The effect of Mg-doping and Cu nonstoichiometry on the photoelectrochemical response of CuFeO₂

Abstract

We report the tuning of CuFeO₂ photoelectrodes by Mg doping and Cu deficiency to demonstrate the effects of carrier concentration on the photoresponse. Carrier type and concentration were quantitatively assessed using the Hall effect on pure, Mg-incorporated, and Cu-deficient pellets (CuFe_1−xMg_xO₂ and Cu_1−yFeO₂, x = 0, 0.0005, 0.005, 0.02, and y = 0.005, 0.02) over the range of thermodynamic stability achievable using solid-state synthesis. The same samples were used in a photoelectrochemical cell to measure their photoresponse. We find that the material with the lowest p-type carrier concentration and the highest carrier mobility shows the largest photoresponse. Furthermore, we show that increasing the p-type carrier concentration and thus the conductivity to high levels is limited by the delafossite defect chemistry, which changes the majority carrier type from p-type to n-type near the Mg solubility limit (x = 0.05) and at high Cu defect concentrations.

---

1 Introduction

Semiconductor oxides offer unique advantages as photoelectrodes in photoelectrochemical cells (PEC). Unlike III–V semiconductors, which can degrade easily,^1–5 oxide materials have been shown to exhibit long-term stability and easy processability.^6–8 The oxides that yield photoreduced products from water or dissolved CO₂, indicating alignment of solid/liquid electronic energies, tend to have wide band gaps,^9–11 and thus they are not optimal materials for PEC applications. Their large band gaps and consequently lower absorption in the solar spectral region lead to decreased photoactivity and, ultimately, lower product yields.¹²

While chemical doping has been employed as an avenue to reduce the band gap,^13–16 this strategy shows discrepant results. Both increased^17,18 and unchanged^18–20 photoresponse with doping have been reported. A rarely approached avenue to optimize materials for photoelectrochemical applications is to use doping to interrogate the influence of a semiconductor photoelectrode's carrier concentration on its photoresponse. Here, we examine the influence of changing the carrier concentration on the photoresponse of the emergent photoelectrode CuFeO₂.

Doping is known to increase a material's carrier concentration (N), and thus enhance its conductivity. When BiVO₄ was doped with Mo¹⁷ or W,²¹ for example, enhancement in photoactivity was explained by a dramatic increase in the electron concentration. Similarly, doped Fe₂O₃ showed an increase in photoresponse that was attributed to greater conductivity.¹⁸ Conversely, increasing N can also hinder the performance of a photoelectrode. Large N can reduce photocurrents because the depletion width (W_SC) is inversely proportional to N. Smaller W_SC allows for a lesser volume from which photogenerated carriers can be formed, reducing the photoeffect.^22,23

Thus, an optimal carrier concentration for photoresponse is expected. Here, we investigate this balance point for CuFeO₂, a promising photoelectrode, by systematically changing the carrier concentration with doping. CuFeO₂ is currently of interest due to its relatively small bandgap (E_g = 1.15 eV),^24,25 favorable band structure for hydrogen production,²⁶ thermodynamic stability, and fabrication from earth-abundant, nontoxic constituents. In addition, CuFeO₂ has recently been reported to reduce CO₂ to formate under illumination at an underpotential.²⁷ Despite these properties, the reported photocurrents are low, 0.16 mA cm⁻² at 100 W cm⁻².²⁶ Although photoelectrochemical evaluation of intrinsic CuFeO₂ has been conducted,^28,29 the effect of dopants on its photoresponse have not been investigated.

In the CuFeO₂ delafossite, a layer of FeO₂ is formed by edge-sharing FeO₆ octahedra that are separated by monovalent copper sticks.³⁰ The intrinsic material is p-type, and the reported electrical conductivities are low.³¹ Although the conductivity has been reported to improve for non-stoichiometric CuFeO_2+x,³¹ the room temperature value was reported to be 2 S cm⁻¹, only a ∼1.5-fold improvement from the intrinsic material.³² It has been shown that substituting a divalent metal (i.e. Mg²⁺) for the trivalent metal (i.e. Fe³⁺) renders CuFeO₂ more p-type and increases its conductivity.^24,33,34 p-Type electrodes are useful photocathodes for reduction reactions, particularly for CO₂ reduction and H₂ evolution, further supporting interest in CuFeO₂.^26,35,36

Here, within range of thermodynamic stability achievable using solid-state synthesis, we show that the photoresponse of Mg-doped CuFeO₂ varies with both p-type carrier concentration and mobility. For comparison, a second source of p-type carriers, Cu-deficiency, is explored to garner information about the correlation of the hole mobility to the material's defect chemistry, and ultimately its photoresponse. We find that increasing the impurity concentration by either Mg doping or Cu deficiency decreases the photoeffect in CuFeO₂, accompanied by lower majority carrier mobilities, and that the system eventually crosses over from p-type to n-type at high defect concentrations.

