Evaluating Charge Carrier Transport and Surface States in CuFeO2

Article pubs.acs.org/cm

Evaluating Charge Carrier Transport and Surface States in CuFeO2 Photocathodes Mathieu S. Prévot,† Xavier A. Jeanbourquin,† Wiktor S. Bourée,† Fatwa Abdi,‡ Dennis Friedrich,‡ Roel van de Krol,‡ Néstor Guijarro,† Florian Le Formal,† and Kevin Sivula*,† Laboratory for Molecular Engineering of Optoelectronic Nanomaterials, École Polytechnique Fédérale de Lausanne (EPFL), Station 6, 1015 Lausanne, Switzerland ‡ Helmholtz-Zentrum Berlin für Materialien und Energie GmbH, Institute for Solar Fuels, Hahn-Meitner-Platz 1, 14109 Berlin, Germany †

S Supporting Information *

ABSTRACT: Interest in delafossite (CuFeO2) as a candidate p-type photocathode for photoelectrochemical (PEC) solar fuel production has recently been increasing, mainly due to its excellent stability in aqueous environments and favorable light absorption properties. However, its PEC performance has remained poor for reasons that have not yet been specifically determined. Herein, we report a detailed investigation on sol− gel-processed CuFeO2 with a range of spectroscopic, PEC, and microscopy techniques aimed at unraveling the material properties governing photogenerated charge carrier harvesting in this v. An analysis of the bulk transport properties using microwave conductivity measurements reveals a good charge carrier mobility (0.2 cm2 V−1 s−1) and a relatively long lifetime (200 ns) for photogenerated charge carriers. Conversely, systematic PEC measurements with varied redox systems reveal the existence of a high density of surface states (1014 cm−2) positioned 0.35 eV above the conduction band, inducing Fermi level pinning at the semiconductor−liquid junction. X-ray photoelectron spectroscopy suggests the presence of a thin layer of metal hydroxide at the surface of the material. These surface states were found to behave as electron traps, correlated with an inversion of polarity at the surface of the semiconductor, and thereby promoting charge recombination and limiting the photovoltage developed at the junction. These findings suggest that if the detrimental effects of the surface states can be eliminated, CuFeO2 would provide a sufficiently high photovoltage to be combined with other solution-processed and stable photoanodes into an easily scalable tandem PEC cell.

■

INTRODUCTION Direct photoelectrochemical (PEC) solar fuel generation represents an attractive technology for the storage of solar energy in the context of a global carbon-neutral energy economy. A possible device to perform this process is the photoelectrochemicalor photoelectrosynthetictandem cell, which is composed of two complementary light-absorbing semiconducting electrodes (photoanode and photocathode) in direct contact with a liquid electrolyte, typically water, that upon illumination develop photopotential to drive fuel production.1−3 Recently, p-type CuFeO2 has attracted growing interest4−10 as a candidate photocathode for such a device as it possesses favorable properties, including an optical band gap, Eg, of 1.5 eV and a large absorption coefficient (up to α ≈ 107 m−1). Furthermore, its flat-band potential, which is positioned ∼1 V vs the reversible hydrogen electrode (RHE) (∼1.2 eV below its conduction band edge), suggests the capability to develop a high photovoltage for the reduction of water to H2, or CO2 to formic acid or other carbon-based fuels. CuFeO2 is moreover a robust oxide; it can operate continuously in neutral © 2017 American Chemical Society

and basic aqueous environments without any loss of performance at the time scale of days (at least). A few different techniques have been used to produce CuFeO2 photocathodes. Seminal reports using electrochemical deposition4,7 or solid-state synthesis5,10 yielded electrodes photoactive for water or CO2 reduction, but offered only limited control over the morphology of the resulting electrode. More recent studies have employed a sol−gel approach6,8,9 which improved tunability and scalability. Improvements in performance have been made using TiO2 as an electronextracting overlayer6 and CuAlO2 as a hole-extracting underlayer, promoting respectively better charge injection to the electrolyte and better charge separation inside the absorbing layer.8 A microwave treatment has also been reported to notably improve the performances of the material;9 however, this technique also resulted in the presence of Cu(II) inside the Received: March 29, 2017 Revised: May 14, 2017 Published: May 16, 2017 4952

DOI: 10.1021/acs.chemmater.7b01284 Chem. Mater. 2017, 29, 4952−4962

Article

Chemistry of Materials

prepared by dissolving the desired equal quantities of salts in 0.1 M tetrabutylammonium perchlorate in acetonitrile. FeCp2+ and Fe(C5Me5)2+ were prepared by oxidation of FeCp2 and Fe(C5Me5)2 using hexafluoroboric acid and p-benzoquinone following a previously reported procedure.18 All the oxidation states unavailable for direct purchase were produced in situ by reduction/oxidation of the corresponding species in deaerated solutions. The process was stopped when the potential of the electrolyte, measured at a Pt wire, was equal to E1/2. Time-Resolved Microwave Conductivity (TRMC) Measurement. CuFeO2 thin films prepared on fused quartz substrates were mounted in a microwave cavity cell and placed within a setup similar to the one described elsewhere.19,20 X-band (8.4−8.7 GHz) microwaves were generated using a Gunn diode. Experiments were carried out at 8.55 GHz, which was found to be the resonant frequency of the sample-loaded cavity. The sensitivity factor (K) of the sample-loaded cavity was calculated from the resonant characteristic using a sample dielectric constant value of 20, as reported elsewhere for CuFeO2.21,22 During the TRMC measurements, the samples were excited by 6 ns fwhm pulses of a frequency-doubled Q-switched Nd:YAG laser at a wavelength of 355 nm (10 Hz repetition rate). The change in the microwave power reflected by the cavity upon excitation was monitored and correlated to the photoinduced change in conductance and the carrier mobility (see the Supporting Information for details). The laser pulse intensities were adjusted using neutral density filters and varied from 3.3 × 1011 to 1.6 × 1014 photons pulse−1 cm−2. Photoelectrochemical Experiments. PEC experiments were conducted using a three-electrode cell connected to a BioLogic SP-500 potentiostat. For experiments in water, a Ag/AgCl(KClsat) electrode was used as the reference, while for experiments in acetonitrile, a Ag/ Ag(NO3) electrode was used as the reference. In all cases, a Pt wire was used as the counter electrode. The light source was a 450 W Xe arc lamp (Osram) calibrated using a Si photodiode to provide a photon flux equal to the AM 1.5G spectrum for hν larger than the band gap of CuFeO2.6 For electrochemical impedance spectroscopy (EIS) experiments, the frequency range was 50 mHz to 1 MHz. All results reported here were obtained under substrate-side (back) illumination and in deaerated (Ar-purged) electrolytes (unless specified otherwise). To perform the ΔOCP (OCP = open-circuit potential) measurement, we note the importance that the electrochemical potential of the electrolyte is stable over time. To do so, the concentration of both members of each redox couple were kept equal by reducing or oxidizing the electrolyte until the open-circuit potential measured at an independent platinum electrode matched the previously measured E1/2 (see the Supporting Information for full details). Roughness Factor Determination. The actual surface area of CuFeO2 was determined by adsorption of the organic dye orange II. A film of known geometric surface area was immersed in 2 mL of a solution of 1.5 mM orange II in water (pH 3.5) for 15 min. The adsorbed dye molecules were then desorbed using 2 mL of 1 M NaOH, and the concentration of this solution was determined by its optical absorption at 450 nm using a UV-3600 (Shimadzu) spectrometer. The surface area was derived from the calculated amount of desorbed dye, considering that each molecule occupied an area of 0.4 nm2 on the surface of the film.23,24 The roughness factor was calculated as the ratio of the actual surface area to the geometric surface area put in contact with the dye solution. Kelvin Probe Force Microscopy (KPFM). KPFM measurements were performed with a Cypher atomic force microscope (Asylum Research) using Pt:Ir-coated tips (AC240TM, Olympus). The work function of the tip was calibrated using reference samples of Pt, Mo, and Cu. X-ray Photoelectron Spectroscopy (XPS). XPS measurements were carried out using a PHI VersaProbe II scanning XPS microprobe (Physical Instruments AG, Germany). Analysis was performed using a monochromatic Al Kα X-ray source of 24.8 W power with a beam diameter of 100 μm. The spherical capacitor analyzer was set at a 45° takeoff angle with respect to the sample surface. The pass energy was 46.95 eV, yielding a full width at half-maximum of 0.91 eV for the Ag

