Plasmon‐Mediated Solar Energy Conversion via Photocatalysis in Noble Metal/Semiconductor Composites

Plasmonics has remained a prominent and growing field over the past several decades. The coupling of various chemical and photo phenomenon has sparked considerable interest in plasmon‐mediated photocatalysis. Given plasmonic photocatalysis has only been developed for a relatively short period, considerable progress has been made in improving the absorption across the full solar spectrum and the efficiency of photo‐generated charge carrier separation. With recent advances in fundamental (i.e., mechanisms) and experimental studies (i.e., the influence of size, geometry, surrounding dielectric field, etc.) on plasmon‐mediated photocatalysis, the rational design and synthesis of metal/semiconductor hybrid nanostructure photocatalysts has been realized. This review seeks to highlight the recent impressive developments in plasmon‐mediated photocatalytic mechanisms (i.e., Schottky junction, direct electron transfer, enhanced local electric field, plasmon resonant energy transfer, and scattering and heating effects), summarize a set of factors (i.e., size, geometry, dielectric environment, loading amount and composition of plasmonic metal, and nanostructure and properties of semiconductors) that largely affect plasmonic photocatalysis, and finally conclude with a perspective on future directions within this rich field of research.


Introduction
As an optical phenomenon, surface plasmons were fi rst demonstrated by Michael Faraday in 1857. [ 1 ] However, they did not attract tremendous attention from scientists at the time. [ 2 ] Recently, in conjunction with nanoscale science and technology, REVIEW photo-corroded and cannot maintain long-term catalytic performance. [ 15 ] Others, such as Fe 2 O 3, capable of utilizing visible light suffer from short exciton diffusion length (2-20 nm) and extremely low photocatalytic activity. [ 16 ] To address these material limitations, plasmonic photocatalysts have emerged. The combination of a noble metal (i.e., gold and silver) and semiconductor signifi cantly enhances the photoreactivity due to two main features: (i) Schottky junctions and (ii) LSPRs. The prominent advantages of plasmonic photocatalysts are displayed in Figure 2 , and discussed in the second part of this review. A Schottky junction (i.e., Schottky barrier) at the interface of a metal and semiconductor drives electrons and holes induced in or near this area to move in opposite directions. [ 7a ] Noble metals can harness light of long wavelengths to produce 'hot electrons' and facilitate quick charge transfer, [ 7b , 17 ] for example, from Au to TiO 2 . [ 18 ] For LSPRs, plasmon resonance wavelengths in the visible and infrared (IR) regions enable the improved utilization of the solar spectrum. [ 19 ] The enhanced local electric fi elds caused by plasmon effects raise the possibility of electron-hole pair generation. [ 20 ] Furthermore, the increased current density can lead to heat generation which may contribute to faster photochemical reactions. [ 21 ] After a brief discussion that motivates the study of plasmonic photocatalysis as presented above, this review aims to summarize impressive recent developments in the photocatalytic mechanisms of plasmonic photocatalysts and recent progress in different factors affecting metal/semiconductor photocatalysts. A perspective on the future development of plasmonic photocatalysts with rational material combination and improved performance is also provided.

Photocatalytic Mechanisms
The photocatalytic process generally contains three major steps: (i) adsorption of reactants on the surface of the photocatalyst; (ii) absorption of incident photons and creation of electronhole pairs; and (iii) photocatalytic reactions. Concurrently, unfavorable recombination of charge carriers in the bulk or at the surface of photocatalyst occurs, and an ongoing challenge is to minimize these undesirable reactions. The synergistic effect between the semiconductor and noble metal leads to pronounced improvements in photocatalytic effi ciency, contributing largely to these three steps. Noble metals, such as Au, display signifi cant UV light absorption via an electron interband transition from 5d to 6s and 6p, which favors the utilization of light [ 22 ] Additional major benefi ts favoring the generation and separation of induced eletron-hole pairs in such systems are detailed in Figure 2 . Other effects may also positively infl uence the photocatalytic process, including the catalytic effect of the noble metal itself (i.e., Pd for hydrogen generation [ 23 ] ). In this section, we concentrate on the discussion of the favorable synergistic effect of noble metals and semiconductors on photocatalysis.

