Plasmon-driven oxidative coupling of aniline-derivative adsorbates: A comparative study of para-ethynylaniline and para-mercaptoaniline

Relaxation of optically excited free electron oscillations in metallic nanostructures, known as localized plasmons, gives rise to a series of intriguing photophysical effects, such as nonthermal distribution of hot charge carriers, local-field enhancements, and photothermal transduction, all of which can be deliberately harnessed to catalyze interfacial photochemical processes, including both bond-cleaving molecular scissoring and bond-forming coupling reactions.^1–12 Photocatalytic molecular transformations on nanostructured metal surfaces may occur through several distinct plasmon-mediated reaction pathways involving plasmon-enhanced intramolecular electronic excitations,^13–16 hot carrier injection into the molecular adsorbates,^17–21 or photothermal activation of certain chemical bonds.^22–24 In numerous cases, the kinetic profiles associated with plasmon-mediated surface photochemistry are intimately tied to the strong interplay among hot carriers, field enhancements, and photothermal heating rather than being dictated solely by one specific plasmon-derived effect.^7,25–31 Besides their photocatalytic functions, metallic nanostructures can also serve as substrates for plasmon-enhanced Raman spectroscopies, such as surface-enhanced Raman scattering (SERS) and tip-enhanced Raman scattering (TERS).^32–38 Because the hot spots for Raman signal enhancement spatially coincide with the active sites for photocatalysis on the nanostructure surfaces, plasmon-enhanced Raman spectroscopies can be utilized as a unique in situ spectroscopic tool with molecular fingerprinting capability to fine-resolve the detailed structural evolution of the transforming molecular adsorbates,^25,39–43 a key step toward thorough understanding of the plasmon-mediated reaction mechanisms.

Plasmon-driven oxidative coupling of aniline-derivatives represents a paradigm-shifting approach to the synthesis of aromatic azo compounds. The discovery and mechanistic studies of this type of reactions trace back to the 1990s when Osawa et al.⁴⁴ first observed three unexpected peaks at 1438, 1388, and 1140 cm⁻¹ in the SERS spectra of para-mercaptoaniline (pMA) chemisorbed to roughened Ag surfaces, whereas none of these peaks were observable in the normal Raman (NR) spectra of pMA. The origin of these anomalous SERS peaks had been under intense debate^45–51 until 2010 when Huang et al.⁵² convincingly proved that surface-adsorbed pMA underwent a bimolecular coupling reaction to produce p,p′-dimercaptoazobenzene (DMAB) under laser illumination during the SERS measurements. Over the past decade, this reaction has become a testbed for developing detailed understanding of the complex reaction mechanisms underpinning plasmon-driven photocatalysis. The rates and yields of the pMA coupling reactions vary significantly from case to case, depending on the plasmonic characteristics of the photocatalysts, the optical excitation conditions, and the local environments in which the reactions occur.^{47,49,53–56} Despite the seemingly divergent kinetic results reported in the literature, a unified mechanistic picture has started to emerge, built upon the mechanistic insights gained from deliberately designed SERS/TERS measurements. Under anaerobic reaction conditions, the plasmonic hot holes serve as the primary driving force for the oxidative coupling of pMA, and the reactions can be kinetically boosted with the aid of electron acceptors, which effectively promote the charge carrier separation.^57,58 In aerobic reaction environments, however, this oxidative coupling reaction switches to a more efficient, hot electron-driven pathway in which the plasmonic hot electrons are injected into the antibonding orbitals of surface-adsorbed O₂, leading to the formation of highly reactive anionic O₂⁻ radicals that are capable of oxidizing pMA.^56,59 When photoactivated O₂⁻ reaches the steady-state coverage on the photocatalyst surfaces, this reaction obeys the first-order rate law, suggesting that the rate-limiting step is the oxidation of pMA by O₂⁻ rather than the photoactivation of O₂ or the azo bond formation.⁶⁰ The dimerization process imposes certain geometric requirements on the molecular adsorbates regarding their surface orientations and structural flexibility. As revealed by recent TERS studies, structurally flexible pMA molecules on ill-defined Ag nanoparticle (NP) surfaces rapidly couple into DMAB, whereas a structurally ordered pMA monolayer on a well-defined Ag {111} surface remains unreactive under identical photoexcitation conditions.⁶¹ In plasmon-driven dimerization of para-nitrothiophenol (pNTP), which is the reductive analog of the pMA coupling reaction, the reactivity has also been observed to be closely related to the surface orientation and structural flexibility of pNTP adsorbates.^62,63 These observations strongly suggest that we should take not only the plasmonic effects but also the adsorbate structures into careful consideration when designing metal–adsorbate systems for plasmon-driven photocatalysis.

While organothiols constitute so far the most intensively studied molecular adsorbate system, molecules can also use a variety of alternative chemical moieties, such as diazonium,^64–71 isocyanide,^72–77 N-heterocyclic carbene,^78–83 and alkyne,^84–87 as the anchoring groups to form covalent interactions with metal surfaces. The plasmon-mediated bimolecular coupling reactions are not restricted to thiolated aniline-derivatives only but may occur between nonthiolated aniline-derivatives as well.⁸⁸ However, it remains unclear how the nonthiolated aniline-derivatives behave differently from their thiolated counterparts due to lack of rigorous comparative studies in the literature. In this work, we systematically compare the plasmon-driven transforming behaviors of an alkynylated aniline-derivative, para-ethynylaniline (pEA), to those of the benchmarking thiolated aniline-derivative, pMA. Analogous to pMA, pEA may also undergo plasmon-driven oxidative coupling reactions to form an azo compound, p,p′-diynylazobenzene (DYAB), on Ag NP surfaces, as illustrated in Scheme 1. While the Ag–thiol interactions lead to the formation of Ag–S bonds through σ electron donation, the interactions between Ag and an ethynyl group involve both σ-donation and π-back donation, forcing the alkynylated adsorbates to adopt adsorption conformations remarkably different from those of their thiolated counterparts.^85,86 The difference in bonding nature may also result in very different local structural flexibilities when the molecules are chemisorbed to metal surfaces. Therefore, pEA and pMA represent two contrasting types of aniline-derivatives that are well-worthy of detailed comparative studies. Here, we used SERS as an in situ spectroscopic tool with the aid of density functional theory (DFT) calculations to correlate the kinetic characteristics of the plasmon-driven coupling reactions to the detailed structures of the transforming adsorbates on Ag NP surfaces. Insights gained from our systematic comparative studies shed light on the crucial roles of the metal–adsorbate interactions in determining the adsorbate conformations and molecule-transforming kinetics during plasmon-driven photocatalytic reactions.

