Hot-Hole-Induced Molecular Scissoring: A Case Study of Plasmon-Driven Decarboxylation of Aromatic Carboxylates

■ RESULTS AND DISCUSSION

Under our experimental conditions, the plasmon-driven decarboxylation reactions occurred on the surfaces of SiO 2 @ Ag SNPs, which executed unique dual-functions as both plasmonic photocatalysts and SERS substrates under nearinfrared excitations. The SiO 2 @Ag SNPs (Figure 1A) were assembled by decorating the outer surfaces of dielectric SiO 2 beads (1.0 ± 0.014 μm in diameter, Figure S1 in the Supporting Information) with a layer of densely packed Ag nanocubes (36 ± 0.79 nm in edge length, Figure S2 in the Supporting Information) following a previously reported protocol. 40 Strong plasmon coupling among the densely packed Ag nanocubes 31,40 on SiO 2 bead surfaces gave rise to a broad plasmon resonance band in the optical extinction spectrum spanning the entire visible and much of the nearinfrared regions (Figure S3 in the Supporting Information). Upon plasmonic excitations, interstitial hot spots with intense local field enhancements exploitable for both photocatalysis and SERS were created inside the nanoscale gaps between adjacent nanocubes. 15,31,40 The Ag surfaces were saturated with self-assembled monolayers of 4-MBA or 2-MBA by incubating the SiO 2 @Ag SNPs with ethanoic solutions of the thiolated molecules. The SiO 2 @Ag SNPs coated with 4-MBA or 2-MBA could be well-dispersed on a poly-4-vinylpyridinefunctionalized Si substrate with interparticle distances typically beyond a few micrometers (Figure S4 in the Supporting Information). We focused a continuous-wave (CW) laser with an excitation wavelength (λ ex ) at 785 nm onto individual 4-MBA-or 2-MBA-coated SiO 2 @Ag SNPs using a confocal Raman microscope (diameter of focal spot: ∼2 μm), such that the molecule-transforming events could be monitored on one SNP at a time through time-resolved SERS measurements (see the schematic illustration in Figure 1B). The pH of the reaction medium was controlled by delivering aqueous solutions of 1.0 mM K 2 CO 3 (pH 10) or HCl-NaOH mixtures The Journal of Physical Chemistry C

(adjustable pH ranging from 2 to 13) into a reaction chamber assembled on a SNPs/Si substrate. Figure 1C schematically illustrates the hot carrier transfer processes and several transient intermediates involved in the plasmon-driven decarboxylation of 4-mercaptobenzoate. Optical excitation of the Ag plasmons at λ ex of 785 nm led to the generation of hot electrons non-thermally distributed above the Fermi level of Ag, while leaving the energetic hot holes below the Fermi level. 4-Mercaptobenzoate chemisorbed on the Ag surfaces served as the hot hole acceptor, while the hot electrons were transferred to solution-phase protons to produce hydrogen atoms. The hot hole injection into 4mercaptobenzoate led to the formation of a carboxyl radical intermediate, which underwent C-C bond cleavage to yield CO 2 and a thiophenyl radical. The surface-adsorbed thiophenyl radical was then rapidly combined with a hydrogen atom to form the final product, TP. 4-Mercaptobenzoate and TP displayed distinct spectral signatures clearly distinguishable in the SERS spectra, which enabled us to track the reaction progress through in situ SERS measurements. The well-defined peaks corresponding to the S-H stretching modes (centered at 2566 cm -1 for 4-MBA and at 2569 cm -1 for TP) in the normal Raman spectra became completely unobservable in the SERS spectra (Figure S5 in the Supporting Information), indicating that both the reactant and product molecules stayed chemisorbed to the Ag surfaces through covalent Ag-S interactions. Figure 1D shows the time-resolved SERS spectra collected from an individual 4-MBA-coated SiO 2 @Ag SNP exposed to 1.0 mM K 2 CO 3 aqueous solution (pH 10) under continuous laser illumination (785 nm, 1.1 mW). Several snapshot SERS spectra captured at various reaction times are shown in Figure 1E, and the temporal evolutions of the relative intensities of several characteristic SERS modes are plotted in Figure 1F. The Raman modes of 1000 and 1022 cm -1 were the spectral signatures of the mono-substituted benzene ring of TP, 32,36,46,47 both of which progressively increased in intensity as TP was produced. The SERS peak at 1420 cm -1 was assigned to the vibrational mode of COO -, ν(COO -), [48][49][50] whose intensity decreased over time during the reactions. The benzene ring mode, ν(CC ring ), down-shifted from 1588 to 1578 cm -1 upon decarboxylation. 32,36 However, due to the limited spectral shift and the overlap of the two peaks, the ν(CC ring ) mode was far from ideal for the quantification of the reactant and product fractions. The characteristic peak of the C-S stretching mode, ν(CS), was centered at 1074 cm -1 for both 4-MBA and TP, 32,36 and no spectral shift of the ν(CS) mode was observed during the reactions. The intensity of the ν(CS) mode gradually increased during the decarboxylation reactions (Figure 1D and Figure S6 in the Supporting Information), and there were several possible reasons for this: (1) The SERS signals of the ν(CS) mode of TP were intrinsically stronger than those of the 4-mercaptobenzoate.

