Introduction

Plasmonic photocatalysis represents an emerging sub-field of heterogeneous catalysis that can convert solar energy, one of the most abundant sustainable energy forms on earth, into chemical energy facilitated by nanoplasmonic catalysts1,2. Signified by the localized surface plasmon resonance (LSPR) effect, plasmonic nanostructures typically possess versatile optical response and photophysical properties that facilitate their distinct functionalities in catalytic pathways3,4,5,6. As a result, plasmonic catalysts, such as Au, Ag, or Cu nanoparticles, intricately designed non-noble plasmonic compounds, and plasmonic superstructures have been increasingly exploited in the catalytic conversion of CO27,8, H2O9,10, urea11, methanol12,13 among others14,15.

Attaining in-depth understanding on the distinct plasmonic photocatalytic mechanisms is crucial for tuning the reaction conditions, pathways, and products; yet it is highly demanding due to their complex nature. Driven by the non-radiative relaxation of the localized surface plasmons, a number of distinct mechanisms could co-exist and, thus, simultaneously contribute to the catalytic processes16,17. For instance, the rapid dephasing of the plasmons through Landau damping and electron-electron scattering is able to induce the thermalization of ballistic “hot” charge carriers that directly participate in the catalytic pathways6,18,19; the intense resonant electromagnetic near- field (E-field) produced by the coherent oscillation of free electrons at nanoplasmonic surfaces could also sustain resonant energy transfer mechanisms11,20,21. Moreover, through electron-phonon scattering of the hot carriers, plasmonic photothermal effect could also be initiated for highly localized heating22,23. Intensive efforts have thus been made towards identifying the contributions of each distinct mechanism in plasmonic photocatalysis, particularly those of thermal versus non-thermal effects1,24,25,26. However, it remains highly challenging to distinguish and quantify thermal vs non-thermal contributions due to the inaccurate measurement of catalyst surface temperature, or amongst various non-thermal pathways because of the intricate synergies that occur between plasmonic charge and energy transfer mechanisms.

In this work, we have developed a generalized “selective shielding” strategy for identification and quantification of distinct plasmonic photocatalytic mechanisms, particularly amongst various non-thermal reaction pathways, employing a prototypical Au-[Fe(bpy)3]2+ plasmonic nanocatalyst and water splitting as test reaction. Through modifying the connection scheme from a conductive to an insulating layer, and adjusting the resonance conditions between the plasmonic antenna and the reactor component, the charge transport and energy transfer mechanisms were selectively shielded. In addition, thermal contributions were calibrated using in-situ Raman nano-thermometry. It was found that plasmonic charge carrier-mediated photoreduction dominates the photocurrent (~57%) in a reducing, hydrogen evolution environment; whereas resonant plasmonic energy transfer dominates (~54%) in an oxidative, oxygen evolution environment. Our work provides in-depth understandings on the role of plasmon resonance in photocatalysis as well as important design principles for plasmonic photocatalysts towards distinct types of chemical reactions.

Results

Design and characterization of the Au-[Fe(bpy)3]2+ Plasmonic Nanocatalyst

The primordial plasmonic photocatalyst configuration typically involves the use of plasmonic nanoparticles (NPs) as nanoantenna and active catalyst components as the reactors27,28,29. A prototypical plasmonic antenna-reactor (A-R) nanocatalyst was first constructed using Au NP cores (Figs. S1a, b) as nanoantenna and tris(2,2′-bipyridine)iron(II) ([Fe(bpy)3]2+) (Fig. S1c) as the catalytic component, whose intramolecular metal-to-ligand charge transfer (MLCT) absorption feature at 520–550 nm (Fig. S1d) overlaps spectrally with the LSPR frequency of Au NPs to guarantee efficient resonant enhancement4,30. The intermediate size of the Au nanoantenna used in the nanocatalyst (59.56 ± 7.07 nm) effectively balances the radiative vs dissipative plasmonic energy losses, and potentially enables both plasmonic charge and energy transfer-mediated non-thermal photocatalytic mechanisms11,31. To enable the differentiation of distinct photophysical mechanisms, two different connection schemes between the antenna and reactor components were utilized, including (a) via a thiolated polyethylene glycol (PEG) linker with carboxyl end groups (HS-PEG-COOH, Mw = 456) (denoted as Au-[Fe(bpy)3]2+), which has been reported to facilitate the transport of plasmonic charge carriers11; and (b) via a silica nanoshell of comparable thickness (denoted as Au@SiO2-[Fe(bpy)3]2+), which insulates the charge transfer processes and enables only the resonant field-enhancement16,30 (Fig. 1a). In the Au-[Fe(bpy)3]2+ nanocatalyst, the thiol moieties at the thiolated end of the linker covalently binds to Au32,33, while the ionized -COO- at the other end connects with [Fe(bpy)3]2+ molecules primarily through electrostatic interaction34, as indicated by our Zeta potential measurements (Table S1). In addition, moderate van der Waals (VDW) interactions between the non-polar bipyridine ligands and the alkyl chain in the linker is also present11 (Fig. S2), contributing to the pinning of the [Fe(bpy)3]2+ catalyst at the tip of the linker molecules (Fig. S3a). Similarly, the silane moieties on the Au@SiO2 surface as well as the hydrophobic silica surface sustains electrostatic as well as moderate VDW interactions that collectively integrates the [Fe(bpy)3]2+ molecules35,36 (Fig. S3b).

Fig. 1: Design of the Au-[Fe(bpy)3]2+ plasmonic nanocatalyst.
figure 1

a Synthesis of Au-[Fe(bpy)3]2+ with different connection schemes. b TEM images and (c) STEM image and EDS elemental mappings of Au-[Fe(bpy)3]2+ nanocomposite. d Raman spectra of Au-[Fe(bpy)3]2+ and controls under 633 nm incident laser. Insets: Raman mapping at 1448 (red) and 1492 (green) cm−1. e Au 4f and Fe 2p XPS spectra and fitting. f UV-vis absorbance spectra of Au-[Fe(bpy)3]2+ colloid and controls.