Schottky Junction
When placed in intimate contact, Au and n-type TiO 2 illustrate a typical Schottky junction. TiO 2 possesses oxygen vacancies and consequently excess electrons, which gives it n-type behavior. [ 24 ] Generally speaking, noble metals have a high work function. The Fermi level of Au is located below that of n-type TiO 2 ( Figure 3 a), and Schottky barrier is formed after Au and TiO 2 contact each other (Figure 3  His research interests include electrochemical methods, corrosion and protection, biomaterials and biosurfaces, advanced materials for energies and environments, such as dye-sensitized and quantum dots-sensitized solar cells, photogenerated cathodic protection, surface wettability, water splitting hydrogen production and photocatalytic degradation of organic pollutants.

Mengye Wang received her
PhD in physical chemistry from Xiamen University, Xiamen, China in 2015. Her research interests include advanced materials for environmental and energy application, such as photocatalytic degradation of organic pollutants, water splitting and dye-sensitized solar cells.
and TiO 2 , electrons diffuse from TiO 2 to Au, and a positively charged region is then formed in TiO 2 . Thus, a space charge region with no free carriers is generated. Meanwhile, a negatively charged region is formed close to Au. Hence, an internal electric fi eld is built up from TiO 2 toward Au. When excited by incident visible light, the internal electric fi eld will drive photogenerated electrons near or in the space-charge region to move to TiO 2 and holes to Au, thereby preventing electrons and holes from recombination. Subsequently, electrons and holes will participate in photocatalytic reactions. As a result, electrons and holes at the surface of TiO 2 and Au participate in photocatalytic redox reactions separately.

Direct Electron Transfer
This phenomenon is also known as the LSPR sensitization effect. [ 7a ] The combination of Au and n-type TiO 2 is again used as an example. In the case of direct electron transfer, the noble metal (Au) and semiconductor (TiO 2 ) are in intimate contact. As shown in Figure 3 b, when Au comes into contact with TiO 2 , the equilibration of Fermi levels causes the bending of the conduction band of TiO 2 at the interface and a Schottky barrier is thus formed. Usually, if the metal nanoparticle is suffi ciently large (i.e., above quantum confi nement) and cannot induce size-dependent bandgaps, the electron states of the metal are continuous and follow the Fermi-Dirac distribution. [ 7a ] Upon illumination, if the plasmon band overlaps the interband transition of Au, electrons oscillating collectively may create a uniform probability for electrons to transfer to the energy level between E f (Fermi level energy of Au) and E f + hν . [ 25 ] Due to electron-electron scattering, electron energy is redistributed and creates a non-equilibrium Fermi-Dirac distribution within 10 femtoseconds. [ 18,25 ] Meanwhile, electrons transfer from the noble metal (i.e., Au) to the semiconductor (i.e., TiO 2 ) through two different ways. The fi rst one is a coherent process. After excitation, the generated electrons directly inject into the

Enhanced Local Electric Field
When illuminated, the coherent free electrons of noble metals oscillate to produce a local electric fi eld. If the incoming light is around the plasmon resonance frequency of the noble metal, the incident photons are absorbed. Due to the electron cloud oscillation, the noble metal acts as a re-emitting light at the same frequency. The excited plasmonic noble metal not only scatters some of this emitted light but also behaves as a concentrator for the local fi eld. [ 26 ] Thus, the intensity of the local electric fi eld is further boosted. As a consequence, the interband transition rate of the semiconductor is radiatively increased (if the energy generated by LSPR is higher than that of the semiconductor band gap) by the enhanced local electric fi eld ( Figure 5 a). [ 27 ] As such, photo-activity of the photocatalyst is improved. It is known that the electric fi eld strength decays as a function of distance from the noble metal (Figure 5 b). [ 28 ] Therefore, enhanced local electric fi elds exist only within a certain distance from noble metal.

