High-throughput parallel testing of ten photoelectrochemical cells for water splitting: case study on the effects of temperature in hematite photoanodes

High-throughput testing of photoelectrochemical cells and materials under well-defined operating conditions can accelerate the discovery of new semiconducting materials, the characterization of the phenomena occurring at the semiconductor–electrolyte interface, or the understanding of the coupled multi-physics transport phenomena of a complete working cell. However, there have been few high-throughput systems capable of dealing with complete cells and applying variations in real-life operating conditions, like temperature or irradiance. Understanding the effects of the variations of these real-life operating conditions on the performance of photoelectrode materials requires reliable and reproducible measurements. In this work, we report on a setup that simultaneously tests ten individual, identical photoelectrochemical cells whilst controlling temperature. The effects of temperature from 26 to 65 °C were studied in tin-doped hematite photoanodes for water splitting – as a reference case – through cyclic voltammetry and electrochemical impedance spectroscopy. The increase of surface-state-mediated charge recombination with temperature mainly penalized the energy conversion efficiency due to the reduction of the photovoltage produced. For parallel measurements in the ten individual cells, standard deviations from 20 to 60 mV for the onset potentials and less than 0.2 mA cm−2 for saturation current densities quantified the reproducibility of the results.


Supplementary Note 1 -UV-visible spectroscopy and scanning electron microscopy
Reflectance and transmittance spectra of Fe 2 O 3 photoelectrodes were measured with a Shimadzu UV-2600 UV-vis-NIR spectrophotometer with an integrating sphere (ISR-2600 PLUS Shimadzu).Absorbance spectra were then calculated and Tauc's equation was fitted supposing an allowed direct optical transition, obtaining the Tauc's plot of Figure S1.Fitting the linear part of the curve, in which parabolic band dispersion can be assumed, the optical bandgap of the Fe 2 O 3 photoelectrodes is calculated to be 2.11 eV, which is coherent to the values of 1.9-2.2eV reported in literature 19 .The morphology of the Sn:α-Fe 2 O 3 thin films deposited on FTO was observed by scanning electron microscopy (Fig. S1).The film resulted fairly homogeneous and compact as in previous reports of α-Fe 2 O 3 thin films obtained by spray pyrolysis 20,21 .The grain size of the film varied between 50 and 150 nm (Fig. S1

Supplementary Note 2 -LEDs characterization
Due to the need of operating ten light sources simultaneously, Light Emitting Diodes (LEDs) were chosen for their high efficiency, low cost and long lifespan.LEDs' emitted power was measured at different distances from the light sources through an Ocean Insight FLAME-S-XR1 spectrometer (optical resolution 1.7 nm FWHM) connected to an optical fiber (core diameter 300 µm, length 1 m) and a cosine corrector (CC-3-UV-S, OceanInsight), used to measure the hemispherical irradiance.The spectrometer was calibrated with a deuterium -tungsten halogen lamp (Ocean Insight DH-3P-BAL-CAL).
The spectral hemispherical irradiance at different distances from the light source of the ten LEDs used for the experimental campaign is reported in Fig. S3.The results of the characterization are summarized in Table S2.
Figure S3: Average of ten LEDs spectral hemispherical irradiance measurements as a function of wavelength and error bars (in lighter colour) at distances from 30 to 90 mm from the light source.
The LEDs had the peak of spectral hemispherical irradiance at 442 nm, in the violet-blue region of visible spectrum, and a full width at half peak (FWHP) of 19 nm at every distance from the light source.The total hemispherical irradiance non-linearly decreased with distance.
The distance from the light source was chosen to simulate the irradiation from the global standard spectrum (AM 1.5G) despite the use of monochromatic light sources.To do so, we assumed: i) all the photons with energy lower than the energy bandgap of the semiconductor (E < E g ) are not absorbed by the photoelectrode; ii) for photons with energies larger or equal than the semiconductor energy bandgap (E ≥ E g ), the only loss considered is the thermalization loss, i.e. the useful energy of the absorbed photon is E g ; iii) the blue LEDs used are ideal monochromatic light sources with characteristic wavelength λ LED,c = 442 nm; iv) the function between two measured values of total hemispherical irradiance is linear; and v) the effect of reflection or diffraction of light due to the quartz window and the electrolyte are G AM 1.5G,usef ul is the total irradiance from the AM 1.5G spectrum which contributes, under the aforementioned assumptions, to a useful effect with the Fe 2 O 3 photoanodes [W m −2 ]; G λ,AM 1.5G is the spectral irradiance from the AM 1.5G spectrum [W m −2 nm −1 ], obtained from the PV Lighthouse database 22 ; G LED,usef ul is the total irradiance from the blue LEDs which contributes to a useful effect with the Fe 2 O 3 photoanodes [W m −2 ]; G LED is the total irradiance from the blue LED [W m −2 ], λ is photons wavelength [nm]; λ g = 590 nm is the wavelength of the photon with energy equal to the semiconductor energy bandgap (E g = 2.1 eV).
Under these assumptions, the calculated irradiance from the LED is G LED = 348 W m −2 and it is obtained at a distance from the light source of 81 mm, as shown in Figure S4.
To prove that this approach is consistent, the number of photons reaching the photoelectrode with AM 1.5G spectrum and with the blue LEDs are compared.The energy of the photons emitted by the LEDs is 2.804 eV, i.e. its wavelength is λ LED,c = 442 nm.With an irradiance of 348 W m -2 , the flux of photons Φ LED = 7.75 • 10 20 m −2 s −1 , which is 0.9% more than the cumulative photon flux with energies higher than the semiconductor energy bandgap from AM 1.5G spectrum, Φ AM 1.5G,E>Eg = 7.68 This confirms that it is a good assumption to place the photoelectrode at a distance of 81 mm from the LED light source to simulate the AM 1.5G spectrum.

