Silver Nanocubes Coated in Ceria: Core/Shell Size Effects on Light-Induced Charge Transfer.

Plasmonic sensitization of semiconductors is an attractive approach to increase light-induced photocatalytic performance; one method is to use plasmonic nanostructures in core@shell geometry. The occurrence and mechanism of synergetic effects in photocatalysis of such geometries are under intense debate and proposed to occur either through light-induced charge transfer (CT) or through thermal effects. This study focuses on the relation between the dimensions of Ag@CeO2 nanocubes, the wavelength-dependent efficiency, and the mechanism of light-induced direct CT. A 4-mercaptobenzoic acid (4-MBA) linker between core and shell acts as a Raman probe for CT. For all Ag@CeO2 nanocubes, CT increases with decreasing excitation wavelength, with notable increase at and below 514 nm. This is fully explainable by CT from silver to the 4-MBA LUMO, with the increase for excitation wavelengths that exceed the Ag/4-MBA LUMO gap of 2.28 eV (543 nm). A second general trend observed is an increase in CT yield with ceria shell thickness, which is assigned to relaxation of the excited electron further into the ceria conduction band, potentially producing defects.

Introduction

Sensitized photocatalysts are gaining increasing attention with the growing concerns about the state of the environment and, with it, the desire to use more sustainable energy sources. Nanostructures that support local surface plasmon resonances (LSPR), which is the photoinduced collective oscillation of electrons, can be used to sensitize semiconductors through decoration, incorporation, or hierarchical construction such as core@shell particles.[1−3] Core@shell nanostructures are attractive because they offer controllable overlap of the enhanced electric field with the semiconductor and even the boundary of the semiconductor with the external media by controlling the thickness of the shell.

Plasmonic enhancement of photocatalysts can be divided into four mechanisms: plasmon-induced resonance energy transfer (PIRET),[4−6] direct charge transfer (CT) during photoexcitation or before plasmon dephasing, indirect CT after plasmon dephasing,[4,7−9] and as a local heat source.[10] PIRET requires coupling of transition dipoles and is strongly dependent on the distance and spectral overlap of the plasmonic donor and the energy acceptor but can occur through space.[4,5] Unlike PIRET, both CT mechanisms need electronic coupling. Direct CT can be divided into three categories based on wavelength dependence and association with plasmonic modes. (1) Plasmon-induced interfacial CT transition (PICTT): a new electronic transition that requires strong coupling between the donor and acceptor across the interface of the plasmonic particle and semiconductor. (2) Direct interfacial CT transition (DICTT): a CT transition not associated with a plasmonic mode, but instead with a weaker metal–adsorbate transition possible above a threshold photon energy.[8] (3) Chemical interface dampening (CID): a new adsorbate-induced plasmon dephasing pathway, with a quantum yield in agreement with the LSPR absorption.[8,11] For indirect CT there is a sensitivity to the photon energy as excited electrons must overcome the Schottky barrier prior to relaxation through heat generation in the lattice.[4,5,7−10]

As well as acting as a sensitizer,[2,3] the LSPR is also a sensor for the local environment: changes to the local refractive index (n) cause spectral shifts[12] and enhance Raman intensities.[13,14] The LSPR spectral properties depend on the particle material, shape, size, and local environment.[15,16] Silver nanocubes support several LSPR modes; the most relevant to this study is the dominant corner mode (red-most mode for an isolated cube). This mode projects the furthest from the surface and is most sensitive to the local refractive index for supported cubes.[12,15−17] The distance-dependent refractive index sensitivity (RIS(t), with t being the distance), determined from finite-domain time-difference (FDTD) simulations, allows for determination of the shell coverage.

CeO2 is a well-known catalyst with demonstrable potential as a photocatalyst,[18−22] which supports Ce(IV) and Ce(III) states forming reactive oxygen vacancies[23,24] and catalyzes reactions such as reducing CO2 into CO.[25] CeO2 has a bulk direct band gap of 3.6 eV and an indirect band gap of ∼2.4 eV;[26] in addition, trap states may induce light absorption throughout the visible regime. Silver nanoparticles are an attractive choice to sensitize ceria as they have high quality plasmonic modes,[27] a high refractive index sensitivity,[28] and, through geometric control, light absorption throughout the visible and near-infrared regime.[29] Additionally, use of a core@shell geometry is of interest, as the metal core can be isolated—either to prevent reactions, such as silver dissolving, or to promote selectivity by only having the semiconductor as an available active site. There have been studies on Ag/CeO2 (nano)composites,[30−34] but only few studies report on core@shell geometries,[35−39] despite their promise as photocatalysts. Silver nanocubes are attractive because of the spatial localization of modes[17] and high |E|2 enhancement,[40] while preserving a high degree of symmetry. Generalizable fundamental understanding of structural effects on the photophysics is important, specifically on the occurrence, mechanism, and wavelength dependence of photoinduced CT.