Plasmon Resonant Energy Transfer
Different from the enhanced local electric fi eld mechanism, plasmon resonant energy transfer refers to a non-radiative dipole-dipole energy transfer process ( Figure 6 ). [ 27b ] During this process, the dipole moment of the electron-hole pair of the semiconductor (e.g., Cu 2 O) is coupled with the plasmonic dipole moment of the noble metal (e.g., Au). [ 27b ] Through the relaxation of the localized surface plasmon dipole, electrons and holes are generated in the semiconductor.   . Direct electron transfer mechanism. a) Electrons are excited to higher energy states under illumination and are fed to the conduction band of the n-type semiconductor (e.g., TiO 2 ). b) The energy states of the noble metal are redistributed to form a Fermi-Dirac distribution via collision but at a higher Fermi level. c) Electron states revert back to the standard Fermi-Dirac distribution. Reproduced with permission. [ 7a ] Copyright 2013, Institute of Physics. the plasmon resonance and the conduction band. The transfer rate k transfer can be expressed as Equation ( 1) .
where τ noble metal is the donor lifetime in the absence of the acceptor, and r is the distance between the noble metal and the semiconductor. R 0 can be expressed as Equation ( 2) .
where K is an orientation factor and usually equal to 2/3, n is the refractive index, ϕ 0 is the quantum yield of the donor, and J represents the normalized overlap integral between the noble metal spectrum F D and the semiconductor spectrum (Equation ( 3) ).
A clear understanding of the mechanisms described above can be achieved through comparison. (A) The direct electron transfer mechanism is limited by the intimate contact between electron donor and acceptor, as well as electronic band alignment between the metal and semiconductor. In addition, electron-hole consumption equilibrium is also necessary in the photocatalytic process for direct electron transfer. (B) The enhanced local electric fi eld mechanism is only applicable to the radiative process when the energy created by LSPR is above that of the semiconductor band gap. [ 27b ] (C) If the band gap of the semiconductor overlaps the LSPR band of the noble metal, the resonant energy transfer process can introduce charge carriers in the semiconductor at energies both above and below the band gap due to non-radiative coupling (unlike enhanced local electric fi elds) with optically inaccessible and optically ineffi cient states at the band edge.
[ 27b ] However, it is not affected by whether or not the semiconductor intimately contacts the noble metal or there is a charge consumption equilibrium problem (unlike direct electron transfer) as the plasmon energy transfer occurs via a near fi eld electromagnetic interaction.

Scattering and Heating Effects
The scattering mechanism is typically relevant for large plasmonic photocatalysts whose diameter is usually larger than 50 nm. In addition to absorbing the incident photons, these large plasmonic nanostructures scatter the incident light to raise the light utilization probability. As shown in Figure 7 , the diameter of the Au nanoparticles should be larger than 100 nm in order to maximize the scattering effect in photocatalysis. [ 7b , 29 ] A heating effect is another consequence of the plasmonic phenomenon. A portion of energy produced by resonant oscillation is dissipated into the metal. However, this part of the electromagnetic energy will be consumed as follows. The negative change rate in electromagnetic energy within the volume of the noble metal equals the energy which fl ees through the boundary surfaces of the volume V per unit time plus the energy loss rate by absorptive dissipation. [ 30 ] Typically, there is no magnetically dispersive medium. Thus, the time-averaged dissipative energy density can be written as Equation ( 4) . [ 30 ] where ε ω ε , and 0 represent the permittivity of free space, resonance frequency, and relative dielectric permittivity of the medium, respectively. The energy of the absorbed light is converted to heat (Equation ( 5) ).
In general, the generation of heat by this effect is comparatively small and cannot signifi cantly alter the temperature of the photocatalytic system. [ 31 ]

Noble Metal/Semiconductor Material Systems for Plasmonic Photocatalysis
For most plasmonic photocatalysts, the noble metal serves as a sensitizer enabling the absorption of visible light by capitalizing on LSPR. Greater utilization of the solar spectrum is one of the advantages of plasmonic photocatalysts, which enables improved photocatalytic performance. In this section, the optical properties of plasmonic photocatalysts altered by modulating the size, geometry and surrounding environment of noble metals are discussed. Some other effects that mediate photoactivity are also included in this section.

