$Mg_xZn_{1-x}O$ contact to $CuGa_3Se_5$ absorber for photovoltaic and photoelectrochemical devices

$CuGa_3Se_5$ is a promising candidate material with wide band gap for top cells in tandem photovoltaic (PV) and photoelectrochemical (PEC) devices. However, traditional CdS contact layers used with other chalcopyrite absorbers are not suitable for $CuGa_3Se_5$ due to the higher position of its conduction band minimum. $Mg_xZn_{1-x}O$ is a transparent oxide with adjustable band gap and conduction band position as a function of magnesium composition, but its direct application is hindered by $CuGa_3Se_5$ surface oxidation. Here, $Mg_xZn_{1-x}O$ is investigated as a contact (n-type buffer or window) material to $CuGa_3Se_5$ absorbers pretreated in $Cd^{2+}$ solution, and an onset potential close to 1 V vs RHE in 10 mM hexaammineruthenium (III) chloride electrolyte is demonstrated. The $Cd^{2+}$ surface treatment changes the chemical composition and electronic structure of the $CuGa_3Se_5$ surface, as demonstrated by photoelectron spectroscopy measurements. The performance of $CuGa_3Se_5$ absorber with $Cd^{2+}$ treated surface in the solid-state test structure depends on the Zn/Mg ratio in the $Mg_xZn_{1-x}O$ layer. The measured open circuit voltage close to 1 V is promising for tandem PEC water splitting with $CuGa_3Se_5$/$Mg_xZn_{1-x}O$ top cells.

absorber in a tandem PEC cell, CuGa3Se5 could potentially lead to STH efficiency as high as 22.8% [12]. Initially a CuGa3Se5/ZnS/Pt device with current output of 8 mA/cm 2 (0V vs RHE, 3electrode) was reported [13], and subsequently the photocurrent density was increased up to 9.3 mA/cm 2 for a mixed phase of CuGaSe2 and CuGa3Se5 with CdS-modified surface and Pt catalyst [14]. More recently, 17 days of continuous water splitting operation for a bare CuGa3Se5 absorber photocathode with ∼12 mA/cm 2 photocurrent at −1 V vs RHE have been demonstrated by our team, suggesting promising durability [9]. These encouraging outcomes clearly point out the need for further fundamental study of this absorber and its interface with other materials.
Traditional chalcopyrites CuGaSe2 and ordered-vacancy CuGa3Se5 absorbers are often interfaced with CdS as a contact material (also known as n-type 'buffer' or 'window') to create the pn-junction. However, CdS contact layers suffer from short wavelength absorption and instability in electrolyte solution. In addition, CdS has a cliff-like ~0.2 eV conduction band (CB) offset with stoichiometric CuGaSe2, and a similar CB offset is expected for CuGa3Se5 [15]. If the conduction band minimum (CBM) of the contact is lower than that of the absorber (a 'cliff' type offset), the device suffers from lower photovoltage, increased interface recombination, and other detrimental effect to device performance [16] [17] [18]. If a CBM of the contact is more than 0.3 eV above that of the absorber (a 'spike' type offset), the resulting barrier impedes the collection of photo-generated carriers [19] [20] [21].
To address these challenges, MgxZn1-xO (MZO) [22] [23] with tunable conduction band position as a function of Mg content has been proposed as an attractive n-type contact layer. MZO thin film was demonstrated with up to x = 0.46 grown by RF co-sputtering without any phase segregation resulting in a bandgap of up to 4.2 eV, while ZnO has a bandgap of 3.24 eV [22].
Combinatorial studies explored the composition spreads of MZO with different deposition methods, such as pulsed laser deposition [24] [25] and chemical vapor deposition [26]. In a previous combinatorial study, we showed that the conduction band position could be tuned by 0.5 eV as Mg concentration changes from 4 to 12% [27], suggesting that it might be a suitable contact to CuGa3Se5 absorber. Integration of MZO as contact material resulted in significant efficiency improvements in different solar cell device technologies such as CdTe [28] and CIGS [29] [30], as well as CuGaSe2 [31] that likely has similar CB position to CuGa3Se5.
In this study, MZO was investigated as the contact layer material for CuGa3Se5 absorberbased PV and PEC devices. Structural, optical and electrical properties of MZO thin films were studied as a function of different experimental conditions such as Mg composition, Ga doping, substrate temperature and deposition ambient. MZO depositions were performed by combinatorial radio frequency (RF) sputtering. For functional CuGa3Se5/MZO PV device, absorber surface pretreatment with Cd 2+ solution was crucial because it removed surface oxidation, led to Cd incorporation, and possibly changed the surface conductivity type. MZO deposition and the surface pretreatment conditions were optimized for solid state solar cell performance. The outcome was a significant improvement in open circuit voltage (up to 920 mV) compared to conventional CdS-contact CuGa3Se5 devices (~730 mV). Replacement of CdS also improved quantum efficiency in the blue region of the spectrum. The PEC characteristics of CuGa3Se5/MZO as photocathode are tested with hexaammineruthenium (III) chloride sacrificial redox electrolyte, which exhibited an onset potential near 1 V vs reversible hydrogen electrode (RHE). These outcomes indicate that CuGa3Se5/MgxZn1-xO could serve as an efficient top cell for tandem PV and PEC water splitting devices.