Figure S1 :
Figure S1: Tauc's plot of Fe 2 O 3 photoanodes and linear fit of the region in which parabolic band dispersion can be assumed.

Figure S2 :
Figure S2: Scanning electron microscopy images of a Sn:α-Fe 2 O 3 thin film on FTO after deposition by spray pyrolysis and annealing: (a) front view and (b) false-colour cross section view to highlight the two layers.

Figure S4 :
Figure S4: Total hemispherical irradiance of the LEDs as a function of the distance from the light source and (in red) graphical method to identify the distance from the light source for which the irradiance corresponds to the one of the AM 1.5G spectrum.

Figure S7 :
Figure S7: Detail of one of the two sides of the setup with five PEC cells in front of five LEDs, each mounted on a heat sink.

Figure S8 :
Figure S8: Temperature distribution in the ten PEC cells during the photoelectrochemical tests.Contrarily to the rest of the manuscript, the error bars are obtained from the temporal standard deviation of the single measurements, the dashed lines are the average temperatures across the ten cells and the colourbands are their error bars obtained from the spatial standard deviations of the ten measurements.

Figure S12 :
Figure S12: Photocurrent density-photovoltage characteristic curves with error bars obtained from ten parallel measurements with Sn:Fe 2 O 3 thin films in blue light (solid line) at (a) 26°C, (b) 36°C, (c) 44°C, (d) 56°C, (e) 65°C.Dotted lines represent the measured photocurrent density at low photovoltages and dashed lines the photocurrent densities extracted linearly fitting the saturation regions of the curves.

Figure S13 :
Figure S13: (a) Sn:α-Fe 2 O 3 thin films applied bias photon-to-current efficiency (ABPE) as a function of applied potential at different temperatures.(b) Maximum ABPE (left axis) and potential at which the maximum ABPE is measured (right axis) as a function of temperature.(c) Sn:α-Fe 2 O 3 thin films apparent photon-to-current (PCE) efficiency as a function of photovoltage at different temperatures.(d) Maximum apparent PCE (left axis) and photovoltage at which the maximum apparent PCE is measured (right axis) as a function of temperature.In (b)-(d) the dashed lines are the regressions of the experimental data.

Figure S14 :
Figure S14: Resistances and capacitances extracted fitting the electrochemical impedance spectroscopy curves at direct potentials from 0.8 to 1.5 V vs. RHE in blue light with the equivalent circuit of Figure 1 (b) .

Figure S15 :
Figure S15: Coefficient of variance of the total resistance R tot (a) averaged in temperature as a function of applied potential and (b) averaged in applied potential as a function of temperature.

Table S2 :
Wavelength of LEDs' spectral hemispherical irradiance peak, full width at half peak (FWHP) and total hemispherical irradiance at distances from 30 to 90 mm from the light source.