Ultrafast Charge Carrier Recombination and Trapping in Hematite Photoanodes under Applied Bias

Transient absorption spectroscopy on subpicosecond to second time scales is used to investigate photogenerated charge carrier recombination in Si-doped nanostructured hematite (α-Fe2O3) photoanodes as a function of applied bias. For unbiased hematite, this recombination exhibits a 50% decay time of ∼6 ps, ∼103 times faster than that of TiO2 under comparable conditions. Anodic bias significantly retards hematite recombination dynamics, and causes the appearance of electron trapping on ps−μs time scales. These ultrafast recombination dynamics, their retardation by applied bias, and the associated electron trapping are discussed in terms of their implications for efficient water oxidation.


S2
The UV-vis spectrum of hematite changes with applied electrical bias; such spectroelectrochemical studies by our group have previously been reported in detail. 6,7

fs-Transient Absorption Spectroscopy (TAS):
Transient absorption spectra were collected on fs-ns timescales using a regeneratively amplified Ti:sapphire laser system (Solstice, Spectra-Physics) and Helios spectrometers (Ultrafast Systems). The Solstice laser system generates 800 nm, 92 fs width, 1 kHz pulses. The excitation (pump) pulse is generated from a fraction of the 800 nm beam via an OPA (TOPAS Prime, Spectra-Physics) and frequency-mixer (NirUVis, Light Conversion) to select the wavelength, 355 nm (for comparison to µs-s TAS). The pump beam diameter is approximately 1 mm at the sample. The intensity of the pump beam was modulated using neutral density filters, and was measured using a 500 µm diameter aperture and power meter (PD10-V2-ROHS, OPHIR Photonics). The probe pulse is delayed with respect to the excitation pulse by a motorised translational stage, which changes the path length of the probe beam. A visible or NIR white light continuum (WLC) is used as the probe, generated by attenuating and focussing a fraction of the 800 nm pulse onto one of two Ti:sapphire crystals of different thicknesses. In order to reduce noise, the WLC is split into two beams: one passes through the sample while the other is monitored by the reference spectrometer. The beams are focussed into two fibre-optic-coupled multichannel spectrometers (CMOS or InGaAs sensors for visible and NIR wavelengths, respectively). A synchronised chopper (500 Hz) is used to block alternate pump pulses; the absorbance change is calculated from adjacent pulses (pump blocked and unblocked). Difference spectra were typically averaged over 2 seconds for each time point recorded; each spectrum was collected several times then averaged. Spectra were corrected for chirp (group velocity dispersion) using Surface Xplorer software (Ultrafast Systems) by fitting the rise of the spectrum. Spectra were timezero corrected such that t 0 occurs at the half-amplitude of the spectrum rise. Individual decay kinetics shown are averaged over ~10 nm. Due to fluctuations of the laser intensity over time during these measurements, normalized kinetics are compared.

µs-s Transient Absorption Spectroscopy (TAS):
The home-built transient absorption spectrometer employed for these studies has been described previously. 8 The third harmonic (355 nm, 200 µJ.cm -2 .pulse -1 , 0.33 Hz, "EE" frontside excitation) of a Nd:YAG laser (Big Sky Laser Technologies, Ultra CFR Nd:YAG laser system, 6 ns pulse width) is employed as the pump, while the probe beam is the monochromated output of a 100 W tungsten lamp (Bantham IL1). The laser light is transmitted to the sample via a liquid light guide. A number of long-pass filters and a bandpass filter (Comar Instruments) were placed between the sample and the detector, in order to attenuate the scattered laser light. The transmittance of photons through the sample is measured by a Si photodiode detector (Hamamatsu S3071). Collected data were filtered and amplified (Costronics amplifier), and were then recorded with an oscilloscope (Tektronics TDS 2012c) on the timescale of µs -ms, and with a DAQ card (National Instruments, NI USB-6211) on the timescale of ms-s. All data were acquired by home-programmed Labview software. Data shown are the average of 300-500 laser shots, with the laser scatter subtracted.