Full size image

The Au-[Fe(bpy)3]2+ plasmonic photocatalyst was next used for structural and photophysical characterization. Transmission electron microscopy (TEM) was first used to evidence its successful preparation. A distinct layer of the thiolated PEG-COOH linker with a thickness of ~2.5 nm could be observed around the Au nanoantenna (Fig. 1b), which is consistent with the contour length of HS-PEG-COOH at corresponding molecular weights37. In proof of successful [Fe(bpy)3]2+ binding, Au and Fe elements were mapped across an individual nanoparticle by Energy-Dispersive X-Ray Spectroscopy (EDS) in scanning TEM (STEM) mode (Fig. 1c). The distribution of the two elements shows them to co-localize within the boundary of the particle, but with Fe preferentially residing at the perimeter. Moreover, spectroscopic characterization provides further evidence on the successful formation of Au-[Fe(bpy)3]2+. The Raman spectrum of Au-[Fe(bpy)3]2+ displays signature peaks both of [Fe(bpy)3](PF6)2 (1322, 1492, 1608 cm−1 etc.)38 and Au NPs (1448 cm−1 corresponding to the surface ligands39) (Fig. 1d). Further Raman mapping at 1448 cm−1 (Au) and 1492 cm−1 ([Fe(bpy)3]2+) demonstrates the overlapping distribution of the two components (Fig. 1d, insets). In our X-ray photoelectron spectroscopy (XPS) measurements, the Au 4 f characteristic peaks of the Au-[Fe(bpy)3]2+ nanocatalysts and those of Au nanoparticles show significant differences (Figs. 1e and Fig. S4). In addition to the Au0 features (87.0 and 83.4 eV), Au-[Fe(bpy)3]2+ exhibits a distinct set of peaks at high energies (88.8 and 84.8 eV) attributed to the Au1+4f7/2 and 4f5/2 peaks derived from the covalent bonding of the thiolated Au atoms on the surface of the nanoantenna, with the extent of the shifts highly consistent with previous reports40,41. In summary, all results confirm the successful formation of the Au-[Fe(bpy)3]2+ nanocomposite.

As a result of the resonance between Au LSPR and [Fe(bpy)3]2+ intramolecular MLCT transition, an evident increase in the absorbance of the Au-[Fe(bpy)3]2+ nanocatalyst is observed in the range of 450–550 nm, which surpasses that of the Au colloids and [Fe(bpy)3]2+ combined, as well as that of a physical mixture of the two components at identical concentrations (Fig. 1f). Notably, the absorbance of Au-[Fe(bpy)3]2+ almost doubled compared to [Fe(bpy)3]2+ at the absorbance maximum of 524 nm, after the absorbance of Au colloids was subtracted. This phenomenon evidences the resonant, synergistic enhancement in absorbance through plasmon-mediated energy transfer11,42. Based on the light extinction behavior, we have also optimized the amount of [Fe(bpy)3]2+ in the hybrid nanocomposite. With an increasing amount of [Fe(bpy)3]2+ (5.65 mM), the absorbance of corresponding Au-[Fe(bpy)3]2+ also exhibited a gradual increase until an input amount of 150 µl, when the increase showed an evident halt (Fig. S5a). This suggests that the binding capability of the linker on the Au surface has reached saturation, and is consistent with dynamic light scattering (DLS) results (Fig. S5b), where a higher input of [Fe(bpy)3]2+ than 150 µl led to evident particle aggregation due to neutralization of the negative charges from [Fe(bpy)3]2+ binding, lowering the electrostatic repulsion between monodispersed nanocomposites. Thus, an optimized Fe concentration of 12.56 ± 2.10 ppm in Au-[Fe(bpy)3]2+ colloid was obtained was chosen for the following characterizations (Table S2).

Identification of the Resonant Plasmonic Energy and Charge Transfer Mechanisms

Plasmonic near field-mediated resonant energy transfer and hot carrier transport are two plausible mechanisms of plasmon-enhancement associated with the Au-[Fe(bpy)3]2+ nanocomposites30,43,44. To identify the resonant energy transfers, Finite-difference time-domain (FDTD) simulations of the E-field around a single Au NP were first performed. Upon resonant illumination, noticeable enhancement of the E-field occurs in a region within 10 nm around the particle (Fig. 2a, inset). To provide experimental evidence on the effect of the resonant electromagnetic near-field, we compared the absorbance spectrum (ABS) and the single-particle dark-field scattering (SCA) spectrum of the Au-[Fe(bpy)3]2+ nanocomposite as well as a Au nanoparticle only control (Figs. 2a, see also Fig. S6). Both scattering spectra showed a redshift of the scattering maxima compared to the absorbance spectrum due to the increased refractive index on the substrate. Notably, the SCA peak position of Au-[Fe(bpy)3]2+ demonstrates an evident redshift by 12 nm compared to Au single particles, which suggests effective dampening of the Au plasmons through resonant energy transfer from Au into [Fe(bpy)3]2+ 45,46. It could also be observed that the SCA full width at half maxima (FWHM) of Au-[Fe(bpy)3]2+ (93.17 nm) is evidently larger than that of Au NPs control (79.34 nm), which could also be attributed to the energy loss of Au plasmons47,48, and further corroborate the above energy transfer pathway in Au-[Fe(bpy)3]2+ nanocomposite. In addition, steady-state photoluminescence (PL) spectra of Au-[Fe(bpy)3]2+ and controls demonstrate significant changes to the emission pattern (Fig. S7), consistent with the resonant energy transfer from Au towards [Fe(bpy)3]2+ that promotes the photoexcited state energy of the latter49,50,51.

