1. Introduction

Photoelectrochemical (PEC) water splitting is an effective method to convert solar energy to chemical energy that can be stored and used when desired. Hematite (α-Fe₂O₃) is a promising photoanode material for PEC water splitting due to its stability (in water and electrolyte), low-cost, abundance and near optimal band gap of 2.1 eV [1]. However, hematite suffers from a mismatch between the hole diffusion length (~20 nm) and the light penetration length (~120 nm at 550 nm) [1]. Only holes generated in the proximity of the hematite/water interface (also called the semiconductor liquid junction (SCLJ)) can contribute to water splitting, the rest is lost through recombination with photogenerated electrons. The reported photocurrents produced by hematite photoanodes are far lower than the theoretical maximum of 12.6 mA cm⁻² [2] because of the significant recombination loss associated with the discrepancy of the light absorption and the hole diffusion length scales.

Many studies have tried to overcome the challenge by shortening the distance a hole needs to travel and enhancing photon absorption near the SCLJ. One route is to use hematite films with thicknesses on the order of the wavelength of incident sunlight and feature sizes on the order of 10¹ nm, including random nanoparticle aggregations [3, 4], ordered cauliflower [5], wormlike [6], nanowire [7] and nanorod [8, 9] structures. These thick films have near total photon absorption. However, these structures do not have well-engineered geometries and morphologies for hole transport [1, 10], hence, hematite photoanodes with these structures still exhibit excessive bulk recombination losses [11].

The other route is to use ultrathin hematite films where all of the hematite is in close proximity of the SCLJ. To enhance the photon absorption in such ultrathin films, one approach is to employ nanophotonic structures. Wang et al. [12] and Qiu et al. [10] coated ultrathin hematite films on fluorine-doped tin oxide (FTO) two-dimensional nanophotonic crystals. These host-guest structures have a graded refractive index that can reduce the total reflectivity and thus, increase the photon absorption in hematite films. Kim et al. [13] illustrated a hematite photoanode with a nanobeam photonic structure that supports Mie resonances to enhance the photons absorbed near the SCLJ. However, these nanophotonic structures have both high surface and interfacial areas between the hematite layer and electron collection layer which increases the amount of surface and interface recombination [14, 15]. In addition, nanopatterning typically requires chemical etching, laser ablation or catalytic vapor deposition, which creates surface defects [16]. These factors increase recombination possibilities in nanophotonic structures. Therefore, studies have also sought to use ultrathin planar hematite films to reduce both bulk and surface recombination losses. To enhance the photon absorption in the ultrathin planar hematite film, surface plasmon resonances supported by noble metal nanoparticles [17, 18] have been suggested. However, these ultrathin planar hematite films have shown reduced photon absorption and thus lower photocurrents compared to hematite films with nanophotonic structures [10]. For example, Iandolo et al. [18] reported a photocurrent of only 0.38 mA cm⁻² at 1.6 V against the reversible hydrogen electrode (1.6 V_RHE). Recently, Dotan et al. [11] coated a 26 nm thick hematite film on a silver-gold alloy substrate and achieved a photocurrent of 3.02 mA cm⁻² at 1.6 V_RHE. The photon absorption in the ultrathin planar hematite film is enhanced due to the destructive interference between the hematite film and the metallic substrate. The interference can be realized in a film much thinner than a quarter wavelength attributed to non-trivial phase shifts of propagating electromagnetic waves at the water/hematite and hematite/metallic substrate interfaces [11, 19]. They then combined this effect with a V-shape cell configuration to achieve a photocurrent of 3.98 mA cm⁻² at 1.6 V_RHE which is the state-of-the-art photocurrent for ultrathin planar hematite photoanodes. Nevertheless, this photocurrent is still lower than the best reported photocurrent of hematite photoanodes with nanophotonic structures (4.36 mA cm⁻² at 1.6 V_RHE, reported by Qiu et al. [10]).

Here, we first quantitatively investigate the spectral absorption feature of ultrathin hematite films on various metallic substrates. We show that destructive interference resonances in ultrathin hematite films can only be achieved in short wavelengths (below 450 nm). Then we propose an ultrathin planar hematite/silver nanohole array/silver substrate structure for solar photoelectrochemical water splitting. We theoretically show that this structure has both high broadband photon absorption comparable to nanophotonic structures due to supported destructive interference and surface plasmon resonance, and lower recombination loss, characteristic of planar ultrathin films. Thus, this structure promises to exceed the state-of-the-art performance of ultrathin hematite photoanodes.

