Manipulation of charge transfer and transport in plasmonic-ferroelectric hybrids for photoelectrochemical applications

Utilizing plasmonic nanostructures for efficient and flexible conversion of solar energy into electricity or fuel presents a new paradigm in photovoltaics and photoelectrochemistry research. In a conventional photoelectrochemical cell, consisting of a plasmonic structure in contact with a semiconductor, the type of photoelectrochemical reaction is determined by the band bending at the semiconductor/electrolyte interface. The nature of the reaction is thus hard to tune. Here instead of using a semiconductor, we employed a ferroelectric material, Pb(Zr,Ti)O3 (PZT). By depositing gold nanoparticle arrays and PZT films on ITO substrates, and studying the photocurrent as well as the femtosecond transient absorbance in different configurations, we demonstrate an effective charge transfer between the nanoparticle array and PZT. Most importantly, we show that the photocurrent can be tuned by nearly an order of magnitude when changing the ferroelectric polarization in PZT, demonstrating a versatile and tunable system for energy harvesting.

To supplement the morphology characterization of the prepared nano-Au array, atomic force microscopic (AFM) was utilized. The relevant results are presented in Supplementary Fig. 1.
Identical parameters to the ones obtained by the SEM measurements were retrieved. Importantly, the thickness was gauged as 60 nm, providing a set of accurate parameters for the FDTD simulation. Supplementary Fig. 2a presents the absorbance spectrum of the ITO/nano-Au structure obtained by the FDTD method. The absorbance is in a good agreement with the experimental data (see Fig. 2b in the main paper), showing a well-defined LSPR near 800 nm.
To simulate the absorbance of the PEC electrodes, we used the geometrical parameters of the nano-Au arrays that closely match the ones of the real samples. The parameters for individual nanoparticles are shown in Supplementary Fig. 2b.

The reasons for introducing periodic plasmonic nanostructures.
(1) Periodic nano-Au array with well controlled geometrical parameters provides a good platform to investigate the absorbance properties and charge transfers of the system both experimentally and theoretically. (2) The geometrical parameters are well defined. The control over the shape, size, distance between nanoparticles and thickness of the ferroelectric layer enables control over the absorption and thus provides further optimization tools. Therefore periodic patterned structures are necessary for further optimization. (3) To directly demonstrate the advantages of the periodic patterns, we have performed a simulation to show how linear absorbance is affected by randomness. As shown in Supplementary Fig. 3, the film with periodic nano-Au array has a much higher LSPR absorbance as in the case nanoparticles being randomly oriented. Moreover, the resonance is much sharper. (4) In conventional films containing Au nano-particles, which are usually prepared by spin-coating or spay-coating the solution with Au particles, the distribution of the plasmonic particles is almost impossible to control. It is unavoidable to have some aggregations of particles.
Such aggregations affect the absorbance, and could give rise to hot-spots, and making the whole structure also mechanically fragile.

Supplementary Note 2. Characterizations of the prepared PZT polycrystalline films
Supplementary Fig. 4a presents the XRD pattern of the PZT film on ITO coated glass substrate.
The three main diffraction peaks could be ascribed to the reflection of (001), (101) where * , Ф , , and are the effective Richardson constant, potential barrier height, permittivity of free space, and dynamic dielectric constant, respectively. E m is the maximum electric field at the Schottky interface. Equation (1) can be converted in the following way to simplify further analysis: Given that E m is proportional to the applied voltage V, ln vs V 1/2 plots should be linear for a constant T. The intercept at V=0 yields the value: From equation (3), Ф could be easily extracted, since depends linearly on 1/T. Supplementary The absorbance spectra of ITO/PZT/nano-Au/PZT structure under different poling conditions. To make a convincing conclusion that the poling influences only the charge transfer between Au and PZT, we measured the absorbance spectra of ITO/PZT/nano-Au/PZT structure under various poling conditions. As shown in Supplementary Fig. 9, the changes in the equilibrium absorbance are marginal, comparable to the variation of absorbance between different samples within the same batch. These results indicate that poling does not impact the band structure of PZT or LSPR properties of the nano-Au.