(2) There might be some orientational changes of the molecules induced by the decarboxylation reactions, causing changes in SERS intensities. (3) The interfacial reactions might cause changes of the local electron densities on the Ag surfaces, modifying the local plasmonic field enhancement to a certain extent. While the peak at 1000 cm -1 was absent for 4mercaptobenzoate, it became almost equally intense as the ν(CS) peak at 1074 cm -1 , i.e., I(1000 cm -1 ) ≈ I(1074 cm -1 ), when all 4-mercaptobenzoate molecules were fully converted to TP (see Figures 1E andS5B in the Supporting Information). Therefore, the apparent fraction of the product, θ TP , could be tracked as a function of reaction time, t, based on The Journal of Physical Chemistry C the temporal evolution of the intensity ratio between the 1000 and 1074 cm -1 peaks, I(1000 cm -1 )/I(1074 cm -1 ). To investigate the power dependence of the reaction kinetics, the power of the excitation laser focused on the samples, P ex , was tuned in the range from 320 μW to 5.0 mW. At each P ex , the SERS-based kinetic measurements were repeated on multiple SiO 2 @Ag SNPs one particle at a time. Instead of probing the transforming behaviors of a single molecule or a few molecules in a single hot spot, our time-resolved SERS results provided the ensemble-averaged kinetic information encompassing all molecular adsorbates in the plasmonic hot spots on each SNP. The θ TP trajectories collected from individual SNPs might deviate significantly from each other even under exactly the same set of reaction conditions (Figure S7A-E in the Supporting Information) due to the intrinsic structural heterogeneity among the SNPs and the resulting particle-to-particle deviation of the averaged field enhancements. Although this reaction involved multiple elementary steps, the θ TP trajectories collected from different particles at various P ex values could all be well-described by a simple firstorder kinetic model, strongly suggesting that the overall reaction kinetics were determined by a rate-limiting step. Considering the facts that the hydrogen atoms and thiophenyl radicals were both short-lived and the ultrafast hot carrier transfers took place typically on fs-ps time scales, 2,6,10,51,52 the rate-limiting step was most likely associated with the C-C bond cleavage. While the spectral evolution signifying the reactant-to-product conversion was well-resolved, none of the radical intermediates illustrated in Figure 1C were spectroscopically resolvable by SERS due to the short lifetimes of these transient species.

When comparing the ensemble-averaged trajectories of θ TP , ⟨θ TP ⟩, at various P ex values (Figure S7F in the Supporting Information), we observed a general trend that both the reaction rate and yield increased with the P ex . The observed first-order rate constant, k obs , and the maximal reaction yield achievable at an infinitely long reaction time, θ t=∞ , were obtained by fitting the experimentally measured θ TP trajectories with the following rate equation:

Careful analysis of the time-resolved SERS results enabled us to correlate the k obs and θ t=∞ to the average field enhancements at the single SNP level. In Figure 2A, we plotted the values of k obs obtained from individual 4-MBA-coated SiO 2 @Ag SNPs as a function of the initial SERS intensities of the ν(CS) mode at 1074 cm -1 , I 0 (1074 cm -1 ). We consistently observed that k obs was linearly proportional to I 0 (1074 cm -1 ) across the entire P ex range under current investigations. None of the square, square root, or fourth root of I 0 (1074 cm -1 ), however, held such linear proportionality to the k obs (Figure S8 in the Supporting Information). In Raman scattering, the excitation laser impinging on the sample introduces an external electric field, E 0 (ω L ), oscillating at the laser frequency of ω L , which induces a dipole moment in the molecule oscillating at the Raman frequency of ω R . The enhanced Raman intensity of a molecule, I SERS (ω L , ω R ), is related to the Raman polarizability of the molecule, α(ω R , ω L ), the signal-amplifying factor associated with the local field enhancement in the hot spot, G EM , and the power of the excitation laser, P ex (ω L ), as described by the following equation 53,54 ω ω

where C is a coefficient related to the signal collection efficiency of the optical system used for SERS measurements.

