Transport and Interfacial Injection of d-Band Hot Holes Control Plasmonic Chemistry

Harnessing nonequilibrium hot carriers from plasmonic metal nanostructures constitutes a vibrant research field with the potential to control photochemical reactions, particularly for solar fuel generation. However, a comprehensive understanding of the interplay of plasmonic hot-carrier-driven processes in metal/semiconducting heterostructures has remained elusive. In this work, we reveal the complex interdependence among plasmon excitation, hot-carrier generation, transport, and interfacial collection in plasmonic photocatalytic devices, uniquely determining the charge injection efficiency at the solid/liquid interface. Measuring the internal quantum efficiency of ultrathin (14–33 nm) single-crystalline plasmonic gold (Au) nanoantenna arrays on titanium dioxide substrates, we find that the performance of the device is limited by hot hole collection at the metal/electrolyte interface. Our solid- and liquid-state experimental approach, combined with ab initio simulations, demonstrates more efficient collection of high-energy d-band holes traveling in the [111] orientation, enhancing oxidation reactions on {111} surfaces. These findings establish new guidelines for optimizing plasmonic photocatalytic systems and optoelectronic devices.


Introduction
5][6] Practical realizations of hot carrier devices thus require a full understanding of plasmonic hot carrier-driven processes, including plasmon excitation (optical response), hot carrier generation, carrier transport and collection at solid/solid or solid/liquid interfaces.To date, despite significant experimental and theoretical investigations on plasmon excitation and hot carrier generation, electronic processes, i.e. transport and collection, have been less considered.7][18] Carriers generated by plasmon decay impinge upon the surface of a plasmonic nanostructure to be collected, either ballistically or after scattering against other carriers, phonons, or defects in the metal.These ultrafast (a few tens of femtoseconds to picoseconds) scattering processes thermalize the carriers and bring their energy distribution closer to the Fermi level of the metal. 19However, plasmonic hot carrier applications require high energy electrons or holes to efficiently drive ensuing processes.
Plasmon-driven photocatalytic devices typically employ metallic nanoantennas/semiconductor heterostructures due to their efficient hot carrier separation.As a result, hot carrier-driven processes typically involve the complex interplay of carrier collection at both metal/semiconductor and metal/electrolyte interfaces.To gain complete system understanding, the geometry of the nanostructures and their absorption properties need to be precisely controlled.In particular, polycrystalline metal structures or films that have been used in almost all the plasmonic metal/semiconductor devices are not ideal for fundamental studies.Instead, single-crystalline nanoparticles or metal films, e.g gold microflakes 20,21 , can be leveraged to obtain high-quality plasmonic nanoantenna (arrays) with unique optical properties [22][23][24] and well-defined crystallographic surfaces, which exhibit distinct catalytic properties 25,26 as well as distinct hot carrier transport properties. 27Additionally, experimental quantification of the internal quantum efficiency (IQE) of hot carrier collection at each interface is critical to clarify the role of interfaces and their interplay, and to further elucidate the potential opportunities and limitations of hot carrier collection in these devices.Moreover, an appropriate model for hot carrier injection across metal/electrolyte interfaces has remained elusive.Overall, lack of comprehensive experimental approach, a theoretical model, and a highly-controlled nanostructure system have prevented an in-depth investigation on the hot carrier transport and collection processes occurring within hot carrier-driven photocatalytic systems.
In this work, we implement a combination of solid-state photocurrent and liquid-state photoelectrochemical measurements to simultaneously study the transport and collection of hot carriers across the metal/semiconductor and metal/electrolyte interfaces.Uniquely, we leverage highlycontrolled single-crystalline plasmonic gold (Au) nanoantenna arrays on a TiO2 semiconducting substrate.
We use wavelength-dependent scanning electrochemical microscopy (photo-SECM) to locally probe the hot hole-driven photo-oxidation of a redox molecule at the nanoantenna surface under different excitation wavelengths across the entire visible regime (1.4-2.8 eV) and for different thicknesses of the nanoantennas.This spectral range spans across both the intra-and interband regimes of Au, and allows disentanglement of plasmon absorption and interband excitation effects.In combination with solid-state measurements, photo-SECM analysis reveals that the IQE of the photoanode device, in which both hot electrons and hot holes are collected, is controlled and limited by the hot hole collection.Our injection probability model for hot hole collection, retrieved from the combination of experimentally determined IQEs of the photoanodes with ab initio predictions of hot hole generation and transport, reveals the maximum extraction probability of the d-band generated high-energy holes particularly from the top {111} facet.Additionally, we find that the thinnest (16 nm) nanoantennas are most efficient for collecting higher energy d-band holes and importantly, suitable for ballistic collection of hot carriers.Overall, our highly-controlled experimental approach allows quantitative comparison with optical modeling and theoretical calculations, enabling the deconvolution of separate optical and electronic contributions in hot carrier-driven photocatalytic systems.This information and methodology can thus play a critical role towards better design and optimization of plasmonic catalysts.

