High-Throughput Experimental Study of Wurtzite Mn1-x Zn x O Alloys for Water Splitting Applications.

We used high-throughput experimental screening methods to unveil the physical and chemical properties of Mn1-x Zn x O wurtzite alloys and identify their appropriate composition for effective water splitting application. The Mn1-x Zn x O thin films were synthesized using combinatorial pulsed laser deposition, permitting for characterization of a wide range of compositions with x varying from 0 to 1. The solubility limit of ZnO in MnO was determined using the disappearing phase method from X-ray diffraction and X-ray fluorescence data and found to increase with decreasing substrate temperature due to kinetic limitations of the thin-film growth at relatively low temperature. Optical measurements indicate the strong reduction of the optical band gap down to 2.1 eV at x = 0.5 associated with the rock salt-to-wurtzite structural transition in Mn1-x Zn x O alloys. Transmission electron microscopy results show evidence of a homogeneous wurtzite alloy system for a broad range of Mn1-x Zn x O compositions above x = 0.4. The wurtzite Mn1-x ZnxO samples with the 0.4 < x < 0.6 range were studied as anodes for photoelectrochemical water splitting, with a maximum current density of 340 μA cm-2 for 673 nm-thick films. These Mn1-x Zn x O films were stable in pH = 10, showing no evidence of photocorrosion or degradation after 24 h under water oxidation conditions. Doping Mn1-x Zn x O materials with Ga dramatically increases the electrical conductivity of Mn1-x Zn x O up to ∼1.9 S/cm for x = 0.48, but these doped samples are not active in water splitting. Mott-Schottky and UPS/XPS measurements show that the presence of dopant atoms reduces the space charge region and increases the number of mid-gap surface states. Overall, this study demonstrates that Mn1-x Zn x O alloys hold promise for photoelectrochemical water splitting, which could be enhanced with further tailoring of their electronic properties.

Introduction

Photoelectrochemical (PEC) water splitting is one of the options for sustainable hydrogen production. Photoactive semiconductors are critical for the development of PEC water splitting technologies. In particular, the choice of an efficient n-type semiconductor photoanode material is a long-standing objective for PEC solar cell design. The key requirements for a single semiconductor to perform efficient water splitting can be summarized as follows: a band gap in the range of 1.5–2.0 eV (taking into account the 1.23 eV water dissociation energy and overpotential losses), band edge positions that straddle the water oxidation and reduction potentials, chemical and electrochemical stability in the operating environment under illumination, fast transport of the photogenerated carriers, high conversion efficiency, and low cost.[1−5] Such moderate-band gap semiconductors can be also coupled with a suitable low-band gap semiconductor for applications in PEC tandem solar cells that could potentially lead to higher solar-to-hydrogen conversion efficiencies.

Transition-metal oxides are promising candidates for water splitting because of their low cost and excellent stability in aqueous environments. However, many transition-metal oxides reported in the literature suffer from one of the following problems: their band gaps are too large for efficient light absorption, their band edges do not align with Nernst potentials for water oxidation/reduction, the materials are highly susceptible to chemical dissolution (especially under localized acidic conditions), and/or they exhibit poor semiconducting properties. For example, TiO2 is the most widely studied photoelectrode material, but its wide band gap (∼3 eV) severely limits its efficiency for PEC fuel generating applications unless interfaced with low-band gap semiconductors at either an anode or cathode.[6−9] Exciting progress has been made over the past years on alternative binary oxide materials such as ZnO,[10] WO3,[11] Fe2O3,[12] Cu2O,[13] NiO,[14] Ta2O5,[15] Nb2O5,[16] etc. However, most of these materials also suffer from intrinsic limitations for PEC applications, especially related to corrosion or to their band edges, which are not well aligned for the production of hydrogen. BiVO4 has also shown promise as a photoanode material,[17] but this material is still prone to photocorrosion and its performance is often limited by poor electron transport and by surface recombination.[18−20] Therefore, new semiconductors need to be discovered and developed with specific functional attributes including band gap, band edge alignment, and improved transport properties.

Our computational research on d5 transition-metal oxides has shown the possibility of creating semiconducting manganese(II) oxide (MnO).[21] This MnO compound can exist in the rock salt (RS) or wurtzite (WZ) structures. The WZ structure is metastable compared to the RS structure, which is the ground state. However, the RS structure presents a large band gap (3.5 eV) and self-trapping of holes, whereas the WZ structure is predicted to have an appropriate band gap (2.1 eV) for water splitting and band-like hole transport instead of hopping. Hence, the question arises on how to make MnO structurally stable in the wurtzite phase and therefore take full advantage of its properties for water splitting applications.

It is well known that the properties of a semiconductor can be tuned through alloying, that is, substitution of one element for another at tens of percent level. There are several reports in the literature on MnO doping in ZnO that focused on substitution of only a few percent of Mn for Zn. These studies focused mainly on the structural, electronic, and magnetic properties of thin films fabricated by sputtering,[22] pulsed laser deposition,[23−25] and molecular beam epitaxy.[26] However, there are only few reports with a particular emphasis on the alloying of MnO with ZnO and/or their application in water splitting. One experimental study[27] looked at the interaction of MnO and ZnO at different temperatures and pressures and determined the experimental conditions at which Mn1–ZnO (0.3 < x < 0.7) was formed in the rock salt structure. Another paper has reported on rock salt MnO:ZnO microflowers (5% of Mn ions in the ZnO matrix), finding that the emission of F+ centers at the oxygen vacancies could enhance the performance of the material in water splitting.[28] Other theoretical studies have demonstrated that MnO:ZnO alloys (Mn1–ZnO) could potentially be used for water splitting,[29−32] reporting on tunability of the band gap,[29] efficiency of different n- and p-type dopants,[30] overpotential for water photo-oxidation on (001) surfaces,[31] and their interfaces with water[32] but all assuming the rock salt structure of Mn1–ZnO.

