Single-particle catalytic monitoring of interfacial charge dynamics during light-driven CdS shell growth on Au nanocubes

Abstract

Noble metal nanomaterials, with localized surface plasmon resonance (LSPR) properties, have shown great promise in photocatalysis. However, understanding plasmon-induced interfacial charge transfer dynamics and optimizing photocatalytic efficiency remain significant challenges. Here we show the precise control of the growth of CdS shells through light-driven deposition while monitoring interfacial charge dynamics by using Au nanocubes (Au NCs) as plasmonic substrates. We discover a non-monotonic relationship between the CdS shell thickness and photocatalytic activity. An optimal shell thickness maximizes the hot-electron transfer efficiency, achieving a 2.62-fold enhancement in catalytic activity compared to bare Au NCs by effectively balancing carrier separation and recombination. Single-particle analysis, through synchronized dark-field scattering and single-molecule catalytic imaging, reveals heterogeneous catalytic behaviors associated with atomic-scale structural variations and dynamic charge transfer mechanisms. The localized electromagnetic field enhancement around Au NCs facilitates hot-carrier generation for CdS deposition, and interfacial Schottky barriers extend hot-carrier lifetimes for redox reactions. This work provides a framework for designing dynamic metal-semiconductor interfaces. It not only deepens our fundamental understanding of plasmon-induced energy conversion but also serves as a guiding principle for the development of high-efficiency solar fuel systems.

Plasmonic nanoparticles show promise for photocatalysis, yet the dynamics of charge transfer remain poorly understood. Here, the authors demonstrate the light-driven growth of a CdS shell on Au nanocubes, correlating shell thickness with charge transfer efficiency at the single-particle level.

Introduction

Solar energy has emerged as a pivotal solution to the global energy crisis due to its accessibility and renewability. Since the groundbreaking work by Honda and Fujishima in 1972 on semiconductor-based photocatalytic water splitting, photocatalysis has garnered significant attention for its capacity to convert solar energy into clean chemical fuels^1,2. Current research focuses on enhancing energy utilization and conversion efficiency through structural optimization of photocatalysts, including tailoring morphological configurations, modifying chemical compositions, improving light absorption capabilities, and facilitating the separation/exploitation of photogenerated hot-carriers^2–4. Nevertheless, the solar-to-chemical conversion efficiency of most semiconductor photocatalysts remains unsatisfactory, necessitating the development of more efficient materials^1,5.

Among various photocatalysts, noble metal nanomaterials have attracted considerable interest owing to their localized surface plasmon resonance (LSPR) properties^6–9. The resonance wavelength and intensity of LSPR can be precisely tuned by adjusting nanoparticle size, morphology, and composition, enabling spectral overlap with solar irradiation for optimized energy harvesting^10,11. For instance, rationally designed plasmonic nanostructures can concentrate incident light through electromagnetic field enhancement at the particle surfaces, generating abundant hot-carriers for catalytic reactions^12,13. However, plasmon-excited high-energy electrons predominantly undergo radiative scattering or thermal dissipation rather than participating in chemical reaction processes^6,14. Moreover, the ultrafast relaxation dynamics (sub-picosecond scale) of hot-carriers severely limit their practical utilization^7,12,15. A promising strategy to prolong carrier lifetime involves constructing metal-semiconductor heterostructures, which exhibit superior catalytic performance compared to individual components^16–19. Upon photoexcitation, interfacial hot-electrons can transfer to the semiconductor conduction band or adsorbate molecular orbitals to drive redox reactions. Thus, elucidating hot-electron transfer mechanisms in hybrid nanostructures becomes crucial for advancing photocatalytic efficiency. Nevertheless, fundamental understanding of plasmon-induced interfacial charge transfer dynamics and its correlation with catalytic activity remains incomplete^20–22.

The performance of plasmonic heterostructures critically depends on the metal-semiconductor interface design¹⁵. Optimal configurations facilitate Schottky barrier formation, promoting hot-carrier generation/separation while preventing metal core oxidation, aggregation, or chemical poisoning during catalysis¹⁰. For example, the strong Au-S affinity enables selective CdS deposition on Au nanoparticles, forming uniform core-shell architectures (Au NC@CdS) that overcome lattice mismatch challenges while enhancing electron mobility^23–25. Conventional synthetic methods, however, often yield structurally heterogeneous composites, complicating mechanistic studies of structure-activity relationships. Particularly, the lack of in-situ characterization tools hinders the evaluation of catalytic activity evolution during semiconductor loading. These dynamic chemical insights are essential for guiding the design of more efficient photocatalysts.

Addressing this challenge requires developing real-time in-situ monitoring techniques to unravel the critical factors governing plasmon-enhanced photocatalysis. Among existing methods, surface-enhanced Raman spectroscopy (SERS) has been widely employed for tracking molecular “fingerprint” variations during chemical reactions, leveraging plasmon-induced electromagnetic field amplification²⁶. Nevertheless, SERS inadequately resolves nanoscale structural transformations. Currently, single-molecule fluorescence imaging provides single-turnover resolution for catalytic activity assessment by eliminating ensemble-averaging effects^27–32. Meanwhile, single-particle dark-field scattering (DFS) microscopy enables real-time monitoring of resonance wavelength shifts during semiconductor growth on plasmonic substrates^33–36. Integrating these techniques offers opportunities to correlate structural evolution with catalytic performance at the single-particle level.

In this work, we employed in-situ microscopic spectroscopy to investigate light-driven deposition of CdS shells on plasmonic Au NCs, achieving control over semiconductor shell formation while monitoring catalytic activity dynamics. Au NCs with low-coordination edges/corners generate intense localized electromagnetic fields, making them ideal substrates for functional nanostructure assembly. To address the dual challenges of “photon harvesting” and “carrier utilization”, we strategically utilized plasmon resonance effects to drive photodeposition of CdS shells, systematically investigating carrier separation-dependent catalytic efficiency. The structural evolution was monitored in real-time via DFS spectral shifts, while single-molecule fluorescence imaging quantified catalytic activity changes before/after semiconductor loading. Intriguingly, catalytic activity exhibited a non-monotonic trend (initial enhancement followed by gradual decline) during shell growth. This in-situ approach circumvents ensemble-averaging limitations inherent to solution-phase syntheses, providing single-particle resolution insights into structure-activity correlations. Our findings establish a framework for designing more efficient photocatalysts and offer a visualization-guided strategy for controllable nanomaterial synthesis.