Results and Discussion
We fabricated plasmonic photocatalytic devices (photoanode) consisting of an array of Au nanodisks (Au NDs) with sizes of the order of tens of nanometers on top of a TiO2/ITO film on a fused silica substrates (Figure 1a).Uniquely, our nanoantennas are made from single-crystalline Au microflakes (SC Au MFs) 20 (see Methods) to exclude the influence of grain boundaries and thus reduce ohmic losses.Importantly, their atomically-smooth and well-defined {111} crystallographic surfaces allows us to deconvolute the effect of different crystallographic facets from other effects in plasmonic photocatalysis.We use TiO2 because of its large optical band gap (EG ~ 3.2 eV 28 ), preventing visible-light absorption within the TiO2 film, and we note it forms Schottky barrier (ФB) across the Au/TiO2 heterojunction, enabling hot-electron collection.The Au NDs are in contact with an electrolyte containing a reversible redox molecule, Fe(CN)6 4- (Ferrocyanide, the reduced form, Red) which enables hot hole collection through photoelectrochemical oxidation reaction (Figure1.a).The oxidaton of this molecule proceeds via one-electron-transfer, outersphere mechanism with a fast kinetic. 29Moreover, the chosen molecule does not absorb visible-light.For complementary solid-state studies, we also work on a plasmonic photodetector device (Schottky photodiode) consisting of an array of SC Au stripes on the same TiO2/ITO film (see Methods), providing an ideal experimental platform for studying the hot electron collection from the same metal with the same semiconductor (Figure1.b).The stripe geometry is dictated by the need of a direct electrical contact to the plasmonic nanoantennas but we note that the electron injection, controlled by the properties of the interface, is identical in both devices.The nanoantennas were fabricated from Au MFs with varying thicknesses of 14 to 33 nm to study the effect of nanoantenna thickness on hot carrier generation, transport, and collection processes within our device.3][34] For photoelectrochemical experiments, we performed a series of photo-SECM measurements on Au NDs arrays with disk thicknesses of 16, 25, and 33 nm and average diameters of 68, 80, and 80 nm, respectively.We present the implemented photo-SECM approach in We determined absorption spectra of the structures experimentally and by simulation under front illumination in ambient air condition (see Methods and Figure S4).Measured absorptions are in good agreement with numerical simulations (Figure S4, dashed lines).To determine the absorption spectra with the illumination condition required for photoelectrochemical experiments, we implemented simulations for an aqueous medium and back illumination condition (see Methods).Calculated absorption spectra are plotted in Figure 2.f.All the Au NDs array heterostructures show a dipole plasmon resonance mode in the intraband region (1.57 to 1.75 eV) which enables us to disentangle plasmon absorption and interband excitation effects.[15] For photoelectrochemical experiments, we used an aqueous electrolyte solution containing 4mM Fe(CN)6 4-and 0.25M KCl, as supporting electrolyte and positioned a 2.4 µm Pt UME tip at 2.5 µm away from the substrate.We performed photo-SECM measurements in competition mode, where an oxidation reaction proceeds at both the UME tip and the substrate (Figure 2.a).During the competition SECM experiment, we applied a potential of 0.4 V vs Ag/AgCl at the Pt UME tip corresponding to electrooxidation of Fe(CN)6 4-at a diffusion-controlled rate (see Figure S2.b) while the substrate was illuminated from the bottom with a tunable, monochromatic light at different photon energies (see Methods).The substrate was at open circuit condition and grounded to take away the accumulated electrons in the ITO layer (Figure 2.a).We focused the laser beam on the Au NDs with a spot diameter of 30 µm to drive the photo-oxidation reaction only at the illuminated area of the substrate upon hot carrier generation and hot hole transfer at the Au/electrolyte interface.We measured the current through the tip as a function of excitation power to monitor changes in the local concentration of Fe(CN)6 4-and thus the photoelectrochemical dynamics at the plasmonic substrate.Figure 2.e shows the time-trace of the tip current (iTiP) upon illumination of a 25nm thick ND array at an excitation wavelength of 450±10 nm, where intensity was modulated up to ~ 14 W/cm 2 .To show the repeatability of the results, we performed the measurements three times at each illumination intensity.We note that once the light is turned off, iTiP returns to the baseline current (iTip,dark), indicating the rapid diffusion and charge transfer of the redox molecules within the tip-substrate gap that enables achieving a steady-state response easily.We observe a linear decrease in the magnitude of the iTip relative to the dark current (iTip/iTip,dark) by increasing the excitation intensity (Figure S8.b).This decreasing trend indicates that the local concentration of Fe(CN)6 4- at the tip-substrate gap decreases due to a hole-driven oxidation reaction at the plasmonic substrate which gets enhanced in kinetics by illumination intensity.To confirm that the observed enhanced photooxidation at the Au NDs originates from the hot carrier generation and collection rather than photothermal heating effects in the range of the studied excitation intensities, we implemented a SECM approach 35,36 for probing the photothermal heating at a plasmonic substrate.Local heating at a plasmonic substrate results in enhanced mass transfer rates of the redox molecules, as well as a shift in the formal potential of a redox couple. 36Our SECM measurements showed no sign of the temperature increase and thus local heating effects on the 25 nm-thick Au NDs array under illumination at its plasmonic resonance and excitation intensities up to 64 W/cm 2 (Figure S6). was observed at shorter wavelengths region by increasing the excitation intensity (Figure S9).To exclude the contribution from TiO2 substrate, the TiO2 /ITO photocurrent was subtracted from the Au/TiO2 /ITO photocurrent (see Methods).Therefore, the substrate photocurrent plots in Figure S8.d, which are after the subtraction correspond solely to the hot hole-driven photo-oxidation reaction at Au NDs surfaces.The external quantum efficiency (EQE) of the photoelectrochemical reaction is determined from the slope of the linear fit to the isub,photo vs illumination power curves for each excitation wavelength using equation S14.The linear relationship of isub,photo vs power shows different slopes, clearly reflecting that the quantum efficiencies of the hot hole-driven oxidation reaction are wavelength-dependent.The measured absorption allows us to calculate the IQE, defined as the probability of a chemical reaction per absorption photon.The obtained EQE and IQE spectra are plotted in Figure 2.g and h for all the ND arrays (see Figure S10 for normalized EQE and IQE spectra).We observe a common feature in all EQE spectra: a peak associated with the characteristic peak plasmon absorption energy in each structure.1][12] Comparing the thickness-dependent IQE spectra, we find a significant difference between the 16 nm and thicker ND structures, in particular in the interband region.The IQE of 16 nm ND structure increases monotonically from 2.4 to 2.8 eV, while the IQE of 25 and 33 nm ND structures rises up to a peak at 2.6 eV, and then drops to higher photon energies up to 2.8 eV.This is possibly because of an increased charge recombination at the Au/TiO2 interface due to the concurrent charge generation in TiO2 at higher photon energies, as also observed in previous studies. 37,38netheless, this interfacial recombination dose not impact the performance of the 16 nm-thick NDs.
Thus, these observations suggest that decreasing the nanoantenna thickness would alter the hot carrierdriven processes across the interfaces.As both hot electron and hot holes are collected at different surfaces it is not clear how the efficiencies of these interfacial collection processes are affected by decreasing the thickness.Therefore, to clarify these two interface contributions, we first focus on the solid-state part to understand the role of metal/semiconductor interface.To study hot electron collection in solid-state, we fabricated a plasmonic Schottky photodiode device consisting of an array of Au stripes with a thickness of 14 nm, close to the thickness of the best performing ND device on the same TiO2/ITO film.We present the implemented solid-state approach in Figure 3.a.A resonance peak at 665 nm (1.85 eV) appears in both spectra.This feature disappears when light polarization is parallel to the stripes (see Figure S5).Such behavior confirms that the photocurrent originates from optical excitation of the dipolar plasmon mode in the nanoantennas. 1e observed similar EQE trend with the absorption spectrum indicates the plasmon excitation manifests itself in an enhanced EQE of the device.Interestingly, consistent with the previous solid-state works 1,2 , we observed that the IQE spectrum (Figure 3.j) exhibits a spectral feature peaking at 550 nm (2.25 eV).The asymmetry between energy distributions for hot carriers generated by interband transitions 19 (peak of distribution at lower energies close to the Fermi level (EF) for hot electrons), combined with our Schottky barrier height (1.25 eV) has consequences for hot electron collection efficiency across Au/TiO2 interface.

