Unassisted photoelectrochemical water splitting exceeding 7% solar-to-hydrogen conversion efficiency using photon recycling

Various tandem cell configurations have been reported for highly efficient and spontaneous hydrogen production from photoelectrochemical solar water splitting. However, there is a contradiction between two main requirements of a front photoelectrode in a tandem cell configuration, namely, high transparency and high photocurrent density. Here we demonstrate a simple yet highly effective method to overcome this contradiction by incorporating a hybrid conductive distributed Bragg reflector on the back side of the transparent conducting substrate for the front photoelectrochemical electrode, which functions as both an optical filter and a conductive counter-electrode of the rear dye-sensitized solar cell. The hybrid conductive distributed Bragg reflectors were designed to be transparent to the long-wavelength part of the incident solar spectrum (λ>500 nm) for the rear solar cell, while reflecting the short-wavelength photons (λ<500 nm) which can then be absorbed by the front photoelectrochemical electrode for enhanced photocurrent generation.

The complex refractive index including index of refraction (n) and extinction coefficient (k) was used to take the absorption of light by each layer into account. Then, the calculated reflection and transmission coefficients were: where '1' and 'l' denote the first and last layer of multilayer films. GA optimization calculations begin with the generation of a population of multilayer cDBR films with a fixed 10 number of layers whose thicknesses and compositions are randomly generated, followed by the evaluation of the FOM of each member, as described in the main text.
First, cDBR stacks using only ITO films were designed. Because ITO films with different porosities could have different refractive indices 2 , this single material could allow the fabrication of the cDBR. For GA optimization, the measured refractive index as well as the extinction coefficient profiles of the dense (θ OAD = 0°) and porous (θ OAD = 70°) ITO layers fabricated by OAD were used (Supplementary Fig. 1a). cDBR structures with varying numbers of ITO layers (1-7 layers) were designed in the GA optimization, with the 7-layer specimen showing slightly superior optical properties (Supplementary Fig. 1b). However, the interface between the dense layer and the porous layer was not completely distinct when the designed 7-layer structure was fabricated by OAD using electron-beam evaporation ( Supplementary Fig. 1c). This indistinct interface led to inconsistencies between the experimental and simulated transmittances 3 , as shown in Supplementary Fig. 1d.
Therefore, we re-designed the hybrid cDBR structures composed of only a few conductive ITO layers stacked on a conventional dielectric DBR framework consisting of two different dielectric thin films with a high refractive index contrast. TiO 2 (n=2.40 at =550 nm) and SiO 2 (n=1.45 at =550 nm) thin films were used for the dielectric DBR stack located under the ITO film and were optimized by varying the number of alternating layers of TiO 2 /SiO 2 thin films from 4 to 6 to compare their performances, as shown in Supplementary Fig. 2.
The dielectric DBR with 6 layers of TiO 2 (Supplementary Fig. 3b). As a result, the hybrid cDBRs with 2-layer and 4layer ITO stacks were fabricated and used in this work, as described in the main text.

Supplementary Note 2: Theoretical studies of the optical function of the hybrid cDBR
Theoretical studies were conducted to investigate the optical function of the hybrid cDBR used in the tandem system. This study is mainly based on the significant linear relationship among light absorption, reflection and transmission using relevant calculations.

Photoanode
In the supplementary note 1, the absorption of cDBR is considered to get more accurate theoretical data and reduce the variances between the simulated and experimental curves. In this part, to make the expression more clearly the cDBR is regarded as an ideal dielectric light filter, which means the light that cannot be transmitted is completely reflected. This assumption will not affect the final estimation of the integration ratios for photoanode and DSSC as shown below. Therefore, in this work, we name the experimental data of the When the value of n trends towards positive infinity, the term [(1 − ) in Supplementary Equation (4) will be zero. Thus the curve of s , shown in Supplementary Fig. 5c, could be drawn based on the data in Fig. 2 and

Supplementary Figs 5a and b.
Considering the intrinsic property of the photoanode, the charge separation efficiency η sep and charge transfer efficiency η tran are not affected by the addition of the DBR stack.
Therefore, the PEC performance of the photoanode with (J sum ) and without (J anode ) the DBR is dependent on J abs , which is the theoretical maximum value after considering the light harvesting ability as shown in Supplementary Equation (5) J abs can be obtained by the integration of the product of the photo flux (corresponding to the intensity of 1.5G AM) and the absorption efficiency (Supplementary Fig. 5a A anode , and Supplementary Fig. 5c A sum ).
The charge density curves obtained from the absorption efficiency curves multiplied by the photo flux for the photoanode with/without DBR are shown in Supplementary Fig. 7a. with the photoanode shown in Fig. 3a. Besides, in order to verify the spectral response with/without cDBR and understand the function of cDBR to photoanode experimentally, the external quantum efficiency (EQE) has been checked and shown in Supplementary Fig. 5d.
The results basically agrees well with the theoretical analyses.

DSSC
In the same way, the absorption curve of Dye JK-306 for the in situ measurement can be named as (Supplementary Fig. 6a). When the photoanode is placed in front of the DSSC, only the light penetrating through the photoanode and glass can be absorbed, which is named as 1 and calculated as o ss (Supplementary Fig. 6b). Note that since has included the transmittance of glass, while the data of o has excluded the glass, we have to re-multiply the ss here. The o and ss used for the calculation are shown in Supplementary Fig. 5b and Supplementary Fig. 6d Considering an ideal case, the absorption of the DSSC after the addition of both the photoanode and DBR, 2 is, J abs can be obtained by the integration of the product of the photo flux and the absorption efficiency (Supplementary Fig. 6b A rear1 , and Supplementary Fig. 6c A rear2 ).
The charge density curves obtained from the absorption efficiency curves multiplied by the photo flux for the two varieties of DSSC performance are shown as Supplementary Fig. 7b.
The ratio of the two areas, (C+D)/C, which corresponds to the value of J rear1 :J rear2 , is approximately 8:7. This value is consistent with the experimental value obtained for the DSSC shown in Fig. 3a.
The theoretical analyses presented above verified the correction of the experimental data obtained in Fig. 3a and demonstrated the authentic function of the DBR, which could be used in the tandem system on the basis of the optical theories.