2 Experimental

CuFe_1−xMg_xO₂ (x = 0, 0.0005, 0.005, 0.02) powder was prepared by grinding stoichiometric amounts of CuO (Sigma Aldrich 99.999%), Fe₂O₃ (Alfa Aesar 99.99%), and MgO (Alfa Aesar 99.99%), according to previously reported procedures.^24,37,38 Roughly 10 g of sample was prepared for samples with lower concentrations of Mg (x = 0.0005, 0.005). The ground powder was heated to 950 °C in a covered alumina crucible under flowing Ar atmosphere for 192 hours with intermittent grinding for x = 0, 0.0005, and 0.005 samples. The x = 0.02 sample was heated to 850 °C under the same reaction conditions. The resultant pellets were black. The reaction progress was monitored using powder X-ray diffraction (XRD) on a Rigaku Miniflex equipped with a Cu Kα X-ray source and a diffracted beam graphite monochromator. After CuFe_1−xMg_xO₂ showed a single delafossite phase in the XRD pattern, the samples were filtered through a 325 mm sieve, and ground with stearic acid in a 16 : 1 mole ratio. Addition of ethanol under low heat dissolved the stearic acid in the mixture, which was then dried, reground, and pressed into a pellet (0.4 mm thickness, 0.785 cm² area) with 4 tons of pressure. The pellets were first heated at 500 °C in air for 4 hours and then sintered for 96 hours under Ar atmosphere at 925 °C. The obtained pellets (≥85% of theoretical density³⁰) were then used for subsequent characterization. Cu_1−yFeO₂ pellets were synthesized using the same methodology with CuO and Fe₂O₃ as starting materials.

A Quanta 200 Field Emission Gun Environmental-Scanning Electron Microscope (SEM) equipped with an integrated Oxford System was employed for Energy Dispersive X-ray (EDX) analysis. The X-ray penetration depth was 1–2 mm. X-ray photoelectron spectroscopy (XPS) spectra were collected under 10⁻⁹ Torr using an electron spectrometer (VG ESCALAB MkII) in constant hemispherical energy analyzer mode with pass energies of 50 and 20 eV for survey and detailed scans, respectively. A Mg Kα X-ray source centered at 1.254 keV was employed. The C 1s peak at 285 eV of adventitious hydrocarbon was used as an internal binding energy reference.

Hall effect data and electrical resistivities were collected using a Quantum Design Physical Property Measurement System. The samples were cut from the pellets into 0.4 mm-thick rectangles, measuring approximately 1.9 mm by 0.6 mm. Hall effect data were obtained at 300 K by sweeping the magnetic field from a negative value to the equivalent positive value at a constant current of 0.1 mA through the sample. The carrier concentration, N, was estimated by assuming a single dominant carrier type and the relationship,

(e – electron charge, R/B – slope obtained from the measured Hall resistance vs. magnetic field measurement, and t – sample thickness). Resistivity was measured over a range of 100 K to 300 K. Both Hall and resistivity four point probe measurements were facilitated by attaching platinum wires to the samples using Ag paint.

Photocurrent density measurements were carried out using the delafossite pellets as the working electrode, Pt mesh (Sigma Aldrich) as the counter electrode and standard calomel electrode (SCE, Accument) as the reference electrode in 0.1 M NaHCO₃ (≥99%, Fisher) aqueous electrolyte, pH 8.1. In the photoelectrochemical cell (PEC), constant incident light (75 W Xe-arc lamp, USHIO UXL 151H, 350 nm-1350 nm) was illuminated onto the working electrode at a fixed distance (∼9 cm) and monitored using a CdS photocell (100 W readout, corresponding to 700 ± 50 mW cm⁻², PerkinElmer) 6.7 mm below the working electrode. The PEC was open to air. The PEC cell was made out of Teflon, and quartz windows were secured opposite to the working electrode, which was fitted in a 0.7 cm diameter slot created to fit synthesized delafossite pellets after protecting the edges with Teflon tape to prevent leakage of the electrolyte. The delafossite pellets were 0.4 mm thick, and they were back-side Au-sputtered (100 nm, Angstrom Metal Sputter), attached to Teflon-coated copper leads using Ag epoxy, and secured with water-tight epoxy (Epoxy Technology H31). The contacts were ohmic (Fig. S1†), and the PEC design allowed for only the front side of the pellet to contact the electrolyte and to be illuminated.