material, which lowered the faradic efficiency for water reduction to 70−90%, indicating the presence of secondary electrochemical processes. Overall, the studies conducted on CuFeO2 to date have reported solar (1 sun) photocurrents of up to 2.5 mA cm−2.8,9 While encouraging, this is far below the performance anticipated from the absorption spectrum of the material (giving a maximum possible solar photocurrent of ca. 15 mA cm−2). Indeed, poor incident photon-to-current efficiency (IPCE)lower than 25% in the visible rangeis typically reported.8 Therefore, CuFeO2 still suffers from important limitations reminiscent of, for instance, hematite11−14 or pyrite15two materials that have also displayed PEC performances poorer than anticipated on the basis of their band gap. Conventionally, low photoelectrode performance is rationalized by the presence of either surface or bulk defect (trapping) states. There is an important distinction between these two possibilities, as bulk and surface defects require different treatment strategies. While a high density of bulk defects based on intrinsic material properties can permanently prevent application of a material as a photoelectrode, surface states are arguably easier to tackle with adequate postprocessing treatments such as applying overlayers16 or etch/regrowth17 repairing techniques. Therefore, to clarify whether the poor performance observed in CuFeO2 originates from bulk or surface defects, we present here the results of a comprehensive investigation based on a variety of PEC, microscopic, and spectroscopic techniques.

■

EXPERIMENTAL SECTION

CuFeO2 Electrode Preparation. The following spin-coating recipe was used for the standard Cu/Fe sol−gel6 in ethanol: 2000 rpm for 60 s followed by 6000 rpm for 10 s. The resulting film on fluorine-doped tin oxide (FTO)/aluminoborosilicate glass substrates was annealed at 450 °C inside a muffle furnace for 1 h, and the deposition/annealing processes were repeated two more times to increase the thickness of the film. The resulting “three-layer” film was annealed at 700 °C under flowing Ar inside a fused quartz tube furnace for 1 h after a temperature ramp of 10 °C min−1. The resulting CuFeO2 electrodes were finally allowed to cool to room temperature under flowing Ar and were directly used for subsequent experiments. Electrolyte Preparation. The following materials were purchased from their respective suppliers: potassium ferricyanide(III) (99+%, Sigma-Aldrich), potassium hexacyanoferrate(II) hydrate (99+%, Sigma-Aldrich), tris(2,4-pentanedionato)ruthenium(III) (TCI), hexaamineruthenium(III) chloride (98%, Sigma-Aldrich), hexaamineruthenium(II) chloride (99.9%, abcr), methyl viologen hydrate (98%, Acros), ferrocene (98%, Acros), chloranil (98%, TCI), bis(pentamethylcyclopentadienyl)iron (99%, abcr), potassium phosphate monobasic (99+%, Acros), potassium phosphate dibasic trihydrate (99+%, Acros), sodium tetraborate decahydrate (99.5%, Sigma-Aldrich), tetrafluoroboric acid (48 wt % in H2O, SigmaAldrich), and p-benzoquinone (98+%, Alfa Aesar). Aqueous redox electrolytes were prepared by dissolving the corresponding oxidant and reductant salts in equal quantities in the desired buffer. The buffer solutions were prepared according to the following recipes in 100 mL of deionized water: pH 4, 38.6 mL of 0.2 M K2HPO4 + 61.4 mL of citric acid; pH 5, 51.4 mL of 0.2 M K2HPO4 + 48.6 mL of citric acid; pH 6, 87.7 mL of 0.2 M KH2PO4 + 12.3 mL of 0.2 M K2HPO4; pH 7, 39 mL of 0.2 M KH2PO4 + 61 mL of 0.2 M K2HPO4; pH 8, 5.3 mL of 0.2 M KH2PO4 + 94.7 mL of 0.2 M K2HPO4; pH 9, 50 mL of 0.025 M borax + 5 mL of 0.1 M HCl, completed to 100 mL with water; pH 10, 50 mL of 0.025 M borax +18 mL of 0.1 M NaOH, completed to 100 mL with water; pH 11, 25 mL of 0.1 M K2HPO4 adjusted with 0.1 M NaOH; pH 12, 0.01 M NaOH in 0.1 M Na2SO4. After the preparation, the pH was measured using a pH meter. The organic electrolytes were 4953


Article

Chemistry of Materials 3d 5/2 peak. For the depth profile analysis, the material was etched with a beam of Ar+. Curve fitting was performed using the PHI Multipak software. Microscopy and Raman Spectrometry. Optical microscopic images were obtained with a Nikon H550L, and scanning electron microscopy was performed with a Zeiss Merlin microscope. Raman spectra were obtained with a LabRam spectrometer (Jobin Yvon Horiba). The excitation line was provided by an Ar laser (532.19 nm).

■

RESULTS The CuFeO2 thin films investigated in this study were prepared using our previously reported Pechini-type sol−gel method.6 The morphology of the resulting film was monitored by optical microscopy and scanning electron microscopy (SEM), as shown in Figure S1, Supporting Information. We observed the expected nanoporous morphology, arising from the hightemperature crystallization, along with a film thickness of 200 nm. The preparation conditions were carefully optimized to ensure that no secondary phase, such as hematite25 or CuO, was present, as supported by microscopy (Figure S1a,c) and Raman data (Figure S2, Supporting Information). A typical linear sweep voltammetry (LSV) curve of the CuFeO 2 electrode (on F:SnO2-coated glass) in O2-purged 1 M NaOH (Figure S3, Supporting Information) illustrates the PEC performance for (sacrificial) O2 reduction. The observed photocurrent is reproducible in the 1.3−1.5 mA cm−2 range at 0.4 V vs RHE. As a first step with these electrodes, we set out to determine if the performance of CuFeO2 is limited by its intrinsic bulk properties using time-resolved microwave conductivity (TRMC). Time-Resolved Microwave Conductivity Measurement. TRMC is a contactless technique for monitoring charge carrier mobility and lifetime in semiconductors.26−28 In this pump−probe method, a nanosecond laser pulse first excites the material, and the resulting transient of the reflected microwave power is subsequently probed. The measured change in reflected power is directly proportional to the sum of the mobilities, ∑μ, of free carriers in the film. We note that while TRMC does not offer a direct way to discriminate between the mobilities of majority and minority carriers, deconvolution is possible using their different effective masses (see the Supporting Information for more explanation). TRMC transients were collected for a typical CuFeO2 film at different pulse intensities (see Figure S5, Supporting Information). For each transient TRMC signal, the peak value corresponds to a lower bound of Φ∑μ, where Φ represents the quantum yield of photogenerated charges. This peak value of Φ∑μ is plotted against the absorbed photon flux in Figure 1 for two similarly absorbing CuFeO2 films (see Figure S4, Supporting Information) prepared on quartz substrates. At high photon densities, Φ∑μ decreases with the log of photon density, due to fast nongeminate electron−hole recombination. The slope of the plot in this range (Φ∑μ ∝ Iabsα−1, with α the reciprocal of the order of recombination28) yields α = 0.7, indicating a combination of first- and secondorder recombination at high illumination intensities. On the other hand, at lower photon densities, Φ∑μ is approximately constant, indicating a first-order recombination. Under 1 sun illumination (equivalent to ∼109 cm−2 pulse−1), the CuFeO2 is expected to operate under the first-order recombination regime, while the presence of a second-order rate at higher light intensities is likely due to recombination processes more prominent at high charge carrier densities (e.g.,

Figure 1. Peak TRMC signal values measured for two (duplicate) CuFeO2 thin films (red and blue markers, respectively) at different fluxes of incident photons per pulse. Inset: typical TRMC signal produced by a thin film of CuFeO2 after a pulse of 8.15 × 1012 photons cm−2. The red trace represents the biexponential fit. The pulse wavelength was 330 nm for all measurements. AM 1.5 illumination is equivalent to ∼109 cm−2 pulse−1.

