Plasmons are collective oscillations of conduction electrons whose characteristics are determined by their spatial extent and coupling conditions^[26]. In metals, bulk plasmons arise near the plasma frequency as longitudinal charge-density oscillations, whereas surface plasmons appear at metal-dielectric interfaces as lower-energy modes confined to the boundary^[27]. When these surface modes couple to electromagnetic radiation at planar interfaces, they form surface plasmon polaritons (SPPs), which propagate along the interface while decaying evanescently into both the metal and the dielectric^[28]. The dispersion relation of SPPs at a planar metal-dielectric interface can be obtained by solving Maxwell’s equations with appropriate electromagnetic boundary conditions and is given by the following equation^[28,29]:

(2) $$ k_{S P P}(\omega)=k_{0} \sqrt{\frac{\varepsilon_{m}(\omega) \varepsilon_{d}}{\varepsilon_{m}(\omega)+\varepsilon_{d}}} \\ $$

where k₀ is the free-space wavevector. The imaginary component of k_SSP determines the propagation length of SPPs along the interface^[30], which defines the spatial range over which plasmonic energy can interact with the semiconductor depletion region. This spatial overlap critically governs whether non-thermal carriers generated during plasmon decay can be separated by built-in electric fields before thermalization^[31]. Because direct phase matching between free-space photons and SPPs is prohibited by momentum mismatch, excitation typically relies on prism-based coupling schemes^[28] or periodic grating structures^[32]. In periodic architecture, reciprocal lattice vectors compensate for the momentum mismatch and enable resonant coupling through hybridization between diffractive orders and surface-bound modes, thereby extending the effective interaction length between optical fields and semiconductor charge carriers^[32].

In contrast to propagating SPPs, LSPRs arise from the quantum confinement of conduction electrons within subwavelength metallic nanostructures^[33]. Under the quasi-static limit, LSPR excitation concentrates electromagnetic energy into nanoscale volumes, generating intense near fields that decay within approximately 10 ~ 20 nm from the metal surface^[34]. Consequently, effective photoelectrochemical enhancement requires that the photoactive semiconductor region be positioned within this near-field zone^[35]. In the quasi-static limit, the resonance condition of dipolar localized surface plasmons is governed by the Fröhlich criterion derived from the dipolar term of Mie theory and is given by the following equation^[33,36]:

(3) $$ \operatorname{Re}\left[\varepsilon_{m}(\omega)\right]=-2 \varepsilon_{d} \\ $$

which highlights the central role of the metal dielectric function and the surrounding environment in determining resonance energies. Notably, the Fröhlich condition can also be satisfied in heavily doped or defect-engineered metal oxides, where increased free-carrier densities and Fermi-level shifts induce a Drude-like dielectric response, thereby enabling LSPR beyond noble metals^[37,38].

Collective plasmonic modes in periodic nanoparticle arrays bridge propagating and localized resonances through lattice-mediated coupling. Near diffraction conditions, hybridization between dipolar LSPR modes and in-plane diffractive orders form surface-lattice resonances (SLRs), which exhibit reduced radiative damping and enhanced spatial coherence across the array^[39,40]. For photoelectrochemical applications, such collective modes enable spectrally selective enhancement when resonance conditions coincide with semiconductor absorption and carrier-collection probability under near-normal illumination^[41]. However, because solar-driven systems benefit from broadband absorption, lattice-coupled resonances are most effective when integrated with broadband absorbers or multiplexed architectures that balance spectral selectivity with overall photocurrent generation^[42].

Taken together, propagating SPPs, localized LSPRs, and lattice-coupled collective modes provide complementary pathways for manipulating light absorption, carrier generation, and interfacial charge transfer across distinct spatial and temporal scales^[31,35,41]. By selecting and integrating these plasmonic modes according to device geometry, junction energetics, and material loss characteristics, plasmonic architectures can be rationally designed to address the key bottlenecks of photoelectrochemical water splitting^[43].

When metal nanostructures are excited by light, plasmons store energy as collective electron oscillations. This energy relaxes via two pathways on different time scales: (i) far-field photon emission and (ii) dissipation into electronic and lattice degrees of freedom. These pathways connect near-field concentration [Figure 2A] to hot-carrier generation and, ultimately, to thermal relaxation [Figure 2B-D]^[44,45].

Accordingly, the total plasmon damping rate can be expressed as the sum of radiative and non-radiative contributions^[46,47]:

(4) $$ \Gamma_{\text {tot }}=\Gamma_{\text {rad }}+\Gamma_{\text {mad }} \\ $$

The spectral linewidth is quantified by the FWHM, γ, which is inversely related to the dephasing time T = 2ℏ/γ^[46].

The radiative quantum efficiency is defined as the ratio of radiative to total damping rates^[47]

(5) $$ Q_{R Q E} \equiv \frac{\Gamma_{\text {rad }}}{\Gamma_{\text {tot }}} \\ $$

For lattice-coupled nanoparticle arrays, the quality factor of SLR is commonly defined as^[40]:

(6) $$ Q_{S L R} \equiv \frac{\lambda}{F W H M} \\ $$

Particle shape and the optical environment determine the Γ_rad-Γ_nrad partition, thereby fixing γ and T^[47]. Immediately after excitation, the carrier population is strongly non-thermal (τ_nth < 100 fs) owing to Landau damping, while subsequent electron-electron scattering (10^-13 ~ 10^-12 s, τ_el) drives rapid redistribution toward a quasi-thermal state - processes summarized in [Figure 2B and C]^[48]. Size- and surface-induced losses can be approximated as^[49]:

(7) $$ \Gamma_{\text {nrad }} \approx \Gamma_{\text {bulk }}^{(e-\mathrm{h})}+A \cdot\left(\frac{v_{F}}{L_{\text {eff }}}\right) $$

where v_F is the Fermi velocity, L_eff is the effective mean free path set by particle size/roughness, and A is a geometry-dependent constant^[49].

Hot-carrier collection and injection at the metal/semiconductor interface are governed by distance and energy. Defining the ballistic length l_b≡v_Fτ_e-e, carriers can arrive before thermalization if the metal-semiconductor separation satisfies d < l_b^[48]. Injection further requires the energy condition E - E_F ≥ Φ_B - qΔV, where Φ_B is the Schottky barrier and ΔV is the interfacial band bending under illumination or bias^[50]. In practice, a moderate Q_SLR is targeted: an excessively large Q_SLR over-suppresses Γ_rad, whereas a moderate value preserves sufficient radiative coupling while L_eff, d, and Φ_B are co-tuned so that injection occurs within the τ_nth, τ_el time window^[40].

During thermal relaxation (τ_ph ≈ 10^-10 ~ 10^-8 s), electron-phonon coupling equilibrates the electron and lattice temperatures, followed by interfacial conductance and diffusion that remove heat to the surroundings - see [Figure 2D]^[51]. At this stage, the energy manifests primarily as photothermal effects that modulate the interfacial microenvironment but no longer contribute directly to hot-carrier injection.

The radiative pathway can be engineered via lattice coupling of localized plasmons. Near the Rayleigh anomaly, hybridization between dipolar LSPRs and in-plane diffractive orders forms subradiant SLRs [Figure 2E]^[40]. Maximizing Q_SLR requires λ ≈ λ_RA in a nearly symmetric refractive-index environment, whereas hot-carrier injection favors an intermediate Q_SLR that preserves radiative coupling while satisfying d < l_b and E - E_F ≥ Φ_B - qΔV^[40]. However, material losses ultimately limit performance: Au supports narrower linewidths and larger Q_SLR in the red/near-IR due to its smaller imaginary permittivity, whereas TiN exhibits higher non-radiative losses and broader linewidths even under lattice coupling [Figure 2E]^[52-54]. Moreover, chemical-interface damping introduces additional structure-dependent losses, emphasizing the need for surface-chemistry control during fabrication^[55].