The chopped light experiments were performed using the above experimental setup and carried out using a CHI760 Electrochemical Potentiostat in linear sweep voltammetry mode with a scan rate of 100 mV s⁻¹. The sweep was performed at reducing potentials (0 V vs. SCE to −1.5 V vs. SCE) for p-type samples and at oxidizing potentials (0 V vs. SCE to 1.5 V vs. SCE) for n-type samples. The frequency of the chopped light was monitored using a Thorlabs-Optical Chopper MC1000A at 198 Hz. Shorter chopping frequencies did not give a detectable measure of the photocurrent, and longer chopping frequencies did not change the amount or yield of the photocurrent.

3 Results and discussion

3.1 CuFe_1−xMg_xO₂

Fig. 1 shows the powder X-ray diffraction (XRD) patterns for the sintered CuFe_1−xMg_xO₂ pellets. The samples show pure delafossite patterns. Samples for x = 0.05 and 0.07 were also prepared; the higher Mg-doped samples contained Cu₂O impurities in agreement with previous studies.²⁴ Thus, x = 0, 0.0005, 0.005 and 0.02 samples were used for subsequent characterization. The EDX analysis of the as-synthesized CuFeO₂ pellet, Fig. S2,† shows roughly 1 : 1 ratio of Cu to Fe and did not yield any other elements besides O (in anticipated 2 : 1 ratio with both Cu and Fe) and adventitious carbon, suggesting that impurities below the detection limit of EDX (up 0.05%) are not present in the material and that the delafossite stoichiometry is preserved. Furthermore, additional EDX analysis of CuFe_1−xMg_xO₂, Table S1,† display incorporation of Mg content ranging from 0 to 1 atomic%, mirroring the increasing amount of prepared dopant from sample to sample. For x = 0.0005, the Mg content was not visible because the doping amount was below the detection limit of the instrument (0.05%). For x = 0.005 and 0.0005, it is noted that the dopant amount may be equal to that of the impurity of the precursor (Fe₃O₄, 99.99% metal basis), however possible impurities present in the precursor material itself were not detectable in the pure x = 0 sample. Significant changes observed in transport properties, shown below, suggest that Mg-doping is present in the lowest doped sample. XPS analyses were performed to probe Mg on the surface: no detectable precipitates or other inhomogeneities in Mg distribution were seen at the surface (Fig. S3†).

Fig. 1 XRD patterns of sintered CuFe_1−xMg_xO₂ pellets for x = 0, 0.0005, 0.005, and 0.02 (A–D), and Cu_1−yFeO₂ pellets, y = 0.005 and 0.02 (E and F). All show a pure CuFeO₂ delafossite pattern, as indicated by the purple database peaks.

Representative Hall resistivities of CuFe_1−xMg_xO₂, x = 0.005 and x = 0.02 at 300 K are shown in Fig. 2. The pellets synthesized by our method produced a reliable linear fit of the Hall data (R² ≥ 0.92) to applied field, allowing for carrier concentration to be determined. The hole and electron concentrations are listed in Table 1.

Fig. 2 Hall resistivity data for CuFe_1−xMg_xO₂. (A) x = 0.0005 depict a positive slope, indicating p-type majority carriers. (B) x = 0.02 depict a negative slope, indicating n-type majority carriers.

Table 1 Carrier concentration (N), type, resistivity (ρ), conductivity (σ), majority carrier mobility (μ), and cathodic photoactivity (j_ph,c) of sintered pellets of CuFe_1−xMg_xO₂ and Cu_1−yFeO₂

Sample	x = 0	x = 0.0005	x = 0.005	x = 0.02	Cu deficient, y = 0.005	Cu deficient, y = 0.02
N (cm^−3)	(8 ± 2) × 10^15	(8 ± 2) × 10^18	(1 ± 0.2) × 10^18	(1 ± 0.3) × 10^17	(2.2 ± 0.6) × 10^16	(2 ± 0.5) × 10^18
Carrier type	p	p	p	n	p	n
ρ at 300 K (Ω cm)	9 ± 3	0.8 ± 0.3	0.6 ± 0.1	0.2 ± 0.05	5 ± 1	5 ± 2
σ (S cm^−1)	0.1 ± 0.04	1 ± 0.5	2 ± 0.4	6 ± 2	0.2 ± 0.05	0.2 ± 0.05
μ n, μp (cm^2 V^−1 s^−1)	90 ± 40	1 ± 0.4	10 ± 3	400 ± 100	50 ± 20	60 ± 15
j ph,c (mA cm^−2)	0.43 ± 0.01	0.20 ± 0.01	0.16 ± 0.01	NA	0.19 ± 0.01	NA

---

The Hall effect measurements show that substituting x = 0.0005 Mg to nominally replace an equivalent amount of Fe in pure CuFeO₂ leads to increased hole doping over the undoped material. Assuming that Mg²⁺ substitutes for Fe³⁺ (see section 3.3), more p-type carriers are expected as the Mg content increases. We find that at x = 0.005, however, the increased Mg content leads to compensation of the p-type carriers by the introduction of n-type carriers. Eventually, in the vicinity of the solubility limit at x = 0.02, the material crosses over to n-type, consistent with the trend suggested by the x = 0.005 sample. This observation indicates the existence of a more complex underlying defect chemistry in this compound.