SEM image in
As a result, we observe an abrupt drop in the solid-state IQE at energies above the interband threshold (gray area in Figure 3.j).On the other hand, due to the generation of high-energy d-band holes, there is still possibility of hole collection at Au/electrolyte interface, resulting in a continued IQE growth in the entire interband regime for the best performing device (c.f. Figure 2.j, orange dashed curve, 2.4 to 2.8 eV).Comparing the liquid-state photoelectrochemical and solid-state photocurrent measurement results, we find a different magnitude and trend between the experimentally obtained IQEs for the two devices.
Higher magnitude of the IQE for the hot electron photodetector device and a very different IQE trend in particular in the interband regime imply that the IQEs of the photoanodes are not limited by the hot electron collection, rather by hot hole collection.(i) Experimentally determined IQE spectrum of the same heterostructure.The gray shaded areas depict the purely interband region and the dashed lines are a guide to the eye in panels (i) and (j).
To be able to understand how the hot hole collection limits the IQE in the photocatalytic system, we leverage theory to disentangle the mechanisms controlling the generation, transport, and injection of the hot carriers across the metal/electrolyte and metal/semiconductor interfaces.Figure 4.a shows the calculated spatially-resolved absorption 2D profiles, i.e. hot carrier generation profiles, in Au NDs with thicknesses of 16 and 33 nm on TiO2/ITO/glass substrate at their plasmon resonance (intraband regime) and at photon energy of 2.75 eV (450 nm, interband regime).We observe a uniform hot carrier generation across the whole volume of the 16 nm ND close to the both solid-solid and solid-liquid interfaces while a localized generation at the solid-solid interface of the 33 nm ND both for on-resonance and off-resonance excitations.We then employed ab initio simulations 41 to elucidate the impact of carrier transport and separate the contributions from carriers based on the number of times they are scattered before collection (N) (see Methods).The energy-resolved flux, FN (E), of hot carriers with energy above (hot electrons) and below (hot holes) the  that can reach all interfaces was calculated at each photon energy.
As the fabricated structures are single crystalline, no orientational averaging was applied on the ballistic carrier fluxes.We calculated the carrier fluxes for {111} top and bottom and {110} side facets 20 separately.or -2eV, to highlight the presence of the high-energy hot holes) for 16 and 33-nm thick Au NDs at 450 nm and their plasmon resonance wavelength.Under on-resonance illumination of the structures in the intraband regime, the maximum flux of the holes that can reach the interface for the 16 and 33 nm NDs comparable, however, we see a clear difference in their spatial distribution.For the 16 nm ND, the flux is well distributed across the interfaces, while a localized flux is observed on the side surface close to bottom interface of the thicker ND, as is evident from their corresponding calculated energy-resolved carrier fluxes at different surfaces shown in Figure S11.At energies above the interband threshold, accessible by the 450 nm illumination, the carrier flux become more uniform also for the 33 nm ND due to the improved generation profile (Figure 4.a at 450 nm).But still, 33 nm ND exhibits a much smaller maximum flux for both energy thresholds compared to the 16 nm ND, in particular for high-energy holes due to the losses in the transport.Therefore, due to the observed non-uniform carrier generation and flux distribution, as well as the increased distance that holes must travel before reaching the Au/electrolyte interface in the thicker nanoantennas, a slower IQE growth was observed compared to the 16 nm ND (Figure 2.j).This is more critical for collecting the high-energy d-band holes, where the mean free-path of hot holes decreases to just a few nanometers 18,40 , leading to a nonmonotonic behavior of thicker ND IQEs upon from partially to entirely interband excitation (Figure 2.j, 2.1 to 2.8 eV).Finally, we combined our experimental IQE results with our generation and transport model to extract the injection probability across the Au/electrolyte interface.Specifically, by assuming the injection probability, Pinj (E), is constant for all samples (as the same surfaces are exposed), we can state that for the NDs:

𝐼𝑄𝐸 = 𝐹 (𝐸). 𝑃 (𝐸)𝑑𝐸 (1)
Where E is the energy of the hot holes with respect to EF and N is the number of the scattering events we include in modeling.As we know IQE and FN (E), we developed a stochastic fitting model to extract Pinj (E).
Specifically, we carried out a fitting procedure several hundred times to calculate the average and possible spread of Pinj (E) (see Method for further details).The resulted Pinj (E) from this fitting approach for the top and side surfaces is shown in initial and homogenized carriers is the same at energies below -2 eV (Figure S13).We observe the extraction probability approaches an exponentially growing curve for increasing hole energy, with around 5 times higher probability of -2.9 eV holes collected from the top compared to the ones collected from the side surfaces.A very low upper-bound injection probability of ~ 0.1 % is obtained for intraband generated holes with energy of -1.46 eV, although they could have sufficient energy to participate in HOMO electron transfer due to the very low oxidation energy of Fe(CN)6 4-(-0.19eV, Figure S2).We propose that higher energy holes are more strongly oxidizing and more likely to participate in this reaction due to a tunneling barrier that they must overcome in order to reach the physisorbed molecule on the surface. 421][12] Therefore, it is expected that these higher energy holes are mostly collected ballistically, otherwise they immediately lose their energy under scattering events and would have insufficient energy to drive the reaction.Our transport model and fitting approach for injection probability reproduced our experimental data well, in particular our best performing device, 16 nm ND heterostucture.Almost the same computed IQE trend was obtained for the scattered hot holes (Figure S14).Based on our analysis of the injection probability, we also calculate the maximum possible IQE for our system that we could achieve for two upper-bound assumptions.First, we assume that the illumination and chemical reaction happen on the same side (e.g.bottom surface).The maximum IQE, in this case, is 0.12 % (Figure S15.a), which is still close to the experimentally obtained maximum value of 0.094 % in our system (Figure 4.e).Second, we assume all the generated holes immediately reach the top surface and are collected before any scattering events without any transport losses.The maximum IQE, in this case, is 5 % (Figure S15.b), which is ~ 50 times higher than our obtained value.This emphasizes that transport is one of the main limiting factors in the performance of hot carrier devices which should be taken into account in future designs.