Results

Spectral tracking of CdS shell growth on Au NCs

To achieve controlled formation of metal-semiconductor core-shell nanostructures, we systematically investigated the conditions for in-situ photodeposition of semiconductor shell on plasmonic nanoparticles. Initially, Au NCs at an optimized concentration were injected into the reaction channel in the absence of semiconductor precursors. The plasmon-driven photodeposition of a CdS shell onto metal nanoparticles is a photochemical process initiated by the resonant excitation of LSPR in the metal core, Fig. 1a^37–39. Upon illumination, the decay of plasmonic oscillations generates energetic hot-carriers, which can inject into adsorbed sulfur precursors (e.g., thioacetamide, TAA) at the metal-solution interface and catalytically decompose them to release S²⁻ ions. This creates a localized high concentration of reactive anions at the nanoparticle surface, which immediately reacts with Cd²⁺ ions in solution. The resulting CdS selectively undergoes heterogeneous nucleation on the metal surface, as it provides a lower-energy barrier compared to homogeneous nucleation. Continued illumination sustains this cycle, leading to the controlled, layer-by-layer growth of a uniform and contiguous semiconductor shell. The spatial specificity of the reaction is dictated entirely by the light-induced plasmonic activity, making this a method for fabricating well-defined metal-semiconductor core-shell nanostructures under mild conditions. Without precursor solution (CdCl₂ + TAA), even under 532 nm laser irradiation for 20 min, the dark-field microscopy (DFM) images and scattering spectra of Au NCs remained stable, with no significant changes in scattering intensity or spectral peak position (Supplementary Fig. 1). These observations confirmed that Au NCs alone exhibited no intrinsic spectral instability under laser illumination.

Fig. 1. Exploration of in-situ photodeposition conditions of CdS shell on Au NCs.

a Schematic diagram of photodeposition. b Single-particle scattering intensity of Au NCs at 0 and 20 min without 532 nm laser irradiation. c Single-particle scattering intensity of Au NCs under 532 nm laser irradiation with precursor solution (CdCl₂ +  TAA, 4.0 mM) at 0 and 20 min, respectively. Data in (b, c) are presented as mean values ± SD from independent experiments (n = 10). The scale bars in the inset images represent 0.5 μm. d Single-particle scattering intensity changes of Au NCs at different time intervals from 0 to 30 min with precursor solution (CdCl₂ + TAA, 2.0 mM) under 532 nm laser irradiation. e In-situ DFM images of Au NCs at different time intervals from 0 to 20 min with precursor solution (CdCl₂ + TAA, 4.0 mM) under 532 nm laser irradiation. The scale bar in the inset image represents 0.5 μm. The scale bar applies to all inset images. f Single-particle scattering spectra at different times. The corresponding time-dependent spectral changes at the peak value (g) and scattering intensity as a function of time (h). i DFM image with localized laser irradiation for photodeposition of CdS shell on Au NCs. Laser irradiation in the area inside the white circle, with no laser exposure outside the area. Images are representative of three independent experiments with similar results in (i). Source data are provided as a Source Data file.

Subsequently, the possibility of spontaneous semiconductor shell deposition was examined by introducing the precursor solution without laser irradiation. The DFM images and peak of scattering spectra of Au NCs remained unchanged over 20 min (Fig. 1b and Supplementary Fig. 2), indicating that CdS could not spontaneously deposit onto Au NCs in the dark. It is worth noting that, under this condition, a minor change in the scattering spectrum was observed, which can be attributed to the adsorption of precursor ions on the Au surface, resulting in plasmonic damping⁴⁰. Strikingly, upon 532 nm laser irradiation, rapid color transitions from green to red were observed in the DFM images, accompanied by evident scattering intensity enhancements and spectral red shifts (Fig. 1c and Supplementary Fig. 3). These results unequivocally demonstrated that the coexistence of Au NC in solution and laser excitation is indispensable for triggering and sustaining the photodeposition of CdS. To confirm this argument, we performed the following control experiments. When heating the precursor solution (TAA + Cd²⁺) without Au NCs at 80 °C in the dark for 10 h, no large particle was observed in the DFM image (Supplementary Fig. 4). Meanwhile, when irradiating a similar sample with a laser for 20 min, the same result was observed. These results together confirm that rapid CdS deposition requires both Au NCs and optical excitation.

To discriminate thermal effect in the photodeposition process, Au NCs were encapsulated with a thin SiO₂ shell (Au NC@SiO₂) to screen hot-carrier transferring from Au (Supplementary Fig. 5). The SiO₂ layer can permit heat transfer but inhibit the transmission of hot-carriers. Under light irradiation, no CdS was deposited on Au NC@SiO₂, as evidenced by the absence of color change in the DFM images (Supplementary Fig. 5c). In contrast, the sample (Au NC@SiO₂) under external heating condition (80 ^oC, 10 h) was applied to simulate a purely thermal effect, which yielded Au NC@SiO₂@CdS core-shell structures with a distinct color change in the DFM image (Supplementary Fig. 5d, g). These results suggest that while CdS formation is thermodynamically feasible at elevated temperatures, such thermal conditions are not operative during our photodeposition experiment. This is because, according to theoretical calculations, the plasmonic thermal effect induced temperature increase (<0.1 K) around Au NCs can be neglected under the photodeposition condition due to the mild laser power (for calculation, see Methods)^25,39,41,42. The above calculated result is further confirmed by infrared thermal imaging outcomes, where no temperature change was observed during the reaction process (Supplementary Fig. 6). From these results, one can conclude that the formation of Au NC@CdS core-shell structures in the photodepositon experiment is mainly ascribed to plasmon-driven photoreaction, which also agrees well with earlier reported results^37–39. As further evidence to verify this argument, we performed deposition experiments under different laser wavelengths. With 532 nm excitation—matching the plasmon resonance of Au NCs—the most significant scattering spectral redshift were observed, indicating the fastest growth rate. In contrast, off-resonant wavelengths (473 and 633 nm) excitation only caused slight spectral changes, suggesting that only few CdS was deposited on the Au NCs surface (Supplementary Fig. 7). This wavelength dependence, with maximal efficiency at resonance, further confirms that the process is a plasmon-driven reaction.

To elucidate the role of precursor concentration in deposition kinetics, photodeposition experiments were conducted under fixed laser intensity (8.45 mW). At low [Cd²⁺] (0.04 mM), single-particle scattering spectra remained stable, retaining the initial green color and unchanged intensity (Supplementary Fig. 8). Increasing [Cd²⁺] to 0.4 mM still failed to induce effective deposition within 30 min (Supplementary Fig. 9). At 1.0 mM [Cd²⁺], partial particles exhibited gradual color transitions (green → yellow → red) accompanied by progressive scattering intensity enhancement (Supplementary Fig. 10). However, asynchronous color evolution among particles indicated nonuniform shell growth, with only a subset of irradiated particles undergoing deposition. This heterogeneity would compromise catalytic reproducibility. Raising [Cd²⁺] to 2.0 mM improved uniformity: Particles displayed synchronized color shifts from green to red within 15–30 min of irradiation (Fig. 1d and Supplementary Fig. 11), suggesting CdS shell formation. Nevertheless, a prolonged activation time was required. To optimize the deposition efficiency, the precursor concentration was increased to 4.0 mM. At this condition, green Au NCs began color transitions within ≈3 min, evolving to red within 20 min (Fig. 1e). Correspondingly, single-particle scattering peaks red-shifted from 552 to 627 nm, with intensity increasing (48.6 → 321.2 arb.u., Fig. 1f–h). DFM intensity kinetics further revealed concentration-dependent saturation behavior: Higher [Cd²⁺] accelerated growth rates and shortened saturation onset. Based on these findings, [Cd²⁺] = 4.0 mM was selected as the optimal concentration for subsequent catalytic studies.