The temperature dependencies of the electrical resistivities are plotted in Fig. 3A. The exponentially decreasing resistivities with increasing temperature indicate that CuFe_1−xMg_xO₂, x = 0, 0.0005, 0.005, 0.02 are semiconducting. The room temperature (300 K) resistivities are listed in Table 1. Significant differences from previously reported resistivities²⁴ are noted; these differences may be attributed to variations in sample preparation. All measurements reported here are made on the same set of samples, and thus the data are self consistent.

Fig. 3 (A) Resistivity vs. temperature plots of CuFe_1−xMg_xO₂ and Cu_1−yFeO₂, depicting semiconducting behaviour for all samples. (B) Resistivity measurements of CuFe_1−xMg_xO, x = 0.0005, pellet, plotted as log of resistivity vs. inverse temperature, yielding thermal activation energies.

Based on the equation,

(ρ – sample resistance, ρ₀ – preexponenial term, E_a – thermal activation energy, k_B – Boltzmann's constant, T – the sample temperature), a linear semi-logarithmic fit can be constructed to estimate E_a values. A representative fit is shown in Fig. 3B. For CuFe_1−xMg_xO₂x = 0, 0.0005, 0.005 and 0.02, the E_a values obtained were 0.053 ± 0.003 eV, 0.034 ± 0.002 eV, 0.021 ± 0.001 eV, and 0.012 ± 0.001 eV, respectively. Given the fact that the measured bulk optical band gap is near 1 eV (see UV-Vis-NIR absorption spectrum, Fig. S4†), this behavior is attributed to the presence of defect-derived impurity bands, localized near the valence band or conduction band edge depending on the dominant carrier type.

As a useful further characterization, the mobilities of the majority carriers were estimated by assuming a single dominant carrier and band-like carrier motion. Using the measured quantities and the relationship

σ = Neμ (3)

(σ – conductivity at 300 K, μ – mobility of the carrier, e – the electron charge, and N the carrier concentration), the mobilities were calculated (Table 1). The majority carrier mobility estimated in this fashion does not follow a simple trend. For the samples dominated by p-type behavior in CuFe_1−xMg_xO₂, x = 0, 0.0005 and 0.005, hole carrier mobilities were estimated as 90 ± 40 cm² V⁻¹ s⁻¹, 1 ± 0.4 cm² V⁻¹ s⁻¹ and 10 ± 3 cm² V⁻¹ s⁻¹, respectively. The relatively high mobilities observed for p-type CuFeO₂ materials motivate further studies to understand possible band-like transport in these materials, however, contemporary band structure calculations³⁹ are complicated by the known antiferromagnetic ordering of this lattice.^40,41 For the x = 0.02 sample, the carrier mobility for transport dominated by electrons was significantly higher, at 400 ± 100 cm² V⁻¹ s⁻¹. The high mobility shown by the n-type samples is not surprising, since in general electrons are known to have higher mobilities than holes.⁴² Assuming a rigid band structure of the host CuFeO₂, the reduction in hole mobility with Mg dopant for x = 0.0005 and x = 0.005 is attributed to changes in the scattering centers induced by the dopants. We note that an extreme low mobility of ≤ 1 cm² V⁻¹ s⁻¹ has been interpreted as evidence for charge transport by polarons in similar delafossite compounds.^24,43

With the electronic characterization in hand, the pellets were used as photoelectrodes in an aqueous PEC. Under similar conditions to those employed in this work, this system has been reported to reduce water to H₂ under illumination.^26,35Fig. 4 shows the behavior of CuFe_1−xMg_xO₂x = 0, x = 0.0005, x = 0.005, x = 0.02, and Cu deficient samples under chopped light illumination at reducing potentials. Reducing potentials, 0 ≥ E ≥ −1.5 V vs. SCE, were chosen because most of the prepared samples exhibited p-type behavior, a key characteristic of photocathodes. The n-type material x = 0.02 does not display a photoresponse at reducing potentials.

Fig. 4 Linear sweep voltammetric measurements of Au-sputtered CuFe_1−xMg_xO₂ pellets with varying x and Cu_1−yFeO₂, y = 0.005 recorded in 0.1 M NaHCO₃, pH 8.1. The scan rate was 100 mV s⁻¹ and the direction of scan was negative. During the measurement, the pellets were periodically illuminated using a 2 Hz chopping frequency by a Xe-arc lamp. The lamp intensity was held constant by monitoring it with a CdS cell.