When plasmons decay, energy is distributed between radiative emission and electronic excitations. Within the latter, intraband processes dominate at longer wavelengths, whereas interband d → sp transitions become increasingly prominent at shorter wavelengths^[56]. Interband transitions arise from d → sp excitations whose energetic depth is governed by the proximity of the d band to the Fermi level, leading to deeper hole generation in noble metals [Figure 3A]^[52,57]. $$ \varepsilon_{I m} $$ quantifies Ohmic losses; its magnitude governs the intrinsic dephasing rate, the LSPR linewidth, and ultimately the hot-carrier yield. Primary optical-constants measurements show that Ag and Au exhibit a monotonic decrease in $$ \varepsilon_{I m} $$ beyond ~ 550 nm, consistent with Drude scaling^[58], whereas Pt, Pd, and Ni maintain large $$ \varepsilon_{I m} $$ due to overlapping interband transitions, which broadens the resonance and weakens near-field enhancement for comparable geometries^[59]. In Figure 3B, Cu displays an interband shoulder around ~ 550 nm that overlaps the LSPR near ~ 670 nm, broadening photon absorption to ~ 450-800 nm within single nanoparticles^[60]. In contrast, Au shows a more distinct dipolar peak because its interband threshold lies at shorter wavelengths^[52]. This spectral overlap provides a design lever for the simultaneous generation and extraction of intraband electrons and interband holes^[61].

Particle size and modal content further regulate the intraband-interband balance. As shown in Figure 3C, ~20 nm Au spheres present a narrow dipolar peak (FWHM ≈ 80 nm) and primarily supply intraband electrons^[62,63]. Increasing the radius to ~ 50-90 nm broadens the linewidth to ~ 180 nm, allowing the interband shoulder to merge with the plasmonic mode; this enables concurrent electron-hole generation and increases the extractable carrier flux^[64]. At radii ≥ 140 nm, higher-order modes and strong scattering emerge; a sparse array of large particles then functions as a light-trapping antenna, extending the internal optical path in ~ 1 µm TiO₂ films^[65].

Synthesizing these primary reports yields a working picture for plasmonic photoelectrodes: intraband electrons tend to favor H₂ evolution, whereas interband holes facilitate O₂ evolution; their relative contributions are tunable via wavelength (material selection), particle size/shape, and junction design [Figure 3B and C]^[66].

Overall, this section establishes the fundamental plasmonic modes, relaxation pathways, and carrier-transition mechanisms that govern how optical energy is converted into extractable charge carriers, thereby providing the physical basis for the plasmon-enhanced photoelectrochemical processes discussed in subsequent sections.

Plasmonic enhancement in PEC photoelectrodes arises when incident photons efficiently couple into LSPR/SPP modes that localize electromagnetic fields at the reactive metal/semiconductor-electrolyte interface, and when interfacial charge transfer outpaces competing relaxation pathways^[67]. Periodic lattices and particle arrays could compensate the photon-SPP momentum mismatch. By compensating energy at the junction, these structures maximize the local optical density of states at the interface. For example, hematite nanorods integrated with an Au nanohole array show order-of-magnitude increases in photocurrent and incident photon-to-current efficiency (IPCE) when an SPP-excited waveguide mode overlaps the absorption band^[68]. Ultrafast spectroscopy on Au/TiO₂ nanostructures indicates ≤ 50-fs electron transfer with high initial yields, establishing that plasmon decay can directly populate the semiconductor conduction band under resonant excitation^[38]. Field-confining antenna-spacer-mirror cavities^[69], gap-plasmon broadband absorbers^[70], and grating-mediated coupling to 2D semiconductors^[71] further amplify light absorption and raise hot-carrier photocurrents by factors of approximately two to ten, underscoring geometry as a primary lever for interfacial transfer without invoking non-plasmonic catalysts. In practice, design and diagnosis proceed in tandem, and optical modes are carefully engineered.

Chemical and electrochemical stability is essential for practical plasmon-enhanced PEC operation^[72]. Under photoelectrochemical conditions, plasmonic metals would suffer corrosion (dissolution and/or surface oxidation), which can be assessed using Pourbaix-diagram-based potential-pH stability windows^[73]. Overlap of PEC operating potentials with corrosion or dissolution regions induces nanostructure evolution and degrades long-term PEC performance^[74].

Chemical Interface Damping (CID) is a phenomenon in which the plasmon damping rate changes when several interfacial processes, such as molecular adsorption, charge transfer, electronic coupling with substrates, and defect states, act simultaneously or in combination at the metal interface^[75].

The extent of CID rises as the adsorbate dipole moment grows; single-particle spectroscopy has demonstrated that adsorption of longer-chain thiols such as 1-decanethiol induces a broader homogeneous LSPR linewidth in Au nanorods compared to shorter analogs such as 1-butanethiol, indicating that stronger interfacial dipoles enhance damping^[76]. This occurs through direct charge transfer from the excited plasmonic metal to adsorbed molecules at the nanostructure surface, leading to increased plasmon linewidth broadening and energy dissipation^[77].

LSPR linewidth broadening can be represented as Γ_tot = Γ_bulk + Γ_rad + Γ_surf + Γ_CID. Bulk damping (Γ_bulk), radiation damping (Γ_rad), electron-surface damping (Γ_surf), and CID (Γ_CID)^[62]. Γ_bulk is damping from inside the metal, Γ_rad is radiation damping (plasmons convert to photons), Γ_surf is surface damping (increases with smaller nanostructure size), and Γ_CID is additional LSPR energy loss due to charge transfer or electron orbital coupling with molecules adsorbed on the metal nanostructure surface.

Temperature variations affect the electronic and optical properties of plasmonic nanostructures, primarily through changes in the dielectric constant of the metal, electron-phonon interactions, and thermal expansion^[78,79]. For instance, in Au thin films, the imaginary part of the dielectric function at 500 °C becomes nearly twice as large as that at room temperature, indicating increased optical losses and reduced plasmonic quality factors^[80]. As a result, the FWHM broadens at a plasmon peak resonance.

According to the Drude-Lorentz model, the damping term (Γ) in the dielectric function increases with temperature due to more frequent inelastic electron-phonon collisions^[80]. In one study, temperature-induced changes in plasmonic properties were experimentally analyzed using UV-VIS-NIR spectroscopy^[81]. This analysis revealed shifts in resonance peaks and variations in intensity, indicating alterations in electron density and plasmonic behavior^[82].

A subsequent temperature rise leads to a localized increase in the dielectric constant, resulting in a decrease in the resonance frequency^[83]. Conversely, blue shifts may occur under specific conditions, precipitated by structural alterations that result in a reduction of the resonance wavelength. Electron-electron scattering has been demonstrated to dissipate energy by increasing the frequency of collisions between free electrons when irradiated with strong light. Electron-phonon coupling has been shown to accelerate energy dissipation due to an increase in lattice vibrations.

