Enhancement of the open circuit voltage of Cu₂O/Ga₂O₃ heterojunction solar cells through the mitigation of interfacial recombination

Cuprous oxide (Cu₂O) is an earth abundant semiconductor suitable for cost-effective deployments at terra-watt scales^1,2 could theoretically deliver power conversion efficiencies (PCE) exceeding 20 % in single-junction solar cells. Open-circuit voltages (V_OC) that are much typically higher than other absorber layers are accessible using Cu₂O-based systems due to its wide bandgap in excess of 2 eV.^3–5 Interfacial recombination is a major loss mechanism in current Cu₂O-based thin film solar cells that limits their efficiency⁶ and is the primary reason for low V_OC in present-day devices.⁷ Recombination at the heterojunction interface is commonly attributed to high densities of defect trap states and to non-ideal band alignments at the interface. Moreover, monolayers of cupric oxide (CuO) are readily formed at the Cu₂O surface during the formation of the heterojunction when the surface is exposed to an oxygen-rich ambient.⁸ This oxidation can happen when the device fabrication process involves unavoidable air exposure of the Cu₂O surface during material transfers between different deposition environments.^9–11 The resulting CuO layer contributes deep-level defect states at the interface leading to enhanced recombination and appreciable performance degradation.^12,13 By applying a suitable n-type buffer layer to construct a heterojunction with an almost flat alignment of the conduction bands, the junction quality can be enhanced.^14,15 Ga₂O₃ buffer layers have proven themselves to be one of the most suitable n-type partners to Cu₂O, leveraging the near-ideal band alignments to achieve V_OC values of up to 1.2 V.¹⁰ In previous studies, the Cu₂O layer has always been exposed to air before the deposition of the Ga₂O₃ buffer layer owing to the limitation of utilizing separate deposition systems to be prepare the two different layers (e.g. electrochemical deposition (ECD) with atomic layer deposition (ALD),¹⁰ thermal oxidation with pulse laser deposition (PLD)¹¹), thereby allowing the undesirable CuO layer to be formed at the crucial junction. Numerous thin film photovoltaic materials including copper indium gallium selenide (CIGS) encounter similar challenges in surface chemistry transformations.¹⁶ Methods have been devised to reduce but not completely eliminate these oxidized layers.¹⁷ When zinc-containing compounds (eg. zinc-tin-oxide) acting as buffer layers are deposited on Cu₂O using ALD techniques, the reactive precursor vapors of diethylzinc can be injected before the start of the buffer layer deposition to reduce the surface CuO back to Cu₂O.¹⁸ Although Ga₂O₃ provides a favorable band alignment with Cu₂O, the ALD precursor for gallium does not completely reduce the CuO layer. It is vital to discover new vacuum-compatible fabrication strategies that mitigate interfacial defects introduced by the surface oxidation of Cu₂O.

Herein, we report a method that successfully circumvents the formation of CuO at the Cu₂O/Ga₂O₃ heterojunction interface via in-situ junction creation. This new approach embraces two compatible vapor deposition processes carried out in the same deposition system that has both ALD and CVD capabilities (Thin film solar cells were fabricated using cuprous oxide (Cu₂O) absorber layers grown by chemical vapor deposition (CVD) and gallium oxide (Ga₂O₃) buffer layers grown by atomic layer deposition (ALD) on the cuprous oxide CVD films. The in-situ formation of heterojunction in the same deposition system without exposure to oxygen-rich ambient was found to be effective in mitigating the creation of detrimental cupric oxide (CuO) at the interface, resulting in a pristine photovoltaic junction capable of delivering an enhanced open-circuit voltage of 1.78 V. Numerical device simulations of a novel two-layer absorber architecture (CVD-Cu₂O on ECD-Cu₂O) showed promising possibilities (theoretical 13.2 % efficiency) for a solar cell combining in-situ junction formation with electrochemical deposition of the absorber layer. (Figure S1). It was found that the pristine p-n junctions suppressed deep-level defects originating from interfacial CuO, resulting in the enhanced V_OC of Cu₂O-based devices. In addition, a numerical simulation model was developed for optimizing a novel two-layer absorber layer device architecture that incorporates the aforementioned in-situ heterojunction formation strategy. Our calculations demonstrate the potential of a device fabrication approach that couples the high V_OC advantages of a junction with high phase purity to the high J_SC delivered by Cu₂O films synthesized using ECD techniques, essentially capitalizing on the best of both approaches.