The Cu : Fe 1 : 1 stoichiometry is preserved even after multiple linear sweep voltammetry measurements, as shown by EDX analyses (Fig. S5A†), with only detectable oxygen and carbon as additional elements. These data, combined with the lack of observable Cu nanoparticles shown in the SEM data after polarization (Fig. S5B†), suggest that corrosive formation of metallic copper does not occur over the short timescale of linear sweep voltammetry measurements and simultaneously collected during electrode illumination. In addition, we note that electrolyses occurring over 8 hours ²⁷ has shown to give rise to XRD diffraction patterns indicative of metallic Cu formation, illustrating that the corrosion processes of CuFeO₂ under reductive potential bias occurs over timescale of several hours.

The observed photoresponse of intrinsic CuFeO₂ is consistent with the presence of the optical direct band gap of 0.9 ± 0.2 eV (Fig. S4†). We report the photocurrent densities from the seventh scan at −0.7 V vs. SCE. This potential value is the thermodynamic reduction potential for water reduction under these pH and electrolyte conditions. Beyond the thermodynamic potential, the observed photocurrent is assisted by an electrode bias. The same number of scans was performed on each electrode sample, and the appearance of significant dark current at −0.9 V vs. SCE remained constant with successive scans.

To a first approximation, the difference in dark and photocurrent with increasing cathodic polarization remains constant beyond −0.7 V vs. SCE, indicative of a saturation photocurrent regime, i.e. the electronic properties of the material govern the observed photocurrent rather than the interfacial electrochemical reaction kinetics. Therefore, while the end point electrochemical reaction remains ambiguous, the nature of the electrochemical reaction at the interface should not be different for the varying dopant concentrations in the current samples, and therefore is not further discussed. Furthermore, the band edge positions of the materials, as examined by the flat band potential and the dark open circuit voltage potential, do not depend on the dopant concentration (Fig. S6†). Because a material's photocurrent depends on the nature of the electrochemical reaction occurring at the interface,²⁶ electrokinetic studies to elucidate the reactions by using varying electrolyte media and atmosphere may be of future interest.

The undoped sample, x = 0, yields the highest photocurrent at 0.43 ± 0.01 mA cm⁻², comparable to that previously reported for undoped thin film and bulk p-type CuFeO₂,^26,27 although direct comparison is not possible because photocurrents and illuminated light intensity do not scale linearly due to recombination effects and varying kinetics of the electrochemical reaction at the interface.^26,44 At the dopant concentrations x = 0.0005 and x = 0.005 lower photocurrents of 0.20 ± 0.01 mA cm⁻² and 0.16 ± 0.01 mA cm⁻², respectively, are observed. We note that the photocurrents of the CuFeO₂ pellets used in this study are lower than prior reports,^7,20,35 which could reflect differences in grain boundary densities between differently prepared materials. The dark currents of the samples increase with doping, which is attributed to their higher conductivities. The n-type x = 0.02 sample is conducting, but lacks a photoresponse at all negative potentials.

As seen in Fig. 4, there is a significant difference between the maximum photocurrent obtained from p-type samples (x = 0, 0.0005, and 0.005) and the maximum dark current of about 3 mA cm⁻² passed through the n-type (x = 0.02) sample. This limited photocurrent of the p-type samples, is attributed to surface recombination processes. Assuming that enough light intensity was applied, a photon-limited current can be ruled out. Additionally, x = 0.0005 and x = 0.005 materials have similar resistivities to that of x = 0.02, so the efficiency of charge mobility is not the contributing factor restricting the photocurrent.

For the CuFe_1−xMg_xO₂ system, we observe no correlation between the mobility of the majority carriers and the photocurrents. Even though doping increases the conductivity of the material, e.g. for x = 0.0005 and x = 0.005, these materials exhibit significantly reduced photocurrents. The p-type carriers for x = 0 are more mobile than those introduced by Mg-dopant in the x = 0.0005 and 0.005 samples. Comparison, however, of these latter two samples to make a correlation of photocurrent with carrier mobility is complicated by the fact that some n-type carrier compensation may already be present for x = 0.005, evidenced by its lower net p-type carrier concentration. To determine the influence of the mobility of the photogenerated minority carriers on the photocurrent density, experiments should be done to determine the mobility of the photogenerated minority carriers by measuring their diffusion lengths and lifetimes.