nongeminate radiative recombination or Auger recombination).28 At low photon densities, Φ∑μ was found to have a value of about 0.225 cm2 V−1 s−1 (average of the two samples measured), which gives us a lower estimate for ∑μ in the film. This value is in agreement with previously reported Hall measurements for CuFeO229 and is 1 order of magnitude lower than for WO3 and Cu2O, 1 order of magnitude higher than for BiVO4, and similar to that of α-Fe2O3.26 Moreover, the time constant associated with the decay is a direct measure of the lifetime of photogenerated carriers inside the active layer. This decay was fit with a biexponential model (see the inset of Figure 1), which yielded time constants of τ1 = 200 ns and τ2 = 4 μs. For comparison, the lifetimes of charge carriers in Cu2O and α-Fe2O3 were reported to be only on the order of picoseconds, while τ can reach a few tens of nanoseconds in BiVO4.30 Therefore, these results indicate that charge carriers are relatively long-lived in CuFeO2. Given that τ2 is typically attributed to trapped carriers, an estimated free carrier diffusion length, L, can be calculated using τ1 as L = (Dτ1)1/2 = 225 nm, where D = kBTμ/e is the diffusion coefficient of the charge carriers, in which kB is Boltzmann’s constant, T is the temperature, and e is the elementary charge. This diffusion length is in good agreement with the optimum film thickness of ∼300 nm previously reported for this material.6 Finally, the results of the TRMC measurement could further be used to get an estimation of the upper limit for the photovoltage achievable by CuFeO2, according to the relation proposed by Lewis:31 VOC,max ≅

kBT ⎛ Jph LnNA ⎞ ⎟ ln⎜⎜ 2 ⎟ e ⎝ eDnn i ⎠

where Jph the photocurrent measured at 0 V vs the reversible hydrogen electrode (RHE), Ln the diffusion length of electrons, NA is the acceptor density, Dn is the diffusivity of electrons, and ni is the density of thermally produced charge carriers inside the film. Using the most conservative estimations for the different parameters (see the Supporting Information for details), Dn = 2.5 × 10−3 cm2 s−1, ni2 ≅ 8.76 × 1011 cm−6, Ln = 225 nm, Jph = 0.6 mA cm−2, and NA = 1018 cm−3, a maximum expected open4954


Article


with the pH: E(pH) = E(pH 0) − 0.059pH (V) at 25 °C. Since none of the redox couples used herein involve the transfer of protons (and are thus not sensitive to the pH), by increasing the pH of the electrolyte by one unit, the redox E1/2 effectively shifts by +59 mV with respect to the band edges of the semiconductor. Consequently, varying the pH allows a range of potentials to be probed using a single aprotic redox couple.39 For this work, the values of E1/2 were measured by cyclic voltammetry, CV (see Figure S6, Supporting Information). In a first series of measurements, CuFeO2 (on F:SnO2-coated glass substrates) was put into contact with two different redox couples in water: Fe(CN)63−/4− (E1/2 = 0.25 V vs Ag/AgCl) and Ru(acac)30/− (E1/2 = −0.49 V vs Ag/AgCl), and the resulting LSV curves under intermittent (1 sun) illumination are shown in Figure 3a,b with respect to E1/2 (where the red star symbols on each curve represent a fixed potential on the RHE scale as indicated). We note that the photocathode does not produce any photocurrent in the absence of these redox couples, and therefore in each case, the photocurrent was fully attributed to the reduction of the redox species introduced inside the electrolyte and water reduction was not considered. When using Fe(CN)63−/4− between pH 5 and pH 7 (i.e., when varying E1/2 between 0.75 and 0.87 V vs RHE), we observed a clear shift in onset potential for the photocurrent with respect to E1/2. However, the onset of photocurrent occurred at a constant potential vs RHE (around 1.0 V vs RHE), in good agreement with the flat-band potential, EFB, of 1.01 V vs RHE previously reported for CuFeO2.6 Since the photovoltage can be approximated from the potential difference between E1/2 and the photocurrent onset, and is directly proportional to |E1/2 − EFB| in this case (i.e., for a ∼180 mV potential range negative of EFB), we deduce that the CuFeO2 SCLJ behaves as expected in this range. Conversely, when tested with Ru(acac)30/− between pH 10 and pH 13 (i.e., when varying E1/2 between 0.31 and 0.49 V vs RHE; see Figure 3b), a shift in the photocurrent onset was observed vs RHE, while the onset appeared unchanged vs E1/2. We note that the smaller photocurrent magnitude in this case was due to the relatively poor solubility of Ru(acac)3 in water, yielding a diffusion-limited photocurrent, as evidenced by the transient photocurrent spikes. More importantly, in this potential range (between 560 and 740 mV negative of EFB), the constant photocurrent onset observed at ca. + 350 mV vs E1/2 indicates a photovoltage independent of |E1/2 − EFB|, in contrast with the linear dependence expected for an ideal SCLJ. To probe whether the observed behavior of the photovoltage was sensitive to the aqueous nature of the electrolyte, e.g., through protonation/deprotonation of the surface of the semiconductor, a similar experiment was performed in MeCN with different redox species: FeCp2+/0 (E1/2 = 0.08 V vs Ag+/Ag), chloranil0/− (E1/2 = −0.30 V vs Ag+/Ag), Fe(C5Me5)2+/0 (E1/2 = −0.44 V vs Ag+/Ag), and MV2+/1+ (E1/2 = −0.71 V vs Ag+/Ag). The resulting LSV curves (under intermittent illumination) are shown in Figure 3c. Note that the lower photocurrents obtained with MV2+/1+ are due to its low solubility in MeCN (inferior to 1 mM). Nonetheless, the onset of photocurrent is seen clearly to be located at ∼0.35 V vs E1/2 for each redox couple despite their different E1/2 values. This consistent behavior compared with the aqueous redox couples suggests that the observed photovoltage limitation is independent of the nature of the electrolyte. The photovoltage can be more directly measured by monitoring the shift in open-circuit potential (OCP) caused

circuit potential of 1.1 V is found (in the absence of surface recombination or shunting across the potential barrier at the junction). Since the water reduction potential is reported at roughly 300 mV below the conduction band edge of CuFeO2,6 a photovoltage of around 0.8 V for water reduction at the surface of a CuFeO2 photocathode is reasonably expected. It is important to note that this analysis assumes a value for Dn based on an estimated mobility of electrons, μe, of 0.1 cm−2 V−1 s−1 (see the Supporting Information). While this values accords with our TRMC measurements and other reports as mentioned, an early report suggests an extremely low electron mobility for Sn-doped n-type CuFeO2 (10−6 cm2 V−1 s−1).32 However, the recent calculations performed on other p-type copper-based delafossites33,34 and our previously reported thickness-dependence study,6 which implies that photogenerated electrons can diffuse hundreds of nanometers and produce a photocurrent under substrate-side illumination, suggest similar effective masses and therefore similar mobilities for holes and electrons (within 1 order of magnitude). The very low electron mobility in the Sn-doped samples could be due to detrimental effects of cation substitution on the crystal structure. Overall, the conservative photovoltage value of 0.8 V we report here indicates that bulk processes should not drastically limit the performance of CuFeO2 and rather suggests that interfacial recombination plays a predominant role in our system. In the following sections, we use a variety of methods to characterize this phenomenon. Photoelectrochemical Properties of CuFeO2. A convenient method to investigate the semiconductor−liquid junction (SCLJ) is to photoelectrochemically “map” the band gap of the semiconducting electrode using a series of redox species with electrochemical potentials varying from the valence band edge to the conduction band edge.35−39 To apply this approach to CuFeO2, we relied on a series of fast outer-sphere electron transfer based redox couples to reduce the influence of charge transfer kinetics as much as possible. These couples, used either in water or in acetonitrile, are shown schematically in Figure 2 with respect to their redox potentials relative to the position of the bands of CuFeO2 under flat-band conditions. Note that, in the case of aqueous electrolytes, ranges of potentials are shown rather than a single value. This is due to a change in electrolyte pH and because the band edges of oxide semiconductors including CuFeO2 display a Nernstian behavior