The SERS signals can be enhanced either by increasing the amplitude of α(ω R , ω L ) through chemical enhancement or by increasing the local electric field enhancement. G EM is related to the local-field enhancements at both the excitation laser and the Raman scattering frequencies, as defined by the following equation 53,54 ω ω

where E Loc (ω L ) and E Loc (ω R ) represent the local field intensities at ω L and ω R , respectively. Due to the broad bandwidth of their plasmon resonances, the SiO 2 @Ag SNPs displayed comparable field enhancements over a broad spectral range covering both the excitation and Raman emission wavelengths. 29,31 In this case, the SERS intensity of a molecule became approximately proportional to the fourth power of the local-field enhancement at the excitation wavelength

The SERS intensities of an ensemble of molecules depended not only on the local field enhancements but also on the numbers and orientations of the molecules being probed. For simplicity, we assumed that the total numbers of molecules and the averaged distributions of molecular orientations in the hot spots were about the same among different SiO 2 @Ag SNPs.

With such an assumption, the measured I 0 (1074 cm -1 ) under our experimental conditions became approximately proportional to the fourth power of the average local-field enhancement, |E Loc (ω L )/E 0 (ω L )| 4 , on each SiO 2 @Ag SNP. The results shown in Figure 2A clearly revealed a linear proportionality of k obs to the average |E Loc (ω L )/E 0 (ω L )| 4 at the single SNP level. The slope, φ, of the linear curve fitting results in each panel of Figure 2A reflected how sensitive k obs was to the variation of the average |E Loc (ω L )/E 0 (ω L )| 4 on individual SiO 2 @Ag SNPs at each P ex . It is worth mentioning that the relationships between reaction rates and local-field enhancements are intimately tied to the detailed reaction mechanisms and, thus, may vary drastically from reaction to reaction. Even for the same plasmon-driven reaction, highly divergent results may be observed upon subtle variation of the photocatalyst structures and detailed reaction conditions, as manifested by the para-nitrothiophenol coupling reactions driven by plasmonic hot electrons. [55][56][57][58] Although a simple linear relationship between k obs and |E Loc (ω L )/E 0 (ω L )| 4 was clearly observed for this plasmon-driven decarboxylation reaction, it may not be readily applicable to other plasmon-mediated molecular transformations. For many plasmon-driven photocatalytic reactions, it is challenging to completely distinguish the effects of hot carriers vs local field enhancements because the abundance of hot carriers is tied to the local field enhancements. In some other cases, the photocatalytic reactions can be purely driven by the enhanced near-fields on the nanostructure surfaces without the involvement of hot carriers. 59,60 The dependency of the reaction rates on the local field enhancements is a critical factor we should carefully consider when designing metallic nanophotocatalysts for specific plasmon-driven reactions, and the time-resolved SERS provides a generic spectroscopic tool enabling us to

The Journal of Physical Chemistry C quantitatively correlate the reaction kinetics to the local field enhancements.

The non-radiative decay of plasmons gives rise to the formation of short-lived, non-thermal hot carriers that undergo rapid relaxation processes over a broad distribution of time scales ranging from femtoseconds to nanoseconds. 2 These energetic hot carriers can either be transferred to molecular adsorbates to drive photocatalytic reactions or relax into lower energy charge carriers distributed fairly close to the Fermi level of the metals through cascading inelastic electron-electron and electron-surface scattering processes. The resulting lowerenergy electrons then get thermalized by coupling with the phonon modes to generate heat, a process known as plasmonic photothermal conversion, which may also kinetically boost molecular transformations on the surfaces of light-illuminated plasmonic nanostructures. However, it has long been challenging to unambiguously delineate the thermal and nonthermal effects involved in plasmon-mediated surface chemistry 20,[61][62][63][64][65][66][67][68][69][70] largely due to the mechanistic complexity of the plasmon-driven reactions and the challenges associated with precise measurement of the local temperature on the photocatalyst surfaces. Whether a specific plasmon-driven reaction is dominated by the non-thermal or thermal effects has been a subject under intense debate, and it has become increasingly evident that some important conclusions drawn from earlier studies are well worthy of cautious re-examination and further scrutiny. 20,[49][50][51][52][53][54][55][56][57] The power dependence of reaction rates provides important implications that may help us distinguish the photothermal from non-thermal photochemical effects. 6,20,61,65 As shown in Figure 2B, the I 0 (1074 cm -1 ) averaged over all SiO 2 @Ag SNPs measured at each P ex , ⟨I 0 (1074 cm -1 )⟩, was directly proportional to P ex , which was fully consistent with eq 5. However, the ensemble-averaged k obs values, ⟨k obs ⟩, exhibited a superlinear relationship to P ex (Figure 2C). We performed least-squares curve fitting to the experimental data with the following equation (8) in which ⟨k obs ⟩ was equal to the nth power of P ex multipled by a scaling coefficient, B. The best fitting result was obtained when n took the value of 1.45. Because of the superlinear The SERS spectra were collected at room temperature with a P ex value of 5.0 mW and a spectral integration time of 1 s. (B) I(1000 cm -1 )/I(1074 cm -1 ) at various temperatures. The intensity ratios shown in panel B were the average values of the results collected on 10 SiO 2 @Ag SNPs at each temperature, and the error bars represented the standard deviations. The values of I(1000 cm -1 )/I(1074 cm -1 ) exhibited a linear relationship to temperature, and the fitting results were shown as a solid orange line in panel B. (C) Representative SERS spectra of TP on individual SiO 2 @Ag SNPs prior to laser illumination, after continuous illumination at a P ex value of 5.0 mW for 30 s in water and after continuous illumination at a P ex value of 5.0 W for 30 s in air. (D) Temporal evolution of I(1000 cm -1 )/I(1074 cm -1 ) during continuous laser illumination (λ ex of 785 nm; P ex of 5.0 W) collected from 10 individual TP-coated SiO 2 @Ag SNPs in water (left panel) and in air (right panel). (E) Ensemble-averaged I(1000 cm -1 )/I(1074 cm -1 ) trajectories of TP-coated SiO 2 @Ag SNPs during continuous laser illumination at a P ex value of 5.0 mW for 30 s in water and in air. (F) Temporal evolution of the temperatures at the SNP surfaces (calculated based on Raman thermometry) under continuous laser illumination at a P ex value of 5.0 W in water and in air.

