Photoinduced radical-mediated atomic dispersion of noble metal nanoparticles

Abstract

The transformation of noble metal nanoparticles into atomically dispersed catalysts has been a long-standing goal to enhance metal utilization and regenerate the activity of agglomerated catalysts. Traditional methods, however, often require high temperatures, specific atmospheres, or complex chemical processes. We present a novel photoinduced strategy for atomic dispersion of noble metal nanoparticles under ambient conditions. Experimental and density functional theory calculations reveal that chlorine radicals (•Cl), together with •O₂^-, promote Pd-Pd bond cleavage. The intermediate [PdCl₄]^2- species formed adsorbs onto TiO₂ via electrostatic interactions and, upon dechlorination, stabilizes into a single-atom Pd₁-N₂O₁ structure. This method is applicable to various noble metals (Pd, Pt, Rh) and different oxide supports (TiO₂ and WO₃), and significantly enhances the catalytic activity of both commercial Pd/C and industrial waste Pd/C catalysts by 17.8-fold and 26-fold, respectively, in the hydrogenation of styrene. This approach offers a simple, green, and sustainable solution for advancing catalytic technologies.

Converting metal nanoparticles into single-atom catalysts remains a challenge. This light-driven method uses chlorine radicals to disperse noble metals at room temperature, dramatically boosting the activity of industrial catalysts.

Introduction

Noble-metal-based catalysts, such as palladium (Pd), platinum (Pt), and rhodium (Rh), are critical in energy conversion^1,2, chemical synthesis³, and environmental remediation⁴ due to their exceptional stability and catalytic performance. However, the scarcity and high cost of these metals necessitate improving their atomic efficiency to reduce production costs and foster the green and sustainable development of related industries^5–7. Recently, dispersing noble metals into ultra-small forms, such as nanoparticles, clusters, and single atoms, has emerged as an effective strategy to enhance both atomic efficiency and catalytic performance^8–10. Nanoparticles and single-atom catalysts exhibit differences in chemical states and coordination environments, which leads to distinct catalytic reaction pathways and different catalytic applications. Among these, single-atom catalysts (SACs) have garnered significant attention due to their unique electronic properties and maximized atomic utilization, leading to exceptional catalytic activity^11–14. Yet, SACs’ high surface energy makes them susceptible to sintering into larger nanoparticles under harsh conditions, which can dramatically reduce their catalytic performance and lead to deactivation^15–20.

To address these challenges, various strategies have been developed to achieve atomic dispersion of noble metal nanoparticles. These include atom trapping, in situ pyrolysis, reaction induction, and ligand assistance. For instance, Datye et al. achieved single-atom dispersion of Pt on CeO₂ by treating Pt nanoparticle-loaded alumina and CeO₂ mixtures at 800 °C²¹. Wei et al. encapsulated Pd nanoparticles in ZIF-8 and, after pyrolysis at 900 °C, observed Pd coordinated as single atoms with nitrogen²². Feng et al. converted Rh nanoparticles into single atoms using a specific atmosphere of N₂ flow and CO/CH₃I at 513 K²³, while Wang et al. employed organic chloride ligands at 400 °C to convert Pd nanoclusters into Pd single atoms²⁴. In these approaches, the nanoparticles are first activated by high temperature treatments or reactive atmospheres, which mobilize metal atoms from the particle surface and cleave metal-metal bonds, and the resulting mobile species are then trapped and stabilized as isolated atoms at strong anchoring sites such as defects or heteroatom coordination environments on the support. In particular, atom-trapping and in situ pyrolysis strategies rely on thermally induced migration or volatilization of metal species that are captured by defect sites, whereas reaction induced and ligand assisted strategies use specific reactants or coordinated ligands to weaken metal-metal bonds and promote the stepwise transformation of metal nanoparticles or clusters into isolated single atom centers. Although these methods have made significant strides in overcoming sintering and improving noble metal utilization, they often require high temperatures, specific atmospheres, or complex chemical operations, which can increase energy consumption, production costs, and environmental impacts. Therefore, developing a simple and energy-efficient strategy for converting noble metal nanoparticles into single atoms under mild conditions is crucial for advancing sustainable catalytic technologies.

Here, we propose a novel photoinduced strategy that achieves atomic dispersion of noble metal nanoparticles under ambient temperature and pressure. Operationally, the dispersion is carried out in an acetonitrile-HCl mixed solution (CH₃CN/HCl, v/v = 1:0.00005) containing nanoparticle and oxide supports, which is simply irradiated with UV light at room temperature, without the need for high temperature calcination or controlled gas atmospheres. This intrinsically reduces energy input and simplifies scale-up. Experimental results and density functional theory (DFT) calculations show that chlorine radicals (•Cl) generated during the reaction reduce the electron density around Pd atoms, weakening the stability of Pd-Pd bonds. This, in combination with the attack by •O₂^-, leads to the cleavage of Pd-Pd bonds. The resulting intermediate species, [PdCl₄]^2-, adsorbs onto the TiO₂ surface through electrostatic interactions, and upon dechlorination, forms a stable single-atom Pd₁-N₂O₁ structure, thus achieving atomic dispersion. This strategy is applicable to a range of noble metals, including Pd, Pt, and Rh, and can be extended to different oxide supports such as TiO₂ and WO₃. Notably, photoinduced treatment significantly enhances the catalytic performance of both commercial and industrial waste Pd/C catalysts in the hydrogenation of styrene. Together, these results highlight both the generality and the scalability of this photoinduced approach.

Results and discussion

The proposed photoinduced strategy for achieving atomic dispersion of noble metal nanoparticles is illustrated in Fig. 1a. At ambient temperature and pressure, Pd_NPs/TiO₂ were dispersed in an acetonitrile-HCl solution (volume ratio 1:0.00005), followed by ultraviolet (UV) light irradiation. After 1 h of exposure to UV light, the Pd₁/TiO₂ product was collected through centrifugation, washing, and drying (see Fig. S1). High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images (Fig. 1b, c) show that, before UV irradiation, Pd nanoparticles were visibly distributed across the TiO₂ surface. Post-irradiation, however, the Pd nanoparticle structure was no longer observable, indicating a significant morphological transformation.