We used Ag NPs as the building blocks to assemble close-packed NP array structures, which played a dual role as both the substrates for SERS and the plasmonic photocatalysts for the coupling reactions. Colloidal Ag NPs were synthesized through seed-mediated nanocrystal growth using cetyltrimethylammonium chloride (CTAC) as the surface-capping molecular ligands. The as-synthesized Ag NPs exhibited a uniform quasi-spherical morphology [Fig. 1(a)] with diameters of 45 ± 4.2 nm [Fig. 1(b)]. The colloidal Ag NPs displayed a well-defined plasmon resonance peak centered at ∼412 nm, which matched the results calculated by the Mie scattering theory in terms of both resonance wavelengths and spectral lineshapes [Fig. 1(c)]. The CTAC ligands on the surfaces of the as-synthesized Ag NPs could be readily displaced by pEA and pMA, both of which were bound to Ag surfaces more strongly than CTAC. The Ag NPs were further assembled into close-packed array structures with a high density of sub-10 nm interparticle gaps [Figs. 1(d) and S1 in the supplementary material] through slow solvent evaporation of aqueous colloidal inks containing CTAC-, pEA-, or pMA-coated Ag NPs on indium tin oxide (ITO)-coated glass substrates. The strong plasmon coupling among neighboring Ag NPs led to the emergence of a broad plasmon resonance band across the visible and near-infrared spectral regions [Fig. 1(e)]. Upon resonant excitations of the plasmons in the Ag NP arrays using a continuous wave (cw) near-infrared laser (785 nm), the local electric fields were enormously enhanced in the interparticle gaps, creating electromagnetic hot spots for not only SERS but plasmon-driven photocatalysis as well. To ensure that the CTAC ligands were completely displaced by pEA or pMA, we incubated the as-synthesized colloidal Ag NPs with 1.0 mM of pEA or pMA for 2 h. The pEA- and pMA-coated colloidal Ag NPs were then centrifuged and redispersed in water before eventually self-assembled into the array structures for SERS measurements. To fully resolve the spectral features of pEA and pMA adsorbates without the interference by any photochemical reactions, the SERS spectra were collected at a low laser excitation power (P_ex) of 0.5 mW with a spectral acquisition time of 1 s while the SERS substrates were exposed to an aqueous solution of 100 mM ammonia borane, a reducing agent that effectively prohibited the oxidative coupling of the aniline derivatives. After ligand exchange, all the SERS spectral features of CTAC vanished, while the characteristic peaks of pEA or pMA, such as the benzene ring breathing (ν_{CC ring}), C≡C stretching (ν_C≡C), C–S stretching (ν_CS), and C–H in-plane scissoring (β_C-H) modes, emerged in the SERS spectra (Fig. S2 in the supplementary material). In particular, the disappearance of the strong Ag–Cl vibrational peak (ν_AgCl) at ∼190 cm⁻¹ together with the emergence of the characteristic ν_AgC (440 cm⁻¹) and ν_AgS (390 cm⁻¹) modes verified that CTAC was fully exchanged with pEA and pMA, which were chemisorbed to the Ag NP surfaces through Ag–C and Ag–S bonds, respectively.