The Journal of Physical Chemistry C power dependence, k obs became more sensitive to the variation of |E Loc (ω L )/E 0 (ω L )| 4 as P ex increased, resulting in larger φ values at higher P ex values (Figure 2D).

As evidenced by ample examples in the literature, the rate of a plasmonic hot-carrier-driven reaction is typically proportional to the excitation power under CW illumination at moderate photon power densities. 6,9,19,20,61,71 However, such a linear power dependence may switch to a superlinear power dependence when reaching a sufficiently high power density regime or under ultrafast pulsed laser illumination as a consequence of multiphoton absorption. Therefore, the transition from the linear to the superlinear power dependence over a certain power density regime has been considered as a hallmark of hot-carrier-driven photocatalysis. 6,9,19,20,61,71 However, the non-linearity of power dependence may also originate solely from the photothermal heating without the involvement of the hot carriers. 61 For a photothermally activated reaction, the rate constant typically obeys the Arrhenius-type temperature dependence. 61,62 Assuming that the temperature elevation due to light absorption was approximately proportional to the incident power impinging on the sample surfaces under our experimental conditions, the relationship between k obs and P ex could be expressed by the following equation

where A was the pre-exponential factor, E a was the activation energy, R was the gas constant, T 0 was the initial temperature before light illumination (295 K in our case), and c was a scaling factor relating the temperature elevation to the illumination power. The experimental results shown in Figure 2C could be fitted almost equally well with eqs 8 and 9 (Figure S9 in the Supporting Information), and it was highly likely that the observed power dependence reflected a superposition of thermal and non-thermal components.

To gain further insight into the relative contributions of the non-thermal hot carriers vs photothermal heating, we first conducted an important control experiment, in which the 4-MBA-coated SiO 2 @Ag SNPs were exposed to an aqueous solution of 1.0 mM K 2 CO 3 maintained at 90 °C for 30 min without laser illumination. After the thermal treatment, SERS spectra were collected on individual SNPs at room temperature. As shown in Figure S10 in the Supporting Information, all of the SERS features of 4-mercaptobenzoate were wellpreserved without any detectable SERS peaks of TP, indicating that the decarboxylation of 4-mercaptobenzoate was essentially a hot-carrier-driven photoreaction rather than a thermally activated reaction. Next, we scrutinized the rationality of the E a and c values obtained from the curve fitting. Fitting the experimental results with eq 9 yielded an E a value of 15.4 ± 0.94 kJ mol -1 , which appeared too low for a reaction involving the cleavage of a C-C covalent bond under thermal conditions. With a c value of 146 ± 39 K mW -1 , the temperature at the photocatalyst surfaces under laser illumination at a P ex value of 5.0 mW would exceed 1000 K, at which the melting of Ag nanoparticles and thermal decomposition of the organic adsorbates were both expected to occur but neither was observed experimentally. Therefore, although the experimental results could be well-fitted with the Arrhenius-type power dependence, this decarboxylation reaction was unlikely to be driven solely by photothermal heating. Last and most importantly, we further developed a semi-quantitative Raman thermometry to probe the local temperature at the photocatalyst surfaces, based upon which we were able to directly evaluate the contribution of photothermal heating.