Fig. 1. Photoinduced atomic dispersion of Pd nanoparticles.

a Schematic illustration of the photoinduced conversion of Pd_NPs/TiO₂ to Pd₁/TiO₂. HAADF-STEM images of Pd_NPs/TiO₂ (b) before and c after photoinduced treatment. d AC HAADF-STEM image of Pd₁/TiO₂. EDS elemental mapping of Pd_NPs/TiO₂ (e) before and f after photoinduced treatment. g 3D intensity plot of Pd₁/TiO₂.

Energy-dispersive X-ray spectroscopy (EDS) mapping (Fig. 1e, f) revealed that, after the photoinduced treatment, Pd was still present on the TiO₂ surface, but its distribution had changed from being concentrated within particle regions to a more uniform dispersion. Inductively coupled plasma optical emission spectrometry (ICP-OES) analysis confirmed that the Pd content remained essentially unchanged before and after the UV treatment (Table S1). Further characterization using aberration-corrected HAADF-STEM (Figs. 1d, g, and S2) showed that Pd atoms exhibited isolated distribution on TiO₂. These results demonstrate that the photoinduced strategy successfully achieved atomic dispersion of Pd nanoparticles.

To investigate the photoinduced atomic dispersion process of Pd nanoparticles (Pd NPs), we subjected Pd_NPs/TiO₂ samples to varying light irradiation times. Prior to light exposure, HAADF-STEM images revealed that the average diameter of the Pd NPs was approximately 2.72 nm (Fig. 2a, S3). After 15 min of irradiation, the average diameter of the Pd particles decreased to about 2.20 nm, indicating a reduction in particle diameter (Fig. 2b, S4). After 30 min, the particles further reduced in diameter, forming clusters around 1.33 nm (Fig. 2c, S5). After 60 min of irradiation, the nanoparticle structure became unobservable, indicating complete conversion into single atoms (Fig. 2d, S6). These observations suggest that the Pd NPs gradually reduced in size, ultimately achieving atomic dispersion.

Fig. 2. Photoinduced transformation of Pd_NPs/TiO₂ to Pd₁/TiO₂.

a–d HAADF-STEM images and Pd particle diameter distributions of Pd_NPs/TiO₂ after photoinduced treatment for 0, 15, 30, and 60 min, respectively. The number of Pd particle counted was 274, 139, and 172 for 0, 15, and 30 min, respectively. e–h DRIFTS spectra of CO adsorption at saturation coverage and followed by Ar flushing on Pd_NPs/TiO₂ after photoinduced treatment for 0, 15, 30, and 60 min, respectively. i–l Fourier transforms of the Pd K-edge EXAFS spectra of Pd_NPs/TiO₂ after photoinduced treatment for 0, 15, 30, and 60 min, respectively. The inset in (i) shows the corresponding structural model, with Pd (pink), N (light yellow), O (light orange), and Ti (purple). m–p Wavelet transform of Pd_NPs/TiO₂ after photoinduced treatment for 0, 15, 30, and 60 min, respectively.

To further explore the photoinduced atomic dispersion of Pd NPs, we employed CO adsorption diffuse reflectance infrared Fourier transform spectroscopy (CO-DRIFTS) to characterize the Pd species (Fig. 2e–h). The CO adsorption peaks at ~2170 cm^–1 and ~2119 cm^–1 correspond to the R and P branches of gaseous CO, respectively²⁵. Before light irradiation (Fig. 2e), a distinct absorption peak at 1923 cm^–1 was observed, indicating that CO was bridged on Pd nanoparticles²⁶, which confirmed that Pd predominantly existed in nanoparticulate form. As the irradiation time increased, the CO adsorption peaks gradually shifted. After 15 min of light exposure (Fig. 2f), new peaks appeared at 2085 cm^–1 and 1948 cm^–1, corresponding to CO adsorbed linearly on single atoms and CO bridged on nanoparticles, respectively²⁷. This suggests that some Pd NPs were converted into Pd single atoms. After 30 min of irradiation (Fig. 2g), the CO linear adsorption peak shifts to 2090 cm^–128, indicating further transformation toward single-atom Pd. After 60 min of light irradiation (Fig. 2h), the CO adsorption peak appears only at 2160 cm^–1, corresponding to CO linearly adsorbed on Pd single atoms²⁹, indicating complete conversion of Pd NPs into single atoms.

To gain deeper insight into the atomic structural evolution of Pd species, we conducted quasi in situ X-ray absorption fine structure (XAFS) measurements at the Pd K-edge. The X-ray absorption near-edge structure (XANES) spectra showed a gradual increase in the white-line intensity with longer irradiation times (Fig. S7), suggesting that the oxidation state of Pd increased over time. The XPS spectra of Pd 3 d further show that, with increasing irradiation time, Pd gradually transforms from Pd⁰ (334.81 eV, 3d_5/2) to Pd²⁺ species (337.49 eV, 3d_5/2) (Fig. S8). Together, these results indicate that Pd transitioned from a metallic to an oxidized state under light irradiation^30,31, with Pd atoms likely dissociating from nanoparticles and interacting with TiO₂ surface sites.

The Fourier-transformed (FT) k³-weighted extended X-ray absorption fine structure (EXAFS) spectra for Pd/TiO₂ at various irradiation times are presented in Fig. 2i–l. The characteristic peaks at approximately 1.5 Å and 2.5 Å correspond to Pd-N/O and Pd-Pd coordination shells, respectively³². With increasing irradiation time, the intensity of the Pd-Pd peak gradually decreased and eventually disappeared, while the Pd-N/O peak intensified. This shift indicates the photoinduced dispersion of Pd NPs into single-atom structures, which bind to N/O sites on TiO₂. Notably, during the intermediate irradiation stages, a Pd-Cl peak around 2.0 Å emerged (Fig. 2j–k)²⁴, suggesting that chloride species may promote Pd atom dissociation from the nanoparticles, aiding the formation of single atoms.

