, loading Pt SAs on 2D Co₃O₄ surface and reducing the valence state of Pt can enhance O₂ activation to greatly promote catalytic performance at low temperatures. To elucidate the origin of this enhancement, we investigated the toluene transformation pathways of Pt–A/Co₃O₄, Pt–H/Co₃O₄, and Co₃O₄ using in situ DRIFTS at T₁₀, T₅₀, and T₉₀ (Fig. 5a–c). For Co₃O₄, the key peaks include benzyl alcohol stretching vibrations (1000–1200 cm⁻¹)⁴⁷; the C–O (1300 cm⁻¹) and C=O (1250 cm⁻¹) stretching vibrations of maleic anhydride species⁴⁸; the stretching vibrations of the aromatic ring skeleton (1520 and 1598 cm⁻¹)^49,50; the asymmetric (1410 cm⁻¹) and symmetric (1550 cm⁻¹) vibrations of benzoate⁵¹; and the C=O stretching vibrations of benzaldehyde (1473, 1489 and 1652 cm⁻¹)⁵². The accumulation of these intermediates begins at T₁₀ but slows at T₅₀ and T₉₀, indicating accelerated consumption at higher temperatures. The Pt SACs exhibit similar trends but display three additional peaks: a second stretching vibration of the aromatic ring skeleton (1489 cm⁻¹)⁵³, suggesting altered toluene adsorption at Pt sites; the C−O stretching vibration of phenolate (1215 cm⁻¹)⁵⁴; and the C=O stretching vibration of benzoquinone (1705 cm⁻¹)⁵⁵. These observations indicate that Pt SAs modify toluene adsorption and intermediate formation. Analysis of the reaction tail gas by thermal desorption-gas chromatography-mass spectrometry (TD-GC/MS) at T₁₀, T₅₀, and T₉₀ (Supplementary Fig. 34) reveals trace amounts of gaseous intermediates (acetaldehyde, acetic acid, ethanol, acetone, ethyl acetate, and benzene) alongside toluene. Less intense gaseous intermediate signals are observed with the Pt SACs compared with Co₃O₄, indicating accelerated intermediate depletion and minimal atmospheric release, and the strongest benzene signal from Co₃O₄ also suggests a different ring-opening pathway for benzoic acid compared to Pt SACs.

Fig. 5. Catalytic mechanism of toluene oxidation.

In situ DRIFTS of toluene oxidation on Co₃O₄ (a), Pt–A/Co₃O₄ (b) and Pt–H/Co₃O₄ (c). Toluene (d) and CO₂ (e) signal of toluene TPD-MS spectra of Co₃O₄, Pt–A/Co₃O₄ and Pt–H/Co₃O₄. f Two pathways for the catalytic oxidation of toluene on Co₃O₄ and Pt–H/Co₃O₄. g Comparative relative energy profiles for toluene oxidation into maleic anhydride on Pt–H/Co₃O₄ and Co₃O₄.

Subsequently, toluene adsorption was investigated to explore the possibility that its altered adsorption could lead to an adjustment of the reaction pathway. In toluene TPD-MS characterization, the toluene signal between 60–220 °C corresponds to the physical adsorption of toluene (Fig. 5d) and the CO₂ signal appearing between 200 and 450 °C arises from the reaction of chemically adsorbed toluene with surface O_lat (Fig. 5e)^56,57. Notably, the toluene adsorption capacity is significantly enhanced after Pt SA loading, among which Pt–H/Co₃O₄ exhibits the largest chemisorption amount. Further, in situ DRIFTS spectra (Supplementary Fig. 35) indicate that both Pt–A/Co₃O₄ and Pt–H/Co₃O₄ exhibit new vibrations of the aromatic ring skeleton and phenol compared with Co₃O₄, indicating that the toluene adsorption configuration still changes with no coexisting O₂. DFT calculations (Supplementary Fig. 36) demonstrate the increased toluene adsorption energy (E_{ads, T}) on the Pt SACs, with the aromatic ring positioned closer to the Co_Oh sites. This geometry is more likely to facilitate the interaction of methyl H and ring H atoms with Pt SA-activated ROS. In contrast, the toluene methyl groups adsorb may farther from the ROS on bare Co₃O₄, leading to the reaction path change upon Pt SA loading.