Our recent calculations showed that ZnO is a favorable candidate for stabilizing MnO in the wurtzite structure and this structure has promising properties for water splitting applications.[33] We predicted the band line up, dopability, and absorption in the visible region for wurtzite Mn1–ZnO. We also predicted the structure transition of Mn1–ZnO from RS to WZ (above x = 0.3) with favorable electron and hole transport properties and suitable physical properties for water splitting. This recently published theoretical work also includes some preliminary experimental results on wurtzite Mn1–ZnO alloys for water splitting.[33] In a more recent publication, we have also shown how this basic Mn1–ZnO work led to a new understanding of phase diagrams for heterostructural semiconductor alloys.[34] This new understanding offers a route for predicting the properties of other heterostructural semiconductor alloys and opens up opportunities in the development of new materials with desired functionalities.[35]

In this paper, we present detailed results on the fabrication, characterization, and application of wurtzite Mn1–ZnO alloys for water splitting. We use high-throughput combinatorial synthesis, measurement, and analysis methodologies to rapidly screen the composition–structure–property relationships of Mn1–ZnO alloys and identify candidate samples for a more detailed study in PEC applications. TEM measurements are reported to demonstrate the homogeneity and structure of the fabricated alloys. Mott–Schottky measurements are used to determine the flat band potential, space charge layer, and carrier concentration of the samples. In addition, measurements of this material for water splitting applications are performed as a function of Zn atomic composition, pH, thickness, and Ga doping. These PEC measurements show that undoped samples exhibit a photoresponse current, while heavy doping inhibits the photoresponse due to decreased space charge layer and surface defect states suggested by photoemission spectroscopy.

Results

Structural Properties

Depending on the substrate deposition temperature (T), oxygen partial pressure (pO2), and chemical environment, the manganese oxide system can exist in a variety of oxidation states including MnO [Mn(II)], Mn2O3 [Mn(III)], Mn3O7 [Mn(II,III)], MnO2 [Mn(IV)], MnO3 [Mn(VI)], and Mn2O7 [Mn(VII)], with a number of possible polymorphs for each composition. However, the oxidation state for Mn we are interested in here is +2 as in MnO, since it is relevant to Mn1–ZnO alloys. Therefore, prior to depositing combinatorial libraries of Mn1–ZnO alloys, we need to identify the processing conditions that allow synthesizing MnO single-phase samples in the correct oxidation state. This is done by varying both the total pressure (PTotal) and the oxygen pressure (pO2) during depositions as a function of the substrate temperature gradient obtained on a combinatorial heater. An analysis of the X-ray diffraction peaks allows the determination of various crystalline structures of known oxidation states as a function of pO2, PTotal, and temperature.

A synthesis map determined from XRD patterns is presented in Figure , along with the representative XRD patterns of Mn–O in different oxidation states. A rock salt single-phase (face-centered cubic, JCPDS #07–0230) of MnO is obtained using a fixed PTotal of 5 mTorr and pO2 of 10–5 Torr over a substrate temperature (TS) range of 250–650 °C. The results show that a pO2 < 10–5 Torr and PTotal < 5 mTorr are needed to grow rock salt MnO. A mixture of MnO and Mn3O4, pure Mn3O4, and a mixture of Mn2O3 and Mn3O4 appear successively as a function of PTotal and/or pO2 increase. Single-phase ZnO (JCPDS #653411) was also obtained at the same condition as the single-phase MnO.

Figure 1

(a, b) Oxidation state maps obtained by XRD of the Mn–O system at (a) different total pressures as a function of substrate temperature and (b) different oxygen pressures as a function of substrate temperature. (c) Representative XRD patterns of the different phases, with the same color code as in (a) and (b). Note that the phase of interest, MnO, is in dark green.

Experimental XRD maps for typical Mn1–ZnO alloy samples deposited at 381 and 519 °C and at different MnO concentrations are shown in Figure . Different crystalline phases are observed as a function of ZnO concentration at 381 °C, including rock salt (RS) at low x, wurtzite (WZ) at high x, and a mixture of RS and WZ at intermediate x. At x < 0.2, a shift of the (111) and (002) RS peak positions is observed with increasing ZnO composition (x), indicating that Zn ions substitute Mn sites without changing the RS crystal structure. The (002) WZ peak in the Mn1–ZnO alloy starts appearing at around x = 0.2 and coexists with the RS peaks up to x = 0.2. Substituting Mn with Zn at x > 0.4 caused a position shift of the (002) WZ peak to shift to higher angles. Correspondingly, the c-axis lattice constant of the wurtzite Mn1–ZnO alloy thin films is elongated with increasing Zn composition.

Figure 2

(a, b) XRD scans of Mn1–ZnO alloys with different x values deposited at (a) 381 °C and (b) 519 °C, leading to different rock salt (RS), wurtzite, (WZ), and mixture (WZ + RS) phases. The RS peaks at 35–36° (111) and 40–41°o (200) and the WZ peak at 34–35° (002) are clearly visible.