The aim of this work is to investigate photoinduced CT in Ag@CeO2 core@shell nanocubes through surface-enhanced Raman scattering (SERS) spectroscopy by using the method developed by Lombardi and Birke.[41] Comparing the intensities of totally and non-totally symmetric Raman modes allows for a quantification of the relative enhancement from the LSPR and from CT.[41] We have investigated the role of core and shell dimensions on CT by using a 4-mercaptobenzoic acid (4-MBA) tag between the silver and ceria as the Raman probe. Here, the wavelength-dependent effects of core and shell dimensions on CT have been examined. FDTD simulations have been performed to model the shell coverage of the Ag@CeO2 core@shell nanocubes and determine the distance dependences of the refractive index sensitivity. These give insights into the relationship between Ag@CeO2 core@shell nanostructure design and light-induced CT processes.

Materials and Methods

Synthesis Ag Nanocubes

Ag nanocubes of varied sizes were synthesized by using a method adapted from the polyol synthesis;[42,43] the solvent for all solutions used in the synthesis was ethylene glycol. A round-bottom flask, containing 40 mL of ethylene glycol (Fluka, >99.9%) and 0.4 g of polyvinylpyrrolidone (PVP, Sigma-Aldrich), was heated at 150 °C for 2 h. 0.4 mL of 3 mM Na2S was then added, followed by the dropwise addition of 2.5 mL of 280 mM AgNO3(Alfa Aesar, 99+%). The reaction was followed with a spectrometer (Avantis Starline AvaSpec-2048) and quenched by cooling when the desired spectrum was observed. After quenching, the product was cleaned through centrifugation (9000g) and redispersion into ethanol through sonication five times. The final suspensions were 10 mL and had an attenuance of 4 × 102 at the spectral maximum.

Synthesis Ag@CeO2 Core–Shell Nanocubes

Silver nanocubes with edge lengths of 47 ± 7 and 71 ± 9 nm were coated with conformal ceria shells of different thicknesses by varying the cerium nitrate and sodium hydroxide concentrations. A mild synthesis method to deposit the ceria shell on the plasmonic cube was chosen, as high temperatures or caustic reagents can easily damage or destroy the particles.[44]

Ceria was deposited on silver in a method adapted from bulk precipitation methods.[45] Ce(NO3)3·6H2O(aq) (Sigma-Aldrich, 99% trace metal) was exposed to a molar equivalent of NaOH(aq) (Sigma-Aldrich, >98%) in the presence of 4-MBA-modified Ag nanocubes. Specifically, 1000 μL of the OD ∼ 400 particle suspension was added to 20 mL of water, followed by 100 μL of 3 mM 4-MBA(EtOH), and the suspension was then incubated for 10 min. The amount of ceria precursor and base to particles was varied to achieve different ceria shell thicknesses, where 10, 100, or 5000 μmol of both Ce(NO3)3(aq) and NaOH(aq) was used. Samples were cleaned through centrifugation (4000g) and dispersion into Uvasol ethanol (VWR) three times.

Initial studies did not include 4-MBA, and ceria grew off the PVP on the Ag nanocubes; however, to observe CT the 4-MBA was included. In mildly acidic conditions 4-MBA attaches to silver by the thiol group;[46] precipitating ceria then binds to the free carboxylic group (similarly, the ceria can grow off the oxygen in PVP[37]) and grows off the Ag nanocubes. This allows for the 4-MBA to act as a linker and be in electronic contact with both the silver and ceria.

FDTD Simulations

The program Lumerical FDTD Solutions was used for simulations for theoretical support of the experimental spectral changes caused by the shell deposition. Ag nanocubes were modeled with rounded edges and corners (r = 2.5 nm); ceria shells were modeled with the same structure. As thickness was increased, the radii of edges and corners were increased, such that the particle was spherical when the thickness is equal to the core size. Scattering and absorption were both considered to produce a calculated extinction relatable to experimental results. While the Johnson and Christy[47] model for silver was used, noncomplex indices of refraction were used to model the shell and surrounding media. The refractive index value n for the ceria shell was set at 2.2, and a value of 1.337 for the solvent (water) was used. Using a fixed value for the shell facilitates generalization of the results for use with the Lorentz–Lorenz equation.