Photoelectrochemistry:
For measurements in a working PEC cell, a three-electrode configuration was employed using a home-made PEEK cell with quartz windows. A Pt-gauze counter electrode and Ag│AgCl│sat.KCl reference electrode (Metrohm) were used in 0.1 M NaOH electrolyte, pH 12.8. Electrolyte solutions were prepared from NaOH (Sigma-Alrich, reagent grade, as received) and MilliQ deionised water (Millipore Corp., 18.2MΩ.cm at 25 °C); the electrolyte was not degassed. For fs-TAS measurements of hematite in a 3-electrode PEC cell, the voltage was applied using a Keithley sourcemeter as a pseudo-potentiostat; for µs-TAS a Sycopel ministat was employed. The current was allowed to stabilise before TAS measurements were begun. Herein, the applied voltage is reported versus the reversible hydrogen electrode (RHE), calculated using E RHE =E°+E+0.059pH, where E° is the standard potential of the Ag│AgCl│sat.KCl reference (~0.21 V RHE at 25 °C), and E is the potential versus Ag│AgCl│sat.KCl. The photoanode (working electrode) was illuminated in the EE (electrolyte-electrode or "front side") direction. Current/voltage curves were acquired using a Metrohm ministat, employing a 75 W Xe lamp (OBB) with two homogenisers and a KG3 filter (Thorlabs). A silicon diode and a 650nm shortpass filter (Edumnd optics) were used to adjust the light intensity such that the number of photons per square centimeter at <650 nm (approximately the absorption edge of hematite) reaching the sample was equivalent to that of AM 1.5.

Estimation of Space-Charge Layer Width
The width of the space-charge layer (depletion region, W) can be estimated: where V is the applied voltage, ε 0 is the permittivity of free space, and e is the charge on an electron. For hematite, taking the relative permittivity ε = 80, flatband potential V FB = 0.5 V RHE , and the donor density N D = 1.5x10 20 cm -3 , this gives W = ~7 nm at 1.4 V RHE for the APCVD Si-Fe 2 O 3 photoanode. 2 It is noted that literature values for the relative permittivity of hematite vary from 12 to 120, 9 thus this value of W should be taken as a ballpark figure.
Given the short hole diffusion length of a few nanometers in hematite, 10-12 essentially all carriers generated outside the SCL will undergo rapid bulk recombination. The nanostructured APCVD α-Fe 2 O 3 photoanodes employed in the studies reported in the main paper have a dendritic nanostructure, with large particles close to the FTO substrate and small nanoparticles at the surface. 2 Although the smallest particles may be completely depleted under strong anodic bias, a significant fraction of the photoanode will be in the bulk region.

Fig S2:
Transient absorption (TA) spectra of nanostructured Si-doped APCVD hematite in an inert atmosphere (N 2 ), under relatively low excitation density (355 nm, 500 Hz, 170 uJ.cm -2 .pulse -1 ; ~2.75x10 14 photons abs.cm -2 .pulse -1 ). The legend indicated time after excitation by the pump pulse. There is a fast decay of the spectrum within the first 10 ps, which is not observed for TiO 2 (see below). The legend indicates probe wavelength. The signal exhibits a rapid decay which begins within the time resolution of the measurement (~200 fs). It is extremely unlikely that this initial decay phase is an artifact, since artifacts are symmetrical about t = 0, 13 and a fast initial transient optical decay is widely reported for hematite produced via several different synthetic routes. [14][15][16][17][18][19][20] Thus the TA signal is due to additional absorption by photogenerated charge carriers; the decay of this signal indicates the loss of charge carriers, either by recombination and/or trapping (relaxation).  Figure 1B in the main paper also fit to a power law, with α=0.11 (fit between 5 ps-6 ns) for an excitation intensity of 82 µJ.cm -2 .
Power law fits to the ultrafast TAS decays of Si-doped nanostructured hematite reported herein have 0.10<α<0.35. Even under strong anodic bias, a sizeable proportion of photogenerated carriers in hematite recombine within 6 ns, which represents a significant inherent limitation to the maximum possible solar energy conversion efficiency of hematite. µJ.cm -2 .pulse -1 ). The legend indicates probe wavelength; the inset shows the first 10 ps of the normalized decay kinetics. The signal probed at ~575 nm decays faster than at other wavelengths, and bleaches after ~500 ps, due to oxidisation of trap states within the spacecharge layer formed at the semiconductor-liquid junction. The signal probed at 650 and 750 nm exhibits a similar fast decay to that observed in N 2 ( Figure 1 in the main paper and Figure S4).