Figure 2. Schematic representation of the potential regions probed by the different redox couples used in this study shown next to the band position of CuFeO2, under flat-band conditions, in water (left side) and in acetonitrile (right side). MV = methyl viologen, DMFc = decamethylferrocene, and EFB = flat-band potential of CuFeO2. 4955


Article


Figure 3. (a−c) LSV curves of CuFeO2 under intermittent 1 sun illumination (10 mV s−1 scan) in (a) 0.05 M Fe(CN)64−/3− and (b) ∼1 mM Ru(acac)30/− (saturated) at different pH values, referenced to the redox potential of the electrolyte. For each curve, the red star represents a fixed potential on the RHE scale to help visualize the apparent Fermi level pinning (FLP). For (c) the LSV curves were obtained for different redox systems at concentration 0.01 M in CH3CN, except for MV2+/1+, which saturated at ∼1 mM. (d) ΔOCP = OCP10 suns − OCPdark measured as a function of the redox potential of the electrolyte, referenced to the RHE. The expected ΔOCP value, in the absence of FLP, is shown as the broken line that intercepts the x-axis at E = EFB.

density of these states located close to the valence band edge prevents the photovoltage from reaching the value anticipated from an ideal SCLJ (represented by the broken line in Figure 3d) as the Fermi energy in the bulk of the semiconductor equilibrates with the energy of these states rather than with the electrolyte, causing a so-called Fermi level pinning (FLP).35,36,38,40 In this situation, the barrier height at the junction is no longer related to E1/2. Rather, since EF is pinned to the surface states, any additional voltage provided by a more negative E1/2, or by an external applied potential, results in the charging of the surface states rather than an increase in band bending in the semiconductor. To illustrate this process, sketches of the band diagrams at the SCLJ in the absence and presence of surface states causing FLP, along with the expected PEC behavior in each situation, are provided in the Supporting Information (Figure S8). The photoelectrochemical measurements presented in this section suggest the presence of FLP caused by surface states located roughly 0.35 V negative of EFB. However, the concentration and the chemical nature of the surface states remain unknown. Since the relative density of surface states compared to the density of bulk carriers primarily determines the extent of FLP, we next turned to additional techniques to gain more insight into the concentration and distribution of these states. Electrochemical Impedance Spectroscopy. Although EIS data have been reported before for CuFeO2,6,9 only the dielectric capacitance41,42 corresponding to the transfer of charges from the bulk of the absorbing layer to the electrolyte through the SCLJ has been evaluated, while the existence of any chemical capacitance associated with the accumulation/transfer of charges through potential surface states (at low frequencies) has been neglected. Here, we propose to refine the initially reported model (a Randles circuit) to include the contribution of the surface states suggested by the observation of FLP. Photoelectrodes displaying surface states have been studied with EIS before and are typically modeled by the equivalent circuit shown in Figure 4a.11,14,43 The following elements are included (following previously established nomenclature11): a series resistance Rs associated with charge transfer through the external circuit and the bulk of the semiconductor, a capacitance Cbulk, such as Cbulk−1 = CSC−1 + CH−1, where CSC

by illumination of the electrode. We note that in the dark at equilibrium the Fermi level of the semiconductor is aligned with the redox potential of the electrolyte even in the presence of intra band gap surface states, in which case the Fermi level of the semiconductor is aligned with the surface states, themselves in equilibrium with the electrolyte.37,40 The measured shift in the open-circuit potential under high-intensity illumination (ΔOCP) is a direct measure of the shift in Fermi level and therefore a measure of the photovoltage at the junction, assuming no voltage loss through the film. Three redox couples were used to “map” the band gap of the CuFeO2 electrode with this method: Fe(CN)63−/4− (E1/2 = 0.25 V vs Ag/AgCl), Ru(NH3)63+/2+ (E1/2= −0.18 V vs AgCl/Ag), and MV2+/1+ (E1/2 = −0.55 V vs Ag/AgCl). By using a range of pH values for each couple (limited by the chemical stability of the molecules involved), we were able to scan the following potential ranges: 0.77−0.90 V vs RHE (using Fe(CN)63−/4− between pH 5.2 and pH 7.5), 0.33−0.49 V vs RHE (using Ru(NH3)63+/2+ between pH 5.0 and pH 7.9), and 0.07−0.38 V vs RHE (using MV2+/1+ between pH 6.8 and pH 12.1). The open-circuit potentials collected in the dark and under illumination for each condition (Figure S7, Supporting Information) show equilibration at E1/2 in the dark as expected. Furthermore, in the potential range scanned using Fe(CN)63−/4− (the closest to EFB), we observed a linear progression of OCP under illumination with the pH. The associated slope was found to be −60 mV pH−1, confirming the Nernstian behavior of the electronic bands of CuFeO2 over the potential range of 0.77−0.90 V vs RHE. On the contrary, in the potential ranges scanned with the two other redox species, no linear behavior was observed. In the case of Ru(NH3)63+/2+, a constant OCP was obtained under illumination, independent of the pH, while in the case of MV2+/1+, the OCP also showed very little variation, even showing a slight increase with the pH. An overview of the OCP measurements is represented in Figure 3d, where ΔOCP is plotted against E1/2, referenced to the RHE. Clearly, a plateau in photovoltage was observed for E1/2 more negative than 0.7 V vs RHE, with a generally constant ΔOCP of ∼0.18−0.28 V in this region. A reasonable explanation for the observed dependence of the photovoltage on |E1/2 − EFB| is the presence of surface trapping states with energy levels inside the band gap. A sufficient 4956


Article


the Bode plot of the phase displayed asymmetric peaks, indicating the presence of two processes (see Figure S10, Supporting Information), hence justifying the use of the model chosen in the present study. Increasing the light intensity induces a decrease in the size of the semicircles in the Nyquist plot, indicating a lower value of Rtrapping + Rct,trap, due to the CuFeO2 photoconductivity. To obtain more insight into the trapping of charges at the surface, the values of the parameters Ctrap and Rct,trap extracted from the model are displayed in Figure 4b,c. In the dark, the presence of a Gaussian peak in the plot of Ctrap is observed to coincide with a drop in Rct,trap. The same behavior was observed under illumination, although Ctrap was 1 order of magnitude higher. There is no clear influence of the light intensity on the parameters associated with charge transfer from the trap, similar to what was previously observed with hematite photoanodes.11 Regardless, the measurements of Ctrap confirm the presence of the previously hypothesized density of surface states, located around 0.7 V vs RHE, and the model involving transfer across the SCLJ occurs through these surface states fits appropriately the experimental data. On the other hand, to examine the behavior of Cbulk with the potential, we used the Mott−Schottky (MS) relationship, which relates the change in CSC−2 with the applied potential to the acceptor density NA and EFB. The MS plots in the dark, as well as under illumination, are shown in Figure 5a. All plots showed a similar

Figure 4. (a) Typical Nyquist plots representing EIS data for a CuFeO2 electrode collected at 0.7 V vs RHE, in 1 M NaOH, in the dark and under 0.5 and 1 sun illuminations, along with the equivalent circuit used to fit them. Markers represent experimental data, and lines represent the fit. (b, c) Extracted values of Ctrap and Rct,trap, respectively, as a function of the applied potential, referenced to the RHE.

corresponds to the depletion layer inside the semiconductor and CH to the Helmholtz layer inside the electrolyte, a parallel resistance Rtrapping associated with charge trapping/detrapping at surface states, and a pair Ctrap/Rct,trap corresponding, respectively, to the capacitive effect induced by the accumulated charges in surface states and to the resistance of charge transfer to the electrolyte through the surface states. We note that direct electron transfer from the conduction band to the electrolyte is not considered in this model (nor is it observed as only two semicircles fit the obtained EIS data).43 Typical Nyquist plots obtained in 1 M NaOH, at 0.7 V vs RHE, in the dark and under two different illumination intensities, are shown in Figure 4a. In all cases, the entire plot could not be properly fit with only one semicircle (see Figure S9, Supporting Information), and