To fabricate the solar cells, a Si wafer with 300-nm-thick thermal oxide (SiO₂) was cleaved into 1 × 1 inch² substrates and coated with a stack of Ti (5 nm)/Au (200 nm) layer by e-beam evaporation to form the back contacts. A 1.5-μm-thick Cu₂O absorber layer, a 20-nm-thick Ga₂O₃ buffer layer and a 80-nm-thick ZnO:Al window layer were deposited successively in the same growth chamber with and without air exposure between deposition runs. (N,N′-di-sec-butylacetamidinato)dicopper(I) (Strem Chemical Company) and deionized water were chosen to prepare cuprous oxide by thermal CVD. The copper(I) precursor was sublimed from a bubbler maintained at 98 °C into a 100 sccm flow of purified nitrogen gas. Water was evaporated from a bubbler maintained at 25 °C into a 5 sccm flow of purified nitrogen gas. Both reactants were mixed together with a 40 sccm flow of purified nitrogen before injection into the process chamber. Following the growth of the Cu₂O absorber at 200 °C and 10 Torr, the Ga₂O₃ overlayer was deposited in the same chamber at 120 °C by ALD. Bis(μ-dimethylamino)tetrakis(dimethylamino)digallium (Strem Chemical Company) and deionized H₂O were maintained at 107 °C and 25 °C, with estimated exposures of 3 and 5 Torr·s, respectively. ZnO:Al was deposited by ALD using trimethylaluminum, diethylzinc and deionized water. The 500-nm-thick Ag top electrodes were deposited by e-beam evaporation. Surface morphologies and film thicknesses of the devices were analyzed using an Ultra 55 FESEM (Zeiss). J-V characteristics of the devices with a cell area of 0.24 cm² were measured by using a Keithley 2400 source meter under the standard 1-sun illumination generated by a Newport Oriel 91194 solar simulator with a 1600 W ozone-free Xe-lamp with a AM1.5G filter and a Newport Oriel 6895 flux controller calibrated by a NREL-certified Si reference cell equipped with a BG-39 window. K-alpha XPS (Thermo Scientific) was used to perform XPS measurements. Hall mobility, carrier concentration and resistivity were measured using a Hall effect measurement system with thin films of the material deposited on 1 x 1 cm² Si substrates with 300-nm-thick SiO₂. These results were utilized in our models for simulation of the devices. The solar cell simulator SCAPS-1D was employed to perform simulations of the Cu₂O-based devices.¹⁹ 

The surfaces of Cu₂O films are expected to be readily oxidized in ambient air to form a nanometer thick CuO layer. During XPS investigations (Figure 1a), distinct peaks are observed at 932.2 and 934 eV, which are attributed to the presence of Cu¹⁺ (Cu₂O) and Cu²⁺ (CuO), respectively.²⁰ Broad peaks in the 940-945 eV region are typically identified as satellite peaks characteristic of Cu²⁺. XPS depth profiling (Figure 1b) shows that the extent of oxidation is limited to several monolayers on the film surface. Figure 1c shows the XPS spectra of the Cu-2p core levels for three Cu₂O films with a 3-nm-thick Ga₂O₃ overlayer and various treatments applied to the surface of Cu₂O to study the oxidation states of copper at the absorber-buffer interface. Bilayer samples were fabricated by depositing a 1.5-um-thick Cu₂O film by CVD whereas the 5-nm-thick overlayer of Ga₂O₃ was coated on the Cu₂O film by ALD. The first Cu₂O/Ga₂O₃ bilayer sample was deposited entirely in the same chamber without breaking vacuum. The surface of Cu₂O is exposed only to purified nitrogen, to the gallium ALD precursor whose reducing nature assists in the preservation of Cu₂O, and to water vapor that has been processed to contain minimal dissolved oxygen. Under these conditions, no CuO is expected to form at the Cu₂O/Ga₂O₃ interface, as established from the XPS Cu-2p scans showing no detectable peaks corresponding to Cu²⁺ states. The second sample was deposited in the same chamber but with both depositions separated by the additional step of cooling the sample chamber to room temperature, followed by exposure of the Cu₂O film surface to oxygen present in the air for 1 minute, then evacuating the chamber before finally reheating in purified nitrogen to the deposition temperature of the Ga₂O₃ ALD process. A 1-nm-thick CuO layer is estimated to have formed on the surface of Cu₂O when exposed to ambient air at room temperature. The deleterious cupric oxide layer is exacerbated in the third Cu₂O/Ga₂O₃ sample by exposing the Cu₂O surface to ambient air at an elevated temperature of 200 °C for 1 minute to create a ≈3-nm-thick CuO layer. All three samples show peaks corresponding to Cu¹⁺ at 932.2 eV. The sample with an intentional heating of the surface in air displayed the strongest Cu²⁺ peaks at ≈934 eV with increased prominence of the 940-945 eV satellite peaks characteristic of Cu²⁺. The second sample that was exposed to air at room temperature had smaller peaks attributed to CuO. Although exposure to ALD metal precursor vapors has been reported to reduce the exposed surface if the reactions are energetically favorable,¹⁸ the presence of trace amounts of CuO is indicative of the gallium precursor being less effective than diethylzinc at reducing Cu²⁺ to Cu¹⁺. To investigate the effect of surface oxidation of the absorber on the solar cell performance, devices with three different surface treatments similar to the bilayer samples in our XPS studies were fabricated. The device stack of the solar cells comprises of Au back contact (500 nm)/Cu₂O absorber (1.5 μm)/Ga₂O₃ buffer (20 nm)/ZnO:Al window (80 nm)/Ag top electrode (500 nm) as shown in the cross-sectional scanning electron microscopy (SEM) image (Figure 2a). J-V measurements demonstrates a strong dependence of device efficiency and V_OC when varying the integrity of the heterojunction interface (Figure 2b). Forming the Cu₂O/Ga₂O₃ interface in the same chamber without a vacuum break between both depositions yielded the best performing device exhibiting a PCE of 2.36 % and a V_OC of 1.78 V. When the Cu₂O surface is exposed to air, XPS analysis detected the presence of Cu²⁺ on the Cu₂O surfaces that were intentionally oxidized but not on the sample where the air exposure was avoided during in-situ junction formation. Heat treatment during air exposure increased the amount of surface oxidation, which further reduced the V_OC and significantly degraded the fill factor (FF) until an almost linear J-V characteristic was observed. The J_SC remained constant for all devices as the introduction of interfacial defects through air exposure and heat treatment had negligible impact on their photogeneration capabilities.