Under photoelectrochemical operating conditions, plasmonic nanostructures are continuously driven far from thermodynamic equilibrium^[84]. In this context, Pourbaix-type frameworks are adopted as a thermodynamic mapping tool to capture illumination-induced electronic shifts together with localized photo-thermal effects^[85]. Plasmonic excitation leads to local heating and redistribution of photoexcited carriers, which dynamically modulate the effective electrochemical potential of plasmonic metals and alter the operational stability window. Such photo-thermal perturbations enhance plasmon damping and resonance broadening, ultimately contributing to time-dependent PEC performance degradation^[85,86]. In photoelectrochemical operation, light absorption and localized photothermal heating dynamically shift the effective electrochemical potential of plasmonic metals, thereby modifying the accessible stability window during operation.

While the SQ framework establishes the efficiency ceiling for single-junction solar conversion, photoelectrochemical water splitting is further limited by the thermodynamic requirement of the hydrogen evolution process^[89,90]. Nonetheless, in the context of solar-driven water splitting, it is imperative to consider additional factors such as Faradaic efficiency^[91]. Smaller bandgaps sustain excessive photocarrier thermalization losses, while larger bandgaps limit photon absorption and thus reduce photocurrent, leading to the characteristic peak at the intermediate bandgap and efficiency decline toward both ends of the spectrum [Figure 5A]^[92].

Due to overpotential constraints, PEC efficiency emerges only within a limited bandgap window and declines outside this range, whereas dual and multi-junction designs mitigate these limitations through bandgap balancing across junctions [Figure 5B-D]^[93,94]. The incorporation of Au nanoparticles, which exhibit LSPR, enables unique charge generation and energy transfer processes that extend beyond conventional photovoltaic mechanisms^[95].

Plasmonic photoelectrochemical systems enhance light utilization and carrier dynamics through non-equilibrium processes that are not captured by conventional photovoltaic models^[96]. In this context, the SQ limit defines the thermodynamic benchmark for equilibrium-based solar conversion and is employed here as a reference framework for interpreting how plasmon-mediated effects, discussed in the following sections, extend the fundamental limits of water-splitting photoelectrodes.

LSPR-excited metal nanoparticles undergo two principal decay pathways [Figure 6A]^[97]. Hot-carrier generation is dominated by non-radiative plasmon decay, which directly converts plasmon energy into energetic charge carriers, whereas radiative decay primarily re-emits photons into the far field and contributes minimally to carrier injection^[98,99]. As schematically illustrated in metal-semiconductor junctions, carriers with kinetic energies exceeding the Schottky barrier height can be injected into the semiconductor conduction band [Figure 6B]^[97].

The excitation pathways underlying hot-carrier generation are reflected in the IPCE spectra of Cu/TiO₂ nanodiodes. Figure 6 illustrates the band alignment at the Cu/TiO₂ junction, where intraband excitation produces a broad low-energy carrier distribution, while interband excitation corresponds to 3d to 4s electronic transitions in Cu, selectively enabling high-energy electrons to participate in interfacial charge transfer^[60]. These distinct excitation channels extend light harvesting beyond the intrinsic bandgap of the semiconductor absorber, enabling additional carrier generation within the photoelectrode^[100].

Hot-carrier injection (HCI) requires appropriate energetic alignment between the metal Fermi level and the semiconductor band edges, enabling hot electrons generated near the metal surface to overcome the Schottky barrier and inject into the conduction band of an n-type semiconductor such as TiO₂^[101]. Simultaneously, hot carriers originating from d-band to sp-band transitions can participate in oxidation reactions at the metal-electrolyte interface, thereby maintaining charge neutrality during interfacial transfer processes^[102].

In plasmonic metal-semiconductor hybrids, enhancement of IPCE is therefore closely associated with efficient hot-carrier injections across the interfacial barrier^[103,104]. The IPCE spectra of Cu/TiO₂ nanodiodes display distinct band features, confirming contributions from both excitation pathways to photocurrent generation. Plasmon-assisted processes are identified by deviations from Fowler’s law in the IPCE-photon energy spectrum, which defines the energy window where surface plasmon effects dominate carrier generation [Figure 6D]^[105]. Comparisons between spherical and nanorod-decorated Au/TiO₂ photoelectrodes further reveal that the longitudinal LSPR mode in nanorods extends the IPCE response toward longer wavelengths [Figure 6E]^[105,106]. Accordingly, combining different particle geometries enables broadband optical absorption, while reducing particle size enhances hot-electron injection efficiency [Figure 6F]^[107]. These observations indicate that charge generation and injection are optimized when the plasmon resonance is spectrally aligned with interband transition energies^[108].

Under realistic operating conditions, IPCE serves as a primary metric for evaluating optical-to-electrical energy conversion efficiency in photoelectrochemical systems by capturing deviations from the idealized SQ framework arising from plasmon-mediated non-equilibrium carrier pathways^[109]. Among these pathways, hot carrier injection directly converts plasmon decay into photocurrent by enabling energetic electrons to cross the metal-semiconductor barrier, thereby extending photoelectrochemical water splitting into sub-bandgap spectral regions and enhancing IPCE under visible light illumination.

In photoelectrochemical water splitting, Local Electro-Magnetic Field-enhancement (LEMF) intensifies light-matter interactions beyond the intrinsic absorption of semiconductor photoelectrodes^[74]. Plasmonic nanostructures generate electromagnetic near-fields that increase the local photon density by several orders of magnitude within nanometer-scale regions adjacent to the semiconductor surface^[110], thereby enhancing electron-hole pair excitation. Unlike macroscopic light-trapping strategies based on scattering or texturing, LEMF compresses electromagnetic energy into sub-wavelength volumes^[43].

LEMF is defined as the resonant amplification of the optical near-field within a few nanometers of a plasmonic nano-object, producing nanometer-scale electromagnetic hot spots^[111,112]. Importantly, LEMF redistributes electromagnetic energy without increasing the total incident light intensity. Finite-difference time-domain (FDTD) simulations show that closely spaced plasmonic nanoparticles, such as a linear five-sphere Au nanochain, generate pronounced near-field hot spots at interparticle gaps, resulting in stronger local electromagnetic fields and enhanced light absorption compared with isolated particles [Figure 7A]^[113,114].

The amplification of the electric field by LEMF is contingent upon the geometry, gap, and material composition of the nanostructure. Comparison of star- and rod-shaped Au antennas shows that sharp, tip-dominated morphologies strongly enhance LEMF through the lightning-rod effect, while simultaneously enabling tunable longitudinal LSPR modes extending into the near-infrared, consistent with experimentally observed field enhancement in nanostar geometries. This localized field amplification increases the effective excitation rate of electron-hole pairs in the adjacent semiconductor, thereby enhancing carrier generation over a broader interfacial area compared to nanorods [Figure 7B]^[115].

This localized effect gives rise to spatially confined zones of enhanced photoactivity, which are sometimes referred to as plasmonic hot-spots^[116]. In these regions, enhanced photon absorption and carrier generation reduce recombination losses by shortening carrier diffusion lengths, while the strong localized field can additionally modulate band bending within the semiconductor^[117]. This, in turn, results in the modification of the built-in electric field that serves as the driving force for carrier separation. This modification has the potential to facilitate the extraction of holes to the electrolyte and enhance the efficiency of water oxidation kinetics^[118].

Figure 7C and D together demonstrate that non-spherical Au nanoparticles exhibit enhanced optical efficiencies because their anisotropic geometries promote stronger localization of the electromagnetic near field. The efficiency factors in Figure 7C increase with shape anisotropy, while Figure 7D directly visualizes the corresponding enhancement of local field intensity, governed by geometry, dielectric environment, and coupling effects^[119,120].

