Transformation-Optics-Designed Plasmonic Singularities for Efficient Photocatalytic Hydrogen Evolution at Metal/Semiconductor Interfaces

Inspired by transformation optics, we propose a new concept for plasmonic photocatalysis by creating a novel hybrid nanostructure with a plasmonic singularity. Our geometry enables broad and strong spectral light harvesting at the active site of a nearby semiconductor where the chemical reaction occurs. A proof-of-concept nanostructure comprising Cu2ZnSnS4 (CZTS) and Au–Au dimer (t-CZTS@Au–Au) is fabricated via a colloidal strategy combining templating and seeded growth. On the basis of numerical and experimental results of different related hybrid nanostructures, we show that both the sharpness of the singular feature and the relative position to the reactive site play a pivotal role in optimizing photocatalytic activity. Compared with bare CZTS, the hybrid nanostructure (t-CZTS@Au–Au) exhibits an enhancement of the photocatalytic hydrogen evolution rate by up to ∼9 times. The insights gained from this work might be beneficial for designing efficient composite plasmonic photocatalysts for diverse photocatalytic reactions.

(2) Sample Fabrication Synthesis of 18 nm Au NPs: A HAuCl4 solution (10 mM) was first prepared by dissolving HAuCl4·3H2O powders in oleylamine. To synthesize the 18 nm Au NPs, the HAuCl4 solution (625 μL, 10 mM) was stirred at 120 ℃ for 12 min and then half of the resulting solution was replaced with pure oleylamine (312.5 μL), followed with another heating at 85 ℃ for 2h. Four aliquots of HAuCl4 (each containing 625 μL, 10 mM) were then repetitively added into the above reaction solution every hour. After that, the reaction solution was cooled down naturally and then washed and redispersed either in CHCl3 for absorption test or in oleylamine for seeding growth of CZTS/Au HNSs.

Synthesis of Wurtzite CZTS NPs:
Wurtzite CZTS NPs were prepared following the reported method with modifications. [1] To be specific, Cu(OAc)2·H2O (32.0 mg), Zn(OAc)2·2H2O (22 mg), Sn(OAc)2 (23 mg) and oleylamine (4 mL) were mixed in a three-necked flask and degassed for 2 hours at room temperature. After that, the temperature was raised to and maintained at 120 °C under N2 bubbling for 30 min. A mixture of 1-DDT (0.05 mL) and t-DDT (0.35 mL) was subsequently injected, followed with a further increase of the temperature to 280 ℃. After 30 min, the product formed and washed by centrifugation three times with ethanol and CHCl3.

Synthesis of f-CZTS@Au, t-CZTS@Au and p-CZTS@Au:
To synthesize f-CZTS@Au, the oleylamine used was all degassed under vacuum for at least 30 min and then stored in N2.
Cu(OAc)2·H2O (14.0 mg), Zn(OAc)2·2H2O (9.6 mg), Sn(OAc)2 (10.1 mg) and oleylamine (1.75 mL) were mixed in a two-necked flask and degassed for 2 hours at room temperature, followed with heating at 120 °C under N2 bubbling for another 30 min. The previously synthesized Au NPs were then quickly injected into the reaction flask, after which the reaction solution was stirred at 120 °C for another 2 min, before the injection of a mixture of 1-DDT (22 μL) and t-DDT (153 μL).
Subsequently, the temperature was raised to and maintained at 280 °C for 30 min. After that, the reaction solution was cooled down naturally. The product was washed three time by centrifugation with ethanol and CHCl3. Finally, the product was redispersed in CHCl3 and stored in dark in a refrigerator for further use. t-CZTS@Au and p-CZTS@Au were synthesized using the similar procedures, with the exception that different amounts of reagents were used. Specifically, 12. Synthesis of t-CZTS@Au-Au and p-CZTS@Au-Au: t-CZTS@Au was used as the seed for the preparation of t-CZTS@Au-Au. Specifically, 1.1 mL of the above t-CZTS@Au, HAuCl4 (10 mM, 1.0 mL) and oleylamine (20 mL) were first mixed in a two-necked flask. The temperature was then raised to and kept at 120 °C for 30 min in dark under N2 bubbling. The product was washed by centrifugation with ethanol and CHCl3, and then re-dispersed in CHCl3. p-CZTS@Au-Au was synthesized following the similar procedures, except that the seed was p-CZTS@Au instead.