Wavelet transform (WT) contour plots for Pd/TiO₂ samples at different irradiation times (Fig. 2m–p) show that at the initial stage (0 min), the maximum intensity at approximately 9.0 Å⁻¹ corresponds to Pd-Pd coordination, confirming that Pd was predominantly in nanoparticle form. As the irradiation time increased, the Pd-Pd bond signal gradually weakened and disappeared, while the signal for Pd-O/N bonds at around 4.5 Å⁻¹ gradually appeared and strengthened³³. This change further supports the transition of Pd from nanoparticles to single-atom structures. Quantitative structural analysis through least-squares EXAFS curve fitting (Table S2, Figs. S9–10) revealed that initially, Pd existed as nanoparticles with a Pd-Pd coordination number of 9.8. As irradiation progressed, the Pd-Pd coordination number decreased, while the Pd-O/N and Pd-Cl coordination numbers increased, indicating the dispersion of Pd NPs with chloride involvement. After 60 min of irradiation, the Pd-Pd and Pd-Cl coordination numbers completely disappeared, and only Pd-O/N coordination remained, with a coordination number of 3. This confirms that Pd NPs were fully dispersed into single atoms, where the Pd single atoms coordinate with three N/O atoms on the TiO₂ surface, forming a stable Pd₁-N₂O₁ structure (Fig. 2l, S2, S11, S23).

In previous characterizations, such as HAADF-STEM, CO-DRIFTS, and XAFS, we observed that under UV light irradiation, the Pd-Pd bonds in Pd_NPs/TiO₂ gradually transformed into Pd-O/N bonds of single-atom Pd. To gain deeper insight into this phenomenon and uncover its detailed mechanism, we designed a series of control experiments to investigate the roles of UV light, acetonitrile, chloride ions, and oxygen in the transformation of Pd nanoparticles into single atoms.

The HAADF-STEM results (Fig. 3a–d and S12–S15) reveal that, in the absence of UV light, the average diameter of Pd NPs remains stable at 2.72 nm, indicating that UV light irradiation is essential for the atomic dispersion of Pd NPs. Under aqueous conditions, the average diameter of Pd NPs increased to 3.76 nm. In the absence of chloride ions, the average particle diameter was 3.0 nm, while in an oxygen-free environment, the average diameter was reduced to 2.62 nm. ICP analysis (Table S3) showed that the Pd loading did not undergo significant changes under these control conditions. These results confirm that UV light, acetonitrile, chloride ions, and oxygen are critical factors in the photoinduced conversion of Pd NPs to single atoms.

Fig. 3. Mechanism of photoinduced atomic dispersion of Pd NPs.

HAADF-STEM images and Pd particle diameter distributions of Pd_NPs/TiO₂ after photoinduced treatment under the following conditions: a without UV light, b without CH₃CN, c without Cl^-, and d without O₂. The number of Pd particles counted was 205, 133, 204, and 115, respectively. e chlorine radicals, f, g superoxide radicals, and alkyl radicals generated during the photoinduced transformation of Pd_NPs/TiO₂ to Pd₁/TiO₂. HAADF-STEM images and Pd particle diameter distributions of Pd_NPs/TiO₂ after photoinduced treatment under the following conditions: h capture of •O₂^- and i capture of electrons (e^-). The number of Pd particles counted was 302 and 150, respectively. j UV-vis spectra of Pd intermediate species. k Zeta potential of TiO₂ dispersed in CH₃CN. l Schematic illustration of the mechanism for the photoinduced transformation of Pd_NPs/TiO₂ to Pd₁/TiO₂.

Having identified these key factors, we further investigated their specific roles in the photoinduced transformation of Pd NPs. During the photoinduced process, TiO₂, as a semiconductor, generates electron-hole pairs upon UV excitation Eq.( 1) (Fig. S16)³⁴. These electrons and holes migrate to the catalyst surface, where the electrons react with oxygen to form superoxide radicals (•O₂^-) Eq.( 2)³⁵, while the holes react with acetonitrile and chloride ions, respectively, generating alkyl radicals (•CH₂CN) Eq.( 3)³⁶ and chlorine radicals (•Cl) Eq.( 4)³⁷. The reactions are as follows:

$${\mathrm{TiO}}_2\to {\mathrm{e}}^{-}+{\mathrm{h}}^{+}$$ 1

$${\mathrm{O}}_2+{\mathrm{e}}^{-}\to \cdot {{\mathrm{O}}_2}^{-}$$ 2

$${\mathrm{CH}}_3\mathrm{CN}+{\mathrm{h}}^{+}\to \cdot {\mathrm{CH}}_2\mathrm{CN}+{\mathrm{H}}^{+}$$ 3

$${\mathrm{Cl}}^{-}+{\mathrm{h}}^{+}\to \cdot \mathrm{Cl}$$ 4

Electron paramagnetic resonance (EPR) spectroscopy (Fig. 3e–g) confirmed the formation of •Cl, •O₂^-, and •CH₂CN under UV irradiation, while no significant EPR signals were observed under dark conditions. This indicates that UV light irradiation triggers the formation of these active species on the TiO₂ surface.

To further explore the roles of these active species in the photoinduced transformation of Pd NPs, we used p-benzoquinone, 2,3-dicyano-5,6-dichlorobenzoquinone (DDQ), and methanol to capture •O₂^-, electrons (e^-), and holes (h⁺), respectively (Fig. 3h–i, S17–S19, and Table S4)³⁶. Experimental results showed that when •O₂^- was captured, Pd particles still existed, with an average diameter of 1.75 nm, and the Pd loading did not significantly change, indicating that •O₂^- facilitated the cleavage of Pd-Pd bonds. When electrons were captured, Pd particles still existed, with an average diameter of 1.64 nm, but the Pd loading decreased. According to Eq.( 2), capturing electrons inhibits the generation of •O₂^-, further confirming the key role of •O₂^- in Pd-Pd bond cleavage. The decrease in Pd loading suggests that electrons may also play a role in the formation of Pd single atoms on TiO₂. Capturing electrons may interfere with the electron transfer between Pd intermediates and TiO₂, affecting the anchoring of Pd single atoms and thus reducing Pd loading. When holes (h⁺) were captured, Pd particles still existed (Fig. S19), with an average diameter of 2.47 nm, and the Pd loading did not significantly change. According to Eq.( 4), capturing h⁺ inhibited the generation of •Cl, highlighting the crucial role of •Cl in the Pd-Pd bond cleavage process.