2. Methods

To harness a broad band of solar energy and reduce recombination losses, we propose the hematite/metal nanohole array/metal substrate planar structure shown in Fig. 1(a). Figure 1(b) shows the optical constants ($m=n+i\kappa$, $n$ is the refractive index, $\kappa$ is the absorption index) of hematite from 300 to 590 nm (300 nm corresponds to the cut-on wavelength of solar spectrum, 590 nm corresponds to the hematite band gap of 2.1 eV) [11]. Hematite is absorbing below 450 nm, while nearly lossless above 450 nm ($\kappa$is less than one tenth of $n$). The photon absorption in an ultrathin hematite film can be enhanced below 450 nm (and approach unity at some wavelengths) through the destructive interference that exists in ultrathin absorbing dielectric/metal structures, as recently described by Dotan et al. [11] and Kats et al. [19]. Above 450 nm, where hematite/metal does not support this interference resonance, the surface plasmon resonance supported by the metal nanohole array can be coupled into the hematite film to enhance the photon absorption.

 

Fig. 1 Ultrathin planar hematite/metal nanohole array/metal substrate structure for solar PEC water splitting. (a) Layer 1 is water, layer 2 is hematite, layer 3 is dielectric (refractive index n = 1.4) filled metal nanohole array with hexagonal lattice and layer 4 is metal substrate. (b) Optical constants ($m=n+i\kappa$) of hematite.

Download Full Size | PDF

To quantitatively optimize the proposed structure, we used the photocurrent as a figure of merit. We assumed that only photons absorbed within 20 nm distance from the SCLJ can contribute to the photocurrent and that each of these photons can generate one electron/hole pair without any recombination loss. Therefore, the photocurrent is expressed as

We also calculated the solar-averaged absorption defined as

We determined α(λ) using the finite difference time domain (FDTD) method [20]. For the nanohole array, we used a hexagonal lattice unit cell and periodic boundary conditions (along x and y).

3. Results and discussion

First, we calculated how the photocurrent varies with hematite thicknesses in a tri-layer structure (shown in the inset of Fig. 2(a)). For the substrate (layer 3), we considered gold (Au), silver (Ag) and aluminum (Al) because they are reflective and can support a surface plasmon resonance in the 300 to 590 nm wavelength range. As shown in Fig. 2(a), the optimal hematite thickness on Au and Ag substrates is 20 nm, while on Al it is 30 nm. We also observed that the photoanode with an Ag substrate produces the best photocurrent, while photoanodes with Al and Au substrates are second and third best, respectively. This is further verified by the solar-averaged absorption variations with hematite thicknesses shown in Fig. 2(b).

 

Fig. 2 Photocurrent and photon absorption of hematite films on Au, Al and Ag substrates. (a) Photocurrent and (b) solar-averaged absorption variations with hematite film thicknesses. Spectral total absorption, α_tot, and the spectral absorption within 20 nm of the SCLJ, α, for hematite films on (c) Au, (d) Al and (e) Ag substrates. (f) Phasor diagrams of the reflected partial waves at 400 and 550 nm for 20 nm hematite/Ag structure.

Download Full Size | PDF

To better understand how metal substrates and hematite thicknesses impact destructive interference and spectral absorption, we calculated the spectral total absorption, α_tot(λ), (in both the hematite and the substrate) and the spectral absorption within 20 nm of the SCLJ for various substrate choices and hematite thicknesses. In the 20 nm hematite/Au structure, strong interference resonances occur below 470 nm (indicated by high total absorption) as shown in Fig. 2(c). However, the Au substrate absorbs a large portion of photons owing to the presence of the interband transition at 470 nm, which reduces the photon absorption within 20 nm of the SCLJ (the absorption in hematite in this case). Therefore, the photocurrent is relatively low. In contrast, most photons are absorbed by the hematite layer in structures with Al and Ag substrates as shown in Figs. 2(d) and 2(e), and consequently these structures reach higher optimal photocurrents.

Figure 2(d) shows that the resonant wavelength (indicated by the location of peak total absorption) is shifted to longer wavelengths with thicker hematite layers. This means photon absorption at longer wavelengths (closer to the band gap), which is desirable, can be enhanced by increasing the hematite thickness. However, the absorption is spread over the entire thickness, such that a lower fraction within 20 nm of the SCLJ is collected. In the case with the Al substrate, the photon absorption within 20 nm of the SCLJ increases at long wavelengths while decreases at short wavelengths with increasing hematite thicknesses. Therefore, an optimal hematite thickness exists by balancing the absorption over the entire spectrum. The optimal hematite thickness is larger on the Al substrate (30 nm) than on the Ag substrate (20 nm), hence, photon absorption within 20 nm of the SCLJ is lower for the hematite/Al structure than for hematite/Ag structure. Comparing Figs. 2(c)–2(e), the hematite/Ag structure clearly achieves the highest overall useful photon current generation of the three structures.