To achieve spatially controlled multi-particle deposition, localized laser irradiation was applied to predefined regions of dispersed Au NCs. As shown in Fig. 1i, particles outside the irradiated zone (white dashed circle) retained their original green color (particles 1 and 2), while irradiated particles (particles 3 and 4) exhibited distinct purple-red shifts. This confirms that deposition occurs exclusively under concurrent laser exposure and precursor availability. Scattering color differences enabled clear discrimination between grown and non-grown particles. Furthermore, this approach permits real-time tracking of intensity and spectral evolution at single-particle resolution, circumventing ensemble-averaging effects and directly probing plasmonic resonance dynamics during shell growth.

Spectrum shift correlates with core-shell formation

To validate the formation of Au NC@CdS core-shell nanostructures via the in-situ photodeposition method, microscopic and spectrometric characterizations were systematically conducted. Scanning electron microscopy (SEM) images revealed a significant increase in particle size after photodeposition compared to pristine Au NCs (Fig. 2a, b). The original smooth surfaces of Au NCs transitioned to a rough morphology post-deposition, indicative of a relatively homogeneous CdS shell growth. This would be due to the fixation of the particle on a polymer layer (polystyrene sulfonic acid) in the flow channel, where the precursor solution can still diffuse into. For the systematic correlation between plasmonic resonance shifts and shell thickness in the core-shell structure, as a working hypothesis for the following analysis, Au NC@CdS with a consistent linear trend of spectral redshift versus CdS shell thickness, as that from the photodeposition method were synthesized in parallel using a solution-phase thermal reduction method and characterized by transmission electron microscopy (TEM), high-angle annular dark-field scanning TEM (HAADF-STEM) and spectroscopic analysis. TEM measurements confirmed the preservation of cubic morphology after CdS deposition, with an average edge length of 45.7 ± 2.2 nm, and shell thickness of 8.0 ± 1.2 nm (Fig. 2c–e). HAADF-STEM elemental mapping unambiguously identified spatially resolved distributions of Au, Cd, and S elements (Fig. 2f). The Au signal maintained the cubic geometry, while Cd and S signals exhibited uniform peripheral enrichment, confirming the conformal deposition of a CdS shell on the Au NCs core.

Fig. 2. Characterization of Au NC@CdS core-shell nanoparticles.

a, b SEM images of Au NCs and Au NC@CdS through the photodeposition method, respectively. c TEM image of Au NC@CdS through solution-based thermal growth. d, e The corresponding size distribution of Au NCs and the shell thickness of CdS in Au NC@CdS analyzed by TEM. f HAADF-STEM and elemental mapping images of Au NC@CdS. g TEM images of Au NC@CdS with different shell thickness. The corresponding UV-vis extinction spectra (h) dark-field scattering spectra i of the nanoparticles through solution-based thermal growth at different time intervals and the shell thickness-dependent peak wavelength variation curves. j Simulated absorption, scattering, and extinction spectra of these nanostructures and the thickness-dependent scattering wavelength variation curve from the simulated results. Data in (h, i) are presented as mean values ± SD from independent experiments (n = 10). Images are representative of three independent experiments with similar results in (a–c, f, g). Source data are provided as a Source Data file.

The correlation relationship between shell thickness and spectral response was then quantitatively investigated using TEM measurement and single-particle scattering spectroscopy based on the thermal growth of nanoparticles for the convenience of statistical analysis. As the CdS shell thickness increased from 0 to 15.8 nm (determined by TEM cross-sectional analysis), the extinction peak red-shifted from 535 to 618 nm (Fig. 2g, h). Correspondingly, single-particle scattering spectra displayed a progressive red-shift from an initial peak at 557 to 627 nm (Fig. 2i), accompanied by an enhancement in scattering intensity. In some cases (e.g., Au@Pt), formation of a shell structure on the nanoparticle surface may reduce the scattering signal due to plasmonic damping⁴³. Interestingly, in this case, the semiconducting CdS shell not only modifies the dielectric environment but also actively participates in charge transfer, leading to a non-monotonic behavior in scattering efficiency. The observed increase in scattering intensity with CdS shell growth would be primarily attributed to the following factors: (1) Increased particle size. (2) The deposition of a dielectric CdS shell modifies the local refractive index around Au NCs, leading to a red-shift and enhancement of the plasmon resonance. This is a well-known effect for metal-dielectric core-shell structures. (3) Charge separation vs. recombination: While non-radiative decay (hot-carrier generation) competes with radiative scattering, the formation of a Schottky barrier at the Au-CdS interface facilitates charge separation, prolonging the lifetime of hot-carriers. This does not necessarily quench the plasmon mode under certain shell thickness. It can lead to sustained plasmonic oscillations under continuous illumination, as evidenced by the enhanced scattering.

As a more confident predictive relationship between shell thickness and spectral shift in photoreduction, experimental single-particle scattering peak positions were compared with finite-difference time-domain (FDTD) simulations from the particles with CdS shell thicknesses of 0, 2, 8, 12, and 16 nm, respectively (Fig. 2j). The measured red-shift trajectory exhibited excellent agreement with theoretical predictions (R² = 0.98), validating that the spectral evolution directly reflects progressive CdS shell growth. These correlations enable real-time, non-invasive monitoring of shell thickness evolution through in-situ scattering spectroscopy, providing a tool for controlling semiconductor deposition at the single-particle level. It is worth noting that the FDTD simulation represents more than just a fitting of the thermal growth data. It is a fundamental physical model designed to calculate the optical response of a defined nanostructure. The agreement between the simulated results and the experimental spectra of thermally grown particles confirms that the model accurately captures the plasmonic behavior of the Au NC@CdS system (Fig. 2i). Meanwhile, from the single-particle scattering spectra, the thermally grown and photodeposited Au NC@CdS with similar shell thickness also exhibit comparable spectroscopic feature, Supplementary Fig. 12. As a result, the “spectral-thickness” relationship can serve as a generalized predictive tool for structure determination with analogous core-shell geometry.