Furthermore, through UV-vis spectroscopy, we successfully detected the reaction intermediate species [PdCl₄]^2- (Fig. 3j)³⁸. Zeta potential measurements of TiO₂ in acetonitrile showed that TiO₂ has a positive surface potential (Fig. 3k), indicating that the intermediate [PdCl₄]^2- ions can adsorb onto the TiO₂ surface through electrostatic interactions (Fig. S20)³⁹. According to the literature, once [PdCl₄]^2- ions adsorb onto the TiO₂ surface, each ion releases two Cl^- ligands, forming a single PdCl₂ unit on the TiO₂ surface³⁹. Subsequently, the •CH₂CN radicals react with the PdCl₂ unit³², removing chlorine and ultimately, through the combined action of nitrogen atoms and lattice oxygen on TiO₂, anchoring the Pd single atom on the TiO₂ surface, forming a stable Pd₁-N₂O₁ structure (Fig. 2l, S2, S11, S23)⁴⁰.

Based on these experimental findings, we propose a mechanism for the photoinduced atomic dispersion of Pd NPs, as depicted in Fig. 3l. This process can be divided into two key steps: (1) Pd-Pd bond cleavage: The •Cl and •O₂^- radicals generated by UV light excitation, along with chloride ions, cooperate to break the Pd-Pd bonds, generating the intermediate [PdCl₄]^2-. (2) Pd single atom formation: The intermediate [PdCl₄]^2- adsorbs onto the TiO₂ surface via electrostatic interactions and releases two chloride ions, forming a PdCl₂ unit. The •CH₂CN radicals then react with the PdCl₂ unit, removing the chloride ions and ultimately anchoring the Pd single atom at the N sites on the TiO₂ surface, forming a stable Pd₁-N₂O₁ structure. During this process, the radicals in the solution and the removed chloride ions continue to interact with the remaining Pd-Pd bonds, until all Pd-Pd bonds are converted to Pd-N/O bonds, achieving complete atomic dispersion of Pd NPs.

To further elucidate the mechanism underlying the transformation of Pd NPs to Pd single atoms, we performed density functional theory (DFT) calculations to simulate the interactions between free radicals and Pd NPs, as well as the formation of Pd single atoms (Fig. 4). For simplicity, a small nanoparticle consisting of 12 Pd atoms (Pd₁₂) was chosen as the model for Pd NPs, and an anatase TiO₂ (101) facet with N sites was used as the support.

Fig. 4. Energies and optimized structures of the Pd-Pd bond cleavage process and the formation of Pd single atoms on TiO₂.

(I) The initial state, where Pd nanoparticles are supported on TiO₂. (I-II) The adsorption of •Cl radicals onto Pd nanoparticles. (II-III) The attack of •O₂^- radicals cooperates with •Cl, leading to Pd-Pd bond cleavage and the formation of the intermediate [PdCl₄]^2- species. (III-IV) The intermediate adsorbs onto the TiO₂ surface, releasing two Cl^- ions and forming the PdCl₂ unit. (IV-VI) The PdCl₂ unit progressively dechlorinates, ultimately forming a stable Pd₁-N₂O₁ structure.

First, we simulated the effect of •Cl on the Pd-Pd bond cleavage process. The DFT calculation results show that after •Cl radicals adsorb onto the Pd atoms at the surface of Pd NPs, the Pd-Pd bond length increases, and the bond energy decreases (Fig. 4, I-II and Fig. S21). The free energy change for this process is significantly negative (ΔG = −3.51 eV), indicating that the adsorption of •Cl is thermodynamically favorable. Additionally, the charge density difference calculation (Fig. S22) reveals that •Cl adsorption reduces the electron density around the Pd atoms, thereby weakening the stability of the Pd-Pd bond. Subsequently, •O₂^- attacks the same Pd atom, cooperating with •Cl to further weaken the Pd-Pd bond, ultimately leading to bond cleavage. Meanwhile, Cl^- in the solution coordinates with the cleaved Pd atom, forming the intermediate product [PdCl₄]^2- (Fig. 4, II-III). The entire Pd-Pd bond cleavage process (Fig. 4, I-III) releases 0.61 eV of energy, indicating that this process is thermodynamically spontaneous. Notably, in the absence of •Cl, the Pd-Pd bond cleavage is endergonic (ΔG = 6.03 eV, Path I-III in Fig. 4), highlighting the crucial role of •Cl in Pd-Pd bond cleavage.

Next, the intermediate product [PdCl₄]^2- adsorbs onto the TiO₂ surface and releases two Cl^- ions, forming the adsorption structure of a single PdCl₂ unit (Fig. 4, III-IV). This adsorption structure corresponds to a Pd-N₂Cl₂ coordination environment. COHP analysis (Fig. S23) shows that the Pd-N bonds exhibit stronger bonding interactions than the Pd-Cl bonds, supporting the stabilization and N-site anchoring of the adsorbed intermediate. The calculation results show that this adsorption process releases 0.84 eV of energy, suggesting that the adsorption of the intermediate product is thermodynamically favorable. Given the high reactivity of •CH₂CN, these radicals readily react with the adsorbed PdCl₂ unit to remove chloride. Our calculations further show that the PdCl₂ unit progressively dechlorinates, ultimately forming Pd single atoms that are anchored at N sites on the TiO₂ surface, resulting in a stable Pd₁-N₂O₁ structure (Fig. 4, IV-VI and Fig. S23). The formation of Pd single atoms (Fig. 4, III-VI) releases 0.8 eV of energy.

Considering the entire reaction pathway, the overall free energy change from Pd-Pd bond cleavage to Pd single atom formation (Fig. 4, I-VI) is ΔG = −1.41 eV, indicating that this reaction pathway is thermodynamically highly favorable.

To assess the universality of the photoinduced strategy, we applied it to TiO₂ supports with varying particle sizes and to different noble metal nanoparticles. TiO₂-S, with a particle diameter of approximately 2 – 5 nm, was used as a support to investigate the photoinduced atomic dispersion effects of Pd NPs, Pt NPs, and Rh NPs. HAADF-STEM images revealed that, prior to light irradiation, the diameter of Pd NPs was approximately 3.43 nm (Fig. 5a and Fig. S24). After irradiation, the Pd NP structure was no longer visible, and EDS analysis indicated that the Pd signal was uniformly distributed across the TiO₂ surface (Fig. 5b and Fig. S25). CO-DRIFTS spectra (Fig. 5c, d) show that before light irradiation, an absorption peak at 1972 cm^–1 was observed, corresponding to CO bridged on nanoparticles²⁶, suggesting that Pd predominantly existed as nanoparticles. After light irradiation, the CO adsorption peak appears only at 2160 cm^–1, attributed to CO linearly adsorbed on Pd single atoms²⁹, confirming the transformation of Pd NPs into single atoms.