LEMF concentrates optical energy near the semiconductor interface, increasing carrier generation where electrochemical reactions occur. In photoelectrochemical water splitting, this near-field localization improves charge separation and reduces recombination losses without increasing incident light intensity.

Resonant energy transfer (RET) is a non-radiative process in which excitation energy migrates from a donor to an acceptor through dipole-dipole coupling^[121]. Plasmon-induced resonance energy transfer (PIRET) is a non-radiative mechanism that transfers plasmonic excitation energy from metal nanoparticles to adjacent semiconductors through near-field dipole coupling, without direct electron transfer^[122]. For instance, in Ag@TiO₂ structures, the spectral overlap between the LSPR and the TiO₂ absorption edge enables PIRET, while the direct metal-semiconductor contact simultaneously allows hot electron injection [Figure 8A]^[122].

To facilitate this interaction, an ultrathin spacer with a thickness of 1-3 nanometers is introduced, ensuring effective coupling without requiring a direct tunneling channel^[123]. In the case of Ag@SiO₂@TiO₂, the insertion of a ~ 10 nm insulating SiO₂ layer blocks hot electron injection, yet PIRET remains operative due to the preserved spectral overlap between the plasmon and the semiconductor [Figure 8B]^[122]. Under these conditions, a substantial increase in the local density of photogenerated electron-hole pairs has been reported within nanometer-scale volumes^[124].

PIRET bypasses the Schottky barrier through resonant interfacial energy exchange, enabling ultrafast charge population and substantial enhancement of photoelectrochemical reaction rates without altering the open-circuit potential^[124,125]. FDTD simulations further confirm that geometric sharpening and reduced interparticle gaps amplify near-field intensity, thereby increasing PIRET efficiency without changing overall device absorbance^[126]. Figure 8C provides an integrated conceptual view, illustrating that plasmonic nanoparticles can simultaneously mediate resonant energy transfer, carrier injection, and radiative processes^[127].

PIRET enables non-radiative energy transfer from plasmonic metals to semiconductors through near-field coupling, bypassing interfacial charge-injection barriers. This mechanism enhances exciton generation near the interface while preserving the electrochemical potential required for water-splitting reactions.

Plasmon-induced interfacial charge transfer transition (PICTT) is a distinct pathway in which plasmon decay directly excites electrons from a metal into an adjacent semiconductor, enabling instantaneous interfacial charge separation while bypassing multistep loss channels^[128]. This contrast to plasmon-induced hot electron transfer (PHET) is schematically illustrated at an Ag/TiO₂ junction [Figure 9A]^[129].

The concept of PICTT was first supported by ab-initio simulation studies. Long and Prezhdo employed time-domain density functional theory with nonadiabatic molecular dynamics to show that upon plasmon excitation in an Au-TiO₂ system, an electron can be injected into the TiO₂ conduction band almost instantaneously^[130]. Crucially, this occurs without requiring the intermediate formation and thermalization of carriers in the metal nanoparticle^[131]. A representative experimental manifestation is shown in Figure 9B, where Au nanoparticles induce efficient charge separation and exciton dissociation at the Au/MoS₂ interface, evidencing interfacial charge-transfer transitions in strongly coupled heterostructures^[132].

Building on this theoretical prediction, subsequent experimental work provided direct evidence for plasmon-induced metal-to-semiconductor interfacial charge transfer transitions and formally established the concept of PICTT^[130]. They studied CdSe semiconductor nanorods decorated with Au tips, a strongly coupled metal-semiconductor heterostructure, and observed highly efficient plasmon-induced electron injection from the Au into the CdSe^[133]. Interfacial charge transfer from Au plasmons into CdSe strongly damps the plasmon resonance and generates a substantial electron population in the semiconductor. The polarization dependence of this yield indicates that the orientation of plasmon oscillation relative to the interface governs the transfer process^[134].

Size and wavelength-dependent competition between PICTT and hot-electron-mediated pathways is quantitatively illustrated by the PHET quantum efficiency map, which shows that PICTT dominates for smaller Au nanoparticles and shorter excitation wavelengths, whereas hot-electron transfer becomes increasingly significant for larger particle diameters and longer wavelengths [Figure 9C]^[107].

Within plasmonic photoelectrochemical water splitting, PICTT directly couples plasmon decay to instantaneous interfacial charge separation, thereby bypassing thermalization losses associated with hot-carrier-mediated pathways. Through this direct charge-transfer channel, plasmonic nanostructures actively participate in driving water oxidation and reduction reactions rather than functioning solely as optical enhancers.

Quantum tunneling injection (QTI) arises when plasmon-generated hot electrons penetrate an ultrathin Schottky barrier via wavefunction tunneling, enabled by the non-thermal carrier population produced through Landau damping during LSPR decay^[135,136]. Compared with carriers generated by direct band excitation, plasmon-induced electrons exhibit broader energy distributions and enhanced out-of-plane momentum components, which favor barrier penetration and render QTI particularly effective in strongly confined Au or Ag nanostructures below 10 nm^[137,138]. The tunneling probability is further governed by interfacial engineering, as depletion-width compression through interfacial states, strain, doping, and dipole modulation enables sub-nanometer barrier profiles in metal-semiconductor heterojunctions^[139,140]. By extracting hot carriers before thermalization, quantum tunneling injections enable sub-bandgap photon utilization and directly enhance the IPCE and visible-light photocurrent of plasmon-modified TiO₂ photoelectrodes^{[108,141-144]}.

The scattering enhancement effect of plasmonic nanostructures is a key mechanism for improving the efficiency of PEC water splitting. The extinction cross-section (σ_ext) of plasmonic nanoparticles is defined as the sum of the absorption (σ_abs) and scattering cross-sections (σ_sca), where (σ_abs) contributes to hot-electron generation and resonant energy transfer, while (σ_sca) enhances light absorption by extending the optical path length within the semiconductor^[36,145]. According to Mie theory, σ_abs is proportional to a³ (where a is the particle radius), while σ_sca is proportional to a⁶, leading small nanoparticles to exhibit absorption-dominant behavior and larger particles to induce strong scattering^[34,146].

As particle size increases, retardation effects induce a transition from dipolar LSPR to Mie-type resonances, enabling strong Mie scattering^[42,146]. For instance, TiO₂ spheres with a diameter of approximately 380 nm exhibit up to a fivefold enhancement in photocatalytic H₂ evolution via improved light trapping^[147].

Periodic plasmonic nanostructures (e.g., gratings, nanohole arrays, and nanowires) enhance PEC performance by inducing scattering and coupling incident light into SPP modes that are re-scattered into guided modes within the semiconductor^[32,148]. In addition, lattice plasmon resonances (LPRs), particularly SLRs, provide high-Q, narrowband forward scattering, with nanorod and nanowire structures serving as representative scattering elements, as illustrated in Figure 10A^[149].

Hybrid architectures combining nanoparticles with nanostructures enable simultaneous optimization of absorption and scattering. For example, Au nanoparticles-decorated BiVO₄ inverse opal electrodes exhibited enhanced PEC activity through the synergistic broadband nanoparticle scattering and slow-photon effects from the inverse opal geometry^[150].

Scattering enhancement can be optimized by controlling particle size, morphology, and spatial arrangement^[151-154]. However, the PEC performance strongly depends on the balance between (σ_abs) and (σ_sca). In Au nanoparticle-modified BiVO₄ photoanodes, particles with a diameter of 30nm achieve nearly a twofold enhancement in photocurrent, whereas larger particles (60 ~ 80 nm) result in reduced performance^[155]. Overall, scattering enhancement is a design-governed mechanism in PEC systems, where optimal performance requires balanced control of absorption and scattering through particle size, morphology, and structural arrangement.