3.2 Cu_1−yFeO₂

An alternative way to introduce holes in CuFeO₂ is by incorporating Cu deficiency. Cu_1−yFeO₂, y = 0.005 and 0.02, were therefore tested for their electronic and photoactive properties (Table 1). The XRD patterns and the temperature dependent resistivity measurements are shown in Fig. 1 and Fig. 3, respectively. Cu_1−yFeO₂, y = 0.0005, did not yield a detectable slope in the Hall data and was therefore omitted from the analysis. The y = 0.02 sample showed n-type behavior, and consequently was not tested for photoactivity at reducing potentials. For y = 0.005, the p-type carrier concentration was larger than that of the stoichiometric material, while the mobility of the majority carriers decreased slightly. This sample showed a decrease in photoactivity (Fig. 4).

3.3 Defect chemistry of doped and nonstoichiometric CuFeO₂

Many theoretical studies have been undertaken to understand the underlying defect chemistry^45–47 in p-type conductive oxides based on Cu₂O and the related CuMO₂ (M = trivalent metal) delafossites.^13,48,49 Using Kröger–Vink notation⁵⁰ the possible point defects within the CuFeO₂ lattice are:

Of these proposed mechanisms, four ((4), (5), (7) and (9)) give rise to p-type behavior. Two key energies are involved: (1) the formation energy of the point defects; and (2) the ionization energy of the defects. A theoretical treatment by Raebiger et al. found, for systems with O–Cu(I)–O coordination (also found in delafossites), that copper vacancies (V_Cu) exhibit the lowest formation enthalpy compared to oxygen interstitials (i.e. ΔH(V_Cu) = 0.7 eV compared to ΔH(O_i) = 1.3 eV) and display small defect ionization energy (i.e. V_Cu ionization is 0.38 eV more shallow than O_i). The resulting holes are released into the valence band.⁴⁷ Similarly, in the CuAlO₂ delafossite, Watson et al. calculated V_Cu and a copper defect on the trivalent metal site (mechanisms (4) and (9) above) to have lower formation energies by 2 and 3.5 eV, respectively, than oxygen interstitials and aluminum vacancies (mechanisms (5) and (7) above).⁴⁵ Even though these calculations were not specifically performed on CuFeO₂, its defect chemistry is expected to be highly analogous. Thus, the p-type conductivity in intrinsic CuFeO₂ is likely to be due to ionized V_Cu defects. The crossover to n-type behavior we observe in copper deficient CuFeO₂ indicates that an n-type defect begins to dominate, the most likely being oxygen vacancy formation (mechanism (6)). Our study indicates that the introduction of holes through the introduction of copper vacancies in nonstoichiometric CuFeO₂ is limited due to the formation of these competing oxygen defects.

Using the same type of analysis the possible additional point defects for Mg-doped CuFeO₂ are:

where Mg²⁺ is in an octahedron in the plane of the Cu¹⁺ atomswhere Mg²⁺ replaces a Cu¹⁺ in the Cu(I)–O sticks

For low doping levels the material is p-type, but then crosses over to n-type at higher doping levels. The dominant defect chemistry in Mg-doped CuFeO₂ cannot be uniquely determined from the current data. Mechanisms (10) and (11) above can be ruled out because there is no room in the Cu layer and because Mg ions do not assume stick configurations with oxygen, respectively. Mechanisms (12) and (13) above lead to p-type doping on Mg substitution. Mechanism (12) is the simple expectation, and mechanism (13) is a more complex associated defect in which a single Mg ion goes into a prismatic interstitial site in the Cu layer, but has to displace three neighboring Cu ions to make room for itself. This mechanism also leads to hole doping.

Increased p-type conductivity on introducing low quantities of Mg dopant (i.e. x = 0.0005) could be due to increased hole carriers generated by mechanisms (12) or (13). Mechanism (13) is motivated by the presence of Cu₂O found as a second phase in this work, indicating thermodynamic instability of the desired CuFe_1−xMg_xO₂ at high Mg doping levels (x ≥ 0.05), which were not tested for photoactivity, and earlier work on the same system.²⁴ To experimentally probe the plausibility of mechanism (13), samples with stoichiometry Cu_1−3xMg_xFeO₂ were prepared. These samples also exhibited a maximum Mg content of x = 0.02, as in CuFe_1−xMg_xO₂ (shown in Fig. S7†). At doping concentrations above x = 0.02, Fe₃O₄ was observed as a second phase rather than Cu₂O (Fig. S8†). Thus, at least from an empirical standpoint, Mg doping seems to lead to defects somewhere in between these two options; perhaps both defect mechanisms (12) and (13) are present. Further theoretical and experimental work is needed to clarify the defect situation. The n-type crossover at x = 0.02 in Mg-substituted CuFeO₂ is attributed to the formation of oxygen vacancies, as is the case in the intrinsic or nonstoichiometric CuFeO₂ host compound (mechanism (6) for undoped CuFeO₂).