Fig. 2: Characterization of Plasmonic Energy and Charge Transfer Mechanisms.
figure 2

a Absorbance spectra (ABS) of Au NPs and single-nanoparticle scattering spectra (SCA) of Au NPs and Au-[Fe(bpy)3]2+. Inset: Simulated E-field around a single Au NP. b Fe K-edge XANES of Au-[Fe(bpy)3]2+ in dark or under 550-nm illumination. c Au 4 f and (d) Fe 2p XPS spectra and fittings of Au-[Fe(bpy)3]2+ in dark or under 550-nm illumination. e, f Raman spectra of Au-[Fe(bpy)3]2+ and controls with resonant (e) and non-resonant incident wavelengths (f). g, h TA spectra of the [Fe(bpy)3]2+ solution (g), of Au NP colloid (g, inset), and of Au-[Fe(bpy)3]2+ (h) with pump wavelength of 515 nm. i DFT calculations of the density of electromagnetic states of Au-[Fe(bpy)3]2+ with and without illumination.

Full size image

The generation and transfer of plasmonic hot charge carriers could also be an important factor in the Au-[Fe(bpy)3]2+ nanocatalyst. This is supported by our simulations of the charge distribution profiles of Au-[Fe(bpy)3]2+, where the presence of distinct electric potentials is found within the Au, linker, and [Fe(bpy)3]2+ layers (Fig. S8). In-situ X-Ray Absorption Fine Structures (XAFS) spectroscopy and in-situ irradiated XPS (ISI-XPS) provide direct experimental evidence for the plasmonic charge transfer from Au nanoantenna towards the [Fe(bpy)3]2+ catalyst. When Au-[Fe(bpy)3]2+ was exposed to 550 nm monochromic illumination corresponding to the red-edge of the Au LSPR absorption band (552 nm) (Fig. S9), the in-situ X-Ray Absorption Near-Edge (XANES) spectra of the Fe rising edge demonstrates an apparent shift towards lower energies compared to without illumination (Fig. 2b), while the Fe-N bond manifested slight elongation (Fig. S10a). Furthermore, with illumination, the Au 4 f XPS bands in the Au-[Fe(bpy)3]2+ nanocatalyst also demonstrate evident shifts towards higher binding energies (Fig. 2c). In particular, the Au1+ features exhibit more significant shifts due to the bridging role of the surface Au-thiol moiety in plasmonic charge transfer. On the other hand, the Fe 2p peaks in Au-[Fe(bpy)3]2+ shifted towards lower binding energies (Fig. 2d). This opposite shift of Au and Fe signatures was also observed for Au-[Fe(bpy)3]2+ upon broad-band irradiation (Figs. S10b, c), but was not seen for Au NP-only and [Fe(bpy)3]2+-only controls (Fig. S11). These results are highly consistent with the light-induced plasmonic charge transfer from Au towards [Fe(bpy)3]2+.

Analysis of the PL and Raman spectra of Au-[Fe(bpy)3]2+ and controls provides further experimental evidence on the transfer of plasmonic charge carriers. Significantly reduced PL intensity was observed for Au-[Fe(bpy)3]2+ compared to the free-standing [Fe(bpy)3]2+-only and the physical mixture controls (Fig. S12). In addition, Au-[Fe(bpy)3]2+ had a longer lifetime (τ1 = 312 ps) than [Fe(bpy)3]2+ solution (τ1 = 173 ps) as well as the control mixture (τ1 = 243 ps) (Fig. S13). The reduction in PL intensity, combined with the prolonged lifetime, demonstrated effective separation and transport of charge carriers in Au-[Fe(bpy)3]2+ 52,. Furthermore, in surface-enhanced Raman spectroscopy (SERS) measurements, with resonant excitation of the drop-casted thin film (Fig. S14), the signature peaks of [Fe(bpy)3]2+ in Au-[Fe(bpy)3]2+ showed apparent shifts towards lower frequencies compared to freestanding [Fe(bpy)3]2+ (Fig. 2e), which was not observed with non-resonant irradiation conditions (Figs. 2f and Fig. S15). This observation indicates lengthening of the Fe-N bonds and decreases in its stretching frequencies due to charge injection from Au nanoantenna under resonant illumination53, and is highly consistent with our XANES analyses above.

To characterize the dynamics of the photophysical responses following photoexcitation, transient absorption (TA) spectroscopy was next performed. TA spectra of [Fe(bpy)3]2+ and Au NP-only controls respectively exhibits excitonic band bleaching centered at around 525 nm (Fig. 2g) and LSPR bleaching with photo-induced absorption bands on both sides, typical of hot carrier-induced plasmon broadening (Fig. 2g, inset). The overlap between both features led to an evident bleaching band between 500 nm-550 nm for Au-[Fe(bpy)3]2+ (Fig. 2h). Bi-exponential decay fitting of the kinetic traces for Au-[Fe(bpy)3]2+ at 525 nm reveals a short lifetime τ1 = 3.1 ps and a longer component τ2 = 680.2 ps (Fig. S16). τ1 is similar to that found for Au NPs (2.9 ps) and is attributed to hot carrier relaxation by electron-phonon (e-ph) scattering, while τ2 is contributed both by the exciton bleaching lifetime of [Fe(bpy)3]2+ (928.6 ps) and phonon-phonon (ph-ph) scattering lifetime in Au NP only solution (567.0 ps)54,55. Intriguingly, after resonant pump excitations of the Au-[Fe(bpy)3]2+ nanocatalyst, new absorption features were observed at ~700 nm and ~850 nm (Fig. 2h), which are not seen in the Au NP-only and [Fe(bpy)3]2+-only controls. These bands are highly consistent with those found previously for triplet and quintet excited states of [Fe(bpy)3]2+ 50,. Close examination and fittings for the latter show a slow buildup with a time constant of around 35 ps and a decay lifetime as long as 1.2 ± 0.3 ns (Fig. S17). The decay lifetime is consistent with the excitonic lifetime found for [Fe(bpy)3]2+, confirming its origin from [Fe(bpy)3]2+ rather than Au NPs. The slow buildup of the photoexcitated states is in contradict to the ultrafast populating (<50 fs) by direct photoexcitation50, thus evidencing the significant role of resonant plasmonic energy and charge transfers in bleaching of the triplet and quintet states.