Conclusion
In summary, we established a fully controlled system to experimentally quantify the IQE in ultrathin (14 to 33 nm) single-crystalline plasmonic Au nanoantenna array Schottky photodiode and photoanode devices that operate via the collection of hot electrons (Au/TiO2) and both hot electrons and holes (Fe(CN)6 4-/Au/TiO2), respectively.All the Au nanoantenna array heterostructures were designed and fabricated to have plasmon resonance in the intraband region, allowing us to uniquely disentangle plasmon absorption and interband excitation effects.Our experimental IQE data combined with ab-initio modeling revealed the role of intra-and interband decay processes and carrier transport over the nanometer size of the antennas, proposing an injection probability estimation for hot holes collected at the metal/electrolyte interface.Comparing the measured IQE spectra for the two devices having very close thicknesses (14nm and 16 nm) showed a different magnitude and trend in IQEs, with an abrupt drop for the photodiode device and a continuous increase for the photoanode device in the interband regime.
The magnitude difference suggested that the efficiency of the photoanodes is indeed controlled and limited by the hot hole collection at the metal/electrolyte interface.Our injection probability model indicated that mostly the d-band generated high-energy holes participated in the oxidation reaction in our photoanode system and particularly are collected from the top {111} surface.Our transport model and fitting approach for injection probability showed that these hot holes and electrons are mostly collected ballistically at both the Fe(CN)6 4-/Au and Au/TiO2 interfaces in our device with the highest collection efficiency for the thinnest nanoantenna photoanode device (16 nm) as compared to the thicker counterparts (25 and 33 nm).Our results and combined approach could reveal mechanistic insights into the generation, transport, and injection of hot carriers in hot carrier-driven photocatalytic systems, and be a guideline for the design of efficient devices operating in ballistic regime, e.g.plasmon-driven artificial photosynthetic systems or optoelectronics.