Photodeposition vs. thermal reduction in solution

Above experimental results highlight evident advantages of in-situ photodeposition method over conventional solution-based thermal reduction for tuning Au NC@CdS core-shell nanostructures in single-particle studies. As demonstrated in Fig. 1e, f, the photodeposition process enables rapid activation of monodisperse Au NCs within the reaction channel, achieving full spectral red shifts (≈70 nm) in 20 min under ambient conditions. In contrast, thermal reduction requires prolonged reaction times (≈20 h) at elevated temperatures to attain comparable shell thicknesses, as evidenced by the gradual color evolution of particles from green (0 h) to yellow (6 h), orange-red (10–20 h), and a corresponding UV-vis extinction peak shift from 536 to 617 nm (Fig. 3). Three major limitations of thermal reduction for in-situ single-particle catalytic performance explorations were identified: (1) Time inefficiency: Photodeposition completes shell growth within minutes, whereas thermal reduction necessitates several hours due to the slow reaction kinetics and energy barriers inherent to heterogeneous nucleation, leading to inconsistent nucleation rates and poor temporal control. It is worth noting that increasing the reaction temperature in thermal reduction would accelerate the reaction rate but may also promote Ostwald ripening, aggregation, or structural degradation. (2) Lack of real-time controllability: Thermal reduction relies on batch sampling at discrete time points for offline characterization, preventing dynamic modulation of shell thickness. Conversely, photodeposition permits real-time tracking of single-particle scattering spectra and dark-field color transitions, enabling precise optical feedback to control shell growth via laser “on/off” switching. (3) Structural heterogeneity: Prolonged thermal reactions amplify stochastic variations in local precursor concentrations and interfacial energies, leading to nonuniform shell thicknesses across the nanoparticle.

Fig. 3. Solution-based thermal reduction method for the deposition of CdS shell on Au NCs.

a Photographs of the color changes from Au NCs to Au NC@CdS after different reaction times. b DFM images of Au NC@CdS with different shell thicknesses at 0–20 h. c The UV-vis extinction spectra of Au NC@CdS varied at different reaction times during the synthesis process. The spectra, colored from green to red to indicate the progression of reaction time (0–20 h) as shown in (b), demonstrate a progressive redshift. d The corresponding time-dependent extinction spectral variations of Au NC@CdS. Scale bar in the inset DFM image in (b) is 0.5 μm. The scale bar applies to all inset images. Source data are provided as a Source Data file.

Photodeposition circumvents these issues by synchronizing light-driven redox reactions at the individual particle surface, ensuring uniform growth kinetics. Statistical analysis of the scattering intensity and peak wavelength distributions at the single-particle level under different reaction times (with comparable UV-vis extinction wavelength) confirmed narrower deviations in photodeposited samples (RSD <8%) compared to thermally synthesized counterparts (RSD >25%), underscoring the enhanced homogeneity achievable with the light-driven method (Supplementary Fig. 13). The spatiotemporal control afforded by photodeposition minimizes interparticle variability, a critical factor for single-particle catalytic studies where structural consistency is paramount. Furthermore, the ambient operational conditions eliminate thermal degradation risks, preserving the integrity of both the Au NCs core and CdS shell. These advantages collectively establish in-situ photodeposition as a robust strategy for fabricating well-defined plasmonic-semiconductor heterostructures with tailored optoelectronic properties.

Semiconductor shell thickness effects on photocatalysis

To elucidate the role of semiconductor shell thickness in regulating photocatalytic performance, we employed the photocatalytic reaction of non-fluorescent Amplex Red (AR) to fluorescent Resorufin (Rf) as a model reaction, enabling exploration of photocatalytic activity at the single-particle level^33,35. The reaction was performed within the flow channel (20 mm × 2 mm × 0.2 mm (L × W × H)) containing H₂O₂ (100 mM) and AR (0.1–5.0 μM) (Fig. 4a). Under illumination conditions, H₂O₂ in the reaction solution can capture photogenerated hot-electrons from the photocatalyst and be reduced to generate hydroxyl radicals (•OH) (i.e., H₂O₂ + e⁻ → •OH + OH⁻). The highly oxidative •OH can subsequently attack and oxidize the AR molecule, ultimately resulting in the dehydrogenation and aromatization of its dihydropyrazine ring (Supplementary Fig. 14). Fluorescence bursts (“on” states) occurred when Rf molecules generated at the nanoparticle surface and excited by 532 nm laser irradiation. The interval between consecutive fluorescence bursts is recorded as “off” states (${\tau }_{\mathrm{off}}$), corresponding to the time required to complete one full catalytic turnover cycle^32,44. The catalytic turnover rates (〈${\tau }_{\mathrm{off}}$〉⁻¹) exhibited substrate concentration dependence, increasing sharply below 2.0 μM AR before saturating, consistent with Langmuir-Hinshelwood (L-H) surface reaction kinetics (Fig. 4b)^31,32. Subsequent experiments used saturated AR conditions (2.0 μM) to eliminate mass transport limitations.

Fig. 4. Single-particle catalytic activity of Au NC@CdS with different shell thicknesses.

a Scheme of the process for the single-particle catalytic imaging of the reaction activity of Au NCs before and after photodeposition CdS shell. b Amplex red concentration-dependent 〈${\tau }_{\text{off}}$〉⁻¹ variation from Au NCs. c Fluorescence intensity versus time trajectories of the catalytic processes from single Au NCs, Au NC@CdS_2.0, Au NC@CdS_8.0, Au NC@CdS_12.0, and Au NC@CdS_16.0, respectively. Data in c were normalized to the maximum value of each dataset. d Distributions of ${\tau }_{\text{off}}$ from the fluorescence turnover trajectories of these five nanostructures. e Distributions of ${\gamma }_{\text{app}}$ from these five nanostructures. The solid lines are Gaussian fitting. f Rate of turnovers as a function of time from these five nanostructures under saturated [AR] concentration. g Turnover rate 〈${\tau }_{\text{off}}$〉⁻¹ and h ${\gamma }_{\text{app}}$ of Au NCs, Au NC@CdS_2.0, Au NC@CdS_8.0, Au NC@CdS_12.0, and Au NC@CdS_16.0 under saturated [AR] concentration, respectively. i Comparison of the statistical heterogeneity index h for these nanostructures. Ⅰ: Au NC, Ⅱ: Au NC@CdS_2.0, Ⅲ: Au NC@CdS_8.0, Ⅳ: Au NC@CdS_12.0, Ⅴ: Au NC@CdS_16.0. Data in (b, g, h, i) are presented as mean values ± SD derived from independent experiments (n = 50, 50, and 10, respectively). Scale bars in the inset DFM images are 0.5 μm. Source data are provided as a Source Data file.

Catalytic trajectories of single Au NC@CdS with varying shell thicknesses (with scattering spectral peak at 568, 585, 603, and 627 nm, which are named as Au NC@CdS_2.0, Au NC@CdS_8.0, Au NC@CdS_12.0, and Au NC@CdS_16.0, respectively) revealed pronounced thickness-dependent activity modulation (Fig. 4c). The determined shell thicknesses for these four particles are 2, 8, 12, and 16 nm based on the shell thickness dependent scattering wavelength relationship, which also agrees well with the SEM measurements (Supplementary Fig. 15). The 〈${\tau }_{\mathrm{off}}$〉⁻¹ from individual nanoparticles initially increased from 0.90 to 4.52 s⁻¹ for bare Au NCs and Au NC@CdS_8.0, and followed by a 71% activity decline after further increasing the CdS shell (Fig. 4g). The non-monotonic catalytic behavior demonstrates that charge separation efficiency, which is governed by the interplay between plasmonic hot-electron injection and semiconductor-mediated hot-carrier recombination, reaches an optimum at intermediate shell thicknesses (8.0 ± 1.0 nm). This is also reflected in the results from single-particle scattering spectroscopic analysis, where the linewidth (FWHM) of the scattering spectra and catalytic activity exhibit a similar correlation (Supplementary Fig. 16). Initial linewidth broadening signifies the efficient utilization of photo-energy for the catalytic reactions. However, further broadening in thick shell sample reflects additional energy damping ways and thus lowers the catalytic activity, revealing a competition between carrier generation and utilization.