To gain deeper insights on the electronic structure of the Au-[Fe(bpy)3]2+ nanocatalyst, density-functional theory (DFT) calculations were next performed. From our density of electromagnetic (DOS) calculations, the d-band centers of Au-[Fe(bpy)3]2+ downshifted under light illumination (Fig. 2i). The surface electronic spin density of the [Fe(bpy)3]2+ molecule was significantly elevated under light irradiation (Fig. S18a), which is consistent with the charge transfer behavior from Au towards [Fe(bpy)3]2+. In addition, the Fermi energy level of Au NPs is higher compared to the calculated LUMO of the catalyst molecule, which presents possibility of plasmonic charge carrier transfer from Au towards [Fe(bpy)3]2+ that could be further promoted by resonant photoexcitation into higher-temperature Fermi-Dirac distribution (Fig. S18b). Overall, these experimental and computational findings provide direct evidence on the occurrence of both resonant plasmonic near-field-mediated energy transfer and plasmonic hot carrier transfer from the Au nanoantenna to [Fe(bpy)3]2+ upon resonant illumination.

Plasmonic catalytic performance and mechanisms

The resonant plasmonic charge and energy transfer mechanisms described above are likely to provide the Au-[Fe(bpy)3]2+ nanocatalyst with favorable photoelectrochemical potential and plasmonic photocatalytic performance14,15,56,57. In this context, we probed its catalytic performance using the water splitting half reactions, hydrogen and oxygen evolution, as test reactions that respectively provide a reductive and oxidative environment. Notably, due to the redox capability of the Fe(II) center in [Fe(bpy)3]2+ via both reductive (Reaction (1)) and oxidative (Reaction (2)) conversions11, it possesses great potentials to facilitate both reductive and oxidative half reactions (Fig. S19), promising overall water splitting photocatalysis.

$${\left[{{Fe}({bpy})}_{3}\right]}^{2+}+1{e}^{-}\rightleftharpoons {\left[{{Fe}({bpy})}_{3}\right]}^{1+}$$
(1)
$${\left[{{Fe}({bpy})}_{3}\right]}^{2+}\rightleftharpoons {\left[{{Fe}({bpy})}_{3}\right]}^{3+}+1{e}^{-}$$
(2)

The catalytic performance was first characterized in 0.05 M H2SO4 towards hydrogen evolution (Fig. 3). In linear scanning voltammetry (LSV) measurements (Fig. 3a), Au-[Fe(bpy)3]2+ demonstrated high current density (−194.47 mA/cm2) as well as strong light response. As an indicator of the photocatalytic performance, a photocurrent density (Jphoto) of −120.92 mA/cm2 was obtained with 520 nm light-emitting diode (LED) irradiation (100 mW/cm2), close to the LSPR wavelength of Au NPs. Broad-band Xe lamp irradiation was also able to induce significant photoelectrochemical response, although to a lower extent than the resonant LED due to the wavelength-specific LSPR excitation. In contrast, control groups including Au NPs, [Fe(bpy)3]2+ solution at identical concentrations, and a physical mixture of both components failed to produce comparable photocatalytic activity due to the absence in the plasmon-mediated photophysical pathways. Au-[Fe(bpy)3]2+ demonstrated much lower Tafel slopes upon irradiation as compared to in dark or control groups under identical illumination (Fig. 3b). A decrease in reaction onset potential was also seen for Au-[Fe(bpy)3]2+ under irradiation (−0.4 V vs Ag/AgCl), which evidences the promising photocatalytic potential of the plasmonic nanocatalysts19,58. Chronoamperometric I-t curves obtained with a constant applied bias of −0.8 V demonstrated much higher photocurrent for Au-[Fe(bpy)3]2+, consistent with LSV and Tafel slope results (Fig. 3c).

Fig. 3: Plasmon-Mediated Catalytic Performance and Mechanism.
figure 3

a LSV curves, (b) Tafel slopes, and (c) chronoamperometric I-t curves at −0.8 V (vs Ag/AgCl) for Au-[Fe(bpy)3]2+ and controls in dark or with irradiation. d Photoelectrochemical hydrogen production of Au-[Fe(bpy)3]2+ under Xe light irradiation at −0.8 V. e Light-modulated photocurrent measured with an applied bias of −0.2 V with or without illumination at 100 mW/cm2. f Wavelength dependence of Au-[Fe(bpy)3]2+ catalytic Jphoto compared to its absorbance. g Plot of Au-[Fe(bpy)3]2+ Jphoto as a function of incident light power. h, i Raman nano-thermometry spectra of Au-[Fe(bpy)3]2+ at different temperatures (h) or after irradiation for 20 seconds (100 mW/cm2) (i). Error bars in all figures represent standard deviations of three independent measurements.