Methods
Synthesis and device fabrication.In order to make a transparent semiconducting substrate, first a 50 nmthick ITO film was deposited on a cleaned 1.5x1.5 cm 2 , 520 µm-thick fused silica substrate in a sputtering system (Pfeiffer SPIDER 600).Subsequently, a 40 nm TiO2 film was deposited onto the ITO surface with Alliance-Concept DP 650 sputtering instrument.The deposited TiO2/ITO film was then annealed in air at 450 °C for 2 hours.Large area SC Au MFs were directly grown on borosilicate glass substrates by a halide and gap-assisted polyol process. 20A PMMA wet-transferring method 43  The reverse direction was applied for the front illumination condition.
Photoelectrochemical and photocurrent measurements.To perform liquid-state photoelectrochemical measurements, a custom-built photo-SECM set-up is integrated by adding to an inverted optical microscope (Nikon Eclipse Ti2), a home-built electrochemical reaction cell, a bi-potentiostat (Biologic SP-300), and an ultramicroelectrode (UME) tip connected to a micro/piezo scanner assembly (MMP1/Nano-F450, Mad City Labs ).Pt UME tips were fabricated by heat-sealing and hard pulling of a Pt micro-wire (25 µm, Goodfellow) within a borosilicate glass capillary (1mm ID, 0.5mm OD, Sutter Instruments) with a laser puller machine (P-200, Sutter Instruments), followed by physical contacting of a Cu wire to the Pt wire using a silver epoxy.The laser-pulled UMEs were perfected on an ultrafine polishing plate (BV-10, Sutter Instruments) under video monitoring.A Pt wire (0.5 mm diameter) and a leak-free Ag/AgCl electrode (LF-1, 1 mm OD, Innovation Instrument) were used as counter and reference electrodes, respectively.
A tunable wavelength filter (SuperK VARIA) was employed to modulate the excitation wavelength and power in a wide spectral range of 450 to 840 nm with the bandwidth of 20 nm.An optical shutter (SH1, Thorlabs) was used to chop the incident light.A combination of two lenses was added to the optical path to provide a bottom illumination on the sample with a 30 µm diameter collimated beam.To define the beam diameter, first a CCD image of the laser spot was recorded and then fitted by a two-dimensional Gaussian.Spot diameter here is as the intensity that falls to of the maximum intensity.The solid-state photocurrent measurements were carried out by utilizing piezoelectric microprobes (Imina Technologies, miBots™) to electrically connect the sample and record the short-circuit current.Note that a polarizer