To determine the position of a phase boundary at a particular temperature on the Mn1–ZnO synthesis diagrams, we use the disappearing phase method. For example, the inset of Figure shows an example of the use of the disappearing phase method to determine the RS/(WZ + RS) phase boundary at 260 °C. In this specific case, the phase boundary location in the Mn1–ZnO alloy is found at x = 0.15, where the WZ XRD peak intensity extrapolates to zero. The composition–temperature phase boundaries that correspond to the solubility limit of Mn in the WZ structure and Zn in the RS structure in Figure are obtained by repeating the disappearing phase analysis at each studied temperature. From the synthesis phase diagram determined over the 180–520 °C temperature range, the solubilities of Mn in the WZ structure and Zn in the RS structure are found to decrease with increasing temperature. Hence, the width of the two-phase WZ + RS region (miscibility gap) increases with increasing temperature. From the thermodynamic point of view, this result may seem unusual, because the miscibility gap is expected to close with increasing temperature due to entropic contributions. This unexpected observation can be explained by kinetic limitations during the thin-film growth, where long-range phase separation is inhibited at low substrate temperatures (200–500 °C), as discussed in more detail in our previous works on Mn1–ZnO[33,34] and related alloys.[35−37]

Figure 3

Mn1–ZnO system phase diagram between 180 and 520 °C. The inset illustrates the disappearing phase method used to determine the WZ/RS phase boundary at 260 °C.

The TEM analysis of the film microstructure is shown in Figure . Bright-field TEM imaging (Figure a) of the Mn0.48Zn0.52O film illustrates a uniform film thickness and columnar microstructure over the 20 μm-wide field of view analyzed in the FIB prepared cross section. As seen in the (01-10) dark-field image (Figure b) and the corresponding selected area diffraction pattern (Figure c), the grain structure of the Mn0.48Zn0.52O film is highly oriented toward (0001) with 10 to 30 nm grains rotated about this axis. The diffraction pattern also shows a single phase that indexes well to the wurtzite-type structure. High-resolution imaging of the film (Figure d) again illustrates the highly oriented film structure with grain rotations about the 0001 direction. Z-contrast scanning TEM (STEM) imaging was completed using a high-angle dark field (Figure a) of the film cross section, and the image illustrates no local changes in elemental distribution. The lack of changes in elemental chemistry is correlated by the EDS spectral image shown in Figure b. The quantitative analysis of the spectral image correlates well to the expected Mn0.48Zn0.52O stoichiometry and is consistent across the film thickness.

Figure 4

(a) Bright-field TEM analysis of the Mn0.48Zn0.52O film microstructure illustrates a uniform film thickness and microstructure over the 20 μm-wide field of view analyzed in the FIB prepared cross section. As seen in the (01-10) dark-field image in (b) and the corresponding selected area diffraction pattern in (c), the grain structure of the Mn0.48Zn0.52O film is highly oriented toward (0001) with 10 to 30 nm grains rotated about this axis. The diffraction pattern in (c) also shows a single phase that indexes well to wurtzite ZnO. (d) High-resolution imaging of the film again illustrates the highly oriented Mn0.48Zn0.52O film structure with grain rotations about the 0001 direction.

Figure 5

(a) Z-contrast scanning TEM (STEM) image of the Mn0.48Zn0.52O film cross section illustrates no local changes in elemental distribution. The lack of changes in elemental chemistry is correlated by the EDS spectral image shown in (b). The quantitative analysis of the spectral image correlates well to the expected Mn0.48Zn0.52O stoichiometry and is consistent across the film thickness.

Optoelectronic Properties

It is important to evaluate the band gap of Mn1–ZnO from the measured absorption edge, because it helps define the efficiency of solar light absorption and therefore the upper limit of the photoconversion efficiency. As previously mentioned, a minimum of 1.23 eV per electron–hole pair generated is needed to drive the water splitting reaction. However, to account for the overpotential losses produced during the electron-transfer processes at the semiconductor/liquid junctions, an average of 2.0 eV is usually needed to drive the reaction.[1] Therefore, it is important for the band gap energy of Mn1–ZnO to be around that value to efficiently split water.

The optical absorption spectra of the Mn1–ZnO for different crystalline structures grown at 381 °C are shown in Figure . Substitution of Zn in MnO shifts the absorption edge toward the higher wavelength region, which indicates a decrease in the band gap energy. As-grown pure ZnO and MnO exhibit an absorption edge at around 3.3 eV. As x in the Mn1–ZnO alloy increases, the optical band gap absorption edge of the alloys shows a redshift down to 2.1 eV for x = 0.5. The color intensity map of optical absorption spectra versus Zn content in Mn1–ZnO is shown in Figure for several different synthesis temperatures. These absorption maps are also compared in Figure with the corresponding XRD color intensity maps. We find that the abrupt change in the optical absorption onset occur at the RS/(RS + WZ) phase boundary, consistent with the appearance of the lower-band gap WZ phase. Furthermore, the Zn composition at which this abrupt change occurs decreases as a function of increasing synthesis temperature, which is consistent with the phase boundary results (Figure ).

Figure 6

(a) MnO, ZnO, and Mn1–ZnO thin films deposited on the glass. (b) Optical absorption spectra for different structures of grown Mn1–ZnO, ZnO, and MnO thin films.