In plasmonic nanostructures, periodicity determines the coupling efficiency between neighboring dipoles and the coupling conditions of diffraction modes^[156]. Well-designed periodicity satisfies the phase-matching condition, thereby strengthening dipole-dipole coupling, reducing attenuation losses, and narrowing the resonance linewidth, leading to the formation of a high Q-factor and enhanced electromagnetic field localization^[157]. As schematically illustrated in Figure 10B, propagating SPP modes and LPRs arise from collective coupling in periodic plasmonic arrays.

Periodic arrays promote the hybridization of SPPs and LSPRs, thereby enhancing light absorption and hot-carrier injection efficiency^[158]. Moreover, periodic arrays enable precise and tunable control of plasmonic responses, allowing optimized periodicity to match the semiconductor bandgap with the resonance peak and thus maximize light-harvesting efficiency^[159,160]. The resulting strong electromagnetic fields induce spatial separation of photogenerated electron-hole pairs^[161].

However, excessively large periods may weaken field focusing or induce non-resonant modes, necessitating balanced designs and motivating a focus on multiscale and three-dimensional periodic plasmonic architectures that optimize light scattering, mode hybridization, and charge dynamics for PEC water splitting^[162-164]. In addition, the integration of refractory and complementary metal-oxide-semiconductor (CMOS)-compatible plasmonic materials (e.g., TiN, HfN) with periodic arrays would represent a meaningful direction for further advancement^[165].

As discussed above, conventional semiconductor photocatalysts are constrained by the SQ limit, narrow absorption range, and low quantum efficiency; however, LPR has recently emerged as a promising strategy to overcome these limitations^[166]. LPR is a collective resonance phenomenon that occurs in periodic arrays of metal nanoparticles that form a much narrower and stronger resonance than individual particles, providing a high Q-factor and strong electromagnetic field amplification^[167]. The unique properties of LPRs can significantly improve the photocatalytic performance of water-splitting reactions in many ways.

First, LPRs extend the light absorption into the visible and NIR, enabling more efficient utilization of photons beyond the intrinsic bandgap absorption limit of semiconductors. Deng et al. demonstrated that Cu-Pt nanoparticle lattices supporting SLRs exhibited up to 60% enhancement in hydrogen evolution reaction activity under white-light illumination compared to conventional LSPR-based catalysts^[168]. Second, the ultraviolet absorption efficiency of wide-bandgap semiconductors can be further enhanced. Recently, TiN nanodisk arrays have been reported to support SLRs with narrower linewidths and more than an order of magnitude higher Q-factors than conventional LSPRs, which can facilitate efficient charge generation and separation in the UV and visible regions^[54]. Third, the collective resonant coupling inherent in LPRs induces a strong electromagnetic field confinement, thereby markedly enhancing charge separation efficiency and suppressing carrier recombination. Li et al. reported a tenfold increase in photocurrent and up to an 18-fold enhancement in IPCE for a hematite nanorod array integrated with an Au nanohole lattice^[68].

LPR-based photoelectrodes hold strong potential for achieving commercially viable, high-efficiency hydrogen production through high-Q resonance control, heterojunction design with diverse semiconductors, and the use of low-cost metals and composite materials^[169], while future research is expected to further expand their commercial feasibility through materials innovation and scalable fabrication strategies.

Conventional PEC photoelectrodes are fundamentally constrained by their narrow spectral absorption and inefficient charge separation^[94]. To overcome these intrinsic limitations, the integration of photonic structures with plasmonic nanostructures has emerged as a promising next-generation strategy^[123]. Figure 10C illustrates various plasmonic-photonic hybrid structures. Photonic architectures enhance light harvesting, charge transport, and catalytic activity by modulating light propagation while plasmonic-photonic hybrid designs further exploit the slow photon effect, internal multiple reflections, and efficient light trapping^[170,171].

The historical development of photonic structures provides an essential framework for PEC applications [Figure 10D]. Specifically, this timeline highlights major breakthroughs: The emergence of the photonic bandgap concept^[172], widespread adoption of Bragg Mirror^[173], discovery of the Honda-Fujishima effect^[174], experimental realization of two-dimensional photonic crystals^[175], development of opal and inverse opal photonic crystals^[176], establishment of woodpile and gyroid architectures^[177,178], and recent advances in plasmonic-photonic hybrid PEC photoelectrodes^{[164,179-182]}.

To elaborate on these key developments, the concept of photonic structures originates from the theoretical work of Rayleigh in 1887, who described the formation of reflection bands in periodic multilayer dielectric media^[172]. This concept was later formalized as the photonic bandgap by Yablonovitch and Jon in 1987^[183]. Subsequently, one-dimensional (1D) multilayer Bragg reflectors became widely adopted in optical and laser devices from the 1970s onward as practical implementations of 1D photonic structures^[173]. In 1996, two-dimensional (2D) photonic crystals based on air-hole silicon slabs were experimentally realized at near-infrared wavelengths^[175]. Since the late 1990s, three-dimensional photonic crystals fabricated via colloidal self-assembly, including opal and inverse opal architectures, have been developed^[176] and were later extended to woodpile and gyroid structures. Woodpile structures enable precise bandgap control, whereas gyroid structures provide isotropic band gaps and enhanced light-matter coupling^[177,178].

The integration of photonic structures into PEC water splitting systems is a relatively recent development. Following the discovery of the Honda-Fujishima effect in 1972^[174], systematic incorporation of photonic architectures into PEC electrodes only began in the 2010s. Early demonstrations included Au-integrated TiO₂ nanotube photonic crystal photoelectrodes in 2013^[182] and Au-decorated Mo-doped BiVO₄ inverse opals in 2014, where strong coupling between plasmonic resonance and the photonic Bragg stopband significantly enhanced photocurrents^[181]. In 2017, three-dimensional FTO/TiO₂/BiVO₄ composite inverse opal photoanodes were developed, simultaneously enhancing charge transport, charge separation, and light harvesting^[180]. While more recently, in 2022, Au/TiO₂ 2D photonic crystal electrodes were reported, experimentally realizing diffractive-plasmon coupling for PEC hydrogen production^[179].

Future research is expected to focus on narrowband spectral matching between SPR and PBG, polarization control, precise structural alignment, and integration of optical absorption with catalytic functionality to maximize solar to hydrogen efficiency. Such plasmonic-photonic hybrid systems offer a promising design pathway for scalable and highly efficient PEC water splitting^[184].

The integration of nanostructured semiconductors into plasmonic systems has emerged as a potent strategy to amplify light-matter interactions and optimize the performance of PEC and photocatalytic processes^[200]. The unique size, shape, and surface-dependent properties exhibited by semiconductors at the nanoscale critically influence light absorption, charge carrier dynamics, and interfacial reactivity, thereby intensifying plasmon-induced phenomena^[201]. Among the most significant benefits is the quantum confinement effect^[202]. When the size of semiconductor particles approaches the exciton Bohr radius, discrete energy levels emerge, enabling tunable band gaps^[203]. For instance, CdS@CdS core-shell quantum dots enable precise spectral alignment with plasmonic resonances through size and thickness control^[204].