Supplementary Note 4. Photoelectrochemical properties of PEC electrodes
PEC performance for the ITO/nano-Au. Supplementary Fig. 7 presents the measurement of the EQE performance of the ITO/nano-Au. Within the error bars the EQE is zero in this energy range.
ITO is a degenerate semiconductor and behaves as a conductor to collect the photo-generated charge carriers. Once ITO is contacted to the nano-Au, there is no band bending at the interface to collect the hot electrons generated in plasmonic nanoparticles. Thus there is no measurable PEC response.
Poling effect and the performance of the ITO/nano-Au/PZT photoelectrode. Supplementary Fig.   8 shows the EQE spectra of the ITO/nano-Au/PZT electrode. Without the poling treatment, the electrode shows an EQE of ~0.2%, higher than that of the ITO/PZT/nano-Au/PZT electrode.
However, after poling treatment, the nano-Au/PZT film was found to deteriorate, demonstrated by the data shown in Supplementary Fig. 8. This is further beneficial for conducting the photo-generated charges to the surface and to drive the PEC reactions. Similarly, the screening charges could also increase the photocurrent of the electrode/ferroelectric/electrode structure, as reported by Qin et al. 9 .
The PEC performance of the ITO/nano-Au/TiO 2 electrode. The ITO/nano-Au/TiO 2 structure was fabricated using a similar method as the one used to prepare the ITO/nano-Au/PZT, except that the TiO 2 was synthesized by a standard atomic layer deposition procedure (ALD) 10 . The corresponding SEM image is shown in Supplementary Fig. 14a. Here the nano-Au array is the same to that in the ITO/nano-Au/PZT hybrids, while the thickness of the TiO 2 was 26 nm. The thickness of the TiO 2 was chosen to be considerably lower as in the case of PZT. A thicker TiO 2 layer would prevent the transfer of electrons to the electrolyte, since, due to the n-type nature of TiO 2 , the band bending at the TiO 2 /electrolyte promotes the conduction of holes to the electrolyte.
Only after the hydrogen catalysts (Pt nanoparticles) are deposited on TiO 2 , which modify the band bending, could the photo-generated electrons be extracted to the electrolyte. As shown in Supplementary Fig. 14b, the EQE is lower than that of ITO/nano-Au/PZT structure, not to mention the lack of charge transfer tuning capability of the PZT films. These results further evidence the advantages of using ferroelectric materials for energy harvesting in plasmonic hybrid nanostructures.

Supplementary Note 5. Analysis of the transient absorbance data
To demonstrate that the transient absorbance data describe the carrier dynamics and transport from nano-Au, we performed control transient absorbance studies on bare PZT films deposited on ITO-glass. Supplementary Fig. 11  Spectral decomposition by means of singular value decomposition. To determine the time evolution of excitations, that give rise to the recorded differential transmission transients, we performed singular value decomposition (SVD) analysis on the data. Since the data reveal different timescales for the low and high frequency ranges, the analysis was performed separately on the two spectral ranges. We should note that the analysis is not affected substantially when shifting the border between the high and low frequency ranges by 0.1 eV.
Supplementary Fig. 12 presents the analysis of the differential transmission transients, ∆T/T(hν,t), in the range between 1.3 and 2 eV (low frequency range).
SVD is a mathematical method which in our case decomposes the data into several spectral components, with each of them described by an individual time trace. In the Supplementary Fig.  main components, whose time traces and their corresponding spectra are shown in panel c) and its inset, respectively. The two spectral components are separately shown in panels d) and e) (note the scale change in panel e), while the residual (the difference between the original data) and the reconstructed data are plotted in panel f  Supplementary Fig. 13 presents the time evolution of the integrated response on as grown, as well as +10 V and -10 V poled samples. One can recognize the trend suggesting that recovery proceeds slightly faster in the sample poled with +10 V. On the other hand, the difference in timescales is rather weak, comparable to the variation between timescales obtained on different samples of the same batch.
As demonstrated in the main text, the +10 V poling potential induces a depolarization electric field (E DP ) with the direction pointing towards the ITO substrate and a downward band bending at electrolyte/PZT interface, which is favorable for the injected hot electrons being transferred to the interface and driving PEC reactions. The -10 V poling potential, however, switches the direction of the E DP . In this case, the injected hot electrons in PZT cannot be transferred to the PZT/electrolyte interface and are trapped in the bulk of the PZT film. As the hot electrons aggregate in the PZT films, the hot electrons injection rate could be slowed down, giving rise to an increase in the recovery time as opposed to the +10 V pre-poled samples.