Figure 7

(a–c) 2D XRD diagram and (d–f) 2D band gap absorption onsets of Mn1–ZnO at different temperatures as a function of composition. The blue dashed lines are theory predictions for the absorption onsets.

It has been found that the introduction of Mn2+ into the ZnO lattice leads to the presence of mid-gap states.[25] The mid-gap states may affect the measured absorption spectra and also influence the determination of the band gap.[38] However, in our case where Zn2+ is introduced into the MnO lattice, there is no evidence of the contribution of mid-gap states as the Zn content is increased in Mn1–ZnO. This is consistent with what has been found in a previous work on undoped Mn1–ZnO.[29] However, the optical results are not sufficient to evaluate the presence of mid-gap states, and further studies using photoelectron spectroscopy are needed.

Most oxides photoanodes for photoelectrocatalytic water splitting in their intrinsic state suffer from poor electronic transport and therefore require doping with electric charge carriers. In this case, solar absorber materials should be n-type (electron carrier)-doped, and the minority (hole) carriers should have a long lifetime and high mobility. Therefore, we have explored the use of Ga as an extrinsic donor dopant for our Mn1–ZnO system. Although it was reported that Ga will not contribute to the carrier or enhance the conductivity of rock salt Mn1–ZnO alloys,[30] we have explored the use of extrinsic n-type doping by Ga because it is known as an efficient dopant in ZnO and wurtzite Zn1–MgO alloys.[39] In addition, our previous band structure calculations predicted a shallow Ga donor state with sufficiently low ionization energy to generate free electrons at room temperature.[33]

As discussed in our previous publication,[33] doping was achieved for the PLD growth by using a ZnO target containing 4 cation % Ga. Thus, in the Mn1–ZnGaO alloys, the Ga doping level varies with the Zn composition x (i.e., yGa = 0.04x). Thus, the conductivity (in S/cm) was normalized by the Ga content per formula unit (i.e., per yGa). Note that the Ga content is 0.02 at the composition x = 0.5. As x and y in Mn1–ZnGaO increase, the conductivity increases from unmeasurable in the RS single-phase region (x < 0.19), to 10 S/cm/Ga within the miscibility gap (0.19 < x < 0.45), to 100 S/cm/Ga observed in the WZ single-phase region around x = 0.5, up to 104 S/cm/Ga approaching the pure ZnO composition typical of Ga:ZnO.[40] Subsequent Hall effect measurements have shown that Mn1–ZnGaO at x = 0.48 (yGa = 0.02) has an electron concentration of n = 3 × 1019 cm–3 and mobility of μ = 0.4 cm2/(V·s), whereas at x = 0.75 (yGa = 0.03), n = 1 × 1020 cm–3 and μ = 3.5 cm2/(V·s), consistent with conductivity values determined by a four-point probe. Interestingly, the work functions of the films measured by Kelvin probe microscopy were not greatly affected by the Zn content in the range in Mn1–ZnO and by the doping concentration of Ga, perhaps due to the contribution of surface dipoles. The band gap of Mn1–ZnO is also not expected to be affected by dilute Ga substitution in a few percent range, where rigid band approximation still holds and band filling effects are negligible.

Discussion

After verifying the dopability of Mn1–ZnO, films without and with different Ga atom percentages (atom %) were fabricated for the photoelectrochemical experiments. In addition to the conductivity, the flat band potential of a photoanode is an important parameter to determine, as it reflects its ability to split water. Both conductivity and flat band potential help elucidate the band structure of a semiconductor when interfaced with a solution. Photoanodes drive the oxygen evolution reaction (OER) at the electrode–electrolyte interface. When the photoanode is illuminated with UV light, electrons and holes are generated, and the photoanode surface shifts the Fermi level of the bulk (measured potential) toward more cathodic potentials. The photogenerated minority carrier holes remain on the photoanode and migrate toward the surface to participate in the OER at the electrolyte interface by oxidizing water or hydroxide to liberate O2. The majority carrier electrons are directed from the photoanode to the Pt catalytic cathode, where they reduce water or protons to form H2 gas.

Mott–Schottky Analysis

In order to understand the intrinsic electronic properties of Mn1–ZnO in electrolyte solutions, we performed electrochemical impedance measurements in the dark. The carrier density and flat band potential (Vfb) at the photoelectrode/electrolyte interface can be quantified from the measured CV curves by the Mott–Schottky equation[41,42]wherein e0 is the electron charge, ε is the dielectric constant of Mn1–ZnO, ε0 is the permittivity of free space, Nd is the donor density, V is the electrode applied bias, Vfb is the flat band potential, and kT/e0 is the temperature-dependent correction term. In this equation, the area factor is included in the measured capacitance, as indicated by the F/cm2 units in the corresponding plots (Figure ). The high-frequency dielectric constant ε of Mn1–ZnO was theoretically (by independent particle approximation)[43] and experimentally (by spectroscopic ellipsometry) determined to be 4 and 4.7, respectively. We used here the experimental dielectric constant. The Mott–Schottky plots collected from different compositions of Mn1–ZnO alloys in PEC cells are presented in Figure . An extrapolation of the linear region (X intercept) in the Mott–Schottky plot (1/C2 versus applied potential V) was used to determine the value of the flat band potential Vfb at 0.5 and 1 kHz. The positive slope indicates that Mn1–ZnO is an n-type material, consistent with Hall effect measurements.