(6) Photocatalytic Hydrogen Evolution Test
To better disperse the nanoparticles in the aqueous reaction solution, the original capping ligand oleylamine was replaced with 3-mercaptopropionic acid using a reported ligand exchange method. [2] Briefly, the concentrated solution of nanoparticles in CHCl3 was mixed with 1 vol.% of 3-mercaptopropionic acid in formamide, followed with vigorous shake to promote the phase transfer in the bi-phase system. After that, the nanoparticles moved into the upper formamide phase, while the clear, colorless CHCl3 solution was discarded. The nanoparticles in the formamide phase were further purified with fresh CHCl3 two times. Finally, the nanoparticles were precipitated with acetone and redispersed in DI water. The photocatalytic hydrogen evolution experiments were conducted in a 100 mL flask with stirring at 8 ℃ under the illumination of a 300 W Xe lamp equipped with an AM1.5G filter (light intensity: ~97 mW/cm 2 , measured by a power meter) for 2 hours. Typically, 5 mg of nanoparticles, determined with the aid of ICP-OES, were dispersed in 50 mL deionized water containing 0.35 M Na2S and 0.25 M Na2SO3 as hole scavengers. Before irradiating the reaction solution, the reactor was thoroughly purged with N2 to remove all oxygen in the headspace of the reactor and dissolved in water. Each sample was tested for at least three times and the product hydrogen was analyzed by GC7900 system.

(7) Recycling Experiment
The recycling stability of the best-performing catalyst, t-CZTS@Au-Au, was evaluated. A total number of 5 cycles, each lasting 6 hours' illumination, was performed. The hydrogen was collected and quantified in each cycle. After each cycle, all the gas was discarded and the reactor was thoroughly re-purged with N2. At the end of the fourth cycle, 10% of the hole scavengers was supplemented to examine its influence on the recycling stability.

(8) Finite Different in Time Domain (FDTD) Simulation
We performed numerical simulations using a FDTD Method. All the FDTD simulations were conducted with a commercial software (LUMERICAL, FDTD Solution). The refractive index of Au was adopted from ref. [3] . The refractive index of the background was set as 1.44, the refractive index of chloroform. [4] The refractive index of CZTS was adopted from ref. [5] . We set the size of the Au nanoparticles as 18 nm and the size of the CZTS nanoparticle as 40 nm, based on the statistical analysis of the TEM images of the corresponding samples. The absorption power density was calculated from Pabs = -(1/2) Re(E×H), wherein E is the electric field and H is the magnetic field.

(9) Transient Absorption/Pump-probe Experimental Details
Transient absorption pump-probe spectroscopy was based on a femtosecond laser with a central wavelength of 1030 nm, and a repetition rate of 5 kHz (Pharos 10 W, Light Conversion).
The 1030 nm-output laser was split into two beams with a beam-splitter. One of the two beams went through an optical parameter amplifier (Orpheus-F, Light Conversion) and was utilized as the probe with a pulse duration <200 fs. The other beam went through another optical parameter amplifier (Orpheus-N-2H, Light Conversion) and was utilized as the pump with a pulse duration <150 fs. The pump pulses were chopped at 500 Hz by a synchronized chopper, and the transient absorption signal was processed by a lock-in amplifier (SR860, Stanford Research System) to suppress the noise. The delay time between the pump and probe was controlled by a moving stage.
The spot of the pump was elliptical, and the average spot diameter (1/e 2 ) was 170 μm. The spot diameter (1/e 2 ) of the probe was 50 μm.

