Hot electron-induced reduction of small molecules on photorecycling metal surfaces

Noble metals are important photocatalysts due to their ability to convert light into chemical energy. Hot electrons, generated via the non-radiative decay of localized surface plasmons, can be transferred to reactants on the metal surface. Unfortunately, the number of hot electrons per molecule is limited due to charge–carrier recombination. In addition to the reduction half-reaction with hot electrons, also the corresponding oxidation counter-half-reaction must take place since otherwise the overall redox reaction cannot proceed. Here we report on the conceptual importance of promoting the oxidation counter-half-reaction in plasmon-mediated catalysis by photorecycling in order to overcome this general limitation. A six-electron photocatalytic reaction occurs even in the absence of conventional chemical reducing agents due to the photoinduced recycling of Ag atoms from hot holes in the oxidation half-reaction. This concept of multi-electron, counter-half-reaction-promoted photocatalysis provides exciting new opportunities for driving efficient light-to-energy conversion processes.

Supplementary figure 2. a, SERS spectra of 4-NTP from colloidal suspension of 20 nm glass shell encapsulated 70 nm Ag nanoparticles. The 4-NTP molecules are pre-coated on the Ag nanoparticle surface before the encapsulation with a silica shell. Due to the shell protection, no SERS signal was detected at OD 2.5 and 5. A very weak signal (~ 50 CCD counts for 30 s integration time) was detected at an extremely high particle density (OD 50), equivalent to about 500 million particles per µl. b, SERS spectra obtained from the sample with OD value 50 at an integration time of 1 s. Only 7 out of 30 spectra exhibit a relatively strong SERS signal. There were about 500 million particles in the detection volume for all the spectra, indicating that even the weak signal from the highly concentrated monomer suspension is from very few clusters. However, after washing (centrifugation and resuspending the particles) the colloid with water, the extinction band shifted back (blue shift). Pink, orange and blue graphs correspond to the samples after 1, 2 and 3 washing steps, respectively. b) Control experiment using Ag superstructures without a molecular SAM. The extinction band red-shifted ~14 nm after the same treatment and washing process.  Figure 4-6). Additionally, the high plasmonic activity is also a requirement for producing hot electrons in the redox process.

Supplementary Note 2
Catalytic activity of Ag in hydride reduction of 4-NTP. When we use noble metals as plasmonic substrates for optical or spectroscopic studies, in particular in single particle experiments, the prize of the metal is typically not relevant. However, in catalysis and in particular with respect to potential industrial applications, it is very important that a cheaper metal can be used to produce the same product or reduce/oxidize the same educt. Silver is much cheaper than Au and Pt: the price of Ag is roughly 1-2 % of the Au or Pt price. Therefore it is advantageous to use Ag instead of Au and Pt for heterogeneous catalysis. In the model reaction from 4-NTP to 4-ATP reduced by sodium borohydride, Au and Pt have already showed their catalytic activity in several previous works [4][5][6] . To test if Ag can also catalyze the same reaction, we first incubated our Ag superstructures with 4-NTP molecules and coated the Ag surface with a SAM of the educt. Then sodium borohydride solution was added to the superstructure suspension as the reducing agent. However, no 4-ATP signal was observed in the SERS spectrum from the reaction (Supplementary Figure 7), indicating that Ag cannot catalyze this hydride reduction. We also used the model reaction from 4-nitrophenol to 4-aminophenol, which cannot be monitored by SERS due to the absence of the thiol surfaceseeking groups. Again no reduction was detected in the UV-Vis absorption spectra after 4 hour-incubation of the reactants and catalysts (Supplementary Figure 8 a).
In a positive control experiment this reaction was easily initiated when 10 nm Au nanoparticles were added (Supplementary Figure 8 b). Therefore we conclude that Ag has no catalytic activity in the classic hydride reduction from 4-NTP to 4-ATP. It has to be mentioned here that the Ag nanoparticles used in these negative control experiments should be synthesized and treated in a careful way, for example using high quality chemicals and solvents for the synthesis and washing steps, to avoid trace amount of catalytically active materials that may lead to false positive results.  Figure 11 a). Experimentally, the extinction maximum of Au/Ag satellite-core superstructure at the ensemble level in colloidal suspension is at 623 nm compared to 600 nm for the Ag/Ag superstructure ( Supplementary   Figure 11 b). The difference between theory and experiment is attributed to the 2D model used in the simulation. We also measured the scattering spectra of single superstructures (Supplementary Figure 11 c). The maximum intensity is between 628 and 642 nm. These results indicate that the LSPR property between Au/Ag and Ag/Ag is not very different and the 632.8 nm laser line is suitable for the excitation of both superstructures. Furthermore, we have tested 785 nm laser excitation for the SERS control experiment using Au/Ag superstructures. The result in Supplementary Figure 12 shows that the Au satellites in the Au/Ag superstructures are not active at both 633 and 785 nm excitation with I − as the electron donor (presence of the symmetric nitro stretching peak of 4-NTP at ~1340 cm -1 ), which indicates that Au, in contrast to Ag, cannot catalyze this six-electron reduction from 4-NTP to 4-ATP.

Supplementary Discussion
In redox chemistry, half-reaction and the corresponding counter-half-reaction are mutually dependent. The hot electrons generated by the non-radiative decay of the plasmons on the metal surface can be used in reduction chemistry. However, it is limited by the high rate of charge-carrier recombination. For example, 4-NTP molecules can be reduced to 4,4'-dimercaptoazobenzene (DMAB) 7-9 on the Ag superstructure due to the hot electrons. In this case each 4-NTP molecule needs 4 hot electrons: Further reduction to 4-ATP (− NH 2 ) is not possible without additional electrons from outside. In other words, the plasmonic Ag surface acts as a water cannon on a fire truck, the hot electrons act as the water and the resonant light is the pump.
Of course a water source such as a hydrant is needed when a large amount of water is required for the fire. In hot electron reduction chemistry, the oxidation half-reaction is the hydrant.
We found by accident that when Cl − was added to the 4-NTP coated Ag superstructure suspension, the nitro groups can be reduced to amino groups ( can be attributed to the much higher photo-sensitivity of AgI compared to other AgXs so that the [Fe(III)(CN) 6 ] 3− can be reduced during the photo-dissociation of AgI (not as a catalyst but as a reducer).
In summary, we tested this recycling mechanism using a different chemical reaction not involving protons, where the halide anion efficacy also trended with I − >Br − >Cl − , but AgX alone cannot catalyze this reduction. These results support our proposed mechanism with the photo-recycling Ag-AgX surface acting as the catalyst. Overall, the mechanism is in principle generally applicable.