Careful comparison of the NR and SERS spectral features in combination with DFT calculations enabled us to gain detailed insights into the chemical nature of the metal–adsorbate interactions. The NR spectrum collected from a solid film of pEA was dominated by three vibrational peaks corresponding to the ν_C≡C, ν_{CC ring}, and β_CH modes, respectively, all of which were also clearly observed in the SERS spectrum of pEA on the Ag NP surfaces [Fig. 2(a)]. Although the ν_{CC ring} and β_CH modes both exhibited similar Raman shifts in the NR and SERS spectra, the ν_C≡C mode at 2095 cm⁻¹ in the NR spectrum downshifted to 1945 cm⁻¹ in the SERS spectrum, indicating that the chemisorption of pEA to Ag resulted in significant weakening of the C≡C bond in the ethynyl group. We also used DFT to calculate the Raman spectra of pEA and a pEA molecule covalently linked to an Ag₁₄ atomic cluster (represented as a unit cell of a face-centered cubic crystalline lattice). When calculating the NR spectrum of pEA, we only included a single pEA molecule without considering the intermolecular interactions possibly existing in the solid-state samples, which might give rise to some minor differences between the calculated and measured NR spectral features. For example, in the DFT-calculated NR spectrum of pEA [Fig. 2(a)], two Raman peaks with comparable intensities were observed at 1158 and 1193 cm⁻¹, which were associated with the asymmetric and symmetric β_CH modes, respectively. Both peaks were slightly upshifted in the experimentally measured NR spectrum of pEA, and the symmetric β_CH mode became significantly weaker than the asymmetric β_CH mode. Although DFT was incapable of accurately predicting the relative peak intensities in the SERS spectra because it calculated the NR spectrum of the pEA-Ag₁₄ compound rather than the SERS of pEA on a metal surface, the Raman shifts of all the major peaks calculated by DFT were in very good agreement with those in the experimentally measured SERS spectrum [Fig. 2(a)]. The significant spectral downshift of the ν_C≡C mode by ∼150 cm⁻¹, which was both experimentally observed and computationally verified, signified the hybrid σπ bonding nature of the Ag–ethynyl interactions. The σ component arose from the overlap of a filled π orbital of the alkyne with the conduction band of Ag, while the π component was caused by the overlap of an unoccupied π* orbital of the alkyne with the valence band of Ag.⁸⁶ The hybrid σπ bonding required the ethynyl group of pEA to be significantly tilted toward the Ag surface, as shown in the DFT-optimized conformation of pEA-Ag₁₄, denoted as pEA-Ag₁₄-o [Fig. 2(b)]. When the pEA molecule chemisorbed to Ag₁₄ through a atop conformation with the C≡C bond aligned vertically to the Ag surface (denoted as pEA-Ag₁₄-v), the Ag–ethynyl interaction was dominated by σ donation without the contribution of π-back donation. The DFT results showed that the ν_C≡C mode of the pEA-Ag₁₄-v conformation had a higher vibrational frequency than that of the pEA-Ag₁₄-o conformation, upshifted from 1945 to 2030 cm⁻¹ (Fig. S3 in the supplementary material). The calculated lengths of the C≡C bond decreased in the order of pEA-Ag₁₄-o > pEA-Ag₁₄-v > unbound pEA (Fig. S4 in supplementary material). These results clearly indicated that both the σ-donation and π-back donation contributed to the experimentally observed spectral downshift of the ν_C≡C mode when pEA was chemisorbed to Ag surfaces. In contrast to the σπ bonding nature of Ag–acetylene interactions, thiolated molecules were chemisorbed to Ag surfaces through σ donation of the lone-pair electrons of S to the conduction band of Ag. The ν_CS mode of pMA slightly downshifted from 1085 cm⁻¹ in the NR spectrum to 1077 cm⁻¹ in the SERS spectrum [Fig. 2(c)], indicating that the formation of the Ag–S bond had rather insignificant influence on the bond length of the C–S bond in pMA, which was further verified by DFT calculations (Fig. S4 in the supplementary material). The DFT-optimized geometry of pMA-Ag₁₄ [Fig. 2(d)] revealed that pMA was chemisorbed to Ag₁₄ at a bridge site with a relatively vertical orientation that was only tilted slightly from the normal to the Ag surface, in line with previous DFT calculations.⁶¹ The single-bond nature of the C–S bond created a certain level of motional freedom of the reactive amine group due to rotation around the surface normal. Therefore, pMA was expected to be structurally more flexible than pEA when chemisorbed to the Ag NP surfaces, especially at the locally curved surface sites. We also used DFT to calculate the binding energies, ΔE_binding, of pEA and pMA to an Ag₁₄ atomic cluster, which corresponded to the energy difference between the molecule–cluster complex (pEA-Ag₁₄ or pMA-Ag₁₄) and the unbound species (Ag₁₄ + pEA or pMA). The values of ΔE_binding were calculated to be −87.8 kcal mol⁻¹ for pEA and −58.9 kcal mol⁻¹ for pMA, respectively.

The pEA and pMA adsorbates on Ag NP surfaces underwent plasmon-driven oxidative coupling reactions in an alkaline aqueous environment containing dissolved molecular O₂. After continuously illuminated by the 785 nm laser at a P_ex of 5.8 mW for 20 s, pEA and pMA dimerized into DYAB and DMAB, respectively, resulting in the emergence of several characteristic SERS peaks of the azo compounds [Fig. 2(e)], including the C–N stretching mode (v_CN) around 1140 cm⁻¹ and three peaks in the wavenumber range of 1350–1500 cm⁻¹, all of which were the spectral signatures of the newly formed azo bond (v_N=N). The DFT results showed that DYAB and DMAB in the trans conformation [Fig. 2(f)] had Raman features drastically different from those of their counterparts in the cis conformation [Fig. 2(g)]. According to our SERS results [Fig. 2(e)], the photochemically produced DYAB and DMAB were both in the thermodynamically favored trans conformation rather than the metastable cis conformation. All the v_N=N modes and the v_CC ring modes of DYAB consistently exhibited larger Raman shifts than those of DMAB in the SERS spectra [Fig. 2(e)], which was a spectral feature fully captured by the DFT calculations as well [Fig. 2(f)].