To statistically compare the catalytic efficiencies of these nanostructures, we analyzed the distribution of turnover time (${\tau }_{\mathrm{off}}$) from individual particles with different shell thickness. Under saturated AR concentration, the probability density functions of turnover time can be expressed as: $f\left({\tau }_{\mathrm{off}}\right)={\gamma }_{\mathrm{app}}\exp \left(-{\gamma }_{\mathrm{app}}{\tau }_{\mathrm{off}}\right)$, where ${\gamma }_{\mathrm{app}}$ represents the apparent rate constant from single nanoparticle^31–33,35. Fig. 4d illustrates the distributions of ${\tau }_{\mathrm{off}}$ from individual nanoparticles with different CdS shell thickness, which all follow an exponential decay. Through analyzing more than 50 particles from each group, the distributions of ${\gamma }_{\mathrm{app}}$ exhibited noticeable static heterogeneity across the particles with different shell thickness (Fig. 4e, h). The mean value of ${\gamma }_{\mathrm{app}}$ from Au NC@CdS_8.0 (1.02 ± 0.05 s⁻¹ per particle) is around 2.62 times higher than that from bare Au NCs (0.39 ± 0.02 s⁻¹ per particle). More interestingly, ${\gamma }_{\mathrm{app}}$ of these particles distributes broadly, indicating the heterogeneous nature of catalytic activity. The static catalytic heterogeneity of these particles can be quantified by the index of h (the ratio between the full-width-at-half-maximum and the center of the ${\gamma }_{\mathrm{app}}$ distribution)^31–33,35. As shown in Fig. 4i, Au NC@CdS_8.0 exhibits the highest h value (0.86 ± 0.10), manifesting the minor variations in CdS deposition critically affect charge transfer dynamics.

The superior activity of Au NC@CdS_8.0 correlates with two factors: (1) Optimal hot-electron utilization: Thin CdS shells enable efficient plasmon-generated electron transfer to adsorbed AR while minimizing “electron-hole” recombination losses. (2) Dynamic activity fluctuations: Time-resolved rate of turnover from single nanoparticles suggests transient “super-active” states in Au NC@CdS_8.0, likely arising from stochastic catalytic hot-spot formation under continuous plasmon excitation (Fig. 4f)^31,45. These findings establish semiconductor shell thickness as a critical design parameter for plasmonic photocatalysts, balancing charge separation efficiency against recombination losses. The single-particle approach decouples intrinsic catalytic properties from ensemble-averaging effects, providing unprecedented insights into structure-activity relationships at the nanoscale.

Mechanism of plasmon-semiconductor synergy in catalysis

To elucidate the origin of enhanced catalytic activity in Au NC@CdS core-shell nanostructures, we systematically investigated the interplay between plasmonic field modulation and hot-carrier dynamics through combined electromagnetic simulations and charge carrier generation analysis. FDTD simulations revealed the dependence of electric field distribution on shell thickness (with the same scattering peak from Au NCs to Au NC@CdS_16.0, Fig. 5a). Pure Au NCs exhibited strong near-field enhancement (|E/E₀| = 2.87) at sharp edges and corners. Upon CdS shell deposition, the plasmon resonance intensity at the original wavelength displayed an initial increase and followed by subsequent reduction. Specifically, the electric field enhancement values of 2.99, 3.32, 2.96, and 2.55 were obtained for CdS shell thicknesses of 2, 8, 12, and 16 nm (from Au NC@CdS_2.0 to Au NC@CdS_16.0, respectively), with maximum field enhancement observed at intermediate shell thickness. In noble metal systems, the intense localized electromagnetic field induced by plasmon resonance facilitates electron excitation from ground state to high-energy states, generating nonequilibrium hot-electrons. A direct positive correlation exists between localized field enhancement and hot-electron generation efficiency.

Fig. 5. Finite-element and DFT simulations of Au NC@CdS.

a FDTD simulations of electric field distributions of Au NCs, Au NC@CdS_2.0, Au NC@CdS_8.0, Au NC@CdS_12.0, and Au NC@CdS_16.0 with light excitation at 532 nm and the corresponding shell thickness-dependent variation curve of the peak electric-field enhancement from Au NC@CdS. b Shell thickness-dependent hot-electron generation rate distributions and local generation rates along the diagonal of Au NCs, Au NC@CdS_2.0, Au NC@CdS_8.0, Au NC@CdS_12.0, and Au NC@CdS_16.0, respectively. c Projected density of states of d-orbital (d-PDOS) for Au NCs (grey curve) and Au NC@CdS (black curve). d Charge density difference analysis of Au NC@CdS. e Schematic diagram of the photo-driven hot-carrier generation and hot-electron transfer in the catalytic process from Au NCs and Au NC@CdS. Source data are provided as a Source Data file.

Finite-element simulations using COMSOL of hot-carrier distribution across various Au NC@CdS configurations demonstrated that the hot-electron generation rate profile exhibited concordance with FDTD-derived electromagnetic field enhancement patterns, both showing higher values for Au NC@CdS (Fig. 5b). The superior photocatalytic activity of Au NC@CdS_8.0 primarily originates from its exceptional hot-electron generation capability. Optimal CdS shell thickness simultaneously enhances hot-electron-hole separation efficiency and improves utilization effectiveness, as evidenced by the diagonal distribution profile of hot-electron generation rates across nanostructures. The observed trends in field enhancement and hot-carrier generation rates precisely mirrored the photocatalytic efficiency hierarchy. While hot-electron generation in plasmonic catalysis remains central, it is worth noting that near-field-assisted excitation may also contribute to lower the activation energy barrier, and synergistically optimize the catalytic process, consistent with recent “antenna–reactor” concepts in plasmonic catalysis³. The strong localized electromagnetic fields from Au NCs not only promote hot-electron generation but also enhance the local electric field, which can polarize and pre-activate reactant molecules near the nanoparticle surface. These results together show that the improved photocatalytic activity is not merely governed by the electromagnetic effect (as evidenced from the 2.62-fold catalytic enhancement of Au NC@CdS_8.0 versus a 1.16-fold field increase (Figs. 4h and 5a)). The charge separation and hot-electron utilization at an intermediate CdS shell thickness represent a balance between efficient hot-electron generation and the subsequent delivery to catalytic sites. Although earlier explorations reported that thermal effects may contribute to plasmonic catalysis, our moderate optical excitation conditions and the flow channel design minimize bulk heating, consistent with the results from previous single-molecule catalytic studies^45,46. The linear relationship in laser power-dependent catalytic efficiency directly confirms this argument (Supplementary Fig. 17).