Kinetic analysis was conducted to further elucidate the toluene oxidation mechanism (Supplementary Fig. 37). The experimental data were fitted using kinetic models corresponding to the Langmuir–Hinshelwood (L–H), Eley–Rideal (E–R), and Mars-van Krevelen (MvK) mechanisms (Supplementary Table 6). The results indicate that the apparent reaction order for toluene is negative, which is inconsistent with both E–R and MvK models, while the L–H model provides a good fit fitting degree, indicating that the L–H mechanism is dominant at low reaction temperatures. Complementary insights were obtained from toluene temperature programmed surface reaction (TPSR) with ¹⁸O₂ labeling (Supplementary Fig. 38) over Pt–H/Co₃O₄. Below 160 °C, signals corresponding to C¹⁶O¹⁸O and C¹⁸O₂ are detected, consistent with the complete toluene oxidation by activated O₂. The emergence of C¹⁶O¹⁸O likely originates from isotopic exchange effect^58,59, instead of direct oxidation by O_lat, as C¹⁶O₂—which arises from oxidation by activated O_lat—is only observed above 160 °C. Together, these results demonstrate that toluene oxidation at low temperatures proceeds predominantly via the L–H mechanism, with ROS derived primarily from O₂ activation.

Based on the above findings, DFT calculations were performed to investigate the relative energy variation during toluene oxidation over Pt–H/Co₃O₄ and Co₃O₄, aiming to demonstrate that this novel toluene conversion pathway occurring on Pt–H/Co₃O₄ is more conducive to enhancing catalytic activity. As shown in Fig. 5f, the oxidation pathway from toluene to benzoic acid proceeds via identical intermediates (benzyl alcohol and benzaldehyde) on both Co₃O₄ and Pt–H/Co₃O₄. Thereafter, the reaction pathways diverge markedly. For Pt–H/Co₃O₄, the transformation proceeds sequentially through salicylic acid, phenol, hydroquinone, p-benzoquinone, maleic acid, and finally maleic anhydride (Supplementary Figs. 39 and 40). In contrast, the pathway on Co₃O₄ involves benzene, cyclohexadienone, 1-butynal, and cis-1,4-butenedial before yielding maleic anhydride (Supplementary Figs. 41 and 42). Maleic anhydride is then oxidized to small VOCs and ultimately mineralized into CO₂ and H₂O. Figure 5g illustrates that, compared to Co₃O₄, the distinct reaction route on Pt–H/Co₃O₄ reduces the relative energy by 2.1 eV for the methyl oxidation process (toluene to benzoic acid, steps 1–6) and by an additional 1.3 eV for the subsequent ring-opening process (benzoic acid to maleic anhydride, steps 6–21). Therefore, the intermediate conversion on Pt–H/Co₃O₄ is more exothermic, thereby underpinning its superior low-temperature catalytic performance.

Discussion

This study demonstrates a strategy to enhance the low-temperature O₂ activation capacity of SACs, offering a pathway to reduce the energy demands of thermo-catalytic oxidation. Through a facile low-temperature reduction approach, we synthesized Pt⁰ SA on two-dimensional Co₃O₄ nanosheets (Pt–H/Co₃O₄), where Pt⁰ SA is stabilized by a strong EMSI arising from the removal of coordinating oxygen atoms. The optimized electronic structure of the Pt⁰ SA facilitates efficient O₂ activation at low temperatures, surpassing the performance of its Pt⁴⁺ SA counterparts, which primarily modulate oxygen vacancies on Co₃O₄ but exhibit inferior O₂ activation capability. As a result, Pt–H/Co₃O₄ achieves superior toluene oxidation activity, operating at significantly lower temperatures than conventional SACs. Mechanistic studies reveal a new reaction pathway enabled by the Pt⁰ SA sites, where enhanced O₂ dissociation and an optimized toluene conversion process contribute to the observed catalytic efficiency. These findings provide fundamental insights into Pt⁰ SA design and highlight its promise in catalytic oxidation, which emphasizes the pivotal role of O₂ activation in developing energy-efficient catalytic oxidation systems.

Methods

Catalyst preparation

Co₃O₄ was synthesized by a solvothermal method. First, 24 mL of ethanol, 84 mL of ethylene glycol, and 14 mL of deionized H₂O were mixed. Then, 1.16 g of Co(NO₃)₂·6H₂O was dissolved in the mixed solution, followed by the addition of 0.42 g of hexamethylenetetramine (HMT) and 0.6 g of polyvinylpyrrolidone (PVP) K30. The resulting mixture was subsequently stirred for 30 min and transferred into a Teflon-lined stainless steel autoclave to react at 180 °C for 18 h. Finally, the obtained material was washed and dried at 100 °C to obtain the precursor, which was calcined in air at 300 °C for 3 h and then in N₂ at 300 °C for an additional 3 h.