(10) Calculation of the Apparent Quantum Efficiency (AQE)
The AQE of each CZTS-based photocatalyst studied in this work was calculated using the following equation: Here, because Au and CZTS are of different chemical natures and the affinity of Au/Au is obviously higher than that of Au/CZTS, when t-CZTS@Au is exposed to the growth solution containing HAuCl4 (Au precursor) and oleylamine (reducing agent), the newly formed Au atoms would preferentially deposit on the partially open surface of the Au NP rather than the surface of the CZTS, inducing the nucleation and overgrowth of the second Au NP. 18 Figure S3. HRTEM images of t-CZTS@Au-Au NPs taken along three orientations, achieved by tilting the same specimen by 25° for each step. The arrows denote that one seemingly spherical Au NP progressively turns into a Au-Au dimer with the specimen tilting.
Note: Due to the random dispersion of the HNSs on the TEM grid, the axis connecting the centers of the two Au NPs and the CZTS NP is not always perpendicular to the TEM beam direction. As a consequence, typically, part of the outer Au NP visually superimposes the inner Au NP and CZTS NP in the TEM image. By properly tilting the specimen, the image, taken when the beam is normal to the axis, can then reflect the realistic geometry of the HNSs, i.e., the outer Au NP being segregated from the CZTS NP (Figure 1c) and the whole Au entity being of a dimeric rather than spherical morphology ( Figure S3).     t-CZTS@Au has a small opening of Au NP without being covered by CZTS, while p-CZTS@Au has a much larger one. Under the constraint of CZTS, the second Au NP, overgrown from the small opening, in t-CZTS@Au-Au has thus a small interfacial contact with the first Au NP. By contrast, the second Au NP in p-CZTS@Au-Au has a much larger interfacial contact with the first Au NP.
As a result, t/p-CZTS@Au-Au have distinct differences, especially in the curvature of the surfaces near the interface between the two Au NPs (please see the schemes in Figure S8). Such differences are also reflected by the distinct morphologies of the Au entity in t/p-CZTS@Au-Au (t-CZTS@Au-Au: calabash-like morphology vs. p-CZTS@Au-Au: wax gourd-like morphology), as revealed by the HR-(S)TEM images in Figure S8.        and continuous Xe light illumination (c), respectively. Note: As suggested by earlier studies, the photoelectrons generated upon illumination participate in the water reduction over the active edge sites of the sulfide structure of CZTS, producing the hydrogen. 1 We employed photoelectrochemistry to study the carrier dynamics, as it is sensitive to probe the light response of the photocatalysts. First, the availability of photoelectrons, which is overall determined by the charge carrier photogeneration, separation and transport, is pivotal for efficient photocatalytic H2 production. 19 We used the photocurrent produced from the photoelectrochemical reduction of ethyl viologen dipercholorate (EV(ClO4)2) as an indicator to evaluate the quantity of photogenerated excitons. 1 Figure S17 reports the chronoamperometric responses of photocathodes made of the CZTS based photocatalysts under xenon light irradiation with periodic light chopping (see details in the experimental section). Again, all the Au/CZTS HNSs exhibit enhanced photocurrent density over CZTS and the enhancement follows the similar trend as observed in the photocatalytic hydrogen evolution rate, i.e., CZTS NPs < f-CZTS@Au < t-CZTS@Au < p-CZTS@Au < p-CZTS@Au-Au < t-CZTS@Au-Au. By conducting photocatalytic and photoelectrochemical tests with only visible light (> 400 nm), we demonstrate that both the UV light and the visible light contribute to the enhanced photocatalysis, despite that the former is more efficiently than the latter in driving the photocatalysis (Figure S18). We further recorded the current density of photocathodes made of CZTS NPs, f-CZTS@Au, p-CZTS@Au and t-CZTS@Au-Au under the laser excitation of three monochromatic wavelengths (i.e., 405, 532 and 808 nm) (Figure S17b). Both the samples and excitation wavelengths were chosen based on their distinct and plasmonic characteristics as displayed in the extinction spectra (Figure 3a).