Figure 5. (a) Mott−Schottky plot of a CuFeO2 electrode in 1 M NaOH in the dark, as well as under 0.5 and 1 sun illuminations. (b) Corresponding plot of the corrected density of surface states (DOS) extracted from Ctrap after removal of the background capacitance. 4957


Article


illumination. This is consistent with the observed inversion region in the dark MS plot. In addition, under illumination NSS is 1 order of magnitude higher (∼1014 cm−2) compared to that in the dark, suggesting that the majority of the states are accessible under illumination only. Overall, the EIS results reinforce the energetic position of the surface states at +0.7 V vs RHE and give an estimation of their density at ∼1014 cm−2. This density is 2 orders of magnitude higher than the minimal density of surface states considered necessary to induce FLP with a bulk acceptor density of 1018 cm−3.35 Thus, the inversion observed in the EIS and the limited photopotential of ∼0.3 V are reasonably caused by these surface states. Finally, the lifetime of carriers trapped in the surface states can be evaluated by measuring τtrapping = RtrappingCtrap (with Rtrapping values obtained from Figure S11, Supporting Information). This yields long lifetimeson the order of several seconds under illuminationcomparable to those measured in hematite.46 We note that this surface trapping process is different from the one probed by TRMC that occurs on a microsecond time scale. Surface Analysis. Further information regarding the physiochemical nature of the surface states was next sought using Kelvin probe force microscopy (KPFM) and X-ray photoelectron spectroscopy (XPS). KPFM afforded the mapping of the Fermi level position at the surface47 in both dark and illuminated conditions on CuFeO2 electrodes in air. As shown in Figure 6b, a 5 μm by 5 μm surface scan revealed a relatively homogeneous surface potential (root-mean-square (rms) deviation of 8.2 meV) compared to the surface roughness (Figure 6a, rms deviation of 12.8 nm). In the dark the Fermi level at the surface was calibrated to −5.20 eV vs vacuum (see Figure S12, Supporting Information), which is equivalent to +0.70 V vs NHE (normal hydrogen electrode) and consistent with the results obtained by EIS assuming FLP at the surface states. Illumination of the sample did not induce a significant shift in average surface potential in the range of light intensity experimentally available (up to ca. 0.1 sun), but the surface remained equally homogeneous (Figure 6c), suggesting a uniform distribution of the surface states on the CuFeO2 and casting doubt on the possibility that the surface states are due to isolated domains of secondary phases. X-ray photoelectron spectra for Fe, Cu, and O recorded on a pristine thin film of CuFeO2 at different etching depths are shown in Figure 7. The Cu 2p signal (shown in Figure 7a) is consistent with the signature expected for Cu(I),48 with two peaks at 932.4 and 952.4 eV, and showed no change as a function of the depth. No Cu(II) could be detected at the

behavior: as the potential was scanned in the negative direction, CSC−2 first increased, as expected in the depletion region, but quickly reached a peak value before decreasing back to its initial value. This decrease is characteristic of the presence of an inversion regime, due to the accumulation of minority carriers (electrons) at the SCLJ, causing an inversion of the polarity of the surface of the semiconductor.44 Using the linear part of the MS plot, it was possible to estimate the bulk acceptor density and the flat-band potential under each condition, using ε = 2021,22 and an electrode roughness f r = 18 ± 3 (calculated on three samples by orange II adsorption, following a method previously reported for Fe2O3 thin films24). We note that while the experimental data yield an estimation of Cbulk−1, we took the measured value of Cbulk−1 as equal to CSC−1, as the assumption that CH ≫ CSC generally holds true for semiconductors due to their low dielectric constant and low carrier density.27 Regarding the inversion region of the MS plot, we note that in all cases the potential range over which the inversion occurred coincided with the energetic density of the surface states, DOS, suggested by Ctrap (as DOS = Ctrap/e) represented in Figure 5b with baseline correction by removing the background capacitance. The total density of surface states, NSS, was calculated by integrating the curves shown in Figure 5b, again taking into account f r. To obtain accurate estimations of the parameters EFB, NA, and NSS, the EIS measurements were performed on three replicate samples. Average values are summarized in Table 1. Table 1. CuFeO2 Semiconductor Parameters Derived from Impedance Spectroscopy Dataa illumination (sun)

EFB (V vs RHE)

NA (1018 cm−3)

NSS (1013 cm−2)

0 0.5 1

1.05 ± 0.01 1.06 ± 0.01 1.07 ± 0.01

1.20 ± 0.42 1.24 ± 0.60 1.49 ± 0.84

0.74 ± 0.64 6.39 ± 0.46 6.80 ± 0.51

Taking ε = 20 and f r = 18. EFB = flat-band potential, NA = bulk acceptor density, and NSS = surface state density. a

No significant influence of the illumination intensity on the measured flat-band potential (EFB ≈ 1.05 V vs RHE) nor on the measured acceptor density (NA app∼ 1018 cm−3) is observed. We note that these values are in agreement with the previously report values using the Randles circuit.6 Interestingly, a significant density of surface states (∼1013 cm−2) was measured in the dark, contrary to what was reported for hematite,11,45 where the surface states were only accessible under

Figure 6. KPFM measurements: (a) height image of the surface of a bare CuFeO2 electrode by atomic force microscopy; (b) dark KPFM mapping of the value of EF at the surface for the same region; (c) shift in EF (ΔEF = EF,light − EF,dark) measured upon illumination of the surface of the sample. 4958


Article


oxygen bonding, while the signal at 531.6 eV can be attributed to metal hydroxyl groups, M−OH.48,49 Moreover, the shift in energy in the Fe signal is consistent with the presence of disordered Fe(III)−OH near the surface of the electrode, suggesting a disruption in the bulk crystal structurewhere iron and oxygen are organized in FeO6 octahedra.49,50 We note that the presence of hydroxide species near the surface could also explain the shift observed in the Cu(I) signal, but the presence of a Cu(I)−OH group is difficult to assess due to the lack of reports on the XPS signature of this particular chemical group. Overall, the XPS analysis suggests that the electronic surface states detected in PEC experiments are likely related to the presence of a ∼10 nm thick hydroxide layer at the surface of the electrode, while the presence of Cu(0) or Fe(II) was discounted.

■

DISCUSSION The different measurements presented here and their interpretation can be summarized in a semiquantitative description of the electronic configuration and charge transport properties of CuFeO2 thin film electrodes. Figure 8 presents a

Figure 7. (a) XPS spectra as a function of the depth of Cu 2p. (b) XPS spectra as a function of the depth of Cu LMM. (c) XPS spectra as a function of the depth of Fe 2p. (d) XPS spectra as a function of the depth of O 1s. The scans were performed at 0, 2, 4, 6, 8, 40, and 60 nm from the surface of the sample, starting from the bottom scan (red trace).

CuFeO2 surface or inside the film given the absence of the signature satellite peak between 940 and 945 eV. However, we note that it is not possible to discriminate between Cu(I) and Cu(0) on the basis of the 2p signal. However, the Cu LMM signature has been reported to be significantly different for Cu(I) and Cu(0).48 Spectra in this range (Figure 7b) showed changes with the depth of the measurement: near the surface, a single peak was observed with a kinetic energy of 916.7 eV. With increasing depth, this peak progressively shifted toward a slightly sharper peak at 918.9 eV. This indicates a change in the chemical environment of the copper centers between the surface and the bulk of the material, supporting the presence of surface states. However, the Cu LMM spectra did not present the features characteristic of Cu(0), i.e., a well-defined peak at 918.7 eV and two well-defined smaller peaks ca. 916.7 and 921.8 eV.48 Thus, rather than Cu(0), the surface states are reasonably Cu(I) centers with a chemical environment different from that of the Cu(I) in the bulk CuFeO2. Similarly, the signals for Fe and O (Figure 7c,d, respectively) show a clear change with the depth. The Fe 2p signature on the surface showed the two features characteristic of Fe(III), with binding energies of 711.2 and 724.8 eV, but upon etching, these peaks progressively shifted to lower energies of 710.2 and 722.6 eV, indicating an analogous change in the environment of the Fe atoms. Moreover, while the signature of the oxygen showed two peaks at the surface (at 529.8 and 531.6 eV), as the depth increased, the peak at 531.6 eV progressively decreased in intensity, vanishing in the bulk. In contrast, the peak at 529.8 eV progressively increased in intensity. This signal is similar to the signature of oxygen in other oxides, consistent with metal−