Furthermore, density functional theory (DFT) calculations were employed to investigate CdS shell-induced electronic structure modifications in Au NC@CdS. Band structure and projected density of states (PDOS) analyses revealed d-band center positions of −3.96 eV for Au NCs and −3.16 eV for Au after formation of Au NC@CdS (Fig. 5c). This shift indicates the changes in the electronic structure of Au and the emergence of hybridized energy levels after Schottky barrier formation, which facilitates electron excitation to conduction bands and consequently enhances photoconversion efficiency. Work function calculations demonstrated that Au NCs (≈5.04 eV) possess a higher work function than CdS (≈4.74 eV) (Supplementary Fig. 18). Under the plasmon-induced nonequilibrium condition, hot-electrons in Au can transiently overcome the Schottky barrier at the Au-CdS interface and be injected into the CdS conduction band^12,15. Charge density difference analysis revealed electron accumulation on CdS and hole accumulation on Au NCs at the interface, creating favorable conditions for photogenerated charge separation (Fig. 5d, e).

Photocurrent response measurements under xenon lamp irradiation demonstrated maximum current density for Au NC@CdS with optimal shell thickness (Fig. 6a, ≈8 nm), indicating superior charge separation efficiency. Open-circuit voltage decay (OCVD) analysis revealed the longest electron lifetime (${\tau }_n$) from the particles with a shell thickness of 8 nm, calculated through the relationship: ${\tau }_n=-\frac{{\mathrm{K}}_{\mathrm{B}}T}{\mathrm{e}}{\left(\frac{\mathrm{d}{V}_{\mathrm{oc}}}{\mathrm{d}t}\right)}^{-1}$, where ${\mathrm{K}}_{\mathrm{B}}$ represents Boltzmann’s constant, T represents absolute temperature, and e represents elementary charge (Fig. 6b)⁴⁷. The observed electron lifetime enhancement directly correlates with photocatalytic performance, confirming that optimized CdS shell thickness maximizes plasmon-mediated charge carrier separation efficiency (Fig. 6c). These comprehensive experimental and computational analyses establish a robust structure-activity relationship in plasmonic photocatalysts, providing insights into hot-carrier manipulation through rational nanostructure design: (1) Formation of a Schottky barrier at the metal-semiconductor interface to suppress back flow of hot-carriers, thereby directly mitigating charge carrier recombination at the Au-CdS interface and, by extension, reducing hot-electron relaxation in Au (via electron-phonon scattering) through enhanced hot-carrier generation by non-radiative decay. (2) Keeping the semiconductor shell thickness within the relevant hot-electron mean free paths to minimize hot-carrier recombination during transport. (3) Geometric field enhancement regions (e.g., cube corners) should be maintained to sustain plasmonic excitation for efficient hot-carrier generation. This work provides a unified framework for rational optimization of hybrid photocatalysts, bridging nanoscale charge dynamics with macroscopic catalytic performance.

Fig. 6. Photoelectrochemical properties of Au NC@CdS.

a Photocurrent measurements from Au NCs, and Au NC@CdS with shell thickness of 2, 8, 12, and 16 nm, respectively, under an applied bias potential of 0 V (vs. SCE). b OCVD analysis, and c average electron lifetime (${\tau }_n$) of Au NCs and Au NC@CdS with shell thickness of 2, 8, 12, and 16 nm under xenon lamp irradiation in Na₂SO₄ aqueous solution, respectively. All measurements were conducted using an AM 1.5 G filter (300 W Xe lamp). “Off” and “On” represent without or with the Xe lamp irradiation, respectively. Source data are provided as a Source Data file.

Discussion

In this work, we establish an approach for engineering plasmonic photocatalysts through precise control of semiconductor shell thickness, integrating in-situ photodeposition with single-molecule fluorescence tracking to resolve structure-activity relationships at the nanoscale. The key innovation lies in demonstrating a non-monotonic dependence of photocatalytic efficiency on CdS shell thickness, where Au NC@CdS_8.0 achieved peak catalytic activity (2.62-fold enhancement over pristine Au NCs). Excessive shell growth suppressed activity by 59%, revealing interfacial charge separation rather than semiconductor presence as the governing factor. Single-particle analysis uncovered significant static heterogeneity in catalytic rates (${\gamma }_{\mathrm{app}}$), with Au NCs@CdS_8.0 exhibiting both the highest catalytic activity and maximum heterogeneity index, highlighting amplified sensitivity to atomic-scale structural variations at optimal thickness. Furthermore, dynamic fluctuations in catalytic trajectories, linked to transient hot-electron dynamics and electromagnetic field effects, were found to synergistically enhance turnover rates. The methodology—combining tunable photodeposition with single-particle spectroscopy—provides a framework for designing heterostructured photocatalysts with tailored charge transfer properties, offering insights for energy conversion and heterogeneous catalysis research.

Methods

Materials

Gold chloride trihydrate (HAuCl₄·3H₂O, ≥99%), cetyltrimethylammonium bromide (CTAB, LOT#A2310376, 99%), cetyltrimethylammonium chloride (CTAC, LOT#F2201132, 99%), sodium borohydride (NaBH₄, 99%), and sodium bromide (NaBr, 99%) were purchased from Aladdin (Shanghai, China). L-ascorbic acid (AA, 98%), cadmium chloride hemihydrate (CdCl₂·2.5H₂O, 99%), thioacetamide (TAA, 99%), hexamethylenetetramine (HMT, 98%), silver nitrate (AgNO₃, 99%), and hydrogen peroxide (H₂O₂ 30 wt.% in water) were purchased from Sinopharm Chemical Reagent Co. Ltd. (Shanghai, China). Amplex Red (99%) was obtained from Invitrogen Ltd. (Paisley, UK). All chemicals were used without further purification. Deionized water was used as the solvent in all experiments.

Synthesis of Au NCs

Au NCs were synthesized via a seed-mediated method following a protocol in ref. ⁴⁸. Initially, CTAB-capped small gold nanoparticles were prepared by adding 50 μL of 10 mM HAuCl₄ to 2 mL of 100 mM CTAB. Subsequently, 160 μL of ice-cold 10 mM NaBH₄ was rapidly injected into the mixture. The solution was manually mixed thoroughly and then incubated undisturbed for 1 h in a 35 °C water bath.

Next, CTAB-capped seeds were grown from the seeds obtained in the previous step. A mixture consisting of 2 mL of 100 mM CTAC, 1.5 mL of 100 mM AA, and 30 μL of the aforementioned cluster solution was prepared and stirred. Then, 2 mL of 0.5 mM HAuCl₄ was added to the mixture, and the resulting solution was stirred for 30 min in a 35 °C water bath at a speed of 500 rpm. As the reaction proceeded, the solution color changed from colorless to pink. After that, the solution was left undisturbed for 2 h in a 35 °C water bath to obtain 10 nm Au NPs, which served as the second-generation seeds.

