Plasmonic enhanced Cu2O-Au-BFO photocathodes for solar hydrogen production

A novel Cu2O-Au-BFO heterostructure photocathode was constructed which significantly improved the efficiency of photo-generated carrier transfer for solar hydrogen production. A BiFeO3 (BFO) ferroelectric film was synthesized on top of a Cu2O layer by a sputtering process. The BFO layer acted to protect the Cu2O layer from photochemical corrosion, increasing photoelectrochemical (PEC) stability. The p–n heterojunction between Cu2O and BFO layers enhanced the PEC properties by suppressing charge recombination and improved interfacial charge transfer efficiency. When Cu2O and BFO are interfaced by Au Nanoparticles (NPs) the PEC performance was further enhanced, due to hot-electron transfer at the plasmonic resonance. After positive poling, the depolarization field across the whole volume of BFO film drove electrons into the electrolyte solution, inducing a significant anodic shift, Vop of 1.01 V vs. RHE, together with a significantly enhanced photocurrent density of −91 μA/cm2 at 0 V vs. RHE under 100 mW/cm2 illumination. The mechanism was investigated through experimental and theoretivcal calculations.


Introduction
The aim of these calculations was to investigate the increased absorption due to Au nanoparticles in the Cu 2 O-Au-BFO system at ~700 nm. The Au nanoparticles are assumed to have the dimensions illustrated in Figure S1. It is also considered that the absorbance peak at 700 nm is caused by a Localized Surface Plasmon Resonance (LSPR) of the Au nanoparticle. The absorption cross-section, σ abs , of an object much smaller than the wavelength, is defined as the power dissipated in the object under plane wave illumination, divided by the power density of the incident wave on the cross-sectional area, Si. Similarly the scattering cross section, σ sca , is defined by considering the power scattered from the object. If we consider a surface totally enclosing the scattering object then we can calculate the normalized absorption (Q abs ) and scattering (Q sca ) efficiencies from reference: 2 Where A is the cross-sectional area and:  (4) In equations (4) the subscript means that the scattered field, rather than the total field is considered. The integral is over a surface that fully encloses the Au nanoparticle.
In the FDTD model the Au was modelled using a Drude-Lorentz model with one Drude and five Lorentz terms in the summation: 3 where ω p is the plasma frequency, α is the strength of the oscillators, ω o is the resonant frequency of each oscillator, j is the imaginary unit and τ is the damping frequency of each oscillator. Table 1 gives the values that were used. The open source FDTD software MEEP was used to run the electromagnetic models. 4 Simulations were carried out on a 2*12 core node (Xeon E5-2692V2) processor with 64GB of memory. The spatial resolution of the FDTD model was 1 nm and the workspace was terminated using perfectly matched layers.
Initially, the field was considered to be an incident plane wave on the upper surface (the top of the dome) of the nanoparticle, as illustrated in Fig. S6. The result obtained for the absorption efficiency is shown in Figure 3. In this model the Cu 2 O and BFO are considered to have a refractive indexes of 2.5 and 2.88, respectively.    S7 shows an LSPR peak at 859 nm, which is a significantly longer wavelength than the absorption peak obtained by UV-Vis measurement for the Cu 2 O-Au-BFO film (Fig. 3d in main paper). Due to the incident angle of the field this LSPR peak is considered to be due to the 50 nm diameter of the semi-ellipsoid particle. The other dimension is 15 nm, so the FDTD model was modified to consider a field incident from the side, as shown in Fig.   S8. Fig. S8. Illustration of the FDTD model where the field is incident to the side of the nanoparticle.
In this model the refractive index around the nanoparticle was considered to be a single material of refractive index 2.55. This was considered to be a reasonable approximation since there is not a large difference in the refractive index of Cu 2 O and BFO, and it enabled a plane wave to be incident on the particle. The result of the calculation using FDTD is shown in Fig. S9.   To consider this further the case where the incident field is at 45 degrees, so there are components of field incident on the top (dome) and the side of the Au was considered. The calculated absorption efficiency is shown in Fig. S11, and the electric field enhancement in Fig. S12. It should be noted here that it is actually easier to "tilt" the particle in the FDTD model as incident fields with obtuse (non-normal) angle of incidence do not have a constant incident angle with frequency.  The FDTD analysis infers that the Au nanoparticle of dimensions shown in Fig. S5, with a field component incident on the 'side' of the particle, would have an absorption peak that corresponds to that of the Cu 2 O-Au-BFO film. The field enhancement is around two orders of magnitude at various points on the surface.

EIS Analysis
The EIS Nyquist plot was fitted to an equivalent circuit consisting of two RC parallel networks. This was accomplished using the free EIS spectrum analyzer software (http://www.abc.chemistry.bsu.by/vi/analyser/ ).
Figures S13-S15 show the results for the three photoelectrodes where the green line is the fitting and the red points the measurements. The low frequency impedance value is R1+R2, where R2 is the charge transfer resistance in the PEC reaction. The high frequency fit is not so good for the Au structures. We attribute this to the charge transfer across the heterojunction. Nevertheless, since the low frequency curve fit is quite good in all cases, illustrating that the Au particles at the heterojunction cause a reduction in charge transfer resistance at the BFO-electrolyte interface.