We used SERS as an in situ monitoring tool to track the progress of the pEA coupling reactions in real time. The SERS-based kinetic measurements were conducted using a confocal Raman microscope, which allowed us to focus the 785 nm cw laser onto a focal spot that was 2 µm in diameter on the samples through a 50× objective (NA = 0.5, WD = 10.6 mm). The SERS signals were collected in a back scattering configuration using the same objective. The plasmon-driven coupling reactions were carried out in a homebuilt reaction chamber⁸⁹ assembled on top of the ITO-supported Ag NP arrays in an aqueous medium containing 100 mM KOH (pH = 13) at room temperature. It has been previously reported that an alkaline environment favors the plasmon-driven coupling of pMA because pMA-to-DMAB conversion requires dehydrogenation of the reacting amine groups.⁵⁴ Figure 3(a) shows the temporal evolution of the SERS spectra collected from a fixed spot on a pEA-coated Ag NP substrate under continuous laser illumination at a P_ex of 3.5 mW. Several snapshot SERS spectra captured at various reaction times are shown in Fig. 3(b). The characteristic v_CN and v_N=N peaks of DYAB emerged in the SERS spectra upon initiation of the coupling reaction and became progressively more intense as the reaction further proceeded. The peak intensity of the v_{CC ring} mode also increased over time because dimerization of pEA into DYAB led to a significant increase in the Raman cross sections of the molecules.⁹⁰ Although the vibrational frequency of ν_C≡C mode was highly sensitive to the surface orientation of the C≡C triple bond, neither detectable peak shift nor any resolvable line shape change in the ν_C≡C mode was observed in the time-resolved SERS spectra, suggesting that the molecules remained chemisorbed to Ag with minimal orientational changes in the C≡C triple bond during the coupling reactions. Under our reaction conditions, only the pEA adsorbates in the hot spots in the interparticle gaps underwent the coupling reactions to form DYAB, while pEA molecules residing in the electromagnetically colder sites outside the hot spots remained unreacted. As schematically illustrated in Fig. S5 in the supplementary material, the coupling reaction may occur either between pEA molecules on the surfaces of two adjacent Ag NPs to form an interparticle DYAB bridging the interparticle gaps or between two pEA molecules on the surface of the same Ag NP to form an intraparticle DYAB. Considering the geometric requirements, it was most likely that the interparticle DYAB molecules were only produced in a tiny region sandwiched between the smallest interparticle spacings in each hotspot, while the intraparticle DYAB constituted the major product of the reaction. Despite their low abundance, the interparticle DYAB molecules might provide a significant contribution to the overall SERS signals because they were produced in the regions with the highest local-field enhancements. However, we were unable to further quantify the exact fractions of the interparticle vs intraparticle DYAB products because these two types of DYAB molecules were not readily distinguishable based on their SERS spectral features.

We studied the reaction kinetics in detail by monitoring the temporal evolution of the reaction progress parameter, Q, which was defined as the intensity ratio between the strongest v_N=N peak at 1450 cm⁻¹ and the v_{CC ring} mode centered at 1594 cm⁻¹,

In our case, Q served as a descriptor of the apparent reaction progress but not necessarily corresponded to the exact product fractions because the SERS intensities were related to not only the number of molecules but also the local-field enhancements in the hot spots. The field enhancements and the molecular packing densities on the Ag NP surfaces varied drastically from sites to sites, making it challenging to quantify the exact numbers of the reactant and product molecules residing in the hot spots. As shown in Fig. 3(c), the evolution of Q over the reaction time, t, could be well-described by a first-order rate law. The values of the apparent rate constant, k_obs, and maximal reaction yield achievable at an infinitely long reaction time, Q_max, were obtained by fitting the Q trajectory with the following rate equation:

We conducted the SERS-based kinetic measurements at various P_exs, and at each P_ex, the kinetic measurements were repeated at ten different spots on the samples under identical reaction conditions. The Q trajectories collected form individual spots and the ensemble-averaged Q trajectories at various P_exs are shown in Fig. S6 and S7, respectively, in the supplementary material. In all cases, the temporal evolutions of Q followed the first-order kinetics, although the k_obs and Q_max values varied from sites to sites on each sample at each P_ex. The site-to-site deviation in reaction kinetics stemmed from the heterogeneities of the local-field enhancements and the molecular distributions in the hot spots. The k_obs and Q_max values at individual spots on the samples at various P_exs are shown as box plots in Figs. 3(d) and 3(e), respectively. In the P_ex range of 1–6 mW, both k_obs and Q_max values increased with P_ex. At P_ex below 0.5 mW, the reactions became kinetically sluggish without any detectable production of DYAB after continuous laser illumination for several hours. As shown in the inset of Fig. 3(d), the ensemble-averaged k_obs, ⟨k_obs⟩, increased linearly with P_ex, which was a typical kinetic feature of photocatalytic reactions driven by plasmonic hot carriers.^2,25,91 The linear relationship between ⟨k_obs⟩ and P_ex also implied that photothermal heating had rather insignificant influence on the reaction rates.^2,91 We chose thiophenol (TP) as a probe molecule with temperature-sensitive SERS features^92–94 to measure the local surface temperatures in the hot spots during the plasmon-driven reactions. As shown in Figs. S8(a) and S8(b) in the supplementary material, the intensity ratios between the SERS peaks at 1000 and 1074 cm⁻¹, I(1000 cm⁻¹)/I(1074 cm⁻¹), of TP chemisorbed to Ag NP surfaces exhibited a linear relationship to the temperature, which provided a working curve for Raman thermometry measurements. Because of the high thermal conductivity of H₂O (∼0.6 W m⁻¹ K⁻¹), the heat generated through photothermal transduction was rapidly dissipated to the surroundings, resulting in limited local temperature elevation when the equilibrium between heat generation and dissipation was established. The steady-state local temperature in the hot spots only reached ∼55 °C even when P_ex was as high as 5.8 mW [Figs. S8(c) and S8(d) in the supplementary material]. At a low P_ex of 0.30 mW, increasing the bulk reaction temperature to 55 °C did not lead to any significant kinetic enhancements (Fig. S9 in the supplementary material), further verifying that the kinetics of the pEA coupling reactions were primarily dictated by plasmonic photochemical effects rather than photothermal heating.