Full size image

Hydrogen production from the Au-[Fe(bpy)3]2+ nanocatalyst was next measured under photocatalytic and photoelectrochemical catalytic conditions. Upon Xenon light illumination (without applied bias), the Au-[Fe(bpy)3]2+ catalyst demonstrated a stable photocurrent density for hydrogen evolution (>200 μA/cm2) over an extended period (Fig. S20a). H2 product measurement under pure photocatalytic conditions shows that a maximum production rate and quantum efficiency of 0.70 μmol g−1 h−1 and 0.047% were reached at 2-hour (Fig. S20b), indicative of the favorable photocatalytic efficacy of the Au-[Fe(bpy)3]2+ nanocatalyst. For photoelectrochemical catalytic conditions under a constant applied bias of −0.8 V, an average HER rate of 2133.55 mmol g−1 h−1 was observed without irradiation (Fig. S21a). The average hydrogen production rate significantly increased to 4160.40 mmol g−1 h−1 with Xe light and 5194.02 mmol g−1 h−1 with 550 nm illumination at 1-sun intensity (Figs. 3d and Fig. S21b). Under both irradiation conditions, the hydrogen production rate gradually increased due to the photothermal effect. In particular, a highest hydrogen production rate of 6379.02 mmol g−1 h−1 was obtained under 550 nm light irradiation in 8 cycles, demonstrating the great potential and stability of the Au-[Fe(bpy)3]2+ nanocatalyst in plasmonic photoelectrocatalysis.

To substantiate the catalyst’s stability, we quantified the Au and Fe elemental compositions in the nanocatalyst before and after photoelectrocatalytic measurements, where only a slight decrease in the relative Fe content was observed in the collected catalyst after reaction (Table S3), suggesting that [Fe(bpy)3]2+ is stably anchored during catalysis. The XRD patterns, TEM images and XPS spectra of the Au-[Fe(bpy)3]2+ catalyst, both before and after the reaction, show almost no detectable changes. In particular, in the TEM images for the nanocatalyst after reaction, the HS-PEG-COOH layer was still observed around the Au NP, suggesting the structural integrity of the nanocatalyst (Fig. S22). No change in the Fe valence state was observed in the Fe 2p XPS spectra before and after reaction (Fig. S23). In addition, in-situ Raman spectroscopy of the nanocatalyst measured in identical conditions as in the PEC measurements also show negligible changes to the vibration signatures over a period of 5 h (Fig. S24). These findings collectively attest to the remarkable stability of the catalyst during prolonged catalytic tests. To obtain further understandings of the photocatalytic performance of Au-[Fe(bpy)3]2+, LSV curves were also obtained in a 0.1 M NaOH electrolyte to assess the oxygen evolution capability. Interestingly, a trend opposite to that in hydrogen evolution was observed. Au-[Fe(bpy)3]2+ demonstrated lower current density and later onset potential upon irradiation as compared to in dark (Fig. S25a). This may be due to the negative role of plasmonic hot electrons in oxidative environment30,59. Chronoamperometric I-t curves indicate steady catalytic performance of Au-[Fe(bpy)3]2+ for up to 10 h (Fig. S25b).

To elucidate the plasmon enhancement mechanisms in hydrogen evolution, we first performed light-modulated photocurrent response measurements with polychromic Xenon lamp or 520 nm and 550 nm LEDs. The modulated photocurrents were recorded in consecutive light-on/light-off cycles under different external biases (−0.2 V and −0.05 V). As shown in Figs. 3e and Fig. S26, the light-modulated photocurrent response in both conditions clearly exhibited two distinct regions: a rapid response current (IRC) region and a subsequent slow-response current (ISC) region, which indicates the presence of both non-thermal and thermal effects and is consistent with previous observations60,61. The proportion of IRC owing to non-thermal photochemical effect is evidently larger, indicating a non-thermal effect-dominating process. Notably, both the Jphoto and the proportion and of IRC was larger with 550 nm illumination, which corresponds to the plasmonic red-edge. The duration of IRC varied in comparison with previous reports, likely due to the prolonged carrier lifetimes and the occurrence of resonant energy transfer-mediated photoreactivity55,62,63. Fitting of the dependence of photocurrent density on light modulation frequencies shows a non-linear dependence (Fig. S27), which is also consistent with the dominance of photochemical effects54.

To further demonstrate the non-thermal enhancement pathways in the Au-[Fe(bpy)3]2+ nanocatalyst, we assessed the dependence of photocatalytic properties on irradiation wavelength and intensity. The photocurrent densities of Au-[Fe(bpy)3]2+ were obtained with irradiation at different wavelengths (all at 1-sun) (Fig. 3f). In general, the dependence profile of Jphoto on incident wavelength closely resembles the absorbance spectrum of Au-[Fe(bpy)3]2+, with only a slight redshift due to the aggregation of the nanocomposites on the electrode surfaces. Radiation at the red-edge of the Au plasmons (550 nm) led to the highest Jphoto (−124.71 ± 35.53 mA/cm2). In terms of the dependence of photocatalytic performance on irradiation light power, it has been previously reported that a linear relationship is typically associated with non-thermal photochemical effect-dominated pathways, whereas a super-linear relationship indicates photothermal effect-induced pathways17,64,65. Notably, the Jphoto obtained by both the 520 nm LED (y = 47.55x + 2.45, R2 = 0.99) and the Xe lamp (y = 7.98x + 23.86, R2 = 0.99) up to 300 mW/cm2 showed a linear dependence on the incident light intensity (Fig. 3g), consistent with a non-thermal effect-dominated reaction pathway largely driven by plasmonic hot carrier transfer as a result of desorption induced by electronic transition (DIET) pathway66.