Fig. 5. Generality of the photoinduced atomic dispersion strategy.

HAADF-STEM images and EDS elemental mappings of Pd_NPs/TiO₂ (a) before and b after photoinduced treatment, respectively. DRIFTS spectra of CO adsorption at saturation coverage and followed by Ar flushing on Pd_NPs/TiO₂ (c) before and d after photoinduced treatment, respectively. HAADF-STEM images and EDS elemental mappings of Pt_NPs/TiO₂ (e) before and f after photoinduced treatment, respectively. DRIFTS spectra of CO adsorption at saturation coverage and followed by Ar flushing on Pt_NPs/TiO₂ (g) before and h after photoinduced treatment, respectively. HAADF-STEM images and EDS elemental mappings of Rh_NPs/TiO₂ (i) before and j after photoinduced treatment, respectively. DRIFTS spectra of CO adsorption at saturation coverage and followed by Ar flushing on Rh_NPs/TiO₂ (k) before and l after photoinduced treatment, respectively.

A similar transformation was observed for Pt NPs and Rh NPs under photoinduced treatment. Prior to light irradiation, the Pt NPs were approximately 2.43 nm in diameter (Fig. 5e and S26), and after irradiation, the Pt NPs disappeared, with EDS showing uniform Pt distribution on the TiO₂ surface (Fig. 5f and S27). CO-DRIFTS (Fig. 5g and Fig. 5h) indicated a peak at 1839 cm^–1 before irradiation, consistent with CO bridged on Pt nanoparticles⁹. After irradiation, the CO adsorption peak appears only at 2090 cm^–1, associated with CO linearly adsorbed on Pt single atoms⁴¹, confirming the transformation of Pt NPs to single atoms. Similarly, Rh NPs, which were approximately 1.78 nm in diameter before irradiation (Fig. 5i and S28), exhibited complete dispersion after irradiation, with Rh uniformly distributed on the TiO₂ surface (Fig. 5j and S29). CO-DRIFTS (Fig. 5k, l) showed a peak at 1873 cm^–1 before light irradiation, consistent with CO bridged on Rh nanoparticles⁴². After light irradiation, two peaks appeared at 2085 and 2015 cm^–1, corresponding to the symmetric and asymmetric stretching vibrations of Rh(CO)₂ gem-dicarbonyl species at Rh single atom sites⁴³, further confirming the conversion of Rh NPs into single atoms.

To further extend the support scope, we applied the photoinduced strategy to WO₃. For Pd_NPs/WO₃, HAADF-STEM and EDS mapping, together with CO-DRIFTS, confirm the conversion of Pd nanoparticles to atomically dispersed Pd species after irradiation (Fig. S30). In addition, larger Pd nanoparticles prepared by chemical reduction and subsequently loaded onto TiO₂ (Pd_NPs-CR/TiO₂) were also tested under prolonged UV irradiation using the photoinduced strategy. HAADF-STEM/EDS mapping and CO-DRIFTS confirm the conversion of these Pd nanoparticles into atomically dispersed Pd species (Fig. S31). These results demonstrate the broad applicability of the photoinduced strategy to different oxide supports (TiO₂ and WO₃) and diverse noble metal nanoparticles (Pd, Pt, and Rh), highlighting the universality of this approach.

The photoinduced atomic dispersion strategy shows immense potential in enhancing and regenerating the performance of industrial catalysts. Our research further validates the effectiveness of this strategy in improving and regenerating the performance of commercial Pd/C (Comm. Pd/C) and industrial waste Pd/C (Ind. W. Pd/C) catalysts. By applying this strategy to these catalysts, we observed atomic dispersion of Pd NPs under photoinduced conditions, as confirmed by HRTEM and CO-DRIFTS characterization (Figs. S32–S33). This structural transformation significantly increases the exposure of Pd atoms. In the hydrogenation of styrene, the photoinduced treated commercial Pd/C catalyst [Pd₁/C&TiO₂(Comm.)] and industrial waste Pd/C catalyst [Pd₁/C&TiO₂(Ind.W.)] exhibited remarkable performance improvements (Fig. 6a, b). Specifically, compared to untreated catalysts (Comm. Pd/C and Ind. W. Pd/C), the catalytic activity of Pd₁/C&TiO₂(Comm.) and Pd₁/C&TiO₂(Ind.W.) increased by 17.8 and 26 times, respectively. Additionally, the TOF values of the photoinduced treated catalysts showed significant improvement (Fig. 6c), with the TOF values of Pd₁/C&TiO₂(Comm.) and Pd₁/C&TiO₂(Ind.W.) reaching 8933 h^–1 and 7067 h^–1, respectively—an increase of 8.2 and 8.5 times compared to Comm. Pd/C (1088 h^–1) and Ind. W. Pd/C (828 h^–1). Notably, both Pd₁/C&TiO₂(Comm.) and Pd₁/C&TiO₂(Ind.W.) maintained nearly 100% conversion after 5 reaction cycles (Fig. 6d), demonstrating excellent stability and reusability.

Fig. 6. Practical applications of the photoinduced strategy.

a Catalytic performance of Pd₁/C&TiO₂(Comm.) and Pd₁/C&TiO₂(Ind.W.) in styrene hydrogenation over time. b Comparison of catalytic performance of TiO₂, Comm. Pd/C, Ind. W. Pd/C, Pd₁/C&TiO₂(Comm.), and Pd₁/C&TiO₂(Ind.W.) in styrene hydrogenation. c Turnover frequency (TOF) (h^–1) for styrene hydrogenation reactions catalyzed by the materials shown in (b). d Cyclic performance of Pd₁/C&TiO₂(Comm.) and Pd₁/C&TiO₂(Ind.W.) for styrene hydrogenation reactions. e Schematic illustration of the photoinduced atomic dispersion process using a continuous flow reactor. Error bars represent the mean ± s.d. from three independent experiments (n = 3).