Moreover, nanostructures such as nanowires, nanorods, and mesoporous films induce light trapping via internal reflection and multiple scattering, which enhances photon absorption and increases optical cross-section^[203,204]. Another notable advantage of nanostructured semiconductors is their high surface-to-volume ratio, which increases the density of surface reaction sites, promotes plasmon-induced surface reactions, and shortens the diffusion length of photogenerated charge carriers^[205,206]. Furthermore, nanostructured semiconductors offer tunable defects and doping profiles^[207]. Engineering surface defects or introducing selective dopants allows for mid-gap state formation, bandgap narrowing, or enhanced visible light absorption^[208]. For example, Au-decorated BiVO₄ photoanodes with oxygen vacancies establish strong metal-support interactions with Au nanoparticles, facilitate hot electron injection, and reinforce interfacial coupling^[209]. Nanostructured semiconductors provide multifaceted advantages-enhanced light harvesting, improved charge separation, and reinforced interfacial reactivity-that collectively augment the performance of plasmonic architectures.

Ordered and disordered LSPR-based nanostructures constitute two complementary plasmonic design strategies for PEC water splitting. Ordered LSPR-based nanostructures are defined as architectures in which metallic nanoparticles or nanoholes are periodically arranged to induce coupling between individual LSPR modes and diffractive orders^[210]. Such periodicity gives rise to SLRs, with Q-factors on the order of several thousand, thereby enabling high-quality, wavelength-selective resonances. [Figure 13A]^[211]. In addition, ordered Au arrays fabricated on metal oxide photoelectrodes strongly amplify the near-field, suppressing electron-hole recombination and optimizing charge separation and injection pathways [Figure 13B]^[212]. These mechanisms manifested in practical PEC performance metrics, where ordered arrays exhibit several-fold enhancements in photocurrent density and IPCE over disordered counterparts, particularly at selected spectral regions [Figure 13C]^[200].

For instance, a system combining Au nanohole arrays with α-Fe₂O₃ nanorod electrodes exhibited IPCE enhancements of approximately 18-fold in the sub-bandgap region and 13-fold near 425 nm^[212], while ordered Au metasurfaces demonstrated ultrahigh-quality resonances with Q ≈ 2340, which critically contributed to electromagnetic field enhancement and photocurrent amplification^[211].

In contrast, disordered LSPR nanostructures refer to disordered metasurfaces in which particle size, shape, and spacing are non-uniform or randomly distributed^[213]. In two-dimensional arrays, non-uniform distributions can be generated through self-assembly techniques such as random sequential adsorption (RSA), while correlated disorder designs enable the introduction of local irregularities while preserving the average interparticle spacing, allowing continuous tuning of scattering responses between periodicity and disorder^[214-216]. The coexistence of variations in size, morphology, and spacing induces broadband plasmonic resonances, thereby enhancing absorption across the entire visible region (450-750 nm) [Figure 13D]. Au nanoparticles formed on TiO₂ via thermal dewetting exhibited size-dependent SPR peaks; nevertheless, the PEC photocurrent consistently appeared near 600 nm, indicating a broadband response^[217]. Moreover, random arrangements generate electromagnetic hotspots at unpredictable locations, as evidenced by long-lived localized plasmon modes confined within 10-20 nm pores of dealloyed Au-Ag nanosponges [Figure 13E]^[218]. Unlike ordered structures that provide sharp enhancements at specific wavelengths, disordered configurations contribute to a stable enhancement of the photocurrent baseline across the broadband region [Figure 13F]^[219].

In conclusion, ordered and disordered LSPR-based nanostructures provide complementary plasmonic enhancement strategies for PEC water splitting. Ordered architectures enable high Q-factors, wavelength-selective field enhancement, whereas disordered configurations offer broadband absorption and stable photocurrent amplification. Together, these design paradigms underscore the importance of tailoring plasmonic nanostructures for efficient solar-driven hydrogen production.

In PEC water-splitting, SPPs constitute a key concept underlying various nanostructures designed to maximize light absorption and charge transfer and can be broadly classified into ordered and disordered architectures that exhibit fundamentally different light-matter interaction regimes^[43]. Ordered SPP nanostructures employ periodic metallic patterns to selectively excite propagating SPP modes, resulting in narrow resonance bandwidths, high Q-factors, and wavelength-selective, stable optical responses^{[149, 220]}. In contrast, disordered SPP nanostructures, composed of non-periodically distributed metal nanoparticles or clusters, support broadband SPP-like modes with probabilistically distributed electromagnetic hotspots, enabling strong near-field localization over a wide spectral range^[221-223].

These distinct interaction regimes translate directly into different PEC performance characteristics. Ordered architectures, exemplified by Au nanohole/hematite (α-Fe₂O₃) nanorod photoelectrodes, exhibited over an order-of-magnitude enhancement in photocurrent density and pronounced IPCE improvements at specific wavelengths due to SPP-induced waveguide modes^[68]. Conversely, disordered Ag/TiO₂ nanoplates and MgB₂ nanosheet assemblies achieved broadband photocurrent enhancement, ~ 5% conversion efficiency, and stable hydrogen^[224]. Rather than representing opposing strategies, periodicity and randomness should be regarded as complementary design parameters that can be selectively employed or combined depending on targeted PEC performance objectives.

Density functional theory (DFT) constitutes a pivotal tool in the study of plasmonic photoelectrodes, enabling atomic-scale elucidation and quantitative interpretation of electronic structures, interfacial charge transfer, electron density distributions at active sites, adsorption energies, and reaction pathways^[225]. In particular, by providing insights into hot electron injection, band alignment modulation, and the stability of surface intermediates at plasmonic metal-semiconductor interfaces, DFT establishes the fundamental basis for enhancing PEC efficiency^[226]. These insights are typically obtained through analyses of representative reactive units within PEC electrodes, including metal-semiconductor nano-heterojunctions^[226,227], alloy-based nanoparticles^[228], and nanocluster-molecule interfaces^[229].

In the Au/Fe₂O₃ interface, DFT calculations revealed that Au clusters bound to the Fe₂O₃ (001) surface generate new electronic states near the conduction band minimum (CBM), facilitating the migration of photoexcited electrons into the Fe₂O₃ conduction band^[226]. Moreover, for Au nanocluster-H₂O gap structures, linear-response time-dependent DFT (LR-TDDFT) calculations demonstrated that dual-cluster gaps induce stronger electromagnetic field enhancement and direct charge transfer^[230]. In Ag-Au alloy/TiO₂ electrodes, DFT identified (Ag_0.6Au_0.4) as the optimal composition for maximizing electron injection and OER activity^[231]. Thus, DFT functions as a central pillar in the next-generation design of plasmonic nanostructure-based PEC hydrogen production systems, delivering its greatest impact when employed synergistically with experimental validation^[232].

Artificial intelligence (AI) has become a powerful tool for the design and optimization of plasmonic nanostructures in PEC water splitting, addressing the complexity of light absorption, charge recombination, interfacial defects, and nonlinear structure-performance relationships that limit traditional synthesis and physics-based simulations^[233-237]. AI-driven methodologies, particularly machine learning (ML) and deep learning (DL), overcome these limitations by providing rapid predictive capabilities^[238,239].

Deep neural networks (DNNs), trained on large simulation datasets, have learned the intricate correlations between structural parameters and optical responses, predicting near-field electric distributions and far-field scattering spectra with 3-5 orders of magnitude faster computation (10³ ~ 10⁵ times) compared to conventional FDTD or FEM simulations^[240]. Reinforcement learning (RL) and data-efficient strategies such as semi-supervised learning (SSL) further enhance structural optimization in limited-data regimes, balancing performance objectives with experimental or computational costs^[241,242].