Size Effect
The collective charge oscillations are limited by the size of noble metal nanostructures due to surface confi nement. [ 32 ] As the size of the nanostructure increases, charge separation is promoted. Thus, a lower frequency is required for the collective electron oscillation, which can be seen by the red-shift of the plasmonic peaks in the spectrally resolved absorption. [ 33 ] By tailoring the size of the noble metal particles, the light response triggered by plasmon resonance can be varied from the visible to near-infrared region. [    c) The corresponding simulated electric-fi eld intensity for Au nanoparticle-coupled TiO 2 nanowires and Au nanorod-decorated TiO 2 nanowires, respectively. (d) IPCE plots of Au nanoparticles, Au nanorods, and their mixture. Reproduced with permission. [ 35 ] Copyright 2013, American Chemical Society.
onto TiO 2 nanowire arrays, the absorption of the photocatalysts in the entire UV-Visible region from 300 to 800 nm can be effectively enhanced, resulting in an increased incident photo nto-electron conversion effi ciency (IPCE) ( Figure 8 ). [ 35 ]

Geometry Effect
In addition to size, the geometry of plasmonic metal nanostructures exerts a non-trivial infl uence on plasmonic photocatalysts. [ 9a , 36 ] During coherent electron oscillation, nanostructures with sharp corners facilitate charge separation more easily than rounded structures. This reduces the restoring force for the electron oscillation, leading to a longer resonance wavelength. [ 36a ] Additionally, the low-symmetry nanostructures favor enhanced resonances as the asymmetry enables the electrons to be polarized in more than one way. [ 37 ] Zhou et al. designed Ag@c-Si core@shell nanocones for the enhancement of broadband solar absorption ( Figure 9 a). [ 38 ] The metallic core assures strong near fi eld enhancement, providing the increased absorption effi ciency. [ 39 ] Meanwhile, the special nanocone-like structure contributes to a linear gradient through the radius of the c-Si shell. This leads to the production of multiple plasmon resonances and, consequently, enhances the light absorption. [ 40 ] The normalized absorption effi ciency η 〈 〉 abs c-Si AM 1.5G is defi ned as Equation ( 6) . [ 38,41 ] where η abs c-Si , λ , and λ ( ) F are the absorption effi ciency of c-Si, the wavelength of the incident light, and the photon fl ux density in the AM 1.5G solar spectrum, respectively. η abs c-Si is expressed as η = / abs c-Si abs geo C C , where and abs geo C C are the absorption and geometrical cross sections of nanocone, respectively. It is clear that the steeper the nanocone (larger ratio of r 2 /r 1 ) is, the more effi cient the light absorption would be (Figure 9 b). In addition, there is an optimal nanocone length (i.e., L z in Figure 9 b) as the short L z possessed relatively less optically confi ned modes; and when L z is beyond a certain value, the L z becomes the major infl uence on geo C rather than on abs C . [ 38 ] By optimizing these parameters, the nanocone-based geometry is predicted to possess improved photocatalytic performance.