TP chemisorbed to the SiO 2 @Ag SNPs exhibited unique temperature-dependent SERS features, which enabled us to use TP as a probe molecule for the Raman thermometry. Upon thermal heating of the TP-coated SiO 2 @Ag SNPs to higher temperatures, both I(1000 cm -1 )/I(1074 cm -1 ) and I(1022 cm -1 )/I(1074 cm -1 ) decreased and the ν(CS) mode slightly down-shifted (Figure 3A). These spectral changes caused by thermal treatment were found to be irreversible, all of which were well-preserved after the samples were cooled down to room temperature. Such thermally induced changes in the SERS spectra of TP were also previously observed on other nanostructured Ag surfaces, 58,72 which were caused by conformational changes of the surface-adsorbed TP, as suggested by DFT calculations. 58 The values of I(1000 cm -1 )/I(1074 cm -1 ) varied linearly with respect to temperature changes (Figure 3B), providing a working curve for us to calculate the local surface temperatures in the plasmonic hot spots where the TP molecules resided.

The SERS spectra of individual TP-coated SiO 2 @Ag SNPs after exposure to continuous laser illumination at a P ex value of 5.0 mW (the maximal P ex investigated in this work) revealed remarkably different temperature elevation profiles in H 2 O and in air (Figure 3C andD). Upon laser illumination, the surface temperature of the SNPs gradually increased, reaching a steady state value when the equilibrium between heat generation and dissipation was established after approximately 15 s. When the SiO 2 @Ag SNPs were exposed to air, their surface temperature increased from 22 to ∼75 °C after laser illumination for 30 s, whereas a limited temperature elevation by only ∼13 °C was achieved on SNPs exposed to H 2 O (Figure 3E andF). Therefore, the plasmonic heating effects were rather insignificant during the plasmon-driven decarboxylation reactions. The steady state values of I(1000 cm -1 )/I(1074 cm -1 ) in the SERS spectra of TP on SiO 2 @Ag SNPs under continuous laser illuminations remained around 1 over the entire laser power range we investigated, dropping slightly from ∼1.03 at a P ex value of 0.32 mW to ∼0.97 at a P ex value of 5.0 mW. Although the photothermal transduction efficiency was an intrinsic property of the SNPs themselves, the steady state surface temperatures of the laser-illuminated SNPs also critically depended upon heat dissipation, which was influenced by the thermal conductivity of the local medium surrounding the SNPs. 61,73,74 Significantly higher temperature elevation was observed on SNPs in air than those in H 2 O because the thermal conductivity of H 2 O (∼0.6 W m -1 K -1 ) was more than 20 times higher than that of air (0.025 W m -1 K -1 ). We also studied the temperature dependence of the decarboxylation rates under laser illumination at a P ex value of 0.32 mW using a heating plate in contact with the Si substrates to control the bulk reaction temperatures. At such a low P ex , the plasmonic photothermal heating effect became negligible and we assumed that temperatures of the medium in the reaction chambers were well-maintained around the targeted temperatures. When raising the reaction temperature from 22 to 40 °C, the reaction rates and yields only slightly increased by less than 10% (Figure S11 in the Supporting Information), further verifying the insignificant contribution of photothermal heating to the kinetic enhancement even at P ex values up to 5.0 mW. Therefore, the observed non-linear power dependence of k obs was mostly like a characteristic of a hot-carrier-driven

The Journal of Physical Chemistry C reaction involving multiphoton absorption rather than a photothermally activated reaction. Although the total excitation powers incident on the samples were in the range of sub-mW to mW, the photon power densities in the focal point (∼2 μm in diameter) of the confocal laser beam were as high as several tens of kW cm -2 , far exceeding those of the light sources typically used for plasmon-driven photocatalysis. 6,[19][20][21]61,67,68 Because of the high local power densities and intense local-field enhancements, it was not surprising that multiphoton processes became pronounced even under CW laser illumination at mW-level P ex values.