To investigate the influence of the photothermal effect, we first used an infrared (IR) camera and thermocouple to measure the temperature change at the electrode surface or in the electrolyte for Au-[Fe(bpy)3]2+ during the catalytic process. Under 1-sun irradiation, both measurements showed only a moderate change in temperature (ΔT < 2 °C) (Fig. S28). As these macroscopic temperature measurement results can only reflect the average temperature in the catalytic macro-environment, in-situ Raman nano-thermometry was next exploited for the precise characterization of the surface temperature on the Au-[Fe(bpy)3]2+ nanocatalyst (Fig. 3h). Raman nano-thermometry was performed with in situ illumination based on the temperature-sensitive behavior of the vibrational modes in [Fe(bpy)3]2+ 67, and the precise temperature was calibrated through external heating. As expected, the peak intensity of the [Fe(bpy)3]2+ lattice vibration ( ~ 1610 cm−1) scales linearly with temperature (Fig. S29a). To investigate the effect of photothermal conversion, 1-sun Xe-lamp illumination was applied for varying durations (Fig. S29b). The peak intensity increased evidently with increasing illumination times, and the temperature corresponding to the peak intensity after 20 seconds of illumination (which corresponds to the LSV scanning time) can be calculated via linear fitting (Fig. 3i). A surface temperature of 29.73 °C (∆T ≈ 5 °C) was thus measured, comparable to recent reports using similar conditions67. Although the temporal resolution of the Raman nano-thermometry might be insufficient to capture the entire transient processes preceding plasmonic heating, in-situ Raman nano-thermometry is highly relevant to catalytic measurements, as it precisely measures the temperature change at the catalytic active sites. The influence of such electrolyte temperature changes on the catalytic performance was next evaluated by comparing the LSV curves for Au-[Fe(bpy)3]2+ in dark at different temperatures (Fig. S30). Moderate changes to the current densities were observed for Au-[Fe(bpy)3]2+ at temperatures up to 5 °C higher than room temperature. Overall, these results further confirm that the photothermal effect has not contributed significantly to the photocatalytic performance.

Differentiation and quantification of distinct plasmonic photocatalytic mechanisms

Based on the analyses above, it can be determined that the photocatalytic performance of the plasmonic nanocatalyst includes contributions from the following factors: (i) plasmonic charge carrier-induced photochemistry; (ii) resonant energy transfer-induced catalysis mediated by the plasmonic near field; (iii) synergistic effect between (i) and (ii) through resonant field-mediated charge transfer; (iv) the intrinsic photocatalytic performance of [Fe(bpy)3]2+; and (v) moderate contributions from photothermal effect. To differentiate the distinct catalytic mechanisms, particularly the plasmonic charge and energy transfers (mechanisms (i)-(iii)), we next modified the connection route of Au-[Fe(bpy)3]2+ from a conductive to an insulating layer, and adjusted the resonance conditions between the nanoantenna and the reactor component. Importantly, these modifications allow us to selectively shield the charge transport and resonant energy transfer mechanisms, respectively, in the plasmonic nanocatalyst.

To shield the effect of plasmonic charge transfer, we replaced the thiolated PEG-COOH linker in Au-[Fe(bpy)3]2+ with an insulating SiO2 nanoshell (Au@SiO2-[Fe(bpy)3]2+), which possess very similar thickness (~2.5 nm) to the contour length of the linker (Fig. 4). FDTD simulation of the E-field around a single Au@SiO2-[Fe(bpy)3]2+ particle indicated an almost identical field intensity distribution as compared to Au-[Fe(bpy)3]2+ (Fig. S31). Thus, the change in connection type does not affect the resonant field-mediated energy transfer efficiency. TEM and EDS elemental mappings confirm the successful preparation of Au@SiO2-[Fe(bpy)3]2+ (Fig. 4a, b). Notably, Au 4 f ISI-XPS spectra of Au@SiO2-[Fe(bpy)3]2+ exhibit no evident changes of the Au binding energies under broad-band irradiation compared to in dark (Fig. 4c). Similarly, no evident shifts of the characteristic Raman peaks of Au@SiO2-[Fe(bpy)3]2+ could be observed when compared to [Fe(bpy)3]2+ under resonant irradiation (Fig. 4d, see also Fig. S32). Moreover, the charge distribution simulation of Au@SiO2-[Fe(bpy)3]2+ also indicates an absence of electric potential in the outermost [Fe(bpy)3]2+ layer due to charge transfer (Fig. S33). These observations are in distinct contrast to Au-[Fe(bpy)3]2+, and are indicative of the exclusion of plasmonic hot carrier transfer from Au nanoantenna to [Fe(bpy)3]2+. The catalytic performance of Au@SiO2-[Fe(bpy)3]2+ was next measured. As expected, much reduced photocurrent (Jphoto = −2.79 mA/cm2) were observed with Xe lamp illumination compared to Au-[Fe(bpy)3]2+ (Jphoto = −33.26 mA/cm2) due to the exclusion of plasmonic charge carriers-induced photochemistry (Fig. 4e). The hindered charge transfer kinetics was further evidenced by electrochemical impedance spectroscopy (EIS) and Mott-Schottky measurements55,68, where a much larger charge transfer resistance (Rct) as well as reduced carrier mobility was measured for Au@SiO2-[Fe(bpy)3]2+ (Figs. 4f and Fig. S34).

Fig. 4: Differentiation of resonant plasmonic charge and energy transfer mechanisms.
figure 4

a TEM images and (b) STEM and EDS elemental mappings of Au@SiO2-[Fe(bpy)3]2+. c Au 4 f ISI-XPS spectrum and fittings in dark or under Xe-lamp irradiation for Au@SiO2-[Fe(bpy)3]2+. d Raman spectra of Au@SiO2-[Fe(bpy)3]2+ and [Fe(bpy)3]2+ with resonant incident wavelength. e LSV curves and (f) EIS Nyquist plots of Au-[Fe(bpy)3]2+ and Au@SiO2-[Fe(bpy)3]2+ in dark or upon 1-sun Xe-lamp irradiation.