The Infl uence of the Surrounding Dielectric Environment
When LSPR are produced, the incident photons are coupled into surface plasmons, which can be observed as peaks in the absorption spectrum. [ 42 ] For plasmonic photocatalysts, one of the most interesting characteristics is the absorption wavelength caused by LSPR effect. The plasmon resonance frequency, ω, can be written as [ 7a ] ω ω ε ≈ + 1 2 p m ( 7) where ω p represents the resonance frequency of bulk metal as given by ω ε = It has been reported that Au nanoparticles ( D = 50 nm) exhibit an LSPR peak at 543 nm ( Figure 10 c). [ 31b ] After combining Au with TiO 2 to yield Janus and core-shell nanoparticles (Figures 10 a,b), these heterojunction nanostructures demonstrated a red-shift in plasmonic peaks [ 31b ] because the presence of a high refractive index TiO 2 layer leads to an increase in ε m and thus a decrease in ω . The Au nanoparticle in the core@ shell structure was encapsulated by the TiO 2 shell, while half of the Au nanoparticle in the Janus structure was exposed to solution (i.e., isopropyl alcohol) and the other half to TiO 2 . As TiO 2 possesses a higher refractive index than the solution, the LSPR peak of core@shell nanoparticles occurred at a longer wavelength (Figure 10 c). For the metal@semiconductor core@shell nanostructure or thin semiconductor layer-deposited noble metals, the thickness of semiconductor should be carefully  The normalized c-Si absorption effi ciency as a function of r 1 , r 2 , and L z , respectively. The optimum L z shown on the right panel displays peaks. Reproduced with permission. [ 38 ] Copyright 2014, American Chemical Society.
tuned as it greatly affects the LSPR properties of photocatalysts. Since the electromagnetic fi eld will extend into the surrounding environment, the overall ε m is dictated by both the semiconductor and the liquid or gas around the plasmonic photocatalyst if the semiconductor shell or layer is within the electromagnetic fi eld. [ 43 ]

Other effects
In previous work, noble metal nanostructures (e.g., Au nanoparticles, Ag@Ag 3 (PO 4 ) 1x core@shell nanoparticles, etc.) were coupled to the surface of semiconductors (e.g., TiO 2 photonic crystals, ZnO nanorod arrays, dendritic TiO 2 nanorod arrays, etc.). [ 31a , 35 , 43b , 44 ] Notably, the amount of deposited noble metal needs to be well controlled. A high coverage of noble metal on the semiconductor will reduce the light exposure of the semiconductor as well as hinder the reactant access of semiconductor. While low coverage will lead to poor light utilization effi ciency during photocatalysis. [ 31a , 35 , 44c ] Thus, an optimum coverage exists between these two degrees of coverage.
The composition of plasmonic metals greatly affects the properties of photocatalysts as well. Lin et al. deposited Ag@Ag 3 (PO 4 ) 1x core@shell nanoparticles on ZnO nanorod arrays to capitalize on visible light absorption introduced by Ag due to its plasmonic properties as well as Ag 3 (PO 4 ) 1x due to its narrow band gap. This system ultimately achieved oxygen evolution in a nonsacrifi cial electrolyte. [ 44c ] Unfortunately, Ag is easily oxidized [ 45 ] during the photocatalytic process and cannot be sustained for long periods. Therefore, Au is more commonly utilized.