Figure1.
Figure1.Schematic of interfacial hot carrier collection in plasmonic metal/semiconductor heterostructure devices.(a) Au nanodisks (Au NDs)/TiO2 photoanode in contact with a reductant (Red) molecule and band alignment showing hot electron and hot hole collection across the Au/TiO2 and Au/Fe(CN)6 4-interfaces, respectively.(b) Au stripes /TiO2 photodiode geometry and band alignment showing hot electron collection across the Au/TiO2 interface.

Figure 2 .
Figure 2.a.SEM image in Figure 2.b shows one of our fabricated NDs arrays for a 25nm-thick Au MF with a lateral size of 140 µm on a TiO2/ITO-coated fused silica substrate.The SEM and AFM images in Figure2.c and d show the magnified view of the fabricated ND structure.SEM and AFM images of the other fabricated NDs array structures are shown in Figure S3.Thanks to the single crystallinity of the flakes, the fabricated nanoantennas exhibit exact shape, size, and ultra-smooth surfaces.
To investigate the wavelength dependence of the photochemical oxidation rate of Fe(CN)64-, we studied a broad excitation wavelength range of 450 to 832 nm covering the entire intra-and interband transition regimes for hot carrier generation in Au NDs.The measured iTip/iTip,dark responses under different excitation intensities at each wavelength are shown in Figure S8.b.We implemented a diffusion model 13 to simulate calibration curves correlating iTip/iTip,dark, substrate photocurrent (isub,photo), and the reaction rate constant (Keff) of the photo-oxidation reaction of Fe(CN)6 4-at the substrate.The substrate photocurrents were extracted using the calibration curve obtained by the diffusion model (Figure S8.c) and the measured data of iTip/iTip,dark under different illumination intensities (Figure S8.b) for each excitation wavelength, and plotted as a function of power in Figure S8.d.As a control experiment, the same SECM measurement was performed on a bare TiO2/ITO substrate in the absence of Au NDs.A slight enhanced photo-induced iTip

Figure2.
Figure2.Liquid-state photoelectrochemical measurement results.(a) Schematic of the designed plasmonic heterostructure and photo-SECM configuration in a competition experiment mode.A fabricated Au NDs array from a SC Au MF on TiO2/ITO substrate is in contact with an electrolyte contacting 4mM Fe(CN)6 4-(Red) and 0.25 M KCl.A 1.2 µm-radius Pt ultra-micro electrode (UME) tip is positioned 2.5 µm away from the substrate.The UME tip is biased at 0.4 V vs Ag/AgCl (reference electrode, RE), and the substrate is at open circuit and grounded.Light is incident on the plasmonic Au NDs array from the bottom.The same oxidation reaction happens at the tip electrode and substrate surface.The current is measured through the tip electrode (working electrode, WE).A Pt wire is used as a counter (CE) electrode

Figure 3 .
photocurrent measurements by electrically contacting microprobes on a non-patterned flake area and on a sputtered Ohmic contact to TiO2.The photocurrent can be collected while illuminating the sample through the bottom with a tunable, monochromatic light (see the schematic in Figure.3a and Methods).

Figure 3 .
Figure 3.e shows the time-trace of the short circuit current (ISC) upon illumination of the device with wavelengths from 450 to 840 nm at different powers.From the I-V response shown in Figure 2.e, a highly rectifying behavior was observed.Fitting of this plot to the diode equation 40 allowed the estimation of a Schottky barrier of ~ 1.25 eV across the Au/TiO2 heterojunction.We determined the EQE and absorption spectra of the device experimentally by measuring both the wavelength-dependent photocurrent as well as transmission and reflection spectra under the same light polarization condition (see Methods) and plotted them in Figure 3.f.A resonance peak at 665 nm (1.85 eV) appears in both spectra.This feature