In the SERS measurements, we probed the transforming behaviors of molecular adsorbates residing in the plasmonic hot spots rather than all molecules on the photocatalyst surfaces. Because the reaction rates were sensitively dependent on the local-field enhancements, the decarboxylation reactions occurred regioselectively in the nanoscale hot spots, whereas the molecules in the "electromagnetically cold" surface regions remained unreacted. Therefore, the values of θ t=∞ extracted from the SERS results represented the apparent reaction yields of the molecules in the hot spots. Even so, not all 4mercaptobenzoate molecules in the hot spots got decarboxylated within the time duration of our kinetic measurements (Figure S7 in the Supporting Information), resulting in nonunity values of θ t=∞ . Although k obs varied significantly from particle to particle due to the variation in average field enhancements among the SNPs, θ t=∞ was uncorrelated to k obs at the single SNP level, exhibiting very similar values at a given P ex (Figure 4A). θ t=∞ appeared almost independent of I 0 (1074 cm -1 ) at each P ex (Figure 4B), while the ensemble-averaged θ t=∞ , ⟨θ t=∞ ⟩, increased with P ex , reaching more than 90% yields at a P ex value of 5.0 mW (inset of Figure 4B).

At first glance, the non-unity θ t=∞ values, especially those observed at low P ex values, could be possibly interpreted in the context of reversible decarboxylation/carboxylation processes under laser illumination. Increasing P ex selectively boosted the decarboxylation of 4-mecarptobenzoate with respect to the carboxylation of TP, thereby shifting the equilibrium toward TP. However, lack of correlation between θ t=∞ and k obs strongly suggested that the non-unity θ t=∞ values were unlikely to be a consequence of dynamic equilibrium between decarboxylation and carboxylation. To gain further insights into the reaction reversibility, we carried out P ex -programmed kinetic measurements, in which the 4-MBA-coated SiO 2 @Ag SNPs were exposed to 1.0 mM K 2 CO 3 under continuous illumination by 785 nm laser at varying P ex . As shown in Figure 4C, we started with a low P ex value at 0.32 mW and then progressively increased P ex to 5.0 mW in a stepwise manner. At each P ex , we collected the SERS spectra continuously until θ TP reached a plateau. Every time we increased P ex , a higher fraction of TP was produced and θ t=∞ increased until exceeding 90% at a P ex value of 5.0 mW. Upon a decrease in P ex , no reserve reaction was observed within the time windows of our measurements, and θ t=∞ remained above 80% even at a P ex value as low as 0.32 mW. The results of P ex -programmed SERS measurements strongly indicated that the plasmondriven decarboxylation reaction was an irreversible process. We hypothesized that the molecules became activated for the photocatalytic decarboxylation only when the local-field intensities (|E Loc (ω L )|) at the molecule-occupying sites exceeded a certain threshold value, whereas the molecules remained unreactive when residing in the colder regions where |E Loc (ω L )| was below the threshold value. Although the average field enhancements, |E Loc (ω L )/E 0 (ω L )|, in each SNP varied from particle to particle, the threshold |E Loc (ω L )| value required for the molecular activation stayed essentially the same for all SNPs at all P ex values. An increase in P ex resulted in higher fractions of the molecules occupying the surface sites Both the rates and the yields of the decarboxylation reactions changed significantly upon variation in the pH of the medium to which the 4-MBA-coated SiO 2 @Ag SNPs were exposed, exhibiting an interesting volcano-type pH dependence at P ex values of both 1.1 (Figure 5A and Figure S12 in the Supporting Information) and 5.0 mW (Figure 5B and Figure S13 in the Supporting Information). In this set of experiments, the pH of the reaction medium was adjusted in the range of 2-13 using HCl-NaOH mixtures. At a pH of 10, both the reaction rates and yields remained essentially unchanged when switching from 1.0 mM K 2 CO 3 to the HCl-NaOH system (Figures S7F,S12G, and S13G in the Supporting Information). In an acidic environment at a pH below 4, the decarboxylation reactions were observed to be kinetically sluggish with very limited reaction yields below ∼15%. As the pH of the reaction medium went up, both k obs and θ t=∞ increased remarkably, approaching the maximal values in the pH range of 10-11. It has been previously reported that deprotonated 4-MBA on Ag and Au nanoparticle surfaces underwent decarboxylation reactions upon optical excitation of the nanoparticle plasmons, whereas protonated 4-MBA remained unreactive. 32,36 Upon receiving a plasmonic hot hole from Ag, a surface-adsorbed 4-mercaptobenzoate anion was converted to a carboxyl radical, which readily underwent CC bond cleavage to release CO 2 . A sharp increase of k obs and θ t=∞ observed over the pH range of 4-7 correlated well with the deprotonation of 4-MBA in this particular pH window, which was further verified by the SERS spectra collected from 4-MBA-coated SiO 2 @Ag SNPs at different pHs.