Figure 8. Proposed energy band diagram for an isolated CuFeO2 electrode in the dark. ECB = conduction band edge, EVB = valence band edge, EF = Fermi level, NA = acceptor density, τ = carrier lifetime, LD = carrier diffusion length, W = depletion width, and NSS = density of surface states.

band diagram displaying the key features established in this work. First, TRMC data revealed a relatively long photogenerated carrier lifetime of 200 ns and an estimated electron diffusion length of 300 nm. Furthermore, detailed EIS measurements yielded an acceptor density NA ≈ 1018 cm−3 and a flat-band potential EFB at 1.05 V vs RHE (−5.55 eV vs vacuum). Considering an effective density of states in the valence band on the order of 1019 cm−3,51 the difference in energy between the valence band edge EVB and the Fermi level EF in the bulk is estimated to be ∼60 meV at 298 K.52 On the other hand, PEC measurements reveal the presence of surface states (located roughly 0.35 eV negative of EFB) that induce Fermi level pinning at the SCLJ. We note that although the conclusions drawn from the OCP and LSV are generally identical, ΔOCP measurements (Figure 3d) yielded a photovoltage value about 150 mV lower than suggested by the LSV data (Figure 3a−c). This discrepancy could be due to voltage 4959


Article


than 1.5 mA cm−2 in the absence of recombination, which is reasonable compared to the experimental value of 0.7 mA.cm−2 reported in Figure 3 considering the likely presence of recombination outside the depletion region due to the absence of a charge-accelerating electric field. It is worth mentioning that substrate-side illumination was employed in this work due to the varying visible light absorption properties of the redox couples used. Overall, the simple photocurrent model discussed here indicates that if the surface states were successfully passivated or removed, a photovoltage of 0.8 V and reduced surface recombination would reasonably give photocurrents approaching 9−10 mA cm−2 at 1 sun illumination under electrolyte-side illumination in the absence of bulk recombination. To eliminate these surface states, we note that many reported approaches to passivate such states involve surface treatment of the film. Initial attempts in this direction have not yielded a significant improvement to date; however, the considerable variety of possible surface treatments59 and their disparate ways of influencing charge carrier dynamics at the SCLJ (surface passivation, dipole formation, etc.) leave a large field of investigation open to find the adequate treatment. These approaches are under investigation in our laboratories.

losses between the SCLJ (where ΔOCP is produced) and the back of the electrode (where the potential is measured) in the absence of current flow. To support this explanation, we note that the value of EFB extracted from the plot in Figure 3d (i.e., where ΔOCP extrapolates to 0) is close to 0.9 V vs RHE, which is also 150 mV lower than EFB extracted from the MS plot (see Figure 5a). Bulk defects such as grain boundaries may be the physical origin of this loss as they could prevent the complete flattening of the bands even at high light intensity. For this reason, our most confident assessment of the surface state position is centered at −5.2 eV vs vacuum (from LSV and EIS data). The KPFM measurements support this value, and together with the PEC measurements in different electrolytes, KPFM further indicates that the presence of surface states was not related to the electrolyte. XPS analysis rather suggests that the possible nature of these surface states is metal hydroxyl groups, while KPFM indicates their homogeneous surface distribution. As hydroxyl groups have been invoked in the past to explain surface trapping in other oxides,45,53,54 we suggest the presence of a continuous 10 nm layer of metal hydroxide or oxyhydroxide on the surface of the CuFeO2. It is worth noting that photoluminescence measurements employed to verify the lifetime of photogenerated charges independently from the TRMC measurement were unsuccessful due to the lack of measurable luminescence in the range of accessible conditions. This is consistent with the presence of a homogeneous layer of surface traps that effectively quenches luminescence from the bulk. We note further that the above-described picture of the electronics of the CuFeO2 photoelectrodes is in contrast with a previous suggestion6 that inter band gap states in the bulk were possibly responsible for the poor performance. Overall, the data presented in this study favor the conclusion that the charge trapping occurs at the semiconductor−liquid junction. Indeed, EIS analysis further indicated that the surface states act as electron traps (as an inversion regime observed is reasonably due to electrons accumulating in the surface traps) and gave an energetic distribution and density of the surface states that confirms the existence of FLP when compared to the bulk acceptor density. Due to the equilibrium of the bulk EF with the surface states, the depletion width is limited to 30 nm (see the Supporting Information for the calculation)52 and the photovoltage is limited to 0.35 V. The limited depletion width of 30 nm is small compared to the penetration depth of the light in CuFeO2 (δp = 1/α ≈ 100 nm for visible photons). Therefore, a significant portion of the charges will be generated outside the depletion width, even under front (electrolyte-side) illumination. An estimate of the maximum photocurrent expected considering this aspect was calculated using the model developed by Gartner and others.55−58 Using appropriate boundary conditions for either substrate-side or electrolyte-side illumination (see the details in the Supporting Information), the maximum photocurrent produced by a 200 nm thick CuFeO2 layer at different photovoltages was estimated in the absence of recombination. Under front illumination, with a photovoltage of 350 mV, and an associated depletion width of 30 nm, one can expect a photocurrent of ca. 6 mA cm−2. In contrast, the case of an ideal junction with water, yielding a photovoltage of 0.8 V and a depletion width of 45 nm, gives a maximum photocurrent of ca. 9 mA cm−2. On the contrary, substrate-side illumination greatly limits the photocurrent, as experimentally observed previously,6 and in the case of a SCLJ with Fe(CN)63−/4−, the expected photocurrent is no higher

■

CONCLUSIONS We have used the results of a variety of photoelectrochemical, microscopy, and spectroscopy techniques to explore and characterize thin-film CuFeO2 photocathodes. Overall, the study conducted in this paper allowed us to get a semiquantitative picture of the transport and recombination of charge carriers inside the material and at the SCLJ. We measured a high density of surface states (NSS ≈ 1014 cm−2), compared to the bulk acceptor density (NA ≈ 1018 cm−3), which is believed to cause Fermi level pinning at the SCLJ, drastically limiting the photovoltage to roughly 0.35 V, regardless of the electrolyte used to create the junction. We conclude that these states act as electron traps, causing an inversion of the depletion layer upon filling, both in the dark and under illumination, thereby promoting charge recombination at the surface. On the other hand, TRMC measurements suggested that the bulk material presented good transport properties, including a very long lifetime for photogenerated charge carriers (∼200 ns). With an associated diffusion length of about 300 nm and known good absorption properties and high stability of the material under operating conditions, CuFeO2 remains attractive for use as a photocathode, as this study indicates that its bulk properties are not the limiting factor for PEC performances. We finally note that if surface states could be passivated, we estimate the maximum photovoltage the electrode could reach for water reduction to be roughly 0.8 V, based on the TRMC measurements, making it a strong candidate for a combination with an oxide photoanode, based on BiVO4 or Fe2O3, for the creation of an earth-abundant, solution-processed, all-oxide tandem photoelectrochemical cell.