Beyond data-driven models, physics-aware frameworks that integrate DFT and ab-initio molecular dynamics (AIMD) with ML offer accurate predictions of hot-electron generation, band alignment, and interfacial charge-transfer dynamics at plasmonic-semiconductor junctions^[243-245]. Representative demonstrations include ML-based prediction of plasmonic coupling from optical color^[246], AI-assisted atomic-level cluster growth control^[247], and inverse design of broadband metamaterials^[248]. Overall, AI-assisted design frameworks overcome the computational limits of conventional simulations by integrating physics-based insights with data, enabling optimal plasmonic nanostructure design for high-performance PEC water splitting.

In Figure 14A, kelvin probe force microscopy (KPFM) mapping directly visualizes a pronounced surface photovoltage (SPV) at a dimeric Au-TiO₂ junction under illumination, which originates from LSPR-induced hot electron injection from Au into TiO₂ across the Au-TiO₂ Schottky interface. In contrast, the complementary hot holes remain confined within the Au nanoparticle^[249,250]. This SPV response is particularly enhanced in closely spaced Au-TiO₂ dimers, where strong plasmonic near-field coupling within the nanogap generates intense local electromagnetic fields, leading to increased hot-carrier generation and more efficient interfacial charge separation compared to isolated nanoparticles^[251,252].

Figure 14B extends this principle to catalyst ensembles, showing that spatially coupled Au assemblies such as linear nanochains generate stronger near-field enhancement and more efficient hot-electron transfer than isolated nanoparticles, yielding nearly 3.5-fold higher visible-light-driven H₂ evolution on Zn_0.67Cd_0.33S^[114,252]. This spatial-design principle is general, as Au-decorated TiO₂ photoelectrodes also exhibit enhanced oxygen evolution under visible illumination, particularly when combined with a p-type NiO layer that facilitates hole extraction and suppresses recombination^[253-255].

Figure 14C illustrates device-level engineering in Au/NiO/TiO₂-based photoanodes, where the photocurrent increases with Au loading and reaches a maximum at an optimal coverage (Au30/NiO/TiO₂)^[255]. In this tri-layer architecture, Au nanoparticles enhance visible-light absorption via LSPR and inject plasmonically generated hot electrons into TiO₂, while NiO (or NiO_x) facilitates hole extraction and suppresses charge recombination, collectively leading to enhanced photocurrent generation^[256]. Similar synergistic effects are observed in bimetallic Au-Ni cocatalyst systems integrated with TiO₂ nanotube photoanodes, where intimate Au-Ni-TiO₂ interfacial contact enables directional charge separation and yields significantly improved photocurrent density and stability compared with single-metal photoanodes^[257].

Figure 14D emphasizes the importance of nanoscale spatial engineering in plasmon-enhanced photoelectrochemical systems. Periodic Au nanodot arrays with ~ 10 nm interdot spacing form strongly coupled plasmonic hotspots at the Au-TiO₂ interface, generating concentrated electromagnetic fields that enhance local carrier generation via near-field coupling^[114,256]. Consequently, TiO₂ electrodes patterned with dense, ordered Au nanodot arrays exhibit an order-of-magnitude higher photocurrent than those with randomly dispersed Au nanoparticles^[114,258]. In contrast, increased interdot spacing or random nanoparticle distributions weaken near-field coupling and diminish this enhancement, underscoring that precise spatial arrangement-beyond simple loading-is critical for maximizing plasmon-induced charge separation and photoelectrochemical efficiency^[114,259]. In addition to structural engineering, heteroatom doping and oxygen vacancy creation in TiO₂ have also been explored to induce LSPR-like behavior. These strategies introduce free charge carriers or mid-gap states, facilitating visible-light absorption and plasmonic charge dynamics. Notably, such effects have been demonstrated even in traditionally non-plasmonic semiconductors such as TiO₂. These material-level strategies align with recent efforts focusing on plasmonic field engineering and spatial coupling to maximize hot-electron utilization in PEC water splitting^[260,261].

Figure 15A illustrates plasmon-driven charge transfer in an Au-g-C₃N₄ system suspended in triethanolamine (TEOA), where LSPR excitation in Au nanoparticles generates hot electrons injected into the g-C₃N₄ conduction band, while hot holes are scavenged by TEOA, enabling Au to act as both an optical antenna and an electron shuttle for H₂ evolution. Under visible-light illumination (> 420 nm), the Au-g- C₃N₄ composite exhibits nearly an order-of-magnitude higher hydrogen evolution than bare g- C₃N₄, and this activity is retained within a narrowed excitation window of 420 ~ 465 nm, extending into the red-edge region (450 ~ 600 nm) where pristine g-C₃N₄ is inactive, thereby confirming a plasmon-driven rather than band-gap-driven mechanism^[95,262].

Figure 15B demonstrates that alloying Au with Pd, Cu, or Ag systematically tunes plasmonic charge-transfer energetics through changes in the E_CB ~ E_F offset and intrinsic optical constants, highlighting alloy composition as a critical design parameter. Increasing Pd content enlarges the E_CB ~ E_F offset to ~ 1 eV, favoring efficient hot-electron injection, whereas Au-Cu alloys exhibit a plateaued offset and Ag-rich alloys collapse the driving force, thereby quenching plasmon-assisted charge extraction^[263]. These trends originate from alloy-induced modifications in the real and imaginary permittivity, which shift interband transition thresholds and broaden plasmon resonance linewidths, demonstrating that intrinsic optical properties-beyond nanostructure geometry-critically govern plasmonic charge-transfer energetics and must be explicitly considered in plasmonic catalyst design^[264].

Figure 15C compares the extinction behavior and plasmonic functionality of three Au-Pd architectures-antenna-reactor dimers, core-shell particles, and atomically mixed alloys-showing that nanoscale architecture critically governs near-field localization, hot-carrier generation, and photocatalytic performance. Antenna-reactor dimers exhibit a sharp, red-shifted LSPR (~ 780 nm) due to strong dipolar coupling at the Au-Pd junction, generating intense localized near-fields on Pd sites and yielding the highest H₂ evolution activity^[265,266]. In contrast, core-shell structures suffer plasmon damping from Pd shells, while fully alloyed Au-Pd nanoparticles experience stronger nonradiative losses that suppress near-field enhancement, resulting in lower activity^[263,266]. Accordingly, photocatalytic performance follows the hierarchy antenna-reactor > core-shell > alloy, highlighting the importance of rational plasmonic-catalytic junction design for efficient hot-carrier utilization.

Figure 15D shows that increasing the number of embedded Au cores in Au@Cu_1.9S photocatalysts markedly enhances plasmon-driven H₂ evolution, with a five-core yolk-shell structure reaching ~ 115 μmol g^-1 h^-1 after 6 h, while single-core, Au-free, or physically mixed Au-Cu_1.9S systems remain far less active^[267]. This enhancement stems from the increased density of plasmonic excitation centers and expanded metal-semiconductor interfacial area, enabling cumulative hot-electron injection and strong interfacial field localization dictated by the core-shell geometry, thereby demonstrating that intimate nano-heterojunction embedding is essential for efficient plasmon-driven catalysis^[267].

Figure 15E illustrates Simulated near-field enhancement (|E/E₀|) shows that interfacial architecture governs plasmonic energy localization, with Au@Cu₂Se exhibiting a > 5-fold field enhancement at ~ 700 nm (LSPR), whereas Au@Cu₂O displays only a modest response near 550 nm^[268]. Spatial field maps reveal that Au@Cu₂Se yolk-shell nanoparticles form an intense ring-shaped hotspot at the Au-Cu₂Se interface, in contrast to delocalized fields in Cu₂Se and metal-confined hotspots in Au, thereby enabling efficient hot-electron transfer into the Cu₂Se shell and accounting for the superior charge utilization and highest H₂ quantum efficiency^[267,268].