Figure 8

Mott–Schottky plots under dark conditions for Mn1–ZnO electrodes with different compositions, measured at 1 kHz in 0.1 M aqueous K2SO4 solution (pH 10). The dashed lines represent the extrapolated lines from the linear portion of the Mott–Schottky plots. The inset is the magnification of the Mott–Schottky plots under dark conditions for Ga0.02:Mn0.5Zn0.48O at 0.5 and 1 kHz. All films are 132 nm-thick unless mentioned ( data for the 272 nm-thick film is in green).

The results for both the flat band potential and the carrier concentration determined for different Mn1–ZnO samples with x in the 0.4–0.6 range are presented in Figure . The carrier concentration of the undoped samples determined from the Mott–Schottky plot increased slightly with the Zn atomic concentration (Figure a). However, regardless of the Zn composition, the concentration is below 1019 cm–3. The carrier density of the samples increases with Ga doping with values varying between 1019 and 1021 cm–3 depending on the Ga atomic concentration (Figure b). It is worth noting that these carrier density values of are in agreement with the results obtained from Hall effect measurements where the measured carrier densities for Ga-doped samples varied from 1019 and 1020 cm–3. The variation of carrier density with Mn1–ZnO thickness is very small, indicating that the average concentration of donors does not vary much as the film becomes thicker (Figure c). However, if the total charge density does not vary, increasing the thickness should lead to an increase in the total amount of charges and therefore an increase in the flat band potential shift toward more cathodic potentials.[44−46]

Figure 9

(a–c) Flat band potential (vs Ag/AgCl) and carrier concentration results as a function of (a) Zn atomic composition x in Mn1–ZnO, (b) Ga atom % doping in Mn1–ZnO, and (c) Mn1–ZnO film thickness.

Photoelectrochemical Performance

The dependence of the PEC performance of Mn1–ZnO electrodes with 0.4 < x < 0.6 on the thickness is also studied, and the results are shown in Figure . The PEC performance of samples with thicknesses of 136 and 673 nm under chopped light with both front illumination (the Mn1–ZnO film facing the light) and rear illumination is shown in Figure a. All films exhibit an anodic photocurrent, indicating n-type behavior, and the photocurrent increases with the applied positive potential. The front- and back-illuminated current–voltage curves for the 136 nm sample strongly overlap each other, whereas for the 673 nm sample, they differ as discussed in our previous publication.[33] The inset of Figure a shows the generated photocurrent difference (in percentage) between the front and back illumination (measured at ∼1.25 V vs Ag/AgCl) for different photoelectrode thicknesses. For a thickness equal or lower than ∼250 nm, a comparison between the front and back illumination shows an almost identical photoresponse. However, from a thickness equal to 200 nm or higher, this difference becomes obvious and more pronounced with the thickness. A larger photoresponse is observed when the material is illuminated from the front side. Under front-side illumination, most charge carriers are generated near the semiconductor/electrolyte interface where most of the light is absorbed. This suggests that the performance of Mn1–ZnO photoanodes for water splitting is not limited by the diffusion length of the photogenerated charge carriers for a film thickness equal or higher than 200 nm. For other oxide materials, the hole transport is the rate-limiting step for the PEC performance at much lower length scales.[17,47,48]

Figure 10

(a–c) Current density plots for Mn1–ZnO as a function of applied potential for different (a) film thicknesses with front and back illumination in pH 10 (K2SO4) solution, (b) Zn atomic compositions for the 136 nm-thick film in pH 10 (K2SO4) solution, and (c) pH values for the 136 nm-thick film reported against RHE. (d) Carrier concentration and space charge width plot as a function of Ga atomic concentration in Mn0.5Zn0.5O calculated using eq . The insets are current density plots of Mn1–ZnO as a function of the (a) thickness for front and back illumination, (b) Zn atomic composition, and (c) pH.

The effect of the Mn1–ZnO composition on the PEC performance was studied by varying x from 0.3 to 0.7. It was found (Figure b) that with an increase in x, the electrocatalytic activity first increases and then decreases. The highest Mn1–ZnO PEC activity was obtained at x = 0.4, with the PEC activity dropping and remaining relatively flat past x = 0.5. The highest Mn1–ZnO photoresponse, correlating with the lowest flat band potential (−3 eV) at x ≈ 0.4 may be due to the lowest doping at this composition and hence the highest spatial extent of the built-in electric field at the liquid/semiconductor junction. Another potential explanation is the induced internal electric field close to the WZ/(RS + WZ) phase boundary that helps the electron–hole separation and thus enhances the photocatalytic activity of the mixed Mn1–ZnO alloys.[49]

The effect of the pH on the Mn1–ZnO photocurrent was studied in different electrolyte environments of 0.1 M K2CO3 with pH 9.2, 0.5 M K2SO4 solution with pH 10, and a buffer salt (sodium phosphate tribasic + K2SO4) with pH 12. In order to compare the behavior of the Mn1–ZnO material in an environment with different pH values, the applied potential was reported against the reversible hydrogen electrode (RHE) according to the Nernst equationwhere EAgCl is the experimental potential measured against the Ag/AgCl reference electrode and EAgCl0 is 0.198 mV (KCl saturated vs NHE) at 25 °C. The results shown in Figure c suggest that as the solution become more alkaline, the photocatalytic activities of Mn1–ZnO substantially improves. Therefore, changing the pH has the opposite effect as varying the doping level of the semiconductor. A maximum photocurrent of ∼340 μA cm–2 was measured at pH = 12 for a 673 nm-thick film.