The power dependence of ⟨k_obs⟩ shown in Fig. 3(d) further implied that the surface-adsorbed pEA molecules could be activated for the coupling reactions only when the local-field intensities exceeded a certain threshold value. When extrapolating the linear power dependence to k_obs of 0, we obtained a P_ex threshold value of 0.42 mW (corresponding to an excitation power density of 134 W mm⁻²) for these oxidative pEA coupling reactions. An increase in P_ex resulted in higher fractions of the molecules occupying the surface sites with local-field intensities above the threshold value, giving rise to increased Q_max values consequently. The ensemble-averaged Q_max, ⟨Q_max⟩, exhibited a linear relationship to the logarithm of P_ex [inset of Fig. 3(e)], and extrapolation of the fitting result to Q_max of 0 yielded a threshold log(P_ex/mW) value of −0.26, corresponding to a P_ex threshold of 0.55 mW, in agreement with the value extracted from the P_ex-dependence of ⟨k_obs⟩. These results indicated that the plasmon-driven pEA coupling on Ag NP surfaces required P_exs above a threshold value of ∼0.5 mW under our experimental conditions. Such excitation power thresholds have also been previously observed on other plasmon-driven molecule-transforming processes, such as the reductive coupling of pNTP⁹³ and decarboxylation of mercaptobenzoate⁹⁴ on Ag nanostructure surfaces. The threshold P_ex values may vary drastically from reactions to reactions and are sensitively dependent on the excitation wavelengths, local-field enhancements on the photocatalyst surfaces, and the local reaction environments. Takeyasu et al.⁵⁵ previously reported a threshold power density of 18.8 W mm⁻² for the pMA-to-DMAB conversion on aggregates Ag NPs at an excitation wavelength of 532 nm. Apparently, the plasmon-driven pEA coupling under our reaction conditions required a P_ex threshold value remarkably higher than that for the pMA coupling under Takeyasu’s reaction conditions.

Switching the molecular adsorbates from pEA to pMA gave rise to significantly higher reaction rates and more complicated kinetic profiles of the plasmon-driven oxidative coupling reactions. At a P_ex of 0.6 mW, pMA on Ag NP arrays underwent rapid coupling reactions to form DMAB [Figs. 4(a) and 4(b)], whereas pEA on Ag NP arrays remained unreactive without forming DYAB at any detectable level. We identified two sub-populations of surface-adsorbed pMA featured by drastically different coupling kinetics. As shown in Fig. 4(c), the temporal evolution of Q (the intensity ratio between the v_N=N mode center at 1436 cm⁻¹ and the v_{CC ring} mode centered at 1578 cm⁻¹) was essentially the sum of two kinetic components as described by the following equation:

in which ks,_obs is the apparent first-order rate constant of the slow-reacting subpopulation and Qs,_max and Q_F,_max represent the maximal yields achievable through the slow and fast coupling reactions, respectively. While the values of ks,_obs, Qs,_max, and Q_F,_max could all be obtained by fitting the Q trajectories using Eq. (3), we were unable to further obtain the k_F,_obs values because the pMA-to-DMAB conversion for the fast kinetic component was fully accomplished within a time duration even shorter than the time resolution of our SERS measurements (200 ms). We systematically studied the kinetics of the pMA coupling reactions at various P_exs in the range of 0.050–2.0 mW, and the two kinetic components were clearly distinguishable in all cases (Figs. S10 and S11 in the supplementary material). The values of ks,_obs, Qs,_max and Q_F,_max at various P_exs obtained through curve fitting of individual Q trajectories are shown as box plots with the ensemble-averaged values shown as inset panels in Figs. 4(d)–4(f), respectively. The ⟨ks,_obs⟩ increased linearly with P_ex until becoming saturated at P_ex above 1 mW [the inset of Fig. 4(d)]. Extrapolation of the linear relationship to ⟨ks,_obs⟩ of 0 yielded a P_ex threshold value of 7.3 × 10⁻⁴ mW (corresponding to a power density of 232 mW mm⁻²) for the plasmon-driven pMA coupling reactions, about three orders of magnitude lower than that for the pEA coupling reactions. Although the ⟨Q_max⟩ values of pEA coupling reactions increased with P_ex [Fig. 3(e)], both the ⟨Qs,_max⟩ and ⟨Q_F,_max⟩ values for the pMA coupling reactions were observed to be independent of P_ex in the sub-mW P_ex regime. These results indicated that the threshold value of local-field intensities required for the activation of pMA was drastically lower than that for the pEA, and all the surface-adsorbed pMA molecules probed by SERS were successfully activated for the coupling reactions over the entire P_ex range explored in this work.

Our SERS results shown in Figs. 3 and 4 clearly showed that the chemical nature of metal–adsorbate interactions was a key factor determining the kinetics of plasmon-driven oxidative coupling of aniline-derivatives. When altering the bonding nature of NP–adsorbate interactions, the Fermi level of the metal may be shifted with respect to the molecular orbitals of the adsorbates, thereby influencing the metal-to-adsorbate transfer of the plasmonic hot carriers. Under our reaction conditions, the plasmonic hot electrons were injected into the antibonding π* orbital of surface-adsorbed O₂ rather than the unoccupied orbitals of the aniline-derivative adsorbates. Therefore, the band alignment between the Fermi level and the molecular orbitals was unlikely to be the major reason causing the striking differences between pEA and pMA in terms of photochemical reactivities and coupling kinetics. The difference in coupling reaction kinetics could be most reasonably interpreted in the context of structural flexibility of the molecular adsorbates. The hybrid σπ bonding nature of the Ag–ethynyl interactions restricted the motion of the surface-adsorbates, making pEA conformationally less flexible than its thiolated counterpart, pMA. Consequently, the pEA coupling reactions exhibited lower k_obs and Q_max and required a remarkably higher P_ex threshold than those of the pMA coupling reactions. The local surface curvature of the NPs was another key factor determining the structural flexibility of the molecular adsorbates. At the high-curvature surface sites, the molecular adsorbates were conformationally more flexible than those at the low-curvature surface sites, resulting in higher reactivity and faster reaction rates. Therefore, the fast and slow kinetic components resolved during plasmon-driven coupling of pMA reflected the transforming kinetics of pMA molecules residing at high-curvature and low-curvature sites, respectively, on the Ag NP surfaces. Our observations were well in line with the TERS results previously reported by Sun et al.,⁶¹ who found that pMA on locally curved Ag NP surfaces underwent rapid coupling reactions to produce DMAB, whereas pMA on an atomically flat {111} Ag surface was completely unreactive. The plasmon-driven oxidative coupling of pEA only took place at the high-curvature sites when P_ex exceeded the threshold value, whereas pEA molecules residing at the low-curvature sites remained unreactive because the conformational rigidity of the surface-adsorbed pEA inhibited the coupling between the amine groups. Therefore, only a single first-order kinetic component was observed during pEA coupling reactions.