As shown in Figure 5C, the protonated 4-MBA in an acidic environment exhibited a broad SERS peak centered at 1680 cm -1 , which was assigned to the CO vibrational mode, ν(CO), of the protonated carboxylic acid group. [48][49][50] Deprotonation of 4-MBA led to the disappearance of the ν(CO) mode, accompanied by the emergence of the ν(COO -) mode of the carboxylate group at 1420 cm -1 . The major changes in the SERS spectral features of 4-MBA upon deprotonation/protonation were well-reproduced computationally through DFT calculations of the Raman spectra of 4-MBA covalently linked to Ag atomic clusters (Figure S14 in the Supporting Information). According to the evolution of the relative SERS intensities of the ν(CO) and ν(COO -) modes (Figure 5D), the pK a of the surface-adsorbed 4-MBA was determined to be around 5, which was consistent with previously reported values. [48][49][50] Therefore, the pH dependence of k obs and θ t=∞ was directly tied to the relative fractions of the protonated and deprotonated forms of 4-MBA in the plasmonic hot spots.

To gain further insight into the pH-dependent kinetics and reactivity, we used DFT to calculate the frontier orbital energies of protonated and deprotonated 4-MBA covalently linked to Ag n clusters (n refers to the number of atoms in the Ag clusters). As shown in Figure S15 in the Supporting Information, the chemisorption of both the protonated and deprotonated 4-MBA to Ag resulted in significant narrowing of the band gaps between the HOMO and the LUMO, shifting both orbitals closer to the Fermi level of Ag. At a λ ex value of 785 nm (1.58 eV), the plasmonic hot electrons were distributed in the energy range within 1.58 eV above the Fermi level, and thus were not energetic enough to get injected into the LUMO of the 4-MBA-Ag n complexes. Therefore, it was the solution-phase protons rather than the surfaceadsorbed 4-MBA or 4-mercaptobenzoate that served as the The Journal of Physical Chemistry C acceptor of hot electrons during the decarboxylation reactions. On the other hand, a fraction of the hot holes (distributed between the Fermi level and 1.58 eV below the Fermi level) had sufficient energy to transfer to the HOMO of the molecular adsorbates. In comparison to those of their protonated counterparts, the HOMOs of deprotonated 4-MBA-Ag n complexes were further shifted toward the Ag Fermi level. From the band alignment point of view, deprotonation of 4-MBA on Ag was more favorable for the hot hole transfer and the acceleration of decarboxylation reactions than the protonated 4-MBA. According to the DFTcalculated electron density maps of the HOMOs of 4-MBA-Ag n complexes (Figure S16 in the Supporting Information), deprotonation of the carboxylic acid group resulted in increased electron densities near the Ag n clusters, making the deprotonated 4-MBA able to more readily accept the hot holes ejected from Ag than the protonated 4-MBA. As a result, the deprotonated 4-MBA was more reactive than the protonated 4-MBA. An alternative possible mechanism of the plasmondriven decarboxylation of 4-MBA involves the hot-holeinduced oxidation of H 2 O into OH• radicals, which then initiate the decarboxylation of the carboxylates. 75 Considering the fact that the redox potentials of H 2 O/OH• lie more than 2 eV below the Fermi level of Ag, 76,77 the hot holes are not energetic enough to drive the evolution of OH• from H 2 O when the Ag plasmons are excited at 785 nm. Therefore, the plasmon-driven decarboxylation reactions observed under our experimental conditions were essentially induced by hot hole injection into surface-adsorbed 4-mercaptobenzoate rather than being driven by the photogenerated OH• radicals. Interestingly, we observed a drastic decrease in both k obs and θ t=∞ in highly basic environments (pH > 11) as the pH further increased (Figure 5A andB), even though all of the surfaceadsorbed 4-MBA existed in the deprotonated form. The depletion of hot holes in Ag was compensated by the hot electron injection into protons, which constituted the other half reaction of the redox cycle. The standard redox potential for proton reduction, U(H + /H 2 ), at a pH of 0 was 4.5 eV below the vacuum level, which was well aligned with the Fermi level of Ag (4.3 eV below vacuum). U(H + /H 2 ) increased as the reaction medium became less acidic, exhibiting a linear dependence on the pH that could be fully described by the classic Nernst equation. In an alkaline environment, U(H + /H 2 ) was shifted further above the Ag Fermi level as the pH increased, resulting in a declined population of the hot electrons with sufficient energies to get injected into the protons. As schematically illustrated in Figure 5E, the pH dependence of reaction rates and yields observed in this work could be well-interpreted in the context of the relative energies of the HOMO of the hot hole acceptors and the redox potentials of the hot electron acceptors, both of which could be modulated by varying the pH of the reaction medium.