Full size image

Although the comparison of Au@SiO2-[Fe(bpy)3]2+ and Au-[Fe(bpy)3]2+ enables preliminarily differentiation of plasmonic charge and energy transfer mechanisms, yet the synergistic effect between these mechanisms still needs to be identified to achieve precise quantification. For this purpose, we further replaced the nanoantenna component in Au-[Fe(bpy)3]2+ with Ag NPs (Fig. S35), whose LSPR occurs at 420 nm and does not sustain spectral overlap with [Fe(bpy)3]2+ for resonant energy transfer11 (Fig. S36). As the excitation of Ag plasmons and the connection with [Fe(bpy)3]2+ via the thiolated PEG linker could still facilitate plasmonic charge transfer, the Ag-[Fe(bpy)3]2+ control allowed the selective shielding of resonant field-mediated energy transfer mechanisms as well as exclusion of the synergistic, field-mediated charge transport behavior. To exclude the difference between the intrinsic charge carrier-generation efficacy between the Ag and Au nanoantenna, the surface capping of Ag was adjusted to allow for identical surface charge and photocurrent response to the Au antenna (Fig. S37a, b). Identical particle size and [Fe(bpy)3]2+ surface density was also maintained for Ag-[Fe(bpy)3]2+, which led to a slightly reduced photocatalytic performance (Jphoto = −21.56 mA/cm2) compared to Au-[Fe(bpy)3]2+ (Fig. S37c).

Hence, the contributions from all five possible mechanisms could be feasibly quantified normalized to the photocurrent density obtained from the Au-[Fe(bpy)3]2+ nanoreactor (Figs. 5a, see also Fig. S38). For the hydrogen evolution half reaction and under broad-band Xe lamp illumination, the intrinsic photocatalytic performance of the [Fe(bpy)3]2+ molecules (mechanism (iv)) accounts for a small portion of the total photocurrent, and was quantified by assessing the Jphoto of an inert silica core-loaded [Fe(bpy)3]2+ at identical loading concentration (Fig. S39). A proportion of 7.6 ± 1.5% was thus calculated (Fig. S40, Eq. S1), and was subtracted from additional contributions (Eq. S2). The contribution of the photothermal effect (mechanism (v)) could be quantified by assessing the current density at 5 °C above room temperature (Eq. S3), which corresponds to the photothermal surface temperature increase according to in-situ Raman nano-thermometry, and a proportion of 15.5 ± 8.4% was calculated (Fig. 5b).

Fig. 5: Quantification of Distinct Plasmonic Enhancement Mechanisms.
figure 5

a Schematic representation of the plasmonic enhancement mechanisms in the Au-[Fe(bpy)3]2+ nanocatalyst and controls. (i)–(v) respectively represent contributions from plasmonic charge carrier-induced photochemistry, resonant energy transfer, synergistic field-mediated charge transfer, intrinsic reactivity of [Fe(bpy)3]2+, and thermal effects. b Percentage contribution of distinct plasmonic pathways to hydrogen evolution photocurrent.

Full size image

Plasmonic charge carrier-induced photochemistry (mechanism (i)) could be quantified using the Jphoto of Ag-[Fe(bpy)3]2+ (Eq. S4), which accounts for 57.2 ± 8.4% of the total catalytic performance, and represents a dominant factor to the hydrogen evolution process. This is in line with previous reports that the charge carrier-induced reactivity typically has larger contributions in catalysis of reduction reactions9,11. As discussed above, the difference between the photocurrent densities of Au@SiO2-[Fe(bpy)3]2+ and the intrinsic catalytic properties of [Fe(bpy)3]2+ defines the contribution from resonant energy transfer process (mechanism (ii)) (Eq. S5), for which a mere 0.8 ± 0.1% was calculated. Finally, the difference between Au-[Fe(bpy)3]2+ and Ag-[Fe(bpy)3]2+ represents the contribution from resonant field-enhancement, with the synergistic effect of field-mediated charge transport and photothermal effect included (mechanisms (ii), (iii) and (v)) (Eq. S6). Thus, the synergistic effect between plasmon-mediated energy and charge transfer processes through resonant field-mediated charge separation can be quantified by combining Eqs. S3, S5, S6 (Eq. S7), and is calculated to be 18.9 ± 1.1% (Fig. 5b). These results thus indicate significant contributions by resonant plasmonic near field-facilitated carrier effect.

The selective shielding strategy developed in this work is highly generalized towards distinct catalytic reactions and catalyst configurations. To demonstrate this aspect, we first quantified the contributions of various mechanisms in the oxygen evolution process for Au-[Fe(bpy)3]2+ using the above method (Figs. S41, S42). Notably, the charge transfer effect in Au-[Fe(bpy)3]2+ is found to be unfavorable in the oxidation process (Fig. S23a). Resonant energy transfer-mediated photocatalytic pathway contributes to 53.9 ± 1.6% of the total photocurrent density, representing a dominant factor in the catalysis of oxygen evolution. In contrast, intrinsic catalytic performance of [Fe(bpy)3]2+ accounts for 17.3 ± 0.5%, and photothermal heating-mediated reactivity had a more evident contribution of 28.8 ± 13.6%. These results indicate evident variations in the plasmonic photocatalytic mechanisms under different chemical environments. Furthermore, we have also achieved quantification of distinct mechanisms associated with plasmonic catalysts with (a) different antenna components (Au, Ag NPs), (b) sizes (40, 60 nm), (c) catalyst components ([Fe(bpy)3]2+, cobalt porphyrin, or a CdS semiconductor photocatalyst), and (d) catalyst loadings (Figs. S43, S44). With different catalyst type or loading, the contributions from individual mechanism vary significantly. Particularly, increased contributions from hot carrier-induced photochemistry are recorded using smaller nanoantenna cores, or with lower catalyst loadings (Fig. S43), respectively due to the more efficient carrier extraction and less intermolecular charge hopping behavior69,70. In addition, contributions from resonant plasmonic near field-mediated energy transfer to hydrogen evolution photocurrent could be significantly enhanced through the use of Ag nanoantenna and resonant catalyst components including cobalt porphyrin (Ag-CoPTyP) and CdS (Ag@CdS) (Fig. S44), consistent with the augmented near field intensity around Ag nanoantenna71,72. These results corroborate the general applicability of our strategy and are particularly helpful for modulating the plasmonic catalytic mechanisms towards distinct fields of applications.