By gaining a deeper understanding of the microscopic mechanism behind the photoinduced nanoparticle atomic dispersion strategy, we aim to apply this approach on a larger scale using continuous flow reactors to regenerate and improve industrial catalysts (Fig. 6e and S34). This method successfully achieved atomic dispersion of Pd NPs and enhanced catalytic performance on Ind. W. Pd/C under mild conditions (Fig. S35). Notably, we also explored using sunlight as a light source, which successfully facilitated the atomic dispersion of Pd NPs on Pd_NPs/TiO₂, further enhancing the catalytic performance of Pd_NPs/TiO₂ (Fig. S36).

Discussion

In conclusion, we introduce a groundbreaking photoinduced strategy for achieving the atomic dispersion of noble metal nanoparticles at ambient temperature and pressure. By leveraging advanced techniques such as STEM, CO-DRIFTS, EXAFS, and DFT calculations, we provide detailed insights into the microscopic mechanisms underpinning this novel dispersion process. Our experimental findings demonstrate the broad applicability of this strategy to a range of noble metals, including Pd, Pt, and Rh, and to different oxide supports such as TiO₂ and WO₃. Notably, the catalytic performance of commercial Pd/C and industrial waste Pd/C catalysts, after photoinduced treatment, exhibited remarkable enhancements in the styrene hydrogenation reaction. This highlights the immense potential of the strategy for the modification and regeneration of industrial catalysts. These results offer valuable perspectives for the efficient use and sustainable development of noble-metal-based catalysts. Moving forward, we aim to expand the applicability of this strategy to other metals and support systems, while assessing its long-term effectiveness and economic viability in industrial applications.

Methods

Chemicals

Titanium oxide (TiO₂) in two forms was used: TiO₂-L (particle diameter ~25 nm, anatase, 99.8%) and TiO₂-S (particle diameter approximately 2 – 5 nm, anatase, 99%).

Other chemicals included acetonitrile (CH₃CN, HPLC, 99.9%), 2,3-dicyano-5,6-dichlorobenzoquinone (DDQ, AR, 98%), methanol (CH₃OH, GR, ≥99.7%), p-benzoquinone (C₆H₄O₂, RG, 99%), hydrochloric acid (HCl, AR, 36-38%), styrene (C₆H₅CH = CH₂, 99%), ethanol (C₂H₅OH, HPLC, 99.9%), ethylbenzene (C₆H₅C₂H₅, 99%), dimethylformamide (DMF, AR, ≥99.5%), hydrogen (H₂, 99.999%), palladium chloride (PdCl₂, 99.9%), chloroplatinic acid hexahydrate (H₂PtCl₆·6H₂O, RG, 99.9%), rhodium(III) chloride hydrate (RhCl₃·xH₂O, RG, 99%), Tungsten(VI) oxide (WO₃, 99.95%), potassium iodide (KI, AR, 99%), polyvinylpyrrolidone (PVP, M_w ~ 55,000), sodium tetrachloropalladate (Na₂PdCl₄, 99%), formamide (HCONH₂, AR, ≥99.5%), and acetone (AR, 99.5%).

Commercial Pd/C (Comm.Pd/C) with a palladium loading of 5 wt.% was purchased from Adamas-beta (Shanghai, China). Industrial waste Pd/C (Ind.W.Pd/C) with a palladium loading of 10 wt.% was collected after long-term industrial use in hydrogenation reactions. All the chemicals were used as received.

Two different supports, TiO₂-L and TiO₂-S, were used in this study. Among them, TiO₂-L has a particle diameter of approximately 25 nm. As a support, its HAADF-STEM characterization results are relatively clear, enabling more accurate statistical analysis of the diameter distribution of Pd NPs and in-depth exploration of the atomic dispersion mechanism of Pd NPs under photoinduced. On the other hand, TiO₂-S has a smaller particle diameter (approximately 2 – 5 nm) and a larger specific surface area, which provides favorable conditions for studying the universality and practical applications of photoinduced strategy.

Characterization

High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images and energy dispersive X-ray spectroscopy (EDS) elemental mapping were collected on TALOS F200S. Aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (AC HAADF-STEM) analysis was performed on ThemisZ transmission electron microscope (accelerating voltage, 300 kV). The contents of metal on the catalysts were analyzed by inductively coupled plasma optical emission spectrometer (ICP-OES, EXPEC 6500). The X-ray absorption fine structure spectra (Pd K-edge) were collected at beamline BL44B2 at the SPring-8 synchrotron in Japan. The storage rings of SPring-8 were operated at 8.0 GeV with a maximum current of 250 mA. using Si(111) double-crystal monochromator. All spectra were collected at room temperature. The XAFS spectra were recorded with fluorescence using a 19-element Ge solid-state detector mode. X-ray photoelectron spectroscopy (XPS) data were obtained on Perkin-Elmer PHI 5000. The electron paramagnetic resonance (EPR) spectra of radicals during photoinduced reactions were recorded on a Bruker A300 model spectrometer. The photoinduced reaction intermediate species was investigated using a UV-Vis spectrophotometer (UV-7502, Shimadzu). The zeta potential of the catalysts was measured using a zeta potential analyzer (Malvern Zetasizer Nano ZS). The styrene hydrogenation reaction was analyzed using a gas chromatograph (GC-7890B, Agilent).

DRIFTS for CO adsorption

DRIFTS measurements were performed using a Bruker VERTEX 80 v Fourier-transform spectrometer equipped with a Harrick diffuse reflectance accessory with ZnSe and quartz window. After the sample was loaded into the reaction chamber, pure Ar (99.999%) was purged into the chamber firstly for background spectra collection. Subsequently, CO was introduced into the system until CO adsorbed saturation. Then the pure Ar was purged to remove the gaseous CO and the spectra were collected.

Synthesis of Pd_NPs/TiO₂-L

In total, 0.25 g of commercial anatase TiO₂-L was dispersed in 50 mL of ethanol in a sealed quartz vial. The TiO₂-L dispersion was purged with Ar gas for 15 min. The TiO₂-L dispersion was vigorously stirred and simultaneously irradiated under a 100 W, 360 nm UV lamp for 5 min. H₂PdCl₄ solution was diluted to 5 mL using ethanol and then injected into the TiO₂-L dispersion. The mixture was stirred vigorously for an additional 20 min. The Pd_NPs/TiO₂-L product was collected by centrifugation, washing, and drying. The Pd content in Pd_NPs/TiO₂-L was determined to be 0.35 wt.% by ICP analysis. Among them, H₂PdCl₄ is prepared by mixing PdCl₂ with HCl.