Semiconductor nanostructures can also affect the performance of plasmonic photocatalysts. For better utilization of the LSPR property, photonic crystal structures can be employed. [ 44a,b,d ] A photonic crystal is a periodic structure that possesses a photonic band gap, in which light will not pass through but instead be trapped. [ 46 ] Zhang et al. incorporated Au nanoparticles onto a bi-layer TiO 2 structure composed of a photonic crystal and nanorods (denoted as Au/TiO 2 NRPCs in Figures 11 a,b). [ 44b ] When the plasmon resonance frequency of the noble metal matches the photonic band gap of the semiconductor, the LSPR effect is markedly promoted. This leads to higher photoconversion (i.e., light-to-chemical energy) efficiency relative to other photocatalyst systems (Figures 11 c,d). In addition, some studies focused on raising the surface area of semiconductors to enhance the adsorption capability of photocatalysts. Three-dimensional hierarchical dendritic Au@TiO 2 nanorod arrays were synthesized with larger surface area. Compared to pure TiO 2 branched nanorod arrays, they exhibited improved charge separation and charge transfer effi ciency. [ 44e ] In addition, the properties of semiconductors should also be considered. Some semiconductors have small band gaps, but suffer from short charge carrier diffusion lengths, which make them unsuitable for use as photocatalysts on their own. For example, hematite Fe 2 O 3 has a small band gap of 1.9 to 2.2 eV, depending on the fabrication method and crystalline status, which is desirable for the visible light absorption. [ 47 ] However, the short exciton diffusion length (several nanometers without bias [ 48 ] and tens of nanometers with bias [ 49 ] ) leads to severe recombination of photo-induced electron-hole pairs in the bulk of Fe 2 O 3 . Several strategies for solving this problem have been developed. First, other types of semiconductor with high carrier mobility can be intimately combined with Fe 2 O 3 to improve the carrier transfer rate due to the strong internal electric fi eld. [ 50 ] The further incorporation of plasmonic metal nanostructures increases the generation of charge carriers close to the Au/Fe 2 O 3 interface and shortens the charge carrier  diffusion distance to the electrolyte thus enabling a higher pho toactivity. [ 43b , 50,51 ] The second strategy is to incorporate Fe 2 O 3 nanorod arrays into patterned Au nanohole arrays [ 52 ] or craft a thin Fe 2 O 3 layer onto Au nanopillars [ 43b ] to utilize the SPP effect and further boost photocatalytic effi ciency. It is noteworthy that another intriguing group of materials for photocatalytic applications are semiconducting aerogels. These materials have attracted great interest because of their inherent advantages, including high specifi c surface area (≈100-1000 m 2 g -1 ). [ 53 ] This ensures effective internal light scattering and increases photon harvesting. [ 54 ] Au nanoparticles have been incorporated into 3D TiO 2 aerogel networks (3D Au-TiO 2 aerogel) ( Figure 12 a), and show three attractive features. [ 55 ] First, Au nanoparticles were locked into the network and aggregation was prevented even after the transformation of amorphous TiO 2 into a crystalline form. [ 55 ] Second, this arrangement retains the high surface area of the host aerogel network. [ 56 ] Third, the photocatalytic activity is retained even after several years of storage. [ 55 ] Therefore, as shown in Figure 12 c,d and 12d, the 3D Au-TiO 2 aerogel displayed a higher IPCE when compared to Au-TiO 2 aerogel (Au nanoparticles fi lling the pores of TiO 2 aerogel network, denoted DP Au-TiO 2 aerogel, in Figure 12 b) obtained via a depositionprecipitation method. [ 55 ]