It is also worth noting that a direct contact of the Mn1–ZnO photoanode with the K2SO4 electrolyte (pH = 10) for over 24 h under continuous UV illumination (simulated 1 sun illumination) did not show evidence of photocorrosion or degradation and did not substantially affect the measured current density values. Specifically, the current densities were 0.117, 0.115, 0.119 mA/cm2 at 0, 7, and 25 h of measurement, which is within the ±0.005 mA/cm2 error of the technique. This observation is very encouraging for Mn1–ZnO as a viable candidate material for PEC applications. However, it is worth mentioning that the Mn–O Pourbaix diagram is quite complex with multiple species and a range of oxidation states. On the other hand, the Zn–O Pourbaix diagram is rather simple, with ZnO being stable at pH = 10 in a wide range of potentials. Thus, one potential explanation for the Mn1–ZnO stability is that a self-passivating ZnO layer is formed on the surface of Mn1–ZnO alloys.

Regardless of the doping concentration, Mn1–ZnO samples doped with Ga did not exhibit any photoresponse. This observation suggests that increased carrier concentration due to Ga doping may reduce the width of the space charge layer, leading to less efficient separation of the photogenerated charge carriers. The width of the space charge layer (W) at the semiconductor/solution interface can also be derived from the Mott–Schottky plot relationship and be expressed asAs one can see in Figure d, the width of the space charge layer is reduced from 20 to 10 nm as Ga is introduced and from 10 to 5 nm as the Ga doping concentration is increased, which is consistent with the observed increase in carrier concentration. So, even for the undoped samples, 1019 cm–3carrier density values obtained from the Mott–Schottky plots may be already too high for forming a wide depletion region in Mn1–ZnO upon contact with the electrolyte, limiting the PEC performance. Therefore, additional Ga doping of the Mn1–ZnO material can only further decrease the depletion width, which is detrimental to the performance of the material.

The lack of PEC activity for the Ga-doped Mn1–ZnO samples may also result from several other effects. First, a very high doping concentration may lead to the dopants forming recombination centers and result into a rapid electron–hole recombination occurring before the electron and/or hole species have time to carry out catalytic reduction/oxidation.[50] Second, the specific surface area may decrease with the increase in Ga concentration that is unfavorable to obtain a high photocatalytic activity. Third, the band alignment of Ga-doped samples may not straddle the water redox potentials to drive the OER. To evaluate these different possibilities, additional measurements such as XPS and UPS may help better understand the effect of Ga doping on the PEC performance of Mn1–ZnO.

Photoemission Spectroscopy

Figure shows ultraviolet photoelectron spectroscopy (UPS) data for wurtzite Mn1–ZnO (x ≈ 0.5) samples undoped and doped with 0.4 and 4% Ga. The data in Figure b show that there is a change in the O 2p orbital distribution in the He(I) spectra as a function of doping. Figure c gives a magnified view of the onset of occupied states, which were used to determine the ionization energy, as given in Table . Note that, here, we are investigating the O 2p σ orbitals are at ∼4 eV and the O 2p π orbitals are at ∼7 eV, rather than the electronic states of the electrode surface defined by the valence features, since a better way to understand these valence features is through a core-level analysis and connection with theories. The He(II) spectra (Figure d) show the increased sensitivity for the metal cation components of the valence band. This is evident in the comparison between the He(I) and He(II) spectra where Zn 3d orbital contributions emerge at ∼11 eV below the Fermi level.

Figure 11

(a) Work function and (b–d) UPS spectra of Mn1–ZnO samples (x ≈ 0.5) with different doping concentrations of Ga as a function of kinetic energy induced by (b) He(I), (c) He(I) zoomed in the valence band region, and (d) He(II).

Table 1

Work Function, Ionization Energy (IE), and Fermi Level Relative to the Valence Band Minimum (ΔE) for Undoped and Doped Mn1–ZnO (x ≈ 0.5)

composition	work function	IE	ΔE
Mn1–xZnxO (undoped)	4.5	5.3	0.8
Mn1–xZnxO doped with 0.4% Ga	4.5	5.9	1.4
Mn1–xZnxO doped with 4.0% Ga	4.3	5.8	1.5

The UPS data presented in Figure suggests that increasing the doping in wurtzite Mn1–ZnO creates increased surface defects (mid-gap states) near the valence band, which can be observed in the tail states. Given the self-compensation, one can imagine similar tail states near the conduction band. The listed ionization energies (Table ) are the onsets of the tail states, as shown in Figure c (zoomed in the region of the valence band from He(I)). The tail states also appear to correlate with changes in the distributions of the oxygen orbitals at 4–7 eV, which are most evident in the He(II) spectra (Figure d). This indicates that Ga causes a change to the valence surface states and local binding environment of the oxygen. In particular, the 4% Ga sample appears to have the broadest distribution, which is consistent with doping. The mid-gap states are the most easily oxidized states and expected to be the first states to participate in the OER.[51]

In addition to light absorption (related to the band gap), the water oxidation efficiency also depends on two major factors: the probability and rate of charge carrier separation and transfer and the recombination rate of photogenerated electron–hole pairs. The mid-gap states play a role in both cases. They can act as intermediates for indirect charge transfer. In this case, the photogenerated carriers trapped in the mid-gap states predominantly initiate the water splitting process. However, the slow kinetics of the hole transfer to the water oxidation reaction leads to a low rate of reaction, which is detrimental to water splitting process.[52−54] Instead of playing an intermediate role and without a chemical involvement with the water splitting reaction, the mid-gap states may act as trapping centers for photoexcited holes, preventing them to participate in the water splitting reaction by promoting the recombination of holes with electrons.[55,56] Therefore, the mid-gap states contribute to lowering the rate of water oxidation. Although there is no consensus in the literature on which of these two processes is predominant, it is clear that both of them impede the photocatalytic reaction.