Synthesis of M_NPs/TiO₂-S

In total, 100 mg of TiO₂-S was dispersed in 20 mL of ethanol and placed in a quartz bottle, followed by the addition of 3.7 μL of a 1 mol/L H₂PdCl₄ solution. The mixture was vigorously stirred and then sealed with the bottle cap. The uniformly dispersed mixture was irradiated under a 360 nm UV lamp at an intensity of 500 mW/cm² for 0.5 h. After irradiation, the Pd_NPs/TiO₂-S product was collected through centrifugation, washing, and drying. The Pd content in Pd_NPs/TiO₂-S was determined to be 0.38 wt.% by ICP analysis. The synthesis of Pt_NPs/TiO₂-S and Rh_NPs/TiO₂-S followed the same procedure, with H₂PtCl₆·6H₂O and H₂RhCl₆ replacing H₂PdCl₄ as the metal precursors, respectively. The Pt content in Pt_NPs/TiO₂-S was determined to be 0.36 wt.% by ICP analysis. The Rh content in Rh_NPs/TiO₂-S was determined to be 0.37 wt.% by ICP analysis. Among them, H₂PdCl₄ and H₂RhCl₆ are prepared by mixing the corresponding precious metal chlorides with hydrochloric acid.

Synthesis of Pd_NPs/WO₃

In total, 100 mg of WO₃ was dispersed in 20 mL of ethanol and placed in a quartz bottle, followed by the addition of 2.8 μL of a 1 mol/L H₂PdCl₄ solution. The mixture was vigorously stirred and then sealed with the bottle cap. The uniformly dispersed mixture was irradiated under a 360 nm UV lamp at an intensity of 500 mW/cm2 for 0.5 h. After irradiation, the Pd_NPs/WO₃ product was collected by centrifugation, washing, and drying. The Pd content in Pd_NPs/WO₃ was determined to be 0.28 wt.% by ICP analysis.

Synthesis of Pd₁/WO₃

In total, 50 mg of Pd_NPs/WO₃ was dispersed in 20 mL of CH₃CN and placed in a quartz bottle, followed by the addition of 1 μL of HCl. After placing the cap on the bottle, the mixture was vigorously stirred. The uniformly dispersed mixture was irradiated under a 360 nm UV lamp with an intensity of 500 mW/cm2 for 2.5 h. After irradiation, the Pd₁/WO₃ product was collected by centrifugation, washing, and drying. ICP analysis determined the Pd content in Pd₁/WO₃ to be 0.26 wt.%.

Synthesis of Pd nanoparticle

In total, 34 mg KI and 100 mg PVP (M_w ~55,000) were dissolved in 10 mL formamide and heated to 120 °C. Subsequently, 59 mg of solid Na₂PdCl₄ was added, and the mixture was maintained for 10 min under vigorous stirring. After the reaction, the mixture was cooled to room temperature, 120 mL acetone was added, and the product was collected by centrifugation. The precipitate was washed with acetone by repeated centrifugation. Finally, the obtained Pd nanoparticles were redispersed in 21 mL ethanol.

Synthesis of Pd_NPs-CR/TiO₂

In total, 500 mg TiO₂-S was dispersed in 5 mL ethanol under stirring to obtain a homogeneous suspension. Subsequently, 1.5 mL of the as-prepared Pd NP dispersion (prepared by the chemical reduction method) was added, and the mixture was stirred for 1 h. The product was collected by centrifugation, washing, and drying. The Pd loading in Pd_NPs-CR/TiO₂ was determined to be 0.29 wt% by ICP analysis.

Synthesis of Pd₁-CR/TiO₂

In total, 50 mg of Pd_NPs-CR/TiO₂ was dispersed in 20 mL of CH₃CN and placed in a quartz bottle, followed by the addition of 1 μL of HCl. After placing the cap on the bottle, the mixture was vigorously stirred. The uniformly dispersed mixture was irradiated under a 360 nm UV lamp with an intensity of 500 mW/cm2 for 6 h. After irradiation, the Pd₁-CR/TiO₂ product was collected by centrifugation, washing, and drying. ICP analysis determined the Pd content in Pd₁-CR/TiO₂ to be 0.28 wt%.

Synthesis of M₁/TiO₂

In total, 50 mg of Pd_NPs/TiO₂-L was dispersed in 20 mL of CH₃CN and placed in a quartz bottle, followed by the addition of 1 μL of HCl. After placing the cap on the bottle, the mixture was vigorously stirred. The uniformly dispersed mixture was irradiated under a 360 nm UV lamp with an intensity of 500 mW/cm² for 1 h. After irradiation, the Pd₁/TiO₂-L product was collected through centrifugation, washing, and drying. ICP analysis determined the Pd content in Pd₁/TiO₂-L to be 0.35 wt.%. The nitrogen content of the Pd₁/TiO₂ -L sample was quantified using an oxygen/nitrogen/hydrogen elemental analyzer (LECO ONH836) and determined to be 0.23 wt%. The synthesis of Pd₁/TiO₂-S, Pt₁/TiO₂-S, and Rh₁/TiO₂-S followed the same procedure, with Pd_NPs/TiO₂-S, Pt_NPs/TiO₂-S, and Rh_NPs/TiO₂-S replacing Pd_NPs/TiO₂-L as the precursors, respectively.

Synthesis of Pd₁/C&TiO₂(Comm.)

In total, 8.3 mg of Comm.Pd/C and 100 mg of TiO₂-S were dispersed in 40 mL of CH₃CN and placed in a quartz bottle, followed by the addition of 2 μL of HCl. After placing the cap on the bottle, the mixture was vigorously stirred. The uniformly dispersed mixture was irradiated under a 360 nm UV lamp with an intensity of 50 mW/cm² for 0.5 h. After irradiation, the Pd₁/C&TiO₂(Comm.) product was collected through centrifugation, washing, and drying. ICP analysis determined the Pd content in Pd₁/C&TiO₂(Comm.) to be 0.378 wt.%.

Synthesis of Pd₁/C&TiO₂(Ind.W.)

In total, 4.0 mg of Ind.W.Pd/C and 100 mg of TiO₂-S were dispersed in 40 mL of CH3CN and placed in a quartz bottle, followed by the addition of 2 μL of HCl. After placing the cap on the bottle, the mixture was vigorously stirred. The uniformly dispersed mixture was irradiated under a 360 nm UV lamp with an intensity of 50 mW/cm² for 1 h. After irradiation, the Pd₁/C&TiO₂(Ind.W.) product was collected through centrifugation, washing, and drying. ICP analysis determined the Pd content in Pd₁/C&TiO₂(Ind.W.) to be 0.370 wt.%.