Pt SACs supported by Co₃O₄ were prepared by a low-temperature H₂ reduction strategy. First, 0.3 g of 2D Co₃O₄ was dispersed into 20 mL of ethanol by stirring for 10 min. Meanwhile, an aqueous solution of Pt(NH₃)₄(NO₃)₂ (1.25 mg/mL) was added to 10 mL of ethanol. This mixture was then rapidly added to the Co₃O₄ dispersion system and stirred vigorously for 2 h, followed by aging for 2 h. Finally, the precursor reaction system was dried at 60 °C to obtain the precursor of the Pt SACs (named Pt–P/Co₃O₄). Pt SACs containing oxidized Pt (named Pt–A/Co₃O₄) were synthesized by calcining Pt–P/Co₃O₄ in air at 250 °C for 2 h. Whereas, Pt SACs containing reduced Pt (named Pt–H/Co₃O₄) were obtained by multiple gas treatments. Specifically, Pt–P/Co₃O₄ was first purged by Helium flow for 15 min and then treated with 5% H₂/Ar flow at 180 °C for 60 min. Finally, the sample was treated with 1% O₂/Ar flow at 250 °C for 30 min. To evaluate the effect of H₂ reduction on Co₃O₄, H₂ reduction of Co₃O₄ was performed under the same conditions as those used for Pt–H/Co₃O₄, and the resulting sample was named Co₃O₄-H.

Catalyst characterizations

XRD was conducted using an X’Pert Pro powder diffractometer (PANalytical Corp., Netherlands), and Raman spectroscopy was performed using an inVia confocal Raman microscope (Renishaw, UK). ICP–MS was conducted on an Agilent 725 (Agilent, USA). SEM was conducted using a MIRA microscope (TESCAN, Czech Republic) equipped with a Quantax EDS (Bruker, USA), and AFM was performed using a MultiMode 8-HR (Bruker, Germany). TEM was conducted using a JEM-2100F transmission electron microscope (JEOL, Japan), and N₂ adsorption and desorption studies were conducted using an ASAP 2460 N₂-adsorption apparatus (Micromeritics, Norcross, GA, USA). Aberration-corrected STEM was performed in the HAADF mode using a JEM-ARM200F microscope (JEOL), and XPS was performed using an Escalab Xi+ spectrometer (Thermo Fisher Scientific, USA). XAFS spectrum was collected at the 4B9A beamline in the Beijing Synchrotron Radiation Facility. EPR was conducted using a ELEXSYS E500 spectrometer (Bruker, Germany). H₂-TPR was performed using a PCA-1200 (Biaode, China). O₂-TPD was conducted on an AutoChem II 2950 HP (Micromeritics, USA). Toluene TPSR with ¹⁸O isotopic labeling and Toluene TPD was tested on a chemisorption analyzer (Belcat II, Japan) with mass spectrometer (Belmass II, Japan). In situ DRIFTS was performed using a VERTEX 70 FT-IR spectrometer (Bruker, Germany). TD-GC/MS was performed using a 7890B GC-5977B MSD (Agilent, USA).

Catalytic performance evaluation

The activity of catalytic oxidation of toluene was evaluated in a fixed-bed reactor, which was composed of two identical stainless-steel tubes with 6-mm diameter, one with quartz wool placed in the middle as a test tube to support the catalyst, and the other as a bypass tube to regulate the reactant gas. Both stainless steel tubes were mounted in a programmed heating furnace. During the dynamic test, 100 ppm of toluene with a total gas flow of 100 mL min⁻¹ was obtained by diluting toluene standard gas with air (relative humidity is adjustable) and passed into the bypass tube at 120 °C. After that, the gas flow was switched to the test tube to pass through 0.1 g of 40–60 mesh catalyst with a gaseous hourly space velocity (GHSV) of 60,000⁻¹. Detailed calculation methods for conversion efficiency (η), E_a, r_s, TOF and reaction order are provided in the Supplementary Information Appendix.

Theoretical calculations

Detailed information regarding DFT calculations is given in the Supplementary Information Appendix.

Author contributions

Y.H. supervised the project. R.L., Y.H. conceived and designed the experiments. R.L., Y.H. performed the key experiments and theoretical calculations and analyzed the results. D.Z., H.Z., L.C., J.C., and S.L. discussed the data. R.L., Y.H. wrote and revised the manuscript.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

Data availability

Source data are provided with this paper. All data are available from the corresponding author upon request. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains Supplementary material available at 10.1038/s41467-026-70170-3.