Figure3.
Figure3.Solid-state photocurrent measurement results.(a) Schematic of the designed plasmonic heterostructure and measurement configuration.A stripe Au pattern is fabricated from a SC Au MF on TiO2 /ITO substrate together with a 100 nm-thick sputtered Au film Ohmic contact.Light is incident on the plasmonic Au stripe array from the bottom and the photocurrent is collected by two microcontact probes electrically connected to the Au flake and the sputtered Au contact pad.(b) SEM image of a 30x30 µm 2 stripe array from a 14 nm-thick Au MF.(c) Higher magnification SEM and (d) AFM images of the fabricated Au stripe array from the Au MF in (b).The average stripe width and thickness are 70 and 14 nm, respectively.The array periodicity is 230 nm.(e) Time-trace of the short-circuit current (ISC) for the

Figure 4 .
b compares the spatially-resolved carrier fluxes of hot holes with energy greater than a threshold (-0.15eV, close to the highest occupied molecular orbital (HOMO) level of Fe(CN)6 4-(-0.19eV)

Figure 4
.c maps the mean free-path of holes in the top d-band in Au across velocity directions.The unit vector of velocity directions of these holes are distributed non-uniformly across the unit sphere with much higher probability between the [110] and [111] directions.These most probable velocity directions also have mean free-paths, indicating that the anisotropy favors hot hole transport towards the top and sides of our single-crystalline samples (compared to intermediate directions).

Figure 4 .
d.No significant difference was obtained in injection probability between non-scattered (Figure 4.c, N=0) and scattered (Figure S12, N=4) holes, as the flux distribution of

Figure 4 .
e shows the predicted IQE spectra from F0 (E) for the holes that reach to the top and side surfaces and the obtained Pinj (E) in Figure 4.d together with the experimentally determined IQEs for 16 and 25 nm-thick Au NDs heterostructures (see 33 nm-thick NDs in Figure S14).

Figure 4 .
Figure 4. Hot carrier generation, transport and injection in plasmonic heterostructure devices.Calculated spatially-resolved (a) 2D absorption profiles and (b) hot hole fluxes reaching the interfaces greater than 0.15 and 2 eV under illumination at 450 nm and the resonance wavelength of the Au NDs having thicknesses of 16 and 33 nm.Fluxes are normalized to the absorbed photon and unit area of the surface.

(
WP25M-VIS,Thorlabs) was added to the optical path to provide a polarized beam perpendicular to the Au stripes.The current-voltage (I-V) and time-trace of the photocurrent (I-t) curves were recorded a Keithley 2450 SourceMeter.Numerical simulation.The electromagnetic simulations were performed using the RF module of the COMSOL Multiphysics v5.6 to simulate absorption spectra as well as 3D internal electric field distributions across the volume of the nanoantennas.The latter was used as an input in the subsequent hot carrier transport calculation code.A 3D unit cell model, consisting of one Au nanoantenna (disk or stripe geometry) on TiO2/ITO/fused silica substrate surrounded with a top layer of water (for disk) or air (for stripe), was simulated by setting the unit cell width equal to the array periodicity (200 nm for disks and 230 nm for stripes) and unit cell length equal to 200 nm and 300 nm for disks and stripes, respectively.Perfect magnetic conductor and perfect electric conductor boundary conditions were used at the sidewalls of the unit cell.A port boundary condition with the excitation "ON" was set at the bottom of the unit cell for the back illumination with a normal incident plane wave (450-850 nm) with electric field polarization perpendicular to the stripe length axis as well as for recording the reflected wave.A second port boundary condition without excitation was used at the top of the unit cell to record the transmitted wave.The absorbed power was calculated by volume integration of the electromagnetic power loss density over the nanoantenna volume.The wavelength-dependent complex refractive indices for SC Au, TiO2 and ITO were taken from refs44 , 45 , and46 .A 2D diffusion COMSOL model was also implemented to the tip current (ITip) response in photo-SECM experiments.Simulation details are provided in the Table of Contents Graphic