Figure 15F shows that plasmon-driven H₂ evolution is negligible in the dark but strongly enhanced under illumination in Au-Pd satellite assemblies, which produce ≈ 8 nmol g^-1 Pt h^-1 about four times higher than Au@Pd core-shell catalysts^[269]. This enhancement arises from intense plasmonic hotspots in Au-Pd nanogaps and discretely partitioned Pd islands that provide accessible H-adsorption sites while avoiding plasmon damping, enabling efficient hot-carrier transfer from Au to Pd^{[263, 269]}. In contrast, continuous Pd shells damp the Au LSPR and suppress field localization, underscoring cocatalyst partitioning as a key structural parameter for efficient plasmon-driven hydrogen evolution^[270]. These findings are in line with recent comprehensive reviews on single-atom plasmonic hybrids and multichannel charge transfer schemes for bond activation^{[271, 272]}.

Figure 16A shows that loading plasmonic or transition-metal cocatalysts enables wide-bandgap TiO₂ to operate under visible light, where bare TiO₂ is inactive. Among various metals, Cu/TiO₂ exhibits the highest H₂ evolution via efficient hot-electron injection and suppressed recombination, as small Cu⁰/Cu⁺ clusters act as electron sinks and hole scavengers while extending visible absorption through interfacial charge-transfer states. Consequently, Cu/TiO₂ achieves H₂ evolution rates ~ 16× higher under UV and ~ 3× higher under visible light than pristine TiO₂, whereas Fe-, Zn-, and Ce-modified TiO₂ show negligible enhancement^[273].

Figure 16B demonstrates a hierarchical plasmonic-MOF (metal organic framework) composite strategy in which a plasmonic Au core embedded in a porous MOF scaffold [e.g., Materials of Institute Lavoisier (MIL)-101(Cr)] enables efficient light harvesting and charge management. Under visible-NIR illumination, LSPR excitation in Au generates hot electrons that are injected into the MOF framework or Pt cocatalyst sites, leading to markedly enhanced H₂ evolution compared with bare Au, with further improvement upon incorporation of a second metal (Ag or Pt)^[274]. In particular, Au/Pt-MOF architectures sustain strong plasmon-polariton modes and stabilized interfaces that facilitate hot-electron transfer, broaden solar utilization into the NIR region, and deliver superior activity and durability relative to Au- or Pt-only systems^[274,275].

Figure 16C demonstrates the extension of plasmonic design to integrated photoelectrodes using a ternary heterojunction photoanode comprising n-type α-Fe₂O₃, a p-type Co-based MOF overlayer [zeolitic imidazolate framework (ZIF)-67], and interfacial plasmonic Ag nanoparticles. Ag LSPR enhances visible-light absorption, while the Fe₂O₃/ZIF-67 p-n junction drives directional charge separation and suppresses recombination, resulting in the highest photocurrent and improved stability in current density-potential (J-V and chopped-light measurements. These results highlight that plasmonic heterojunction engineering effectively enhances charge separation and oxygen evolution in water-splitting photoelectrodes^[276].

Figure 16D presents the dynamic on/off photoresponse of a plasmonic Z-scheme photocatalyst under chopped illumination, showing that the ternary titanate nanosheets (TNSs)-Au-g-C₃N₄ composite delivers the highest and most stable photocurrent among all configurations^[275]. This enhancement arises from Au-mediated Z-scheme charge transfer, which directs electrons to the reduction side for H₂ evolution while retaining holes on the oxidation side for O₂ evolution, thereby confirming the effectiveness of interfacial plasmonic mediators in promoting both half-reactions^[275].

Figure 16E compares ZnO nanorods with uniform and gradient Au doping, showing that spatially controlled dopant distributions critically influence plasmonic photoelectrode performance. In gradient Au-ZnO, surface-enriched Au maximizes near-field overlap with the photoactive region and strengthens the built-in electric field, yielding the highest photocurrent in J-V measurements by enhancing directional charge separation and suppressing recombination. These results demonstrate that dopant-gradient engineering effectively couples plasmonic field localization with internal electric-field control to improve solar-driven hydrogen and oxygen evolution^[277]. Moreover, such strategies align with broader approaches used to induce plasmonic behavior in wide-bandgap metal oxides like TiO₂ and ZnO, where heteroatom doping or the creation of oxygen vacancies introduces mid-gap states or localized free carriers. These modifications facilitate visible or near-infrared LSPR activation, thereby enhancing light absorption and hot-carrier generation even in originally non-plasmonic hosts.

Figure 16F links photoluminescence metrics and nanoscale coupling to plasmonic photocatalytic performance, showing that higher band-edge-to-trap-state photoluminescence (PL) spectroscopy intensity ratios-achieved via surface passivation-indicate reduced recombination, prolonged carrier lifetimes, and improved hydrogen and oxygen evolution^[278]. In addition, shrinking metallic nanogaps induces hybridized bonding and antibonding plasmon modes whose intensity ratios track near-field coupling strength, as visualized in gold nanoparticle superlattices, where sub-nanometer gaps yield extreme field confinement and enhanced hot-carrier generation that promotes efficient interfacial electron injection^[279].

Figure 16G illustrates plasmon-induced ultrafast charge-transfer dynamics in an Ag₁₉Pt-H₂O cluster, where a single Pt atom introduces multichannel electron-transfer pathways that rapidly populate the LUMO from multiple occupied orbitals upon photoexcitation. These parallel transfer routes drive O-H bond dissociation within ~ 40 fs, whereas the pure Ag₂₀-H₂O cluster lacks such dynamics and shows no ultrafast bond cleavage, highlighting the decisive catalytic role of Pt incorporation. Overall, plasmon excitation in Ag-Pt single-atom systems enables not only hot-electron injection but also multichannel charge-transfer processes that accelerate bond activation^[280]. These findings resonate with emerging strategies utilizing single-atom sites to control multichannel electron transfer for plasmon-activated bond cleavage, as discussed in recent high-level reviews^[271,272].

Here, a summary of current challenges in plasmonic solar water splitting is given. First, there is no unified mechanism of plasmonic photocatalysis; therefore, theoretical ultimate limit is not clear yet. Therefore, in the development, there are ambiguities in the performance. Second, there is an uncertainty in the morphological optimization of plasmonic components as well as their arrangement. As overall plasmonic enhancement comes from the combination of the near-field, diffractive, far-field (Mie-like), and photonic hybridization effects, it is hard to achieve sufficient optimization. For development, it costs more time than other systems. Finally, as efficient interfacial charge-transfer and corresponding damping issues exist, the interfacial chemical structure is essential. However, in the plasmonic system, optimizations are not certain, which hinders the development of plasmonic-enhanced catalysis systems.

By achieving full optimization in the future, which leads to overcoming challenges, the following things are expected as benefits. The conventional band edge-limited absorption would be conquered. As a result, a broadband absorbable photoelectrode for covering AM 1.5G could be realized. This is consistent with the previously suggested theoretical expectations that bypass the SQ limit, which indicates an economically efficient water-splitting system. Partial realization in literature gives more potential to this argument. Moreover, the next-generation hydrogen society could get closer. Besides, the development of the plasmonic photoelectrode stimulates development in adjacent fields, such as solar-to-fuel conversions and sensing.