For the synthesis of Au NCs, 3.4 mL of H₂O, 1.5 mL of 200 mM CTAC, 200 μL of 20 mM NaBr, 40 μL of the seed solution, 585 μL of 2.5 mM AA, and 225 μL of 5.0 mM HAuCl₄ were combined. The mixture was gently shaken and then incubated undisturbed in a 35 °C water bath for 1 h. The final product was centrifuged at 3400 × g for 10 min and redispersed in 1 mM CTAC.

In-situ photodeposition of CdS shell on Au NCs

Au NCs were fixed within a flow channel (20 mm × 2 mm × 0.2 mm in L × W × H) covered with polystyrene sulfonic acid (M_w ≈ 70,000 g mol⁻¹). Unadsorbed particles were rinsed away with deionized water. Then, the precursor solution (TAA and CdCl₂) was injected into the channel for the subsequent reaction. The numerical aperture (NA) of the oil immersion lens was adjusted to its maximum value, and a 532 nm laser beam (with a power density of 8.45 mW) was activated for fixed-point irradiation of varying durations. The area of the focal point is around 1.8 × 10⁻² mm². After each irradiation, the NA was readjusted to its minimum value, and dark-field images and single-particle spectra at the corresponding positions were collected separately.

Synthesis of Au NC@CdS by solution-based thermal reduction

Au NC@CdS were synthesized using a seed-mediated approach as described in ref. ²⁴. About 1 mL of 100 mM CTAB, 1 mL of 100 mM AA, and 1 mL of 100 mM HMT were added to 5 mL of Au NCs solution. Then, different amounts, specifically 300 µL of 1 mM TAA and 150 µL of 1 mM CdCl₂, were injected into the growth solution. The mixture was maintained in an 80 °C oil bath for 1 h. The final product was centrifuged at 1800 × g for 10 min, washed, and dispersed in water for subsequent usage.

Synthesis of Au NC@SiO₂ and Au NC@SiO₂@CdS

Au NC@SiO₂ were synthesized using an approach as described in ref. ^35,49. The as-prepared Au NCs (3 mL) were centrifuged at 1800×g for 10 min. The precipitate was redispersed into water (3 mL). Silica coating was carried out by the sequential addition of CTAB solution (100 mM, 160 μL), TEOS (1.75 wt%, 60 µL in ethanol) and NaOH (0.1 M, 30 µL) under gentle shaking. The reaction was kept at room temperature under gentle shaking for 12 h, during which silica was deposited on the facets of Au NCs. The resultant sample was centrifuged at 700 × g for 10 min. The precipitate was redispersed into water (300 μL) for further usage. Au NC@SiO₂@CdS were synthesized by directly replacing Au NCs in the procedure as-prepared for Au NC@SiO₂, while keeping all other experimental conditions unchanged.

Characterization

The morphologies of the nanoparticles were characterized using a transmission electron microscope (TEM, JEOL, JEM-2100, Japan) and a field-emission scanning electron microscope (SEM, JEOL, JSM-7610FPlus, Japan). UV-vis extinction spectra were acquired with a spectrophotometer (UV-4080, Japan). Dark-field optical microscopic (DFM) images were captured on an upright optical microscope (Winner-Scat, Hunan OPP Tech., China) equipped with a 40× objective.

Single-molecule catalysis experiments

First, the excessive reaction solution in the channel was rinsed with deionized water. Then, reactants with different concentrations (a mixture of 0.1–5.0 μM AR and 100 mM H₂O₂) were injected into the reaction channel at a flow rate of 5 μL min⁻¹. A 532 nm laser beam was turned on to excite the fluorescence of the products. A 100× total internal reflection fluorescence objective lens (NA = 1.49) was used to collect the fluorescence signals of resorufin molecules, which were detected by an electron-multiplying charge-coupled device (EMCCD). For each experiment, more than 2000 frames were captured at a rate of 20 frames per second to track the catalytic trajectory. All captured data were processed and analyzed using MATLAB (R2021b) and ImageJ (version 1.47 v).

Photoelectrical response measurements

All the photoelectrical response measurements were performed using a three-electrode system. Electrode preparation: A suspension of Au NCs or Au NC@CdS catalyst (15 μL, OD = 5) was drop-cast onto an ITO electrode (0.5 × 1.5 cm²) within an effective area of 0.5 × 0.5 cm², dried in an oven at 50 °C and subsequently cooled to room temperature. Thereafter, 1.5 μL of Nafion perfluorinated resin solution was drop-cast onto the sample to form a thin film as work electrode. A Platinum tablet (1 × 1 cm²) and a saturated calomel electrode (SCE, filled with 3 M KCl) were used as the counter and reference electrodes, respectively. Then, 45 mL of 0.5 M Na₂SO₄ was used as the electrolyte, and all photoelectrical response tests were conducted without stirring.

Electrochemical response measurements were conducted on an electrochemical workstation (CHI 660E, CH Instruments, China), using an Xe lamp (300 W) (HSX-F300, Beijing Niubite Technology Co., Ltd., China) equipped with an AM 1.5G filter as the excitation source. Transient photocurrent responses were recorded at a bias potential of 0 V vs. SCE with on/off light cycles of 2.4 s. Transient OCVD measurements were conducted under illumination from a 300 W Xe lamp for 400 s (dark: 0–100 s; light: 100–300 s; dark: 300–400 s).

FDTD simulation

The electric field distribution of nanoparticles was simulated using FDTD Solution 8.0 software. A total-field scattered-field light source in the range of 500–900 nm was employed for the simulation. The geometric parameters of Au NCs and Au NC@CdS were consistent with those observed in the TEM images. For Au NCs, the edge length was set to 45.7 nm, and the radius of curvature of the corner was set to 5.0 nm. For Au NC@CdS, the CdS shell was set as a hollow-shell structure. Its side length was greater than that of the gold nanoparticles, and the thicknesses were set to 2, 8, 12, and 16 nm, respectively. The dielectric constant of Au was obtained by fitting the measurement data of Johnson and Christy. The real part of the refractive index of CdS was set to 2.5, and the imaginary part was set to 0. The electric field strength distribution and enhancement factor of these nanoparticles were calculated with a grid size of 1 nm × 1 nm × 1 nm, and the refractive index of the surrounding medium was set to 1.33.

COMSOL simulation

The COMSOL Multiphysics software (COMSOL Inc., Burlington, MA) was used to calculate the hot-electron generation rate and distribution of Au NCs and Au NC@CdS. In the theoretical model, the wavelength of the incident light was set to 532 nm, and the amplitude of the electric field was 1 V m⁻¹. The electric field propagated downward along z-axis, with the polarization direction along x-axis. The models of the two nanostructures were consistent with those used in the FDTD simulation. The outermost layer was a perfectly matched layer with a thickness set to half of the wavelength. In terms of material parameters, the refractive index of the ambient medium (water) was set to 1.33, the dielectric function of gold was adopted from Johnson and Christy’s data, and the refractive index of CdS was defined as 2.5. The mesh was generated according to the geometric structure of the model, ensuring sufficient refinement in critical regions. After meshing, a transient solver was employed with relative tolerance and a maximum of 25 iterations.