Conclusions and Outlook
Although metal/semiconductor hybrid nano-photocatalysts are in their infancy of development, they effi ciently address the issues of low utilization of the solar spectrum and the inefficient photocatalytic performance of some semiconductors. This review highlights recent progress in the mechanistic understanding and development of plasmon-mediated photocatalysis, and summarizes the effects of various parameters in plasmonic photocatalysts design on the photocatalytic process.
To date, many studies have focused on the utilization of UV and visible regions of the solar spectrum, which accounts for 52% of the total solar spectrum. Infrared light, comprising 43% of the solar spectrum, appears to be an appealing energy source for photocatalysis as well. In recent years, up-conversion materials, usually encountered in biological and clinical studies, have attracted considerable attention by photocatalytic researchers. [ 57 ] Rare-earth-element-doped upconversion materials absorb infrared light and emit UV and visible light via a two-or multi-photon mechanism, which has been welldocumented. [ 58 ] The energy scheme depicting the relevant processes of energy transfer in upconversion materials is shown in Figure 13 . Unfortunately, rare-earth-element-doped upconversion materials generally suffer from low up-converting efficiency due to small absorption cross sections and low energy transfer effi ciencies. [ 59 ] A typical approach to improve the upconversion emission effi ciency is to combine different host matrices with rare-earth atoms. [ 60 ] After the discovery of the infl uence of a dielectric interface on fl uorescence in 1969, [ 61 ] some alternative efforts have been made to enhance the upconversion luminescence through the hetero-integration of rareearth-element-doped upconversion materials with metallic nanoparticles (i.e., Au and Ag). [ 62 ]   In photocatalysis, commonly used upconversion materials are lanthanide-doped materials sensitized by Yb 3+ in which upconversion multicolors (i.e., green, red and blue) are observed due to the presence of Er 3+ and/or Tm 3+ ions. [ 63 ] When upconversion nanoparticles (UCNPs) are combined with metals, the spectroscopic properties of UCNPs are profoundly changed as a result of the local electrical fi eld generated by SPR. [ 64 ] Consequently, the excitation fi eld and the emission of UCNPs are enhanced. Meanwhile, non-radiative relaxation as well as quenching by non-radiative energy transfer from the upconversion materials to metal surface occurs. [ 65 ] Clearly, a competition between these processes governs the fi nal degree of enhancement of upconversion emission.
UCNPs are required to be at an appropriate position in the vicinity of Au because the fi eld enhancement strictly relies on the polarization of the excitation light. [ 64a ] Schietinger et al. attached nanocrystals (i.e., NaYF 4 :Yb 3+ /Er 3+ ) to Au spheres using atomic force microscope to achieve an overall upconverison enhancement factor of 3.8 compared with bare NaYF 4 :Yb 3+ /Er 3+ nanocrystals (Figure 13 b). [ 64a ] The enhancement of green emission was found to be more pronounced than that for red emission (Figure 13 d). This observation can be attributed to the following two possible reasons. The fi rst is the higher emission rate of green light caused by surface-plasmon-coupled emission (SPCE). [ 64a , 66 ] SPCE occurs when the plasmon resonance frequency of Au nanospheres overlaps with the emission bands.
Thus, higher amplifi cation of green emission was observed due to a better plasmonic coupling. The second reason is the rearrangement of the excitation scheme due to the enhanced excitation fi eld. [ 64a ] With increasing excitation power, the Er 3+ ion was excited to make the 4 I 11/2 → 4 F 7/2 transition prior to the nonradiative decay process ( 4 I 11/2 → 4 I 13/2 ). [ 64a ] This procedure accounted for the fi lling of the red-emission level, 4 F 9/2.
[ 64a ] As shown in Figure 13 e, after coupling to Au nanospheres, the rise time of the green and red emissions was reduced due to two factors. First, the excitation intensity around the UCNPs was enhanced because of the plasmonic effect of Au nanoshperes (Figure 13 c). Second, the second energy transfer step to 4 F 7/2 and 4 F 9/2 was amplifi ed owing to the coupling to the plasmon resonance of Au. Moreover, additional non-radiative channels introduced after coupling with Au led to shorter lifetimes for the levels, and thus a faster decrease in fl uorescence (Figure 13 e). [ 64a ] The SPCE is enabled by frequency matching between localized plasmon resonance and the emission band of upconversion materials. Clearly, some factors such as size, spacer distance, dielectric environment, and the shape of the metal nanostructures signifi cantly infl uence the plasmon resonance frequency, and thus play important roles in plasmon-enhanced upconversion emission. [   in Figure 14 a, the emission intensity can be modulated by tuning the silica spacer. Near the Ag/SiO 2 interface, the upconversion intensity quenched rapidly (enhancement factor <1) (Figure 14 b) because the intimate contact between the Ag core and SiO 2 shell resulted in effi cient non-radiative energy transfer from the upconversion material to the metallic surface. [ 65c , 67a , 68 ] Therefore, with an increase in the spacer thickness, SPCE and local fi eld enhancement dominated and the thickness of the silica shell was found to be optimized at 30 nm (Figure 14 b). By adjusting the size of Ag nanospheres, the