Analogous to 4-MBA, 2-MBA, when deprotonated, also underwent a decarboxylation reaction to produce TP on the surfaces of laser-illuminated SiO 2 @Ag SNPs, as schematically illustrated in Figure 6A. Although a strong peak corresponding to the S-H stretching mode of 2-MBA was observed at around 2530 cm -1 in the normal Raman spectrum, it became undetectable in the SERS spectrum (Figure S17 in the Supporting Information), indicating that 2-MBA was chemisorbed to Ag surfaces upon formation of a covalent Ag-S The Journal of Physical Chemistry C bond rather than physically adsorbed on the Ag surfaces. The strongest peak in the SERS spectrum of deprotonated 2-MBA on SiO 2 @Ag SNPs at a pH of 10 was observed to be the ν(C-S) stretching mode centered at 1028 cm -1 . The benzene ring mode, ν(CC ring ), split into a double-peaked spectral feature (1558 and 1578 cm -1 ) due to the structural asymmetry of molecules. The calculated Raman spectra of 2-MBA and 2-MBA-Ag n complexes were in good agreement with the experimentally measured Raman and SERS spectra in terms of the positions of the major peaks (Figure S18 in the Supporting Information); however, the DFT calculations were incapable of precisely predicting the relative intensities of various vibrational modes in the SERS spectra because the simple atomic cluster model we used did not take the surface orientation of the molecule adsorbates and the electromagnetic enhancement effects of SERS into consideration.

Figure 6B shows the time-resolved SERS spectra collected from a 2-MBA-coated SiO 2 @Ag SNP exposed to 1.0 mM K 2 CO 3 under continuous laser illumination at a P ex value of 2.6 mW. Several snapshot spectra at various reaction stages were highlighted in Figure 6C. As the decarboxylation reaction proceeded, the SERS signatures of TP at 1000 and 1074 cm -1 emerged and gradually intensified, while the intensity ratios between the two peaks, I(1000 cm -1 )/I(1074 cm -1 ), remained very close to 1 (Figure 6D). The SERS peak of TP at 1022 cm -1 spectrally overlapped with the ν(C-S) mode of 2mercaptobenzoate at 1028 cm -1 , giving rise to a doublepeaked, broadened spectral feature in the wavenumber range of 1018-1032 cm -1 when both the reactant and product molecules coexisted in the plasmonic hot spots. The ν(CC ring ) modes of TP and 2-mercaptobenzoate also overlapped at 1578 cm -1 . More detailed spectral features in the wavenumber ranges of 975-1175 and 1475-1675 cm -1 were highlighted in Figure S19 in the Supporting Information. During the reactions, several characteristic SERS peaks of 2-mercaptobenzoate centered at 1128, 1116, and 1150 cm -1 all became remarkably stronger (Figure 6E), while both the peak positions and relative peak intensities remained essentially unchanged. Such intensified SERS signals of 2-mercaptobenzoate might originate from the orientational changes of 2-mercaptobenzoate with respect to the Ag surfaces when some of the adsorbate molecules were converted into TP. How the molecular orientations are influenced by the decarboxylation reactions, however, still needs to be further investigated in greater detail. Although the temporal evolution of SERS spectral features allowed us to track the plasmon-driven decarboxylation process, the intrinsic spectral complexity of 2mercaptobenzoate, the variation of the SERS intensities, and the spectral overlap between several major SERS peaks kept us from being able to extract the k obs and θ t=∞ values from the time-resolved SERS results. Therefore, it remains challenging to quantitatively compare the decarboxylation rates of 2mercaptobenzoate to those of 4-mercaptobenzoate under specific reaction conditions.

The decarboxylation of 2-mercaptobenzoate was verified as a plasmon-driven photochemical reaction rather than a thermally activated process because no reaction was observed under thermal conditions even at temperatures up to 90 °C (Figure S20 in the Supporting Information). Similar to 4-mercaptobenzoate, 2-mercaptobenzoate underwent fast decarboxylation reactions on the surfaces of laser-illuminated SiO 2 @Ag SNPs at a pH of 10 but became essentially unreactive at a pH of 2 (Figure S21 in the Supporting Information). In this acidic medium, the intensities of all of the SERS peaks of the protonated 2-MBA remained unchanged under continuous laser illumination, further indicating that the amplification of 2mercaptobenzoate SERS signals observed at a pH of 10 (Figure 6B andC) was associated with the photochemical decarboxylation reactions rather than being photophysically induced by the laser illumination. Analogous to the case of 4-MBA, the deprotonation of the carboxylic acid group shifted the HOMO of the surface-adsorbed 2-MBA closer to the Ag Fermi level (Figure S22 in the Supporting Information), thereby making the deprotonated form energetically more favorable for the metal-to-adsorbate transfer of hot holes than the protonated form.