Nanostructure Characterization

The UV–vis spectra of Ag@CeO2 core–shell nanostructures in suspension were collected by using a ThermoSci EVO600 spectrometer. For refractive index sensitivity measurements, the solvents used were deionized water, ethanol, and ethylene glycol. Characterization of the nanoparticles used a Philips CM300ST-FEG transmission electron microscope (TEM) with acceleration voltage of 300 kV and energy dispersive X-ray analysis (EDX). The dimensions of uncoated Ag nanocubes and Ag@CeO2 core–shell nanocubes were determined from 20 measurements of shells/crystallites on the particles. Scanning electron micrographs (SEM) were recorded by using a JSM-6010LA SEM. X-ray diffraction (XRD) analysis was performed on a Bruker D2 (Cu Kα source) diffractometer, which revealed increased shielding of the silver core with increased shell thickness; the diffractogram is shown in Figure S1 of the Supporting Information. X-ray photoelectron spectroscopy (XPS) was performed with a Quantera SXM, with an Al Kα source (1486.6 eV).

Raman Measurements

Raman measurements were performed on a diluted suspension of Ag@CeO2 core–shell nanocubes in Milli-Q water. Raman spectra at 785 nm excitation were collected with an Avantes AvaRaman probe system, containing 1% ethanol. For Raman measurements with 458, 488, 514, and 568 nm excitations, the excitation source was an Ar/Kr ion laser (Coherent). A single grating monochromator (Jobin Yvon, focal length 640 mm) with a liquid N2 cooled CCD camera (Princeton Instruments) was used for detection. Calibration of the Raman shift for these excitation energies was performed using cyclohexane as a reference. The Raman spectra were invariant with respect to the excitation intensity between 2.5 and 10 mW (Figure S2).

Results and Discussion

Nanostructural and Optical Characterization

Figure shows the TEM micrographs of the uncoated Ag nanocubes (i) and the Ag@CeO2 core@shell nanocubes with different ceria shells (ii–iv), for cubes with an edge length of 47 ± 7 nm (A) and 71 ± 9 nm (B). For amounts of ceria insufficient to coat the Ag nanocubes fully, we observe an incomplete coverage with small CeO2 crystallites (ii); otherwise, conformal CeO2 shells with tunable thickness are formed (iii and iv). See Table S1 for specific values for each sample and Figure S3 for additional TEM and SEM micrographs showing no larger structures. There is no significant difference in the size of the isolated ceria crystallites between the 47 and 71 nm Ag nanocubes coated with ceria, with 2.6 ± 0.3 and 2.3 ± 0.2 nm sized crystallites, respectively. Aggregated crystallites form the thicker ceria shells, resulting in a relatively conformal coating of the Ag nanocubes. The TEM micrographs show that the ceria shells are porous, that is, have voids. This is most likely due to the absence of a chelating agent which would slow down the deposition of ceria.[38] XPS reveals a Ce3+/Ce4+ ratio of 0.09 (see Figure S4 and Table S2). The coverage has been determined from the simulated and experimental extinction properties in different solvents (see below). A summary of shell thicknesses and coverages is provided in Table S1.

Figure 1

TEM of (A) 47 ± 7 nm silver cubes with (i) no shell, (ii) incomplete coverage with 2.6 ± 0.3 nm CeO2 crystallites, (iii) a 4.6 ± 1.3 nm shell, and (iv) 14.1 ± 3.9 nm shell and (B) 71 ± 9 nm cubes with (i) no shell, (ii) incomplete coverage with 2.3 ± 0.2 nm crystallites, (iii) a 3.7 ± 1.1 nm ceria shell, and (iv) a 17.6 ± 3.5 nm conformal shell. The scale bars in all micrographs are 30 nm. Core and shell dimensions have all been determined from no fewer than 20 measurements. Additional micrographs are presented in Figure S3.