The unreactive subpopulation of pEA at the low-curvature sites on Ag NP surfaces could be effectively activated when the conformationally more flexible pMA molecules were incorporated into the adsorbate monolayers. With a mixed pEA and pMA adsorbate layer on Ag NP surfaces, it became possible for a pEA molecule to couple with an adjacent pMA molecule, leading to the formation of a heterodimer product, p,p′-mercaptoethynylazobenzene (MEAB) instead of the DYAB homodimers. We incubated the colloidal pEA-coated Ag NPs with various concentrations of pMA (C_pMA) for two h after which the characteristic peaks of both the v_CS and v_C≡C modes became present in the SERS spectra [Fig. 5(a)], verifying the co-adsorption of both pEA and pMA on the Ag NP surfaces. As C_pMA increased, the intensity ratios of the v_CS to the v_CC ring modes, I(v_CS)/I(v_{CC ring}), went up. We chose I(v_CS)/I(v_{CC ring}) as an indicator for the apparent surface coverage of pMA, θ_pMA,app, based on which an adsorption isotherm was obtained [Fig. 5(b)]. The adsorption behaviors of pMA onto the pEA-coated Ag NP surfaces could be well-described by a simple Langmuir binding isotherm,

in which K_pMA is the equilibrium constant for pMA binding and θ_max is the maximal θ_pMA,app achievable at high pMA concentrations. θ_pMA,app increased with C_pMA until asymptotically approaching a θ_max value of ∼0.28 [Fig. 5(b)]. With a saturated pMA monolayer coverage on Ag NP surfaces, the I(v_CS)/I(v_{CC ring}) value was measured to be 2.3±0.2 based on the SERS results collected from pMA-coated Ag NP arrays [Fig. 2(c)]. Therefore, only a small fraction of pEA (<∼13%) could be displaced with pMA under our ligand exchange conditions. The peak intensity ratio between the v_C≡C and v_{CC ring} modes, I(v_C≡C)/I(v_{CC ring}), gradually decreased as θ_pMA,app increased [Fig. 5(c)]. The I(v_C≡C)/I(v_{CC ring}) value decreased by ∼14% after incubating the pEA-coated Ag NPs with 1 mM pEA for two h.

According to the results of DFT calculations, the three types of azo products, DMAB, MEAB, and DYAB all exhibited similar Raman spectral features in the wavenumber range of 1100–1700 cm^-1 except that the Raman peaks of the v_{CC ring} and v_N=N modes slightly downshifted when switching from DYAB to MEAB and DMAB [Fig. S12(a) in the supplementary material]. After continuous laser illumination of the mixed pMA and pEA adsorbates on the Ag NP substrates (pEA-coated Ag NPs incubated with 300 μM pMA for two h) at a P_ex of 5.8 mW for 30 s, the reactive aniline-derivatives were fully converted into the azo compounds. Because only a small fraction of pEA was exchanged with pMA on the Ag NP surfaces (less than 7%), the probability of forming DMAB was expected to be rather limited. Therefore, the products of the coupling reactions were expected to be dominated by DYAB and MEAB. We observed that the SERS peaks of all the v_N=N modes were downshifted with respect to those of DYAB produced through pEA coupling reactions, while the SERS peak of the v_{CC ring} mode was asymmetrically broadened toward the lower wavenumber side [Fig. S12(b) in the supplementary material], indicating that both MEAB and DYAB were produced. When comparing the SERS spectra with the DFT results (Fig. S13 in the supplementary material), it became apparent that MEAB produced through the pEA-pMA coupling reactions was in the trans conformation rather than the cis conformation. Because of the limited spectral shifts and significant spectral overlap of the peaks, we were unable to further quantify the relative fractions of MEAB vs DYAB based on the SERS results.

Time-resolved SERS measurements enabled us to obtain detailed kinetic and mechanistic information about the plasmon-driven oxidative coupling reactions occurring among intermixed pEA and pMA adsorbates on Ag NP surfaces. Figures 5(d) and 5(e) show the temporal evolution of the SERS spectra collected from intermixed pEA and pMA adsorbates on Ag NP surfaces (pEA-coated Ag NPs were incubated with 300 μM of pMA for two h and then assembled into the close-packed array structures) under continuous laser illumination at a P_ex of 1.5 mW. More detailed spectral features associated with the temporal evolution of the v_{CC ring}, v_C≡C, and v_CS modes are highlighted in Fig. S14 in supplementary material. The SERS peaks of the v_{CC ring} modes of the DYAB and MEAB were centered at 1594 cm⁻¹ and ∼1580 cm⁻¹, respectively. As the coupling reactions proceeded, the v_{CC ring} peak, originally centered at 1594 cm⁻¹, became asymmetrically broadened as a shoulder at ∼1580 cm⁻¹ was progressively developed, indicating that both DYAB and MEAB were produced during the coupling reactions. Both the peak positions and lineshapes of the v_C≡C mode remained essentially unchanged, suggesting minimal orientation changes in the C≡C triple bond in the ethynyl group during the coupling reactions. Although no spectral shift of the v_CS mode was observed, the intensity of the v_CS peak progressively went up during the reactions because MEAB had larger Raman cross sections than those of pMA. As shown in Fig. 5(c), the temporal evolutionary profile of Q could be fully described as the sum of two kinetic components, each of which obeyed a first-order rate law,