fl uorescence intensity can be changed due to plasmon resonance scattering. [ 65c , 69 ] The remarkable emission increase from Ag-50nm@SiO 2 -30nm@Y 2 O 3 :Er showed that the quenching was weaker than the enhancement especially at and beyond the spacer thickness of 30 nm but was not insignifi cant. [ 65c ] The band ratio (green emission band/red emission band) in Figure 14 b further confi rmed a preferential increase of the green emission band. [ 65c ] In addition to metallic nanoparticles, other structures such as nanoshells have also been studied. Recently, Au NPs and continuous Au nanoshells were attached to the surface of UCNPs to explore their plasmonic modulation of upconversion emission. [ 66 ] The procedure for fabricating such Au@UC  hybrid structures is illustrated in Figure 15 a. Poly(acrylic acid) (PAA) possessing good hydrophilicity was fi rst coated on the surface of UPNPs. Poly(allylamine hydrochloride) (PAH) was then attached onto the surface of nanoparticles to create a positive charge on their surface. To couple the Au NPs, the aqueous solution of UPNPs was mixed with negatively charged Au NPs for various times. By adding additional Au precursors and reductant, a Au shell grew around the UPNPs with Au NPs serving as seeds for the nucleation. [ 66 ] During the Au NP seeding, emission intensity increased as more Au NPs were attached onto the surface of upconversion nanostructures, and an enhancement factor of more than 2.5 was obtained (Figure 15 b). [ 66 ] The reason for the enhanced emission has been noted previously. However, a quenching of emission intensity was observed during the formation of the shell (Figure 15 c). This phenomenon was attributed to three factors: (1) the SPR resonance frequency depends heavily on the geometry of Au. [ 9a , 19a ] When a complete shell was formed, the SPR peak was red-shifted into the near-infrared region, which reduced the SPCE; (2) a continuous shell enhanced the scattering of excitation fl ux, and thus decreased the effective excitation light; (3) an integrated shell blocked the emission transmittance from the UPNPs. [ 66 ] It is interesting to note that recently rare-earth-elementdoped upconversion materials have been utilized in photocatalysis. However, plasmon-mediated upconversion materials have rarely garnered attention from photocatalytic researchers. In this context, it is advantageous to harvest as much of the solar spectrum as possible in photocatalysis by exploiting plasmonenhanced rare-earth-element-doped upconversion materials.
Looking to the future, in addition to plasmon-enhanced upconversion materials discussed above, many multifunctional materials composed of noble metals and semiconductors can be crafted for plasmonic photocatalysis. First, a mixture of plasmonic photocatalysts containing complex plasmonic structures (i.e., plasmonic nanoparticles, nanorods and nanowires) can be rationally designed and synthesized. For the dimension of noble metal particles that is larger than the quasistatic limit, the red-shift of plasmonic peaks will occur due to the retardation effect when the particles become larger. Such plasmonic photocatalysts containing a mixture of noble metals is highly desirable as they enable broader light absorption by extending into the near-infrared region.
Second, as plasmonic metal nanostructures with sharp geometries possess lower restoring force for the charge oscillation, the corresponding plasmonic peak shifts towards the longer wavelength. Clearly, photocatalytic materials with sharp geometry are favorable in photocatalysis and merit further exploration.
Third, recent advances in alloyed noble metals have garnered great attention due to their improved stability, chemical properties and tunable plasmonic resonances. [ 70 ] The combination of semiconductors and alloyed plasmonic metals may provide a promising opportunity to create stable plasmon-mediated photo catalysts for long-term use.
Finally, we envision that by judiciously constructing uniform plasmonic photocatalysts with tunable size and complex architectures (for example, core@shell nanoparticles with precisely variable core diameter and shell thickness, [ 71 ] as well as strictly biphasic Janus nanoparticles and BMW-logo-like nanoparticles with precisely controlled diameter), the infl uence of the dimension and architecture of these intriguing nanostructured plasmon-mediated photocatalysts on LSPR  Figure 15. a) Fabrication processes for Au nanoparticle attachment (yellow dots) and Au shell (yellow shell) growth on the UPNPs (pink hexagon). Up-conversion spectra of NaYF 4 :Yb/Tm NPs during b) Au seed growth stage (0-360 min), and c) Au shell formation stage (0-10 min), respectively. Reproduced with permission. [ 66 ] and thus the photocatalytic performance can be systematically evaluated.
Given the short period of time dedicated to the development of plasmonic photocatalysis, further fundamental studies are still required. With rapid advances in synthetic techniques, a greater diversity of materials with good size and shape control are possible. It is notable that such synthetic techniques are useful for not only novel materials but also the rational and deliberate combination of existing well-characterized materials. It is important to underscore that indeed novel material development is essential. However, the nanostructuring and combination of established materials are equally important in pushing new and exciting research into plasmonic photocatalysis with improved photocatalytic performance.