Figure presents the UV–vis spectra of the uncoated and coated 47 nm (A) and 71 nm (B) Ag nanocubes. The silver interband transition around 300 nm red-shifts with CeO2 shell thickness; this is likely due to interband scattering,[48] and, in particular, for the thickest shells, possibly from the CeO2 bandgap (3.6 eV, 344 nm).[49,50] The red-shift of all the spectral features with increased CeO2 shell thickness is due to an increase in local index of refraction, as the plasmonic modes are highly sensitive to the local environment.[15] CeO2 has a refractive index between 2.2 and 2.8 in the visible regime,[49,50] significantly higher than the refractive index of water. The mode centered at 348 nm observed for uncoated 47 and 71 nm Ag cubes is localized on the face of the cube[17] and is highly sensitive to the sharpness of the corners.[51] The retention of this mode during coating suggests that no significant etching of the corners occurs. The mode at 390 nm (47 nm cubes) and 405 nm (71 nm cubes) is localized on the edges, while the intense, lowest energy mode at 460 nm (47 nm cubes) and 485 nm (71 nm cubes) is localized at the corners of the cube.[17] The latter mode is strongly dependent on the cube edge length and is the most sensitive to a change in local index of refraction.[16,17] The broad band around 700 nm for the 47 nm silver nanocubes with a 2.6 nm shell, and to a low extent those around 750 nm for the 71 nm silver nanocubes with a 2.3 nm shell, is most likely due to mild aggregation. It is worth noting that no new electronic transitions indicative of direct CT are observed.

Figure 2

UV–vis spectra of (A) 47 nm and (B) 71 nm silver nanocubes, PVP capped (0 nm), and 4-MBA treated and coated with ceria (shell thickness given in legend), suspended in ethanol. Red-shifting of modes occurs due to the increased local index of refraction caused by the presence of CeO2.

The ceria shell coverage, while accounting for the voids in the shell, has been determined by correlating experimental and simulated changes in the UV–vis spectrum induced by a change in the local refractive index by using (i) different solvents (Figure S5) and (ii) changing the ceria shell thickness. The volume fraction of the shell (φ) depends on the shell coverage (η) and the distance-dependent refractive index sensitivity (RIS(t), with t the distance) according to:due to the |E|2 profile around the plasmonic nanostructure.[15] The value of φ has been determined by using the Lorentz–Lorenz equation:[52]where neff is the effective refractive index, a function of the ith component’s refractive index n and φ the volume fraction. There are only two components considered for this system: the shell and solvent as the PVP/4-MBA capping agent is a constant and needed to prevent aggregation. The effective refractive index of the Ag cubes has been determined from RIS(t) by dispersing uncoated Ag cubes in solvents of different refractive indexes and determining the peak shift of the corner mode relative to n (see Figure S5).

Figure A shows the effect of ceria shell thickness on the peak shift of the corner mode, simulated by FDTD for cubes with an edge length ranging from 50 to 100 nm: for an edge length ≤90 nm, RIS(t) = (0.951 ± 0.009) (see fit included in Figure A and Figure S6). This model of RIS(t) has been validated by comparing all shell thicknesses as determined by TEM with simulated values (Figure B). Applying eq yields an average shell coverage η = 0.67 ± 0.25 (specific values for all samples are shown in Table S1), allowing quantification of the shell thickness through UV–vis spectra alone in future synthesis of Ag@CeO2 nanocubes.

Figure 3

(A) Simulated peak shift of the corner mode with respect to n for cubes of 50–100 nm, normalized from a pure solvent to a pure shell system, including a fit to RIS(t) = (0.951 ± 0.009), with t the shell thickness. (B) Comparison of the shell thickness determined by TEM and UV–vis using the results from FDTD and the average coverage, giving statistically similar results.

Photoinduced Charge Transfer

The 4-MBA linker has been used as a SERS probe for photoinduced CT and facilitates examination of the role of photoexcitation wavelength and core@shell dimensions. The degree of CT (ρCT) has been determined using the theory developed by Lombardi and Birke. Equation compares the selective enhancement of the non-totally symmetric vibrational modes, which are enhanced by CT, relative to totally symmetric modes which are only enhanced by the LSPR. A value of 0.5 implies equal Raman enhancement from LSPR and CT:[41]The parameter k is an index identifying individual molecular bands in the Raman spectrum. I(LSPR) is the intensity of line (k) where only the LSPR contributes to the SERS enhancement, while I0(LSPR) denotes the intensity of a chosen totally symmetric line, also with only LSPR contribution. I(CT) is the intensity of a non-totally symmetric mode where both LSPR and CT contribute. For a non-totally symmetric line I(LSPR) will normally be small or zero, reducing eq intoThis method is only capable of observing direct CT that involves the 4-MBA linker. It is not possible to observe any CT directly from silver to ceria bypassing the 4-MBA. Additionally, there is no information regarding the directionality of the process. However, given the lack of light absorption by the ceria or 4-MBA and the strong absorption by the silver, it can be concluded that CT originating from light absorbed by the silver is the most likely mechanism.