4 Conclusions

Photoelectrodes may provide an avenue to convert photon energy into a desirable chemical fuel. Although semiconductor oxide materials have been found to photocatalytically split water or reduce CO₂, many such materials suffer from low product yields. A possible optimization method for oxide photoelectrodes is through chemical doping. In this work, we examine whether the photoresponse of CuFeO₂ could be improved by changes in carrier concentration (N), experimentally modulated using two different hole formation schemes. We observed that the material with the lowest p-type carrier concentration (about 10¹⁶ cm⁻³) and highest mobility (about 100 cm² V⁻¹ s⁻¹) gave the best photoresponse. Further, both doping studies showed that the p-type carrier concentration in CuFeO₂ could not be tuned to high values due to the dominance of a n-type defect, which we believe is an oxygen vacancy, at high dopant levels. This reflects the complex defect chemistry of delafossites. Our studies provide a basis for further optimization of CuFeO₂ photoelectrodes. Similar studies of p-type oxides with other defect chemistries may be of interest for future photoelectrode optimization.

Acknowledgements

The authors acknowledge support of this research from the Office of Basic Energy Sciences, Department of Energy. The solid-state synthesis and physical characterization of materials were carried out under the direction of R. J. C. under Grant DE-FG02-98ER45706. Photoelectrochemical experiments and analysis were carried out under the direction of A. B. B. under Grant DE-SC0002133. A. W. acknowledges the Princeton University Andlinger Center and the Princeton Environmental Institute for research funds.