Defect densities at both the heterojunction interfaces and in the bulk of the photovoltaic materials are prime factors that determine device performances. To understand the effects of air exposure at the interfaces of our devices and gain quantitative insights into the degree to which the resultant interfacial recombination affects cell parameters, we performed simulations of J-V curves while varying the density of interfacial defects (Figure 3a). As observed, the V_OC of the cells shows a dramatic increase with a corresponding decrease in the interfacial defect density. For simulated devices with interfacial defect densities < 10¹² cm^-3, the J_SC and FF remain almost constant. By further increasing the defect concentration at the interface, the diode characteristic is observed to start significantly deviating away from those of an ideal diode, resulting in the noticeable degradation of both the FF and J_SC. This behavior demonstrates the importance of interfacial passivation to suppress interfacial defects which is shown to be the primary factor affecting the V_OC of our devices. The effects of absorber layer quality on the solar cell parameters are captured in the device simulations by comparing devices with bulk defect densities of 10¹⁴, 10¹⁵ and 10¹⁶ cm^-3 (Figure 3b). The simulated devices show increasing J_SC with decreasing bulk defect densities. Our calculations suggest that there is tremendous room for improvement in J_SC by replacing the CVD-Cu₂O layer with ECD-Cu₂O^10,21,22 or thermal oxidation of copper sheets.^23,24 Simulation predicts doubling our current cell efficiencies with just this single change to our device stack. Based on our device modelling, the performance of devices with no air exposure at the heterojunction interface is bulk recombination dominated. The extracted surface recombination velocity is in the range of 10 cm/s and the bulk defect density in the CVD-Cu₂O absorber is fitted to 10¹⁴ cm^-3.