Nevertheless, the 20 nm hematite/Ag structure still absorbs poorly at longer wavelengths (near the bandgap) as shown in Fig. 2(e), because the destructive interference can only be realized over a narrow spectrum (which is also the situation for two other substrates). This is further verified by the phasor diagrams of the reflected partial waves at 400 and 550 nm for this structure shown in Fig. 2(f). The total reflection coefficient can be expressed in the form of partial waves as,

To enhance the absorption above 450 nm, where hematite/metal does not support destructive interference, we introduced an array of nanoholes in the metal substrate to support a surface plasmon resonance that couples into the hematite film.

We optimized the height, radius and lattice constant of the Ag nanohole array using the particle swarm optimization technique [21]. The thickness of the hematite was kept constant at 20 nm. To make the structure more practical and compatible with common fabrication techniques, the nanohole areas are assumed to be filled with a dielectric material having refractive index of 1.4 (approximately corresponding to the refractive index of polydimethylsiloxane or glass). Although the contact area between the hematite layer and the metal substrate is reduced, the destructive effects on the photocurrent are expected to be minimal [17, 22, 23]. Due to the change from a solid Ag substrate to a dielectric-filled Ag nanohole array, the interference condition is slightly altered and the photon absorption below 450 nm (in the destructive interference region) is reduced correspondingly, as shown in Fig. 3(a). However, above 450 nm, where the solar flux is higher, the photon absorption is significantly enhanced due to the surface plasmon resonance. The optimization procedure maximizes the photocurrent by balancing photon absorption below and above 450 nm. The optimal height, radius and lattice constant of the dielectric-filled Ag nanohole array is 30 nm, 90 nm and 260 nm, respectively. This optimized structure exhibits high photon absorption over a broader spectrum compared to hematite on a solid Ag substrate.

 

Fig. 3 Photon absorption in the hematite/dielectric-filled Ag nanohole/Ag substrate structure. (a) Absorption in hematite within 20 nm of the SCLJ for bare 20 nm hematite film, 20 nm hematite/Ag substrate and 20 nm hematite/dielectric-filled Ag nanohole/Ag substrate. The optimal height, radius and lattice constant of Ag nanohole are 30 nm, 90 nm and 260 nm. The electric field magnitude distributions at 550 nm in y = 0 plane for (b) 20 nm hematite/dielectric-filled Ag nanohole/Ag substrate and (c) 20 nm hematite/Ag substrate.

Download Full Size | PDF

Figures 3(b) and 3(c) show the electric field distribution as evidence of the surface plasmon resonance at 550 nm with the nanohole array. When surface plasmon resonances are generated, the collective charge oscillations at the Ag surface create a sharp field enhancement that is spatially confined at the edge of nanoholes and at the Ag/hematite interface, and decays exponentially into the hematite film. These highly intense electromagnetic fields contribute to the photon absorption enhancement above 450 nm.

Finally, we compared the photocurrent and solar-averaged absorption of our proposed structure with recent results from Dotan et al. [11] (state-of-the-art photocurrent for ultrathin planar hematite photoanodes) and Qiu et al. [10] (state-of-the-art photocurrent for ultrathin nanophotonic hematite photoanodes). We report the photocurrent and the solar-averaged absorption of the structure in Dotan et al. [11] based on spectral photon absorption in the hematite film within 20 nm of the SCLJ as in this work. As shown in Fig. 4, the structure proposed in this work exceeds both the calculated photocurrent and the solar-averaged absorption. A direct comparison with Qiu et al. [10] cannot be abtained because they did not report spectral absorption in the hematite film within 20 nm of the SCLJ. Nevertheless, our proposed structure has comparable photon absorption properties to their ultrathin hematite films with nanophotonic structures. Moreover, our proposed structure should exhibit lower recombination losses because of its planar structure [14–16]. Therefore, our structure promises to exceed the state-of-the-art photocurrent of ultrathin hematite photoanodes.

 

Fig. 4 Experimental and theoretical photocurrents (left axis) and solar-averaged absorption (right axis) comparing prior nanophotonic and planar structured hematite films reported in Qiu et al. [10] and Dotan et al. [11] with the ultrathin planar hematite film proposed and modeled in this work.

Download Full Size | PDF

4. Conclusions

In conclusion, we have theoretically designed and characterized an ultrathin planar hematite photoanode for solar PEC water splitting. We have demonstrated that the use of an optimized dielectric-filled silver nanohole array as a substrate improves broad band absorption due to supported destructive interference and surface plasmon resonance. Our proposed structure has higher solar-averaged absorption compared to previous ultrathin hematite film photoanodes with planar and nanophotonic structures. In addition, the planar nature of our design should enable lower recombination losses. This design has the potential to exceed the state-of-the-art photocurrent of ultrathin hematite photoanodes and aid in making solar PEC water splitting a more efficient and viable renewable energy alternative.