where k_1,obs and Q_1,max are the apparent rate constant and maximal reaction yield for the fast coupling reactions at the high-curvature surface sites, respectively, while k_2,obs and Q_2,max are the apparent rate constant and maximal reaction yield for the slow coupling reactions at the low-curvature surface sites, respectively. While both MEAB and DYAB could be produced at the high-curvature sites, the products at the low-curvature sites were dominated by MEAB only because of limited conformational flexibility and reactivity of surface-adsorbed pEA. At the same P_exs, the pEA–pMA coupling reactions at the low-curvature sites were found to be significantly slower than the pMA–pMA coupling reactions. At a fixed P_ex of 1.5 mW, we systematically studied the kinetics of the coupling reactions among intermixed pEA and pMA adsorbates as a function of pMA surface coverage, and at each θ_pMA,app, the time-resolved SERS measurements were repeated at ten different spots on each sample (Fig. S15 and S16 in the supplementary material). The k_1,obs, Q_1,max, k_2,obs, and Q_2,max values extracted from the curve-fitting results are shown as box plots in Fig. S17 in the supplementary material. The ensemble-averaged k_1,obs and Q_1,max values both increased with θ_pMA,app [Figs. 5(g) and 5(h)], indicating that the incorporation of pMA onto the Ag NP surfaces further promoted the oxidative coupling reactions at the high-curvature surface sites. In contrast, the ensemble-averaged k_2,obs values appeared almost independent of θ_pMA,app [Fig. 5(i)], suggesting that the orientational change in pMA served as the rate-limiting step during the pEA–pMA coupling reactions at the low-curvature surfaces sites. Interestingly, the ensemble-averaged Q_2,max exhibited a volcano-type relationship to θ_pMA,app, and the highest ⟨Q_2,max⟩ values were achieved at θ_pMA,app around 0.15 [Fig. 5(j)]. We hypothesized that as θ_pMA,app exceeded a certain level, pMA molecules might start to accumulate locally and form structurally ordered and rigid domains on the Ag NP surfaces, which caused the decrease in Q_2,max. We also attempted to create intermixed pEA and pMA adsorbates on Ag NP surfaces by incubating pMA-coated Ag NPs with pEA solutions (Fig. S18 in the supplementary material). In this case, only a very limited fraction of surface-bound pMA was displaced by pEA even at pEA concentrations as high as 1 mM (the v_C≡C mode was almost undetectable in the SERS spectrum after ligand exchange). The incorporation of trace amount of pEA into the pMA adsorbate layer did not cause any significant changes in the kinetics of the fast reactions at the high-curvature sites. However, the slow coupling reactions occurring at the low-curvature sites became even slower because the activation energy barrier for the MEAB formation was higher than that for the DMAB formation.

This work focused on the systematic comparison of the plasmon-driven transforming behaviors of an alkynylated aniline-derivative, pEA, and a thiolated aniline-derivative, pMA, chemisorbed to Ag NP surfaces, which allowed us to correlate the conformations and conformational flexibility of the adsorbates to the photochemical reactivity and molecule-transforming kinetics during plasmon-driven oxidative coupling reactions. Time-resolved SERS was used as an in situ spectroscopic tool to fine-resolve the detailed structural evolution of the aniline-derivative adsorbates during the plasmon-driven coupling reactions, and the experimental results were further corroborated by DFT calculations. While the σ-bonding nature of Ag–thiol interactions created a certain level of conformational flexibility of surface-adsorbed pMA, an ethynyl group interacted with Ag through a hybrid σπ-bonding mode, forcing the C≡C bond in pEA to lean toward the Ag surfaces with restricted freedom of conformational changes. Therefore, pEA exhibited decreased photochemical reactivities and slower coupling rates in comparison to pMA under identical photocatalytic reaction conditions. The structural flexibility of the adsorbates was also dependent on the local surface curvature of the NPs. While the pMA molecules underwent rapid coupling reactions at the high-curvature surface sites, the reactions became significantly slower at the low-curvature sites as the adsorbate molecules became more densely packed and thus structurally less flexible. In contrast, the plasmon-driven pEA coupling reaction occurred only at the high-curvature sites, whereas the pEA molecules residing at the low-curvature sites remained essentially unreactive due to limited conformational flexibility. When certain fractions of surface-adsorbed pEA were exchanged with pMA, the originally unreactive pEA at the low-curvature sites became reactivate, forming a heterodimer product, MEAB, through pEA–pMA coupling reactions. The results reported in this work shed light on the crucial roles of the metal–adsorbate interactions in kinetic modulation of plasmon-driven photocatalytic reactions, implying that in addition to the plasmonic characteristics of the photocatalysts, the structures of the molecular adsorbates should also be deliberately tailored when optimizing plasmon-mediated molecule-transforming processes.

Chemicals and Materials. Silver nitrate (AgNO₃, ≥99.0%), pEA (C₈H₇N, 97%), pMA (C₆H₇NS, 97%), sodium borohydride (NaBH₄, 99.99% trace metal basis), ascorbic acid (C₆H₈O₆, AA, ≥95%), and ITO-coated glass substrates were purchased from Sigma-Aldrich. Thiophenol (C₆H₆S, TP, 99+%) and cetyltrimethylammonium chloride (C₁₉H₄₂NCl, CTAC, 96%) were purchased from Alfa Aesar. All reagents were used as received without further purification. Ultrapure water (18.2 MΩ resistivity, Barnstead) was used for all the experiments associated with nanostructure synthesis and sample preparation.