■

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.chemmater.7b01284. 4960


Article


■

(14) Le Formal, F.; Pendlebury, S. R.; Cornuz, M.; Tilley, S. D.; Grätzel, M.; Durrant, J. R. Back Electron−Hole Recombination in Hematite Photoanodes for Water Splitting. J. Am. Chem. Soc. 2014, 136, 2564−2574. (15) Cabán-Acevedo, M.; Kaiser, N. S.; English, C. R.; Liang, D.; Thompson, B. J.; Chen, H.-E.; Czech, K. J.; Wright, J. C.; Hamers, R. J.; Jin, S. Ionization of High-Density Deep Donor Defect States Explains the Low Photovoltage of Iron Pyrite Single Crystals. J. Am. Chem. Soc. 2014, 136, 17163−17179. (16) Le Formal, F.; Tétreault, N.; Cornuz, M.; Moehl, T.; Grätzel, M.; Sivula, K. Passivating Surface States on Water Splitting Hematite Photoanodes with Alumina Overlayers. Chem. Sci. 2011, 2, 737−743. (17) Jang, J.-W.; Du, C.; Ye, Y.; Lin, Y.; Yao, X.; Thorne, J.; Liu, E.; McMahon, G.; Zhu, J.; Javey, A.; Guo, J.; Wang, D. Enabling Unassisted Solar Water Splitting by Iron Oxide and Silicon. Nat. Commun. 2015, 6, 7447. (18) Gray, H. B.; Hendrickson, D. N.; Sohn, Y. S. Magnetic Susceptibility Study of Various Ferricenium and iron(III) Dicarbollide Compounds. Inorg. Chem. 1971, 10, 1559−1563. (19) Kroeze, J. E.; Savenije, T. J.; Warman, J. M. Electrodeless Determination of the Trap Density, Decay Kinetics, and Charge Separation Efficiency of Dye-Sensitized Nanocrystalline TiO2. J. Am. Chem. Soc. 2004, 126, 7608−7618. (20) Savenije, T. J.; Ferguson, A. J.; Kopidakis, N.; Rumbles, G. Revealing the Dynamics of Charge Carriers in Polymer:Fullerene Blends Using Photoinduced Time-Resolved Microwave Conductivity. J. Phys. Chem. C 2013, 117, 24085−24103. (21) Kimura, T.; Lashley, J. C.; Ramirez, A. P. Inversion-Symmetry Breaking in the Noncollinear Magnetic Phase of the Triangular-Lattice Antiferromagnet CuFeO2. Phys. Rev. B: Condens. Matter Mater. Phys. 2006, 73, 220401. (22) Terada, N.; Nakajima, T.; Mitsuda, S.; Kitazawa, H.; Kaneko, K.; Metoki, N. Ga-Substitution-Induced Single Ferroelectric Phase in Multiferroic CuFeO2. Phys. Rev. B: Condens. Matter Mater. Phys. 2008, 78, 014101. (23) Bandara, J.; Mielczarski, J. A.; Kiwi, J. 1. Molecular Mechanism of Surface Recognition. Azo Dyes Degradation on Fe, Ti, and Al Oxides through Metal Sulfonate Complexes. Langmuir 1999, 15, 7670−7679. (24) Kay, A.; Cesar, I.; Grätzel, M. New Benchmark for Water Photooxidation by Nanostructured α-Fe2O3 Films. J. Am. Chem. Soc. 2006, 128, 15714−15721. (25) Bourée, W. S.; Prévot, M. S.; Jeanbourquin, X. A.; Guijarro, N.; Johnson, M.; Formal, F. L.; Sivula, K. Robust Hierarchically Structured Biphasic Ambipolar Oxide Photoelectrodes for Light-Driven Chemical Regulation and Switchable Logic Applications. Adv. Mater. 2016, 28, 9308−9312. (26) Brenot, R.; Vanderhaghen, R.; Drevillon, B.; French, I.; Roca i Cabarrocas, P. Time Resolved Microwave Conductivity Measurements for the Characterization of Transport Properties in Thin Film MicroCrystalline Silicon. Thin Solid Films 1997, 296, 94−97. (27) Dicker, G.; de Haas, M. P.; Siebbeles, L. D. A.; Warman, J. M. Electrodeless Time-Resolved Microwave Conductivity Study of Charge-Carrier Photogeneration in Regioregular poly(3-Hexylthiophene) Thin Films. Phys. Rev. B: Condens. Matter Mater. Phys. 2004, 70, 045203. (28) Abdi, F. F.; Savenije, T. J.; May, M. M.; Dam, B.; van de Krol, R. The Origin of Slow Carrier Transport in BiVO4 Thin Film Photoanodes: A Time-Resolved Microwave Conductivity Study. J. Phys. Chem. Lett. 2013, 4, 2752−2757. (29) Chen, H.-Y.; Wu, J.-H. Transparent Conductive CuFeO2 Thin Films Prepared by Sol−gel Processing. Appl. Surf. Sci. 2012, 258, 4844−4847. (30) Ravensbergen, J.; Abdi, F. F.; van Santen, J. H.; Frese, R. N.; Dam, B.; van de Krol, R.; Kennis, J. T. M. Unraveling the Carrier Dynamics of BiVO4: A Femtosecond to Microsecond Transient Absorption Study. J. Phys. Chem. C 2014, 118, 27793−27800.

Additional optical and electron microscopy images, Raman spectra, TRMC data analysis, photovoltage estimation, determination of E1/2 for redox couples, and estimation of the photocurrent (PDF)

AUTHOR INFORMATION

Corresponding Author

*E-mail: kevin.sivula@epfl.ch. ORCID

Fatwa Abdi: 0000-0001-5631-0620 Kevin Sivula: 0000-0002-8458-0270 Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS The Swiss Competence Centers for Energy Research (SCCER Heat and Electricity Storage) funded by the Kommission für Technologie und Innovation (KTI; Contract No. 1155002545) is gratefully acknowledged for financial support.

■

REFERENCES

(1) Prévot, M. S.; Sivula, K. Photoelectrochemical Tandem Cells for Solar Water Splitting. J. Phys. Chem. C 2013, 117, 17879−17893. (2) Hu, S.; Xiang, C.; Haussener, S.; Berger, A. D.; Lewis, N. S. An Analysis of the Optimal Band Gaps of Light Absorbers in Integrated Tandem Photoelectrochemical Water-Splitting Systems. Energy Environ. Sci. 2013, 6, 2984−2993. (3) Seger, B.; Castelli, I. E.; Vesborg, P. C. K.; Jacobsen, K. W.; Hansen, O.; Chorkendorff, I. 2-Photon Tandem Device for Water Splitting: Comparing Photocathode First versus Photoanode First Designs. Energy Environ. Sci. 2014, 7, 2397−2413. (4) Read, C. G.; Park, Y.; Choi, K.-S. Electrochemical Synthesis of PType CuFeO2 Electrodes for Use in a Photoelectrochemical Cell. J. Phys. Chem. Lett. 2012, 3, 1872−1876. (5) Gu, J.; Wuttig, A.; Krizan, J. W.; Hu, Y.; Detweiler, Z. M.; Cava, R. J.; Bocarsly, A. B. Mg-Doped CuFeO2 Photocathodes for Photoelectrochemical Reduction of Carbon Dioxide. J. Phys. Chem. C 2013, 117, 12415−12422. (6) Prévot, M. S.; Guijarro, N.; Sivula, K. Enhancing the Performance of a Robust Sol−Gel-Processed P-Type Delafossite CuFeO2 Photocathode for Solar Water Reduction. ChemSusChem 2015, 8, 1359− 1367. (7) Kang, U.; Choi, S. K.; Ham, D. J.; Ji, S. M.; Choi, W.; Han, D. S.; Abdel-Wahab, A.; Park, H. Photosynthesis of Formate from CO2 and Water at 1% Energy Efficiency via Copper Iron Oxide Catalysis. Energy Environ. Sci. 2015, 8, 2638−2643. (8) Prévot, M. S.; Li, Y.; Guijarro, N.; Sivula, K. Improving Charge Collection with Delafossite Photocathodes: A Host−guest CuAlO2 /CuFeO2 Approach. J. Mater. Chem. A 2016, 4, 3018−3026. (9) Jang, Y. J.; Park, Y. B.; Kim, H. E.; Choi, Y. H.; Choi, S. H.; Lee, J. S. Oxygen-Intercalated CuFeO2 Photocathode Fabricated by Hybrid Microwave Annealing for Efficient Solar Hydrogen Production. Chem. Mater. 2016, 28, 6054−6061. (10) Wuttig, A.; Krizan, J. W.; Gu, J.; Frick, J. J.; Cava, R. J.; Bocarsly, A. B. The Effect of Mg-Doping and Cu Nonstoichiometry on the Photoelectrochemical Response of CuFeO2. J. Mater. Chem. A 2017, 5, 165−171. (11) Klahr, B.; Gimenez, S.; Fabregat-Santiago, F.; Hamann, T.; Bisquert, J. Water Oxidation at Hematite Photoelectrodes: The Role of Surface States. J. Am. Chem. Soc. 2012, 134, 4294−4302. (12) Du, C.; Zhang, M.; Jang, J.-W.; Liu, Y.; Liu, G.-Y.; Wang, D. Observation and Alteration of Surface States of Hematite Photoelectrodes. J. Phys. Chem. C 2014, 118, 17054−17059. (13) Iandolo, B.; Hellman, A. The Role of Surface States in the Oxygen Evolution Reaction on Hematite. Angew. Chem., Int. Ed. 2014, 53, 13404−13408. 4961