Figure shows the Raman spectra at various excitation wavelengths for the 47 nm (A) and 71 nm (B) Ag nanocubes with 4-MBA linker and different ceria shell thicknesses. The Raman spectra are invariant to the laser intensity between 2.5 and 10 mW (see Figure S2). The mode at 1070 cm–1 used for normalization of the Raman spectra is the C–S stretch and a symmetric ring breathing mode (νCS and 1)[53−55] and has A1 character. The assignment of the 1140 and 1180 cm–1 modes is in debate—however, not the modes at 1370 and 1415 cm–1 which are of B2 character.[54−58] Specifically, the 1415 cm–1 mode is the asymmetric ring mode 18b, and the 1370 cm–1 is assigned to βO–H, νC–ph, 19a, and asymmetric νCO2 vibrations.[53−55]Table summarizes the assignments of the modes. On the basis of these assignments, we have used the 1415 cm–1 mode as I(CT) and the 1070 cm–1 mode as I0(LSPR) to resolve ρCT. For quantification of ρCT, the spectra have been fit with Gaussian functions (see the Supporting Information for details).

Figure 4

Raman spectra of 4-MBA between silver and ceria for (A) 47 nm and (B) 71 nm silver nanocubes, normalized to the intensity of the 1070 cm–1 mode. The degree of charge transfer, ρCT, at different excitation energies for (C) 47 nm and (D) 71 nm Ag@CeO2 nanocubes with different shell thicknesses; the dashed vertical line is at 543 nm, the wavelength of light needed to excite an electron from the silver Fermi level to the 4-MBA LUMO. The ρCT was calculated by using the variance of the 1415 cm–1 mode as compared to the 1070 cm–1 mode.

Table 1

Raman Mode and Symmetry Assignments from the Literature for 4-MBA

wavenumber (cm–1)	band assignmenta	symmetry	ref
1070	νCS and 1	A1	(55, 57, 65)
1140	9a + γOH	A1	(55)
	δCH + ν15	B2	(57, 65)
1180	βOH + 9a	B2	(55)
	δCH + 9a	A1	(57)
	νCH	A1	(57, 59)
1370	βOH + νC–ph + 19a and νasCO2	B2	(55)
1415	18b	B2	(55, 59)
	βOH + νC–ph + 19a and νasCO2	B2	(57)
	νCOO–		(59, 63)
1600	9a	A1	(55)
	νCC	A1	(57, 59, 63)

ν: symmetric stretching; νas: asymmetric stretching; β: bending; δ: deformation.

An important parameter to address is the 4-MBA layer thickness. In the ideal case a monolayer is present, with the 4-MBA molecule likely attached to the Ag core via the thiol group and to the ceria shell via the carboxylic acid group,[46] implying a subnanometer 4-MBA layer thickness. Multilayer structures may be formed through hydrogen bonding between the thiol and the carboxylic acid groups and by van der Waals interactions. However, such a multilayer structure would show a βOH Raman mode which would vary between an incomplete shell to the complete shell; that is, variation of the 1180 cm–1 mode and the 1370 cm–1 mode.[46] As such effect is absent in the Raman spectra (Figure A,B), it can be concluded that a multilayer of 4-MBA is unlikely, which is further supported by the TEM experiments (Figure ) unable to resolve the 4-MBA linker. Incomplete coverage of the Ag surface can however not be excluded, although no PVP Raman signals have been observed.

Figure C shows ρCT as a function of excitation wavelength for the 47 nm Ag nanocubes with three different ceria shell thicknesses. The measurements at 785 nm excitation have 1% ethanol, which gives rise to the extraneous peak at 1460 cm–1. For both nanocube edge lengths, the most pronounced effect on the intensities of the 1370 and 1415 cm–1 modes is observed at 488 and 458 nm excitation. At 785 nm excitation ρCT is close to zero for all three ceria shell thicknesses, as can be expected for photon energies significantly lower than both the LSPR and a potential Ag to 4-MBA electronic transition. The ρCT value significantly increases on decreasing the excitation wavelength, with a notable increase at 514 nm and shorter wavelengths. In addition, ρCT is higher for thicker ceria shells. Figure D, presenting ρCT values for the 71 nm Ag nanocubes coated with different ceria shell thicknesses, shows similar trends. The absence of a band at 1415 cm–1 for Ag cubes coated with 4-MBA only (Figure S7) demonstrates that the ceria shell is crucial for the CT process.