The Wave Optics Module in COMSOL Multiphysics was employed, along with the frequency domain physics interface. Key parameters were defined as follows^50,51:

Input power I₀ was expressed as:

$$I_0=\frac{{E_0}^2}{2\times Z_0}$$ 1

$E_0$ is the incident electric field amplitude, $Z_0$ is the free space impedance. Scattering cross-section ${\sigma }_{\mathrm{sca}}$ was expressed as:

$${\sigma }_{\mathrm{sca}}=\frac{\mathrm{intop}\left(\mathrm{nre}{\mathrm{l}}_{\mathrm{Poav}}\right)}{I_0}$$ 2

The term $\mathrm{intop}\left(\mathrm{nre}{\mathrm{l}}_{\mathrm{Poav}}\right)$ was the component of the Poynting vector in the outward normal direction of the integration boundary. Absorption cross-section ${\sigma }_{\mathrm{abs}}$ was expressed as:

$${\sigma }_{\mathrm{abs}}=\frac{\mathrm{intop}\left(\mathrm{ewfd.Qh}\right)}{I_0}$$ 3

The term $\mathrm{intop}\left(\mathrm{ewfd.Qh}\right)$ calculates the total resistive heating power inside a defined volume by integrating the power loss density. Normal electric field intensity ${\mathrm{E}}_{\mathrm{n}}$ was defined as:

$$E_{\mathrm{n}}=n_x\times \mathrm{ewfd.}{E}_x+n_y\times \mathrm{ewfd.}{E}_y+n_z\times \mathrm{ewfd.}{E}_z$$ 4

In this context, $n_x$, $n_y$, and $n_z$ are the x, y, and z components of the unit normal vector, defining the outward direction of the surface. $\mathrm{ewfd.}{E}_x$, $\mathrm{ewfd.}{E}_y$, and $\mathrm{ewfd.}{E}_z$ are the x, y, and z components of the total electric field vector at the point of interest. The local hot-electron generation rate of the nanoparticles was calculated according to Eq. (5)³³:

$$\mathrm{Rate}=\frac{2}{{\mathrm{\pi }}^2}\times \frac{{\mathrm{e}}^2\times {E_{\mathrm{f}}}^2}{\hslash }\times {\left(\frac{\mathrm{freq}}{\mathrm{h}}\right)}^3\times \mathrm{abs}\left({E_{\mathrm{n}}}^2\right)$$ 5

In this context, $E_{\mathrm{f}}$ is defined as the Fermi energy level of the metal. The term ${\left(\frac{\mathrm{freq}}{\mathrm{h}}\right)}^3$ represents the strong cubic dependence on the incident photon frequency, a signature of the plasmonic density of states and its decay phase space. $E_{\mathrm{n}}$ was a normal electric field intensity. The pre-factors, including $\mathrm{e}$ (the elementary charge) and $\hslash$ (the reduced Planck’s constant), constituted the fundamental physical constants that set the overall scale of the generation rate.

DFT calculation

DFT theoretical calculations of the electronic structures of Au NCs and Au NC@CdS materials were performed using VASP in MedeA software⁵². First, models of Au (111) surface and Au (111)-CdS composite surface were constructed. The number of layers of the Au (111) surface was set to 4, and the number of layers of the CdS surface was set to 1–4 to simulate different CdS shell thicknesses. To avoid the influence of periodic boundary conditions, a vacuum layer of 15 Å was set in the z-axis direction. During the geometric optimization process, ultrasoft pseudopotentials were used, and the cutoff energy of the plane-wave basis set was set to 400 eV. To accurately describe van der Waals interactions, empirical dispersion correction in the Grimme format was introduced. The electronic exchange and correlation effects were described by the generalized gradient approximation (GGA) exchange correlation functional. In addition, the differential charge density (DCD) and the projected density of states (PDOS) were calculated using vaspkit, and the electronic structures, charge transfers, and surface properties of the Au (111) and Au (111)-CdS surfaces were visually analyzed using visualization for electronic and structural analysis (VESTA).

Photo-induced temperature estimation from Au NCs

Under continuous-wave excitation, the steady-state temperature increase ΔT at the surface of a plasmonic nanoparticle is generated by LSPR absorption. The temperature rise was estimated following the guidelines by Baffou et al. and related reports^25,39,41,42: $△T=\frac{{\sigma }_{\mathrm{abs}}\times I}{4\times \mathrm{\pi }\times \kappa \times R}$, where ${\sigma }_{\mathrm{abs}}$ is the absorption cross-section, $I$ is the irradiance (optical power per unit area), $\kappa$ is the thermal conductivity of the surrounding medium, and $R$ is the Laplacian radius of the nanoparticle. The Laplacian radius of an individual Au NC is around 28.4 nm. The absorption cross-section of an individual Au NC obtained from FDTD simulations at the specified excitation wavelength of 532 nm is around 8.5 × 10³ nm². The laser power is 8.45 mW with a focal point size around 1.8 × 10⁻² mm². The average conductivity of κ is around 0.99 W m⁻¹ K⁻¹ (κ(glass) = 1.38 W m⁻¹  K⁻¹ and κ(H₂O) = 0.6 W m⁻¹ K⁻¹). The estimated steady-state temperature increase from a single Au NC is only about 1.13 × 10⁻² K.

The infrared thermal imaging experiment was performed on the microscope with two imaging channels. One of the channel was equipped with a CCD camera for single-particle imaging, and the other channel was equipped with an infrared thermal imaging camera (coupled with a 532 nm notch filter). To measure the temperature change, we first captured an image with particles fixed on the glass slide surface (Supplementary Fig. 6a). After 20 min of laser illumination (the same condition for the photodeposition experiment, i.e., 532 nm laser with excitation power of 8.45 mW), the temperature image was simultaneously captured from the same area with the infrared thermal imaging camera.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

X.F., C.Z., and M.C. carried out the sample synthesis. L.X. supervised and conceived the project. X.F. and C.Z. conceived the project and conducted the single-particle imaging and catalysis experiments. X.F., C.Z., Z.Y., Q.Z., and L.W. analyzed the data. X.F. and L.X. wrote the manuscript. C.Z., M.C. and L.W. checked the manuscript. L.W., Z.Y., and L.X. provided reagents and performed the photodeposition and photo-induced temperature estimation experiment. All authors contributed to the data analysis and discussions.

Peer review

Peer review information

Nature Communications thanks Julian Gargiulo, Masaki Saruyama, and the other, anonymous, reviewer for their contribution to the peer review of this work. A peer review file is available.

Data availability

The data that support the findings of this study are available from the corresponding authors upon request. Unprocessed data were provided as Supplementary Data. 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/s41467-026-71517-6.