While we note that the laser-induced thermal effect may play a non-negligible role in the photocurrents (as reflected by the slow current transient), 20 under all studied circumstances, t-CZTS@Au-Au shows the highest response to light excitation, agreeing well with its broadband and intense absorption. By contrast, f-CZTS@Au and p-CZTS@Au exhibit marked enhancement of current density over CZTS only at the wavelength of 405 and 532 nm. It is noteworthy that a noticeable light response is still observed for t-CZTS@Au-Au at 808 nm, though much weaker than at 405 and 532 nm. The weaker photocurrent density at 808 nm is partly due to the lower absorption of t-CZTS@Au-Au in the near-IR range that decreases the population of the plasmonic hot electrons. In addition, the intrinsic low kinetic energy of the plasmonic hot electrons due to the excitation by the long-wavelength (808 nm) and thus low-energy (1.53 eV) photons can be another key reason. Notwithstanding the weaker photocurrent, the t-CZTS@Au-Au sample shows an impressive enhancement over the pure CZTS sample that produces nearly no photocurrent. With regard to p-CZTS@Au, it presents a higher current density than f-CZTS@Au does at both 405 and 532 nm (Figure S17b), but lower at 808 nm (Figure S19), which is likely because the two wavelengths 405 and 532 nm are closer to the SPR band of the former while 808 nm to the latter.     Note: In the manuscript, we model the contact between the Au nanoparticles as a point (singularity) and the size of the Au nanoparticles identical. In practical applications, point contacts (singularities) are unlikely to be realised due to limitations in synthesis and the surface tension of the metal.
However, the bluntness of contacts can also be modelled by transformation optics, [21,22] demonstrating the absorption enhancement and adiabatic focusing at a broad spectral range.
Meanwhile, the gigantic field enhancement effect on the kissing points always occurs with contacting spheres with different diameters. [23] Besides the previous theoretical work, [21][22][23] here we numerically investigate the absorption spectra considering the parameters from practical geometries, as demonstrated in Figure S21a. We consider two factors; the overlap (δ) and the size difference (D1, D2) between the two contacting spheres. Figure S21b shows the absorption spectra in the two Au nanoparticles when the kissing point (δ = 0) degrades to a contacting area (δ > 0), while Figure S21c illustrates the cases of two Au nanoparticles with different sizes. A prominent absorption improve is realised compared with single Au nanoparticles. Figure S21 also illustrates that size fluctuation may also play an important role in the broadband absorption achieved experimentally, together with the contacts predicted by transformation optics.
However, the unique feature of adiabatic focusing of transformation optics guarantees the strong near-field enhancement around the contacting area at different wavelengths, in spite of the bluntness of the kissing point (cases with δ > 0, shown in Figure S22). Photons at different wavelengths can be simultaneously concentrated at the same region, improving the generation rate of hot electrons. With our design, the area is also the tri-phase region where the chemical reaction takes place, maximising the effect of transformation optics for photocatalysis. Note: The apparent quantum efficiency (AQE) of t-CZTS@Au-Au under the present conditions is 0.044% (see the calculation in the experimental section), which does not show advantages with respect to nowadays the state-of-the-art photocatalytic systems. 24 Nevertheless, since this work focuses substantially on studying how the optical geometric structure fundamentally affects the photocatalytic activity, whereas the AQE can be easily raised by simply increasing the amount of the t-CZTS@Au-Au particles in the photocatalysis, the further enhancement of the photocatalytic performance from the perspectives of optimizing the catalyst design would be more intriguing. In this sense, by increasing the density of the plasmonic singularity through incorporating more Au-Au dimers at the same single CZTS NP or the sharpness of plasmonic singularity via more meticulous chemistry, a further enhancement in the photocatalytic performance can be expected.
Moreover, considering Pt is a prominent H2-evolution catalyst, deposition of Pt NPs onto the CZTS surface of the t-CZTS@Au-Au hybrid nanostructures also holds the potential to additionally elevate the hydrogen production rate (see Figure S23 for conceptual schemes).