Material synthesis and measurements
Ga-doped MgxZn1-xO thin film sample libraries with orthogonal composition gradients of Mg and Ga were deposited by combinatorial RF magnetron sputtering from ZnO, Mg and Ga2O3 targets ( Figure 1a). 50x50 mm Eagle XG glass substrates were cleaned with laboratory grade detergent followed by sonication in warm DI water, acetone and isopropanol. MZO depositions were done at a pressure of 3 mTorr in Ar/O2 atmosphere where the total gas flow rate was fixed at 16 sccm. The chamber base pressure was 5x10 -7 Torr. O2 flow, found to be crucial for good quality transparent films, was varied from 0.5 to 2% of the total gas flow for different depositions. A premixed 5% O2 in Ar cylinder was used as O2 source for precise flow control. A gas ring around the substrate carrying the O2 lines ensured uniform O2 containing environment across the substrate.
The samples were mounted on a temperature calibrated Inconel substrate holder and heated by a radiative heater. The substrate temperature was varied from room temperature to 200 C. The depositions were performed for 120 min, that resulted in film thicknesses in the 80 to 120 nm range.  Each sample was characterized at 4x11 grid points with the following spatially resolved methods ( Figure 1b). X-ray diffraction (XRD) patterns for the MZO thin films were obtained using a Bruker D8 Discover XRD instrument. Electrical resistivity/conductivity data was measured by a custom four-point probe system; the highest measurable sheet resistance by the equipment is Experimental combinatorial data collected in this study were managed, analyzed and displayed using our publicly available COMBIgor software package for Igor Pro [32], and will be made available through the High Throughput Experimental Materials Database (HTEM DB) [33].
Kelvin probe measurement system from KP technology was used to determine the work function for both the MZO (on doped Si substrates) and CuGa3Se5 (on Mo coated glass substrates) films. A gold reference (work function 5.1 eV) in air was used to quantify the absolute value of the surface potentials. X-ray photoelectron spectroscopy (XPS) was performed on the CuGa3Se5 films using monochromatic Al Kα radiation and a pass energy of 29 eV. The spectrometer binding energy scale was calibrated at high and low energy using clean gold and copper foils and known transition energies. Data analysis and peak fitting were performed using a combination of Igor and PHI MultiPak.
Cross-section scanning electron microscopy (SEM) image of the device was taken with a Hitachi S-4800 SEM instrument operated at 2kV. Cross-section transmission electron microscopy (TEM) specimens were prepared using the focused ion beam (FIB) lift out technique with the final Ga + ion milling performed at 3 kV. Ga + ion FIB damage was subsequently removed using low energy (< 1kV) Ar + ion milling in a Fischione Nanomill with the sample cooled by liquid nitrogen.
Scanning transmission electron microscopy (STEM) imaging and EDX mapping analysis were performed in a FEI Tecnai F20 UltraTwin field emitting gun STEM operated at 200 kV and equipped with an EDAX Octane T Optima Si drift detector (SDD) EDX system.

Device fabrication and characterization
Substrates for device fabrication were soda-lime glass. Mo back contact was deposited by Photoelectrodes were made by indium bonding an insulated Cu wire to an exposed part of the Glass/Mo substrate. A number of different electrolyte solutions was tested, including (1M Na2SO4 + pH7 buffer), (NaOH + H2SO4 + 0.5M Na2CO3, pH = 9.6), and (0.5M Na2SO4 + 0.25M KH2PO4 + 0.25M K2HPO4, pH = 6.7), yet the CuGa3Se5/MZO thin films were found to be unstable in all of these solutions with the photocurrent limited due to charge transport. As such, LSV analyses The standard reduction potential (E 0 ) for the hexaammineruthenium redox couple is 0.1 V vs RHE [35].

MgxZn1-xO thin film characterization:
For CuGa3Se5 device integration, combinatorial RF sputtered MgxZn1-xO was deposited with x values in the 0 to 0.15 range, and Ga/(Ga+Zn) atomic ratio from 0% to 15%. Figure 1b shows the composition profile of one such Ga:MgxZn1-xO library, and RBS data for a selected sample is presented in Figure S1.