Discussion

In summary, we have developed a highly generalized selective-shielding strategy that differentiates and quantifies the contributions of distinct plasmonic photocatalytic mechanisms, including resonant near-field-mediated plasmonic energy transfer, hot carrier-induced photochemistry, and photothermal heating-mediated reactivity, using a prototypical Au-[Fe(bpy)3]2+ antenna-reactor photocatalyst. Through modifications of the resonance condition and connection routes, plasmonic charge and energy transfer mechanisms were selectively shielded, allowing for plausible quantification of these mechanisms as well as their synergistic effect. It is found that plasmonic charge carrier-mediated photochemical reactivity dominates the photocurrent (~57%) in a reducing, hydrogen evolution environment, whereas resonant plasmonic energy transfer dominates (~54%) in an oxidative, oxygen evolution environment. Our strategy provides fundamental, quantitative and highly generalized understandings on the role of plasmon resonance in photocatalysis. It also provides important design principles for plasmonic photocatalysts towards distinct reaction types and catalyst configurations.

Methods

Preparation of Au-[Fe(bpy)3]2+ plasmonic nanocatalyst

First, the synthesized Au colloid was thoroughly washed with DI water three times to remove PVP. Then, HS-PEG-COOH aqueous solution (1 mg/mL, 1 mL) and [Fe(bpy)3]2+ solution (1:1 ratio of water to ethanol, 4.6 mg/mL, 150 µL) were added, after stirring for 12 h under dark condition, it was washed twice with DI water to obtain Au-[Fe(bpy)3]2+.

Preparation of Au@SiO2-[Fe(bpy)3]2+ and Ag-[Fe(bpy)3]2+ controls

Au@SiO2-[Fe(bpy)3]2+ was obtained by mixing the synthesized Au@SiO2 colloid (1 mL) with [Fe(bpy)3]2+ solution (1:1 ratio of water to ethanol, 4.6 mg/mL, 140 µL) and stirring for 3 h, the [Fe(bpy)3]2+ molecules are attached to negatively charged SiO2 surface through electrostatic interaction, and then washed three times with DI water. For the preparation of Ag-[Fe(bpy)3]2+, dispersed 0.5 mg of Ag NPs in HS-PEG-COOH aqueous solution (1 mg/mL, 1 mL), then added 150 µL of [Fe(bpy)3]2+ solution (1:1 ratio of water to ethanol, 4.6 mg/mL) and stirred for 12 h under dark condition, it was washed twice with DI water to obtain Ag-[Fe(bpy)3]2+. The Fe concentrations in Au-[Fe(bpy)3]2+, Au@SiO2-[Fe(bpy)3]2+ and Ag-[Fe(bpy)3]2+ and remain consistent according to the ICP-OES results.

Measurement of Raman spectroscopy

The in-text Raman spectra and maps of different samples were measured on drop-casted thin films of Au-[Fe(bpy)3]2+ and controls on silicon wafer substrates (〈100〉, Ferrotec Shanghai) or in an in-situ electrolysis cell with the Renishaw InVia confocal Raman Microscope with 100× air objective (Leica, Numerical Aperture = 1.25). Excitation laser of 532 nm (1800 l/mm diffraction grating), 633 nm (1800 l/mm diffraction grating), and 785 nm (1200 l/mm diffraction grating) were used with 0.5% laser powers and 10 s exposure time.

Measurement of single-particle scattering spectroscopy

Scattered light was collected with a 50× air-spaced objective (Zeiss, NA = 0.8) and passed to a hyperspectral detection system consisting of an imaging spectrograph with a slit aperture (Princeton Instrument, Acton SpectraPro 2156) coupled to a thermoelectrically cooled back-illuminated CCD camera (Princeton Instrument, Pixis 400BR) and mounted on a computer-controlled translation stage (Newport, LTA-HL).

In Situ X-ray absorption fine structure (XAFS) characterization

XAFS spectroscopy at the Fe K-edge was performed on sector 20 BM-B of APS, equipped with an external xenon light source. All samples were ground lightly using a mortar and pestle. The powder was dispersed onto Kapton tape, forming a uniform and homogeneous layer, which was then folded a few times to achieve absorption of at least 20%. Fe K-edge XAFS spectra of Au-[Fe(bpy)3]2+ and [Fe(bpy)3](PF6)2 were recorded in transmission mode by measuring the incident and transmitted photon flux in ionization chambers I0 and It, respectively. Standard Fe foil was used as reference for beamline energy calibration at the Fe K-edge. The background removal, normalization and Fourier transform (FT) of XAFS spectra were performed using Athena software. The extracted EXAFS data were weighted by k3 and then converted to R space by FT to obtain the magnitude plots of the EXAFS spectra.

In situ XPS characterization (ISI-XPS)

X-ray photoelectron spectroscopy was performed on XPS Thermo Escalab 250Xi. During the experimental process, full-spectrum light source with the light intensity of 100 mW/cm2 (HF-GHX-XE-300) were used as the illumination source. The light irradiation was introduced into the analysis chamber through the quartz window, which allowed illumination from the top. Then, the spectra of Au 4 f and Fe 2p were analyzed in dark and under light conditions to investigate the behavior of electron transfer.

Reporting summary

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