References

1. T. G. Deutsch, C. A. Koval and J. A. Turner, J. Phys. Chem. B, 2006, 110, 25297–25307.
2. M. Szklarczyk and J. O. M. Bockris, J. Phys. Chem., 1984, 88, 5241–5245.
3. T. O. M. Singh, in Engineered Ceramics, Wiley-American Ceramic Society, 2015, pp. 457–495.
4. X. Chang, T. Wang and J. Gong, Energy Environ. Sci., 2016, 9, 2177–2196.
5. K. Sivula and R. van de Krol, Nature Reviews Materials, 2016, 1, 15010.
6. K. Rajeshwar, J. Phys. Chem. Lett., 2011, 2, 1301–1309.
7. N. Serpone and A. V. Emeline, J. Phys. Chem. Lett., 2012, 3, 673–677.
8. M. P. Woodhouse and B. A. Parkinson, Chem. Soc. Rev., 2009, 38, 197–210.
9. C. S. Enache, D. Lloyd, M. R. Damen, J. Schoonman and R. van de Krol, J. Phys. Chem. C, 2009, 113, 19351–19360.
10. A. Fujishima and K. Honda, Nature, 1972, 238, 37–38.
11. J. C. Hill and K.-S. Choi, J. Phys. Chem. C, 2012, 116, 7612–7620.
12. J. R. Bolton, S. J. Strickler and J. S. Connolly, Nature, 1985, 316, 495–500.
13. R. Asahi, T. Morikawa, T. Ohwaki, K. Aoki and Y. Taga, Science, 2001, 293, 269–271.
14. A. Kwang-Soon, Y. Yanfa and A.-J. Mowafak, J. Vac. Sci. Technol., B: Microelectron. Nanometer Struct.--Process., Meas., Phenom., 2007, 25, L23–L26.
15. D. Paluselli, B. Marsen, E. L. Miller and R. E. Rocheleau, Electrochem. Solid-State Lett., 2005, 8, G301–G303.
16. M. P. D.-E. G. Bin-Daar, J. B. Goodenough and A. Hamnett, J. Chem. Soc., Faraday Trans. 1, 1983, 79, 1199–1213.
17. W. Luo, Z. Yang, Z. Li, J. Zhang, J. Liu, Z. Zhao, Z. Wang, S. Yan, T. Yu and Z. Zou, Energy Environ. Sci., 2011, 4, 4046–4051.
18. R. Shinar and J. H. Kennedy, Sol. Energy Mater., 1982, 6, 323–335.
19. C. Jorand Sartoretti, B. D. Alexander, R. Solarska, I. A. Rutkowska, J. Augustynski and R. Cerny, J. Phys. Chem. B, 2005, 109, 13685–13692.
20. M. S. Prevot, N. Guijarro and K. Sivula, ChemSusChem, 2015, 8, 1359–1367.
21. M. Li, L. Zhao and L. Guo, Int. J. Hydrogen Energy, 2010, 35, 7127–7133.
22. W. W. Gartner, Phys. Rev., 1959, 116, 84–87.
23. M. A. Butler, J. Appl. Phys., 1977, 48, 1914–1920.
24. F. A. Benko and F. P. Koffyberg, J. Phys. Chem. Solids, 1987, 48, 431–434.
25. Y. Jin and G. Chumanov, RSC Adv., 2016, 6, 26392–26397.
26. C. G. Read, Y. Park and K.-S. Choi, J. Phys. Chem. Lett., 2012, 3, 1872–1876.
27. J. Gu, A. Wuttig, J. W. Krizan, Y. Hu, Z. M. Detweiler, R. J. Cava and A. B. Bocarsly, J. Phys. Chem. C, 2013, 117, 12415–12422.
28. S. Omeiri, Y. Gabès, A. Bouguelia and M. Trari, J. Electroanal. Chem., 2008, 614, 31–40.
29. A. Omeiri, B. Bellal, A. Bouguelia, Y. Bessekhouad and M. Trari, J. Solid State Electrochem., 2009, 13, 1395–1401.
30. R. D. Shannon, C. T. Prewitt and D. B. Rogers, Inorg. Chem., 1971, 10, 719–723.
31. R. D. Shannon, D. B. Rogers, C. T. Prewitt and J. L. Gillson, Inorg. Chem., 1971, 10, 723–727.
32. F. A. Benko and F. P. Koffyberg, J. Phys. Chem. Solids, 1987, 48, 431–434.
33. R. Nagarajan, A. D. Draeseke, A. W. Sleight and J. Tate, J. Appl. Phys., 2001, 89, 8022–8025.
34. Z. Deng, X. Fang, S. Wu, Y. Zhao, W. Dong, J. Shao and S. Wang, J. Alloys Compd., 2013, 577, 658–662.
35. Y. J. Jang, Y. B. Park, H. E. Kim, Y. H. Choi, S. H. Choi and J. S. Lee, Chem. Mater., 2016, 28, 6054–6061.
36. U. Kang, S. K. Choi, D. J. Ham, S. M. Ji, W. Choi, D. S. Han, A. Abdel-Wahab and H. Park, Energy Environ. Sci., 2015, 8, 2638–2643.
37. E. Mugnier, A. Barnabé and P. Tailhades, Solid State Ionics, 2006, 177, 607–612.
38. R. D. Shannon, D. B. Rogers and C. T. Prewitt, Inorg. Chem., 1971, 10, 713–718.
39. V. R. Galakhov, A. I. Poteryaev, E. Z. Kurmaev, V. I. Anisimov, S. Bartkowski, M. Neumann and T.-R. Zhao, Phys. Rev. B: Condens. Matter Mater. Phys., 1997, 56, 4584–4591.
40. S. Kimura, K. Watanabe, T. Fujita, M. Hagiwara, H. Yamaguchi, T. Kashiwagi and N. Terada, J. Phys.: Conf. Ser., 2012, 400, 032039.
41. M. Salavati-Niasari, F. Davar, M. Mazaheri and M. Shaterian, J. Magn. Magn. Mater., 2008, 320, 575–578.
42. A. d. A. Jesus, Nature, 2011, 479, 317.
43. B. J. Ingram, G. B. Gonzalez, T. O. Mason, D. Y. Shahriari, A. Barnabe, D. Ko and K. R. Poeppelmeier, Chem. Mater., 2004, 16, 5616–5622.
44. S. Hamwi, T. Riedl and W. Kowalsky, Appl. Phys. Lett., 2011, 99, 053301.
45. D. O. Scanlon and G. W. Watson, J. Mater. Chem., 2011, 21, 3655–3663.
46. D. O. Scanlon and G. W. Watson, J. Phys. Chem. Lett., 2010, 1, 3195–3199.
47. H. Raebiger, S. Lany and A. Zunger, Phys. Rev. B: Condens. Matter Mater. Phys., 2007, 76, 045209.
48. H. Kawazoe, M. Yasukawa, H. Hyodo, M. Kurita, H. Yanagi and H. Hosono, Nature, 1997, 389, 939–942.
49. H. Yanagi, K. Ueda, S. Ibuki, T. Hase, H. Hosono and H. Kawazoe, Mater. Res. Soc. Symp. Proc., 2000, 623.
50. J. A. Thornton, Annu. Rev. Mater. Sci., 1977, 239–260.

---

Footnote

† Electronic supplementary information (ESI) available: SEM, XPS, EDX, and UV-Vis-NIR measurements of as-synthesized CuFe_1−xMg_xO₂ pellets. Mott–Schottky plots of CuFe_1−xMg_xO₂. SEM and EDX analyses of CuFe_1−xMg_xO₂ after use as a photoelectrode. XRD of CuFe_1−xMg_xO₂ and Cu_1−3xMg_xFeO₂. See DOI: 10.1039/c6ta06504j

---