In an attempt to further understand the correlation between Ga concentrations and poor electrochemical behavior, we constructed a tentative band diagram of the doped Mn1–ZnO materials based on photoemission spectroscopy measurements. If we assume that the optical gap describes the energetic position of the conduction band, then we can conclude that increasing the Ga dopant increases the n-type character of the sample, as shown in the band diagrams in Figure . Specifically, as the Ga concentration increases, the Fermi level approaches the conduction band. However, the validity of such an assumption is questionable for the Mn1–ZnO sample. In Figure a, the Fermi level at the surface is closer to the valence band (ΔE ≈ 0.8 eV), indicating a p-type material. This result to first order appears as a direct contradiction to the n-type behavior demonstrated by Hall effect and Mott–Schottky measurements. This disparity could be explained by the fact that a bulk band gap may be very different at the surface. In addition, in the n-type material with defects, the Fermi level can be pinned at the surface at lower energy than in the bulk, leading to upward band bending and apparent lowering of the Fermi level measured by photoemission spectroscopy. However, assuming the same (surface) band gap, with increasing Ga concentration (Figure b,c), there is an increasing n-type surface character, consistent with bulk measurements. In the future, it may be interesting to compare this band diagram picture derived from photoemission measurements (Figure ) to a similar picture that is constructed from photoelectrochemical measurements.

Figure 12

(a–c) Energy band diagrams of Mn1–ZnO (x ≈ 0.5): (a) undoped, (b) doped with 0.4% Ga, and (c) doped with 4% Ga.

Conclusions

We used a high-throughput screening methodology to study Mn1–ZnO materials in a wide composition range (with x varying from 0 to 1) and identify a suitable composition for water splitting applications. The films were grown by combinatorial pulsed laser deposition. The observed band gap value Egap of 2.1 eV at the x ≈ 0.5 alloy composition is consistent with theory predictions for the WZ structure of the Mn1–ZnO alloy. The as-deposited films are insulating, and doping with Ga improved the conductivity of the films. The electrical conductivity for the 4% Ga-doped Mn1–ZnGaO alloy with x = 0.48 is 1.9 S/cm, with an electron concentration of 3.1 × 1019 cm–3 and mobility of 0.8 cm2/(V·s), consistent with Mott–Schottky measurements.

Mott–Schottky measurements were also used to determine the flat band potential and space charge width of all films and help evaluate the carrier concentration of undoped samples. Water splitting experiments have shown that as-deposited undoped films exhibit a photoresponse under illumination, while doped films with Ga do not produce any photocurrent. This could be explained by the small space charge region and the presence of mid-gap surface states. The maximum photocurrent density of ∼340 μA cm–2 was measured at pH 12 for a 673 nm-thick undoped film with 0.4 < x < 0.6.

Overall, this study demonstrates that the Mn1–ZnO alloys are promising for photoelectrochemical water splitting but need further investigations and optimization to be considered as an efficient material. For instance, a careful control of the electron concentration is needed, with an appropriate trade-off between good transport properties, the absence of mid-gap states, and low recombination losses. Passivation of the mid-gap states or enhancing the OER rate by coupling Mn1–ZnO with an electrocatalyst can also help enhance the photoresponse. Regardless of these future challenges, Mn1–ZnO appears to be a stable and viable candidate for applications in PEC solar cells for hydrogen generation by water splitting.

Methods

Materials Synthesis

Mn1–ZnO alloy thin films were grown on glass (Corning Eagle XG) substrates by pulsed laser deposition (PLD). We employed a thin-film high-throughput combinatorial approach[57−64] to deposit “libraries” with a spatial gradient in both the composition and the substrate temperature. KrF excimer laser (Spectra-Physics) operating at 248 nm with a pulse duration of 25 ns was used. The laser was operated at 40 Hz, and the beam was focused through a 60 cm lens onto a commercial 1 inch MnO and ZnO rotating targets (purity 99.99%) at a 45° angle, with a laser fluence of 4 J/cm2 on the target surface. The target-substrate distance was fixed to 7 cm. Prior to the depositions, the growth chamber was evacuated to a base pressure of 10–9 Torr using a turbomolecular pump then backfilled with an O2/Ar mixture. In order to investigate the effects of different gas ratios and the total pressure, the O2/Ar gas ratio was varied from 10–3 to 0.5 while keeping the total pressure (Ar + O2) at 5 mTorr, and then the total pressure was varied from 5 to 100 mTorr. The deposition was performed at a substrate temperature ranging from 150 to 700 °C. A temperature gradient across the substrate was obtained by introducing a spacer between the holder and the substrate as reported elsewhere.[65] Films with a lateral composition gradient were obtained by depositing sequentially from MnO and ZnO targets with a specific geometry with respect to the substrate.