EPR

EPR measurements were performed using a Bruker A300 spectrometer. The instrument was operated at a center field of 3510.0 G, sweep width of 100.0 G, microwave frequency of 9.853 GHz, microwave power of 19.72 mW, modulation frequency of 100.0 kHz, modulation amplitude of 1.0 G, time constant of 10.24 ms, and sweep time of 46.08 s. For the detection of chlorine radicals (•Cl), 50 mg of Pd_NPs/TiO₂ was dispersed in 20 mL of a CH₃CN/HCl solution (1 μL HCl), and 100 μL of the dispersion was mixed with 100 μL of 100 mM PBN solution. A 60 μL aliquot of the resulting mixture was loaded into a capillary and analyzed using in situ EPR under UV irradiation at 0, 1, and 10 min. For the detection of superoxide radicals (•O₂^-) and alkyl radicals (•CH₂CN), the same dispersion protocol was used, and 100 μL of the catalyst suspension was mixed with 100 μL of 100 mM DMPO solution. A 60 μL sample was loaded into a capillary and analyzed using in situ EPR under UV irradiation at 0, 1, and 10 min.

Catalytic tests

The catalyst, Pd₁/C&TiO₂(Comm.), was first dispersed in ethanol. Certain amount of the dispersion was taken for catalytic reaction, depended on the amount of catalyst needed. To be specific, for styrene hydrogenation with a mole ratio of 1.5:10000 (Pd:styrene), 7.5 × 10^–3 μmol Pd was needed. After the taken dispersion of Pd₁/C&TiO₂(Comm.) was diluted in 10 mL ethanol in a glass pressure vessel, the catalyst dispersion was then sonicated and mixed with 50 μmol styrene by a magnetic stirring bar. The vessel was then charged with H₂ to 1 atm pressure and kept at 30 °C in a oil bath during stirring. The reaction progress and conversion were monitored using gas chromatographic analysis. For the other catalysts [i.e, TiO₂, Comm.Pd/C, Ind.W.Pd/C and Pd₁/C&TiO₂(Ind.W.)], they were also dispersed in ethanol for catalysis. The reaction conditions were kept the same with that for Pd₁/C&TiO₂(Comm.). Catalytic activity data were obtained from three independent experiments under identical conditions, and the average values were reported.

Metal dispersion determination

The metal dispersion was determined for turn over frequency (TOF) calculations by CO titration experiments. The CO titration experiments were carried out on a Micromeritics Auto Chem II 2920 chemical adsorption instrument with TCD detector.

DFT calculations

First-principles calculations based on density functional theory (DFT) were performed using the Vienna Ab initio Simulation Package (VASP.6.1.2)⁴⁴. The exchange-correlation energy was described within the generalized gradient approximation (GGA) with the Perdew-Burke-Ernzerhof (PBE) functional⁴⁵. The projector augmented wave (PAW) method was employed in combination with a plane-wave basis set⁴⁶, with a cutoff energy of 400 eV. Convergence criteria were set to 1 × 10^–4eV for the total energy and 5 × 10^–2eV/Å for the atomic forces. All calculations were carried out in a solvent environment, where the acetonitrile electrolyte (relative permittivity ε = 37.5) was implicitly modeled using the polarizable dielectric continuum model implemented in the VASPsol code⁴⁷. The TiO₂ support was modeled using an anatase TiO₂ (101) slab constructed from the bulk structure and expanded to a 3 × 4 surface supercell (cell parameters: a = 16.3287 Å, b = 18.8800 Å, c = 25.3775 Å). The slab consists of three TiO₂ layers. To eliminate spurious interactions between periodic slab images, a 15 Å vacuum layer was introduced along the z-direction. During geometry optimization, the topmost TiO₂ layer was fully relaxed, while the bottom two layers were kept fixed at their bulk positions to mimic a semi-infinite substrate. All adsorbates and Pd species were fully optimized together with the relaxed surface layer⁴⁸. Additionally, dipole moment correction was applied in all surface structure optimizations to suppress artificial electric fields arising from the asymmetric slab-vacuum configuration⁴⁹. Brillouin zone sampling was performed using the Monkhorst-Pack scheme, with k-point grids generated by VASPkit⁵⁰ following a recommended spacing of 0.04 × 2π Å^-1. van der Waals interactions were accounted for using the DFT-D3 dispersion correction method⁵¹.

The adsorption energy (E_ad) was calculated to evaluate the stability of adsorbed species, following the formula:

$$E_{\mathrm{ad}}=E_{\mathrm{total}}-E_{\mathrm{slab}}-E\left(\mathrm{M}\right)$$ 5

where E_total denotes the total energy of the slab-adsorbate system, E_slab is the energy of the clean slab, and E(M) represents the energy of the molecule (M).

Reaction free energies (ΔG) were computed by considering the enthalpic and entropic contributions at 298.15 K and 1 atm, using the relation:

$$\mathrm{\Delta }G=\mathrm{\Delta }{E}_{\mathrm{DFT}}+\mathrm{\Delta }{E}_{\mathrm{zpe}}-T\mathrm{\Delta }S$$ 6

where E_DFT is the total energy obtained from DFT calculations, E_zpe denotes the zero-point vibrational energy (derived from harmonic frequency analysis), The entropy of adsorbed molecules (TΔS) is calculated from the vibrational frequencies associated with the normal modes in the harmonic approximation.

Author contributions

X.C. and Z.B. proposed the study. X.C. conceived the research, designed the experiments, characterized the samples, and drafted the manuscript. X.C. and K.Z. performed the DFT simulations. Z.B. and X.Q. conceived the research and designed the experiments. Q.Z. and X.Q. were responsible for the STEM characterization. J.Z. and K.Z. conceived the research and performed catalytic evaluation. All the authors discussed the results and participated in writing the manuscript.

Peer review

Peer review information

Nature Communications thanks Xiaoqiang An, Andrew Logsdail, who co-reviewed with Igor Kowalec, and the other, anonymous, reviewer(s) 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/s41467-026-70742-3.