Figure shows an energy level diagram[24,59] and the proposed CT mechanism. If the CT were to occur through PICTT, CID, or indirect CT, then it would be expected that ρCT would follow the LSPR absorption of the silver core.[60] Given that no LSPR dependency is observed here, CT is most likely dominated by the DICTT mechanism. The increase in ρCT at 514 nm and shorter wavelengths can be assigned to CT of an electron from the Ag core to the LUMO of the 4-MBA, which is allowed for photon energies >2.28 eV (i.e., <543 nm) through the DICTT mechanism. The increase with photon energy likely originates from the increasing probability for CT from an electron at a greater depth in the Fermi sea; on average, half the photon energy is left in the hole formed on photoexcitation.[6] Indeed, at 785 nm excitation CT is insignificant for all core@shell geometries. However, all ρCT values at 568 nm excitation exceed those at 785 nm, indicating some CT, although inefficient. The photon energy at 568 nm light (2.18 eV) is 0.1 eV below the energy gap between the Ag Fermi level and the LUMO of the 4-MBA linker.[59] Within the thermal distribution of electrons at room temperature, there is a 2% chance of the state 0.1 eV above the silver Fermi level being populated (Figure S8), which likely accounts for the low ρCT. Additionally, CT from the 4-MBA HOMO to the silver may occur. While this may have some contribution to ρCT, it is not expected to demonstrate the trends seen here, as it would lead to relatively constant ρCT values for wavelengths shorter than 543 nm, which is not observed.

Figure 5

Energy diagram and proposed mechanisms of light-induced CT, the black dashed line is the Fermi energy of silver used as a reference.[59,66] Photoexcitation could promote an electron from the Ag core to the 4-MBA linker, followed by electron transfer from the 4-MBA to the CeO2. The latter process may result in the formation of an oxygen vacancy (Ovac).

Aside from this excitation wavelength effect on ρCT, a second general trend observed is a higher ρCT value with increasing ceria layer thickness. This dependency indicates that not only the 4-MBA intermediate layer affects the CT process but also the ceria outer shell. There are statistically significant increases in ρCT with a thicker shell through the 4-MBA pathway (Figure C,D), which indicate that the electron excited from the Ag core to the 4-MBA can relax into the CeO2 conduction band (CB). Similar observations were made for a Au/4-MBA/TiO2 system, where the metal to 4-MBA CT occurs, and then the electron is transferred to the semiconductor.[60] The ρCT value for, in particular, the thinnest shells may be limited by charge buildup, which impedes further CT.[61] This effect likely decreases with increasing shell thickness, explaining the higher ρCT observed.

Interestingly, a dependency on the core size is not observed here. In other works,[11,60] there has been a significant increase in ρCT values for excitation wavelengths that were in resonance. Here we observed a ρCT dependency purely determined by incident photon energy, apparently independent of the LSPR position, indicating a difference in mechanism (DICTT here, CID in ref (7), and PIRET in ref (8)). Additionally, no significant shift of the Raman peaks that could be induced by a strong electric field[62,63] is observed for the two core sizes studied here. Only a slight shift is seen in the 1070 cm–1 mode for Ag cubes without a ceria shell (Figure S7), indicative of a difference in hydrogen bonding.[64] This Ag@CeO2 system merits further studies through photocatalytic activity testing and ultrafast spectroscopy to investigate photoinduced formation of oxygen vacancies.

Conclusions

This study focuses on the relation between the dimensions of Ag@CeO2 nanocubes and the efficiency and mechanism of light-induced CT. A 4-MBA linker between core and shell has been used as a probe for light-induced direct CT by Raman spectroscopy. The ceria shell coverage has been determined with FDTD simulations, which describe the distance-dependent refractive index sensitivity, and UV–vis spectroscopy.

The degree of photoinduced CT has been determined by comparing the intensities of totally symmetric and non-totally symmetric modes. For all Ag@CeO2 nanocubes investigated, the degree of CT increases with decreasing excitation wavelength, with a pronounced increase around 514 nm. This trend is fully explicable through a direct CT mechanism from the silver Fermi level to the 4-MBA LUMO. Interestingly, there was no trend seen relating to the core size or to the extinction of the silver nanocube core.

A second general trend observed is a higher CT yield with increasing ceria shell thickness, indicating further electron transfer from the 4-MBA to the ceria shell. It may be that electron transfer from 4-MBA to CeO2 becomes more likely with increasing shell thickness. As this process could lead to formation of oxygen vacancies, this shows promise for photocatalytic application. These results provide deeper understanding of light-induced charge transfer in plasmonic@semiconductor nanocubes.