Photoelectrochemical Characteristics:
The PEC characteristics of the CuGa3Se5 photocathodes were investigated with chopped light linear sweep voltammetry (LSV) with a 3-electrode system under simulated AM1.5G illumination. The LSV data is shown in supplement Figure S2, for analysis performed in hexaammineruthenium (III) chloride redox mediator and KCl supporting electrolyte with pH7 buffer. It should be noted that LSV measurement performed with a sacrificial redox agent is not water splitting, however it gives an estimate of the quality of the device characteristics. There was sign of degradation during consecutive testing, however that could be due to scanning near high anodic potentials. The dark current was nearly zero. The photocurrent saturated at more cathodic potentials. The initial current transient could be due to charge transport limitation in the electrolyte and/or at the electrode surface. The photocurrent onset potentials extrapolated from the LSV curves are shown in Table 1.  Figure S2 inset.

Surface and Interface:
To better understand the PEC characteristics of the electrodes, the absorber surface and interface modification due to the Cd 2+ solution treatment was further investigated. Surface sensitive XPS/UPS of the as-deposited and treated CuGa3Se5 revealed how the wet treatment affected the absorber surface ( Figure 4). The XPS survey spectrum of the as-deposited CuGa3Se5 is shown in the supplement Figure S3. The surface doping of as grown CuGa3Se5 appeared stable over a period of months, and was only slightly p-type (the XPS measurement on the as grown sample was repeated after 3 months, data not shown). The valence band position (EF-EVBM) value measured with monochromatic Al K and He I excitation was 0.55 eV and 0.72 eV respectively (supplement Figure S4). UPS results with He I light (21.2 eV) had a probe depth of less than 1 nm in the sample surface, while VBM measurements with Al K (1487 eV) measured up to 10 nm into the bulk of the film due to higher photoelectron kinetic energy. This indicated the presence of a downward band bending at the as-grown surface.
The elemental compositions of the CuGa3Se5 surface with different treatments is shown in Table 2. There were a number of changes that happened with both NH4OH and Cd 2+ (65°C 15 min) treatments ( Figure 4). A high degree of surface oxidation is observed for as deposited absorber, along with Na diffused from the soda lime glass substrate. With both NH4OH and Cd 2+ treatments, surface Na was removed and the O quantity was reduced. The gallium 2p3/2 peak narrowed significantly with Cd 2+ treatment, probably due to the removal of gallium oxides. It is also interesting to note that the cadmium solution treatment (and likely the subsequent air exposure), caused the appearance of oxidized selenium, Se 4+ . For CdTe devices a very thin layer of oxidized tellurium was found to be critical for well-passivated interfaces [36]. Such Se oxidation could also be beneficial for surface passivation of the CuGa3Se5 absorbers studied here.
No Cd could be detected on the Cd 2+ treated CuGa3Se5 films using XRF and SEM/EDX analysis techniques, which are more sensitive to the bulk than XPS. Such comparison of XPS and EDX (probes deeper, up to 100 nm) and XRF (probes up to 1000 of nm) compositions for asgrown films indicated a Cu deficiency on the surface which is likely due to the sites being replaced by Na atoms. Cd 2+ treatment introduced Cd on the absorber surface (supplement Figure S3).
Exposure of the surface to aqueous cadmium sulfate likely caused an ion exchange process between cadmium and copper. Removal of Na atoms by NH4OH could allow Cd atoms to occupy these sites and change surface doping. The (EF-EVBM) values from XPS and work function values from Kelvin probe measurement for differently treated films also suggested a change in surface doping.    -EVBM = 0.28 eV was calculated using the carrier concentration of ~2×10 14 cm -3 from Hall effect measurement and the carrier effective masses were obtained from literature [37]. The attainable CBM range for MZO is from our previous combinatorial study [27]. surface conductivity type towards becoming intrinsic. Cadmium did not appear to dope the surface fully n-type, unlike what was reported for CIGSe or CISe [38] [39]. CIGSe and CISe normally are p-type at the surface and after cation exchange with cadmium solution, become n-type. However, Cd 2+ treatment for CuGa3Se5 only moved the Fermi level upward, closer to being intrinsic.

Photovoltaic characterization:
To understand the influence of the contact CB position on the absorber performance in  The photovoltaic device performance for various configurations are summarized in Table   3. Although significant open-circuit voltage improvement was observed in MZO based devices, the fill factor and photocurrent output were reduced.  The results of this research will facilitate the understanding of CuGa3Se5/MgxZn1-xO interface, and the use of CuGa3Se5 absorbers for both PEC and PV device applications. Figure S1. Rutherford Backscattering Spectroscopy (RBS) to quantify Mg composition of the MZO films.