Photoelectrochemical (PEC) water-splitting is a promising route for the direct conversion of solar energy into chemical energy^[1]. The PEC synthesis of hydrogen uses catalytically active light-capturing semiconductors in contact with an electrolyte forming the semiconductor-electrolyte interface. The semiconductor-electrolyte interface is at the core of photoelectrochemical water splitting devices performance and therefore requires special focus and investigation^[2]. Charge transfer at semiconductor-electrolyte interface is often discussed in terms of the Marcus-Gerischer model which refers to ideal, direct charge transfer^[3]. Nevertheless, deviations from the ideal charge transfer are observed in the presence of interface states which were considered in this work.

We developed a multi-physics 1D model of a planar water-splitting photoelectrode accounting for electromagnetic wave propagation within the electrolyte and the semiconductor, charge transport and conservation within the semiconductor, and charge transfer across the semiconductor-electrolyte interface^[4]. Interface states at the semiconductor-electrolyte interface influence the potential barrier and were considered developing an extended Schottky contact model^[4]. The photoelectrode model was validated with current-voltage measurements using a n-type GaN photoelectrode immerged in 1M H₂SO₄ utilizing additional experimental measurements (impedance spectroscopy and cyclic voltammetry) to determine missing material parameters of GaN, namely flatband potentials, surface lifetimes, surface transfer kinetic velocities and hydrogen evolution reaction exchange current density. The validated multi-physics model was then used to predict the performance of other semiconductor materials with  different semiconductor-electrolyte interface properties through variation of parameter such as electron and hole mobilities, surface lifetimes, flatband potential, permittivity, doping concentration, bulk Shockley-Read-Hall recombination, and hole and electron interface kinetics. The large number of relevant material and interface characteristics renders the identification of the most significant parameters challenging and we used statistical tools, i.e. fractional factorial design (FFD), as an approach for solving this challenge. The statistically identified significant parameters were further investigated and theoretically optimized in a detailed parameter study.

Key factors based on our model system GaN were identified using FFD at two different potentials: 0.3V vs RHE and 1.23V vs RHE. Hole and electron surface lifetimes and doping concentration appeared to be the most significant factors at 1.23V vs RHE. At 0.3V vs RHE, hole and electron surface lifetimes and flatband potential were the most significant factors. The parametric analysis of hole and electron lifetimes and doping concentration for varying potential further showed that the photocurrent resulted from a complex combined influence of the numerical values of all three parameters. The photocurrent, being a minority current, was generally more influenced by the hole surface lifetime in n-type GaN at 1.23V vs RHE. Especially at large hole surface lifetimes, the influence of electron surface lifetime became negligible. The insensitivity of the photocurrent on electron surface recombination lifetime at high doping concentrations resulted from the dominating terms in the recombination, namely the electron concentration and hole surface lifetime. Nevertheless, under low doping and short hole surface lifetimes, the electron surface lifetime showed to still have an impact on the photocurrent at 1.23V vs RHE. This counterintuitive but numerically proven result might be particularly important when explaining measurements of surface-treated photoelectrodes^[5]. These three parameters must be simultaneously considered when optimizing the photocurrent for varying operating potential. For example at an operating potential of 0.3V vs RHE, the photocurrent density increases by 0.25mA/cm² or 69% only by increasing the doping concentration from 10¹⁴cm^-3 to 10¹⁶cm^-3.

The developed methodology uses an experimentally-validated numerical model and statistical analysis to provide understanding of the performance of water-splitting photoelectrodes and to identify most significant material parameters and interface properties. Subsequent in-depth parametric analysis of the most significant parameters allows for the quantification of their effect and subsequent optimization of the device for maximum performance. The presented methodology provides a general approach to identify and quantify main material challenges and design considerations in working PEC devices. The predictive character of the validated model can be further exploited with confidence to approach and investigate morphologically complex electrodes and material classes for which research-related questions are not yet answered.

References

[1]       M. Dumortier, S. Tembhurne, S. Haussener, Energy Environ. Sci. 2015, 3614.

[2]       N. Guijarro, M. S. Prévot, K. Sivula, Phys. Chem. Chem. Phys. 2015, 17, 15655.

[3]       M. N. Latto, G. Pastor-Moreno, D. J. Riley, Electroanalysis 2004, 16, 434.

[4]       Y. Gaudy, S. Haussener, 2015 (under review).

[5]       J. Y. Kim, J.-W. Jang, D. H. Youn, G. Magesh, J. S. Lee, Adv. Energy Mater. 2014, 4, 1.