Synthesis of Ag NPs. Colloidal Ag NPs with an average diameter of 45 nm were synthesized following a previously reported protocol.⁹⁵ Colloidal Ag seeds were first synthesized by adding 25 μl of 100 mM AgNO₃ and 450 μl of 20 mM NaBH₄ into 10 ml of 0.5 mM CTAC in an aqueous solution under magnetic stir. The reactant mixtures were kept at 30 °C for 40 min. Then, 100 μl of the as-prepared seeds and 100 μl of 100 mM AgNO₃ were added to 88 ml of an aqueous solution containing 20 mM CTAC. After keeping the reactant mixture at 60 °C for 20 min, 1 ml of 100 mM AA was added, and the mixture was kept at 60 °C for another three h. Finally, 100 μl of 10 mM Cu(NO₃)₂ was introduced into the reactant mixture to etch the sharp vertices and edges of the Ag NPs. After further reaction for one h, colloidal Ag NPs with a uniform quasi-spherical particle morphology and a narrow size distribution were synthesized. The Ag NPs were centrifuged, washed with water three times, and finally redispersed in water for future use.

Time-Resolved SERS Measurements. The SERS spectra were collected from molecular adsorbates on close-packed Ag NP arrays supported by ITO-coated glass substrates. The as-synthesized Ag NPs were incubated with 1.0 mM of pMA or pEA at room temperature for two h to ensure the complete displacement of CTAC with the target aniline-derivative adsorbates. For the ligand exchange between pEA and pMA, the pEA-coated colloidal Ag NPs were incubated with various concentrations of pMA for two h at room temperature. After the ligand exchange processes, the colloidal Ag NPs were centrifuged and redispersed in water. The Ag NPs coated with various molecular adsorbates were assembled into close-packed arrays through slow solvent evaporation of aqueous colloidal inks on ITO-coated glass substrates at room temperature.

Time-resolved SERS spectra were collected using a BaySpec Nomadic^TM confocal Raman microscope built on an Olympus BX51 reflected optical system under 785 nm laser excitation in the confocal mode (focal spot size: 2 µm in diameter). A 50× dark-field objective (NA = 0.5, WD = 10.6 mm, Olympus LMPLFLN-BD) was used for both Raman signal collection and dark-field imaging. The Raman signals were collected in a back scattering configuration. The laser power was controlled between 0.03 and 6 mW by using neutral density filters, and the integration time under most conditions varied in the range of 0.3–1 s. A homebuilt reaction chamber⁸⁹ was assembled on top of the ITO-supported Ag NP arrays, and an aqueous medium containing 100 mM KOH (pH = 13) was continuously flowed into the reaction chamber at a constant flow rate of 1 μl min⁻¹.

DFT Calculations. DFT calculations were performed using the Q-Chem 5.4 software package with the functional of B3LYP applied in all calculations. The 6-31+G** basis set was used for non-metal atoms, while the def2-ecp basis set was used for the Ag atoms and their effective core potential. The polarizable continuum model (PCM) was used as the solvation model. The medium surrounding the molecules was water, whose dielectric constant was set to be 78.39. The Raman activities obtained from the output files were converted to relative Raman intensities according to the equation in the following:⁹⁶ 

where ν₀ is the frequency of the excitation laser, ν_i is the vibrational frequency of the ith normal mode, h is the Planck constant, c is the speed of light in a vacuum, k_b is the Boltzmann constant, and S_i is the Raman activity of the ith normal mode. f is an appropriately chosen common scaling factor for all peak intensities. In our calculations, the excitation frequency was set at 12 738.85 cm⁻¹, which corresponded to an excitation wavelength of 785 nm. A scaling factor of 1 was used for all the peak intensities, and the temperature was set at 298.15 K. The Raman intensity were expanded with Lorentzian line shape. A full width at half-maximum (fwhm) of 7 cm⁻¹ was used for all the Raman peaks.

Structural Characterizations. TEM images were obtained using a Hitachi HT7800 transmission electron microscope operated at an accelerating voltage of 120 kV. SEM images were taken using a Zeiss Gemini 500 field emission scanning electron microscope. The optical extinction spectra were collected from colloidal NP suspensions or Ag NP arrays on ITO-coated glass substrates using a Beckman Coulter Du-640 spectrophotometer. Optical dark-field microscopy images of the SERS substrates were taken using an Olympus BX51 optical microscope, which was integrated with the confocal Raman microscope.

See the supplementary material for additional figures as noted in the main text, including SERS spectra, detailed kinetic results, and Raman thermometry results.

This work was supported primarily by the National Science Foundation under Grant Nos. CHE-2202928 (chemistry and photochemistry on metallic nanoparticle surfaces) and OIA-1655740 (hierarchical assembly of nanoparticles) and, in part, by the University of South Carolina (UofSC) Office of Vice President for Research through an ASPIRE-I Track-IV award. K.C. was partially supported by a UofSC SPARC Award. The Zeiss Gemini 500 thermal field emission scanning electron microscope used in this work was purchased using the South Carolina STEM Equipment Funds.

The authors have no conflicts to disclose.

K.C. collected the experimental data and did the DFT calculations. K.C and H.W. analyzed the data. H.W. designed the project, supervised the research, acquired funding support, and wrote the paper. Both authors have given approval to the final version of the manuscript.

The data that support the findings of this study are available within the article and its supplementary material.