Article

Chemistry of Materials (31) Lewis, N. S. A Quantitative Investigation of the Open-Circuit Photovoltage at the Semiconductor/Liquid Interface. J. Electrochem. Soc. 1984, 131, 2496−2503. (32) Benko, F. A.; Koffyberg, F. P. Opto-Electronic Properties of Pand N-Type Delafossite, CuFeO2. J. Phys. Chem. Solids 1987, 48, 431− 434. (33) Godinho, K. G.; Morgan, B. J.; Allen, J. P.; Scanlon, D. O.; Watson, G. W. Chemical Bonding in Copper-Based Transparent Conducting Oxides: CuMO2 (M = In, Ga, Sc). J. Phys.: Condens. Matter 2011, 23, 334201. (34) Suzuki, I.; Nagatani, H.; Kita, M.; Iguchi, Y.; Sato, C.; Yanagi, H.; Ohashi, N.; Omata, T. First-Principles Study of CuGaO2 Polymorphs: Delafossite α-CuGaO2 and Wurtzite β-CuGaO2. Inorg. Chem. 2016, 55, 7610−7616. (35) Bard, A. J.; Bocarsly, A. B.; Fan, F. R. F.; Walton, E. G.; Wrighton, M. S. The Concept of Fermi Level Pinning at Semiconductor/Liquid Junctions. Consequences for Energy Conversion Efficiency and Selection of Useful Solution Redox Couples in Solar Devices. J. Am. Chem. Soc. 1980, 102, 3671−3677. (36) Kelly, J. J.; Memming, R. The Influence of Surface Recombination and Trapping on the Cathodic Photocurrent at pType III-V Electrodes. J. Electrochem. Soc. 1982, 129, 730−738. (37) Baglio, J. A.; Calabrese, G. S.; Harrison, D. J.; Kamieniecki, E.; Ricco, A. J.; Wrighton, M. S.; Zoski, G. D. Electrochemical Characterization of P-Type Semiconducting Tungsten Disulfide Photocathodes: Efficient Photoreduction Processes at Semiconductor/Liquid Electrolyte Interfaces. J. Am. Chem. Soc. 1983, 105, 2246− 2256. (38) Lewerenz, H. J. Surface States and Fermi Level Pinning at Semiconductor/ Electrolyte Junctions. J. Electroanal. Chem. 1993, 356, 121−143. (39) Warren, E. L.; Boettcher, S. W.; Walter, M. G.; Atwater, H. A.; Lewis, N. S. pH-Independent, 520 mV Open-Circuit Voltages of Si/ Methyl Viologen2+/+ Contacts Through Use of Radial n+p-Si Junction Microwire Array Photoelectrodes. J. Phys. Chem. C 2011, 115, 594− 598. (40) Nagasubramanian, G.; Wheeler, B. L.; Bard, A. J. Semiconductor Electrodes XLIX. Evidence for Fermi Level Pinning and Surface-State Distributions from Impedance Measurements in Acetonitrile Solutions with Various Redox Couples. J. Electrochem. Soc. 1983, 130, 1680− 1688. (41) Bisquert, J. Chemical Capacitance of Nanostructured Semiconductors: Its Origin and Significance for Nanocomposite Solar Cells. Phys. Chem. Chem. Phys. 2003, 5, 5360−5364. (42) Bisquert, J. Physical Electrochemistry of Nanostructured Devices. Phys. Chem. Chem. Phys. 2008, 10, 49−72. (43) Bertoluzzi, L.; Lopez-Varo, P.; Jiménez Tejada, J. A. J.; Bisquert, J. Charge Transfer Processes at the Semiconductor/Electrolyte Interface for Solar Fuel Production: Insight from Impedance Spectroscopy. J. Mater. Chem. A 2016, 4, 2873−2879. (44) Sato, N. Electrochemistry at Metal and Semiconductor Electrodes; Elsevier Science B.V.: Amsterdam, 1998. (45) Klahr, B.; Hamann, T. Water Oxidation on Hematite Photoelectrodes: Insight into the Nature of Surface States through In Situ Spectroelectrochemistry. J. Phys. Chem. C 2014, 118, 10393− 10399. (46) Klahr, B.; Gimenez, S.; Fabregat-Santiago, F.; Bisquert, J.; Hamann, T. W. Electrochemical and Photoelectrochemical Investigation of Water Oxidation with Hematite Electrodes. Energy Environ. Sci. 2012, 5, 7626−7636. (47) Melitz, W.; Shen, J.; Kummel, A. C.; Lee, S. Kelvin Probe Force Microscopy and Its Application. Surf. Sci. Rep. 2011, 66, 1−27. (48) Poulston, S.; Parlett, P. M.; Stone, P.; Bowker, M. Surface Oxidation and Reduction of CuO and Cu2O Studied Using XPS and XAES. Surf. Interface Anal. 1996, 24, 811−820. (49) Mansour, A. N.; Brizzolara, R. A. Characterization of the Surface of α-FeOOH Powder by XPS. Surf. Sci. Spectra 1996, 4, 357−362.

(50) Grosvenor, A. P.; Kobe, B. A.; Biesinger, M. C.; McIntyre, N. S. Investigation of Multiplet Splitting of Fe 2p XPS Spectra and Bonding in Iron Compounds. Surf. Interface Anal. 2004, 36, 1564−1574. (51) Krishnan, R. Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry. Encyclopedia of Electrochemistry; WileyVCH Verlag GmbH & Co. KGaA: Weinheim, Germany, 2007. (52) Cendula, P.; Tilley, S. D.; Gimenez, S.; Bisquert, J.; Schmid, M.; Grätzel, M.; Schumacher, J. O. Calculation of the Energy Band Diagram of a Photoelectrochemical Water Splitting Cell. J. Phys. Chem. C 2014, 118, 29599−29607. (53) Di Valentin, C.; Pacchioni, G.; Selloni, A. Electronic Structure of Defect States in Hydroxylated and Reduced Rutile TiO2 (110) Surfaces. Phys. Rev. Lett. 2006, 97, 166803. (54) Sivula, K.; Zboril, R.; Le Formal, F.; Robert, R.; Weidenkaff, A.; Tucek, J.; Frydrych, J.; Grätzel, M. Photoelectrochemical Water Splitting with Mesoporous Hematite Prepared by a Solution-Based Colloidal Approach. J. Am. Chem. Soc. 2010, 132, 7436−7444. (55) Gärtner, W. W. Depletion-Layer Photoeffects in Semiconductors. Phys. Rev. 1959, 116, 84−87. (56) Reichman, J. The Current-voltage Characteristics of Semiconductor-electrolyte Junction Photovoltaic Cells. Appl. Phys. Lett. 1980, 36, 574−577. (57) Allongue, P.; Cachet, H.; Horowitz, G. Detailed Analysis of a Redox Stabilized Liquid Junction Solar Cell. Application to the Cell. J. Electrochem. Soc. 1983, 130, 2352−2357. (58) Choo, S. C.; Tan, L. S.; Quek, K. B. Theory of the Photovoltage at Semiconductor Surfaces and Its Application to Diffusion Length Measurements. Solid-State Electron. 1992, 35, 269−283. (59) Guijarro, N.; Prévot, M. S.; Sivula, K. Surface Modification of Semiconductor Photoelectrodes. Phys. Chem. Chem. Phys. 2015, 17, 15655−15674.

4962


Evaluating Charge Carrier Transport and Surface States in CuFeO2

Recommend Documents