Structural, Chemical, and Surface Characterizations

Each of the combinatorial samples was automatically mapped at 44 points (4 × 11 grid) to find the properties of the material as a function of composition. X-ray fluorescence spectra were collected using a Matrix Metrologies MaXXi 5 with W anode, 800 μm collimator, and 300 s scan time. The resulting spectra were modeled using the MTF-FP software to obtain both the chemical composition and the thickness of the samples. The X-ray diffraction patterns of Cu Kα radiation were collected using a Bruker D8 Discover. The diffraction peaks were assigned based on the MnO and ZnO reference diffraction patterns obtained from the Inorganic Crystal Structure Database (ICSD). Kelvin probe mapping with a 1 mm-diameter gold-coated probe was used to determine the work function of Mn1–ZnO combinatorial samples. The samples were placed in ambient atmosphere and grounded during the work function measurements. The Kelvin probe mapping system was calibrated both before and after every mapping run using a highly oriented pyrolytic graphite standard with the reference 4.6 eV work function.

Optical and Electrical Characterizations

Optical spectra were mapped in the 300–2000 nm range using a fiber optics-based spectrometer with normally incident angle of light generated by deuterium and tungsten/halogen light sources and collected using Si and InGaAs detector arrays. The measured transmittance (T) and reflectance (R) spectra were converted into the absorption coefficient α using the relationship α = −ln[T/(1 – R)]/d, where d is the film thickness. The measurement of the electrical conductivity was carried out by using the four-point probe method. The current was applied by a Keithley 6221 DC and AC current source, and the voltage was measured with a Keithley 2182 voltmeter. Spectroscopic ellipsometry (J. A. Woollam M-2000S) was also used to both determine the dielectric functions and verify the thickness accuracy of the films.

Transmission Electron Microscopy

Specimens for transmission electron microscopy (TEM) investigations were prepared site-specifically in the combinatorial library using standard focused ion beam (FIB) thinning and in situ lift-out techniques. Specifically, an FEI Co. (Eindhoven, Netherlands) Nova 200 dual-column instrument operating at 5 keV electron energies was used to first identify a composition of Mn0.48Zn0.52O and then protect the sample surfaces using electron beam-induced Pt deposition to a thickness of approximately 50 nm. A Ga ion accelerating voltage of 30 kV was used to further deposit a thicker layer of Pt and perform completion of rough milling and in situ extraction to a Cu grid. Final thinning was completed at 5 kV Ga ion accelerating voltages. Thinned specimens were subsequently examined in an FEI Co. Talos 200 FX (S)TEM operating at 200 keV. EDS spectral images (1000 × 1000 pixels) were acquired using a nominally 1.5 nA STEM probe at count rates up to 125 kcps using four windowless detectors. Spectral image analysis was completed using the Esprit v.1.9 software (Bruker Inc., Berlin, Germany). Quantification was completed using Cliff–Lorimer correction factors and fixing the oxygen concentration to 50 atom %, assumed from the stoichiometries of ZnO and MnO.

Photoelectron Spectroscopy

X-ray and UV photoelectron spectroscopy were performed on single-composition films with a Kratos Axis Ultra X-ray photoelectron spectrometer with a monochromatic Al Kα source (1486.6 eV) for XPS measurements and He(I) and He(II) excitation sources (21.2 eV and 40.8 eV, respectively) for UPS measurements, all at a base pressure of 10–9 Torr. Photoelectrons were collected in a hemispherical analyzer and detected with a photodiode array. A −10.00 V bias was applied to the sample to enhance the collection of the lowest kinetic energy electrons during UPS analysis. All UPS spectra were referenced to the Fermi level, EF, of a clean polycrystalline gold substrate.

Photoelectrochemical (PEC) Experiments

Following all previous characterizations, single-composition Mn1–ZnO samples for PEC characterizations were fabricated on the FTO substrate at the same conditions as for combinatorial samples, except the substrate deposition temperature that was fixed to 297 °C. For these samples, the Zn composition in Mn1–ZnO was varied with 0.40 < x < 0.60. The electrochemical measurements were performed using a CH Instruments 660A electrochemical analyzer with a conventional three-electrode system that consists of a platinum foil as a counter electrode, Ag/AgCl (saturated KCl) as a reference electrode, and Mn1–ZnO (grown at 297 °C) on FTO glass substrates as the working electrode. The electrolytes used were 0.1 M K2CO3 with pH 9.2, 0.5 M K2SO4 solution with pH 10, and a buffer salt (sodium phosphate tribasic + K2SO4) with pH 12. A 250 W tungsten lamp solar simulator from Newport Instruments was used as a light source (calibrated to global AM1.5 intensity with a reference cell). The samples were illuminated through the back and front sides of Mn1–ZnO photoelectrodes under chopped light. An insulating epoxy (Loctite Hysol 1C) was used to cover the entire Mn1–ZnO sample except for the active area (between 0.27 and 0.42 cm2). Electrochemical impedance measurements were performed in a 0.5 M K2SO4 carbonate buffer solution (pH 10) using an Impedance analyzer (Solartron SI 1260, Ametek, USA) and the same three-electrode system. A sinusoidal (AC) potential with and amplitude of 7 mV was superimposed on a bias voltage, and the measurements were carried out through systematic variation of the bias potential from 1.5 to 0.25 V versus Ag/AgCl at frequencies of 0.5 and 1 kHz. A scan rate of 5 mV/s was employed with a data acquisition rate of 5 s per point. The ZView software version 3.1c (Scribner Associates Inc., Southern Pines, NC, USA) was used for estimating the flat band potential and carrier concentrations through fitting the experimental data.