Simulations of device performance were carried out to predict the optimal conduction band offset (CBO) between the absorber and buffer layer. Figure 3c shows the device favoring lower carrier concentration for the absorber and a higher carrier concentration for the buffer for maximizing the photovoltaic efficiencies. To motivate further studies on overcoming the challenges of increasing the electron density of ALD-Ga₂O₃,²⁵ the results are exhibited as a function of both the CBO and the carrier concentration of the buffer layer while keeping the carrier concentration of the absorber layer constant at the optimal 10¹⁶ cm^-3. CBOs at the junction can possibly be tuned by doping Ga₂O₃ prepared in an ALD process with a suitable dopant such as aluminum or indium to raise and to lower the conduction band, respectively. The simulated behaviors exhibited in Figure 3d can be classified into two clusters with the first populated by devices whose Ga₂O₃ buffer have majority carrier concentrations that lie below 10¹⁸ cm^-3. In this scenario the optimal CBO was predicted to be at +0.1 eV. This positive offset forms a type-I conduction band spike at the Cu₂O/Ga₂O₃ interface which is small enough to allow unimpeded current flow into the Ga₂O₃ layer via thermionic emission. As we increase the CBO, the spike becomes sufficiently large to suppress the collection of light-generated carriers. When a negative CBO is introduced at the junction, the type-II cliff structure increases the interfacial recombination velocity. In these cases where the offset deviates away from +0.1 eV, the device efficiencies are reduced as shown in our numerical calculations. Interestingly the second group distinguished by higher buffer layer carrier concentrations has optimal CBOs in the range of +0.2 to +0.4 eV which are much larger than that of devices with buffer layers of lower electron concentrations. The best performing Cu₂O devices with a 1.78 V (V_OC) was measured to have a carrier concentration of 10¹⁶ cm^-3 for the undoped Ga₂O₃ buffer layer and a negative CBO of -0.2 eV that forms a cliff structure. Such devices have larger sensitivity to CBOs relative to the carrier densities of the Ga₂O₃ buffer layer which are also more challenging to achieve experimentally. Based on our modelling studies, we can most readily achieve the greatest efficiency improvements of our present-day devices by increasing the CBO to be slightly positive with the introduction of aluminum into the Ga₂O₃ buffer layer. Increasing the buffer layer carrier densities to exceed 10¹⁸ cm^-3 not only widens the depletion region in the absorber to reduce interfacial recombination but also narrows the conduction band spike at the interface. This enables the tunneling of photogenerated carriers through the barrier, which in turn enables the use of larger CBOs to further mitigate interface recombination. Additional simulations were performed to provide guidance and to evaluate the potential of fabricating the bulk of the Cu₂O absorber using the ECD technique which produces higher quality materials known to deliver devices exhibiting much higher photocurrents (up to 11.5 mA/cm²) compared to that of the CVD-grown material (2.0 mA/cm²).¹⁰ An ultra-thin CVD layer of Cu₂O is proposed to form the p-n junction with the Ga₂O₃ window layer, leveraging the ability to deposit both layers in the same deposition system without oxygen exposure to maintain the junction integrity for achieving the low interfacial recombination velocities. The predicted J-V characteristics (Figure 2d) clearly highlights the strengths of the two Cu₂O fabrication techniques individually and when combined. Simulated devices using an ECD-Cu₂O absorber were able to achieve a J_SC of 8.4 mA/cm² but suffered from a lower V_OC of 1.2 V. In comparison, the CVD-Cu₂O devices displayed better V_OC of almost 1.8 V at the expense of a lower J_SC of 2.0 mA/cm². Numerical models of the 2-layer absorber system (Figure 2c) demonstrated significant enhancement in device performance, obtaining a cell efficiency of 7.2 %. A small drop of 170 mV in V_OC is observed when progressing from the CVD-Cu₂O absorber to the 2-layer absorber stack explained by the dominance of the smaller optical bandgap of the ECD-grown material. The upside is an increase in the wavelengths that can be absorbed from the incident irradiation, improved photocurrent collection efficiencies from lower bulk defect densities and the opportunity to form the critical junction with the Ga₂O₃ window layer without interfacial defects generated from surface oxidation. Consequently, the use of two different absorber components predicts improved overall device performance. When presented with the compromise of lower voltages in exchange for larger photogeneration, an optimization was done by varying the thickness of the CVD-grown layer (Figure S2). By increasing the film thickness of CVD-Cu₂O, the V_OC experienced a negligible change while the J_SC and FF and hence device efficiency increases initially to achieve a predicted 13.2 % efficiency before decreasing more dramatically as a result of the lower hole mobility of the thicker CVD-grown film relative to the more crystalline Cu₂O grown by ECD. The highest PCE was obtained from a film thickness of 20 nm for the CVD-grown Cu₂O layer.

In summary, it was shown that the Cu₂O/Ga₂O₃ heterojunction without the defective CuO layer at the interface was successfully achieved through the in-situ formation of the heterojunction using two compatible deposition processes (CVD for Cu₂O and ALD for Ga₂O₃) executed in the same deposition system. It is believed that consecutive vapor depositions without air exposure in the cell fabrication process circumvent the oxidation of Cu¹⁺ on the surface of Cu₂O films. This approach is applicable to other material systems to improve the interface quality by mitigating the interfacial recombination originating from the oxidized species, hence achieving enhanced solar cell performance.

Devices simulations predicted improved device efficiencies of over 13 % that can potentially be achieved using the proposed 2-absorber approach and with a computationally optimized thickness of 20 nm for the CVD-Cu₂O layer. This strategy could provide solutions to the shortcomings of both systems: low J_SC with CVD-Cu₂O, and low V_OC with ECD-Cu₂O. Simulations of devices that combine both absorber layers demonstrate a high V_OC by maintaining phase purity of Cu₂O to form a pristine heterojunction between CVD-Cu₂O and ALD-Ga₂O₃ while simultaneously leveraging the low bulk defect densities of the solution-grown Cu₂O for achieving a high J_SC.

See supplementary material for details on the deposition system design, simulation model parameters and optimization results.

This work was supported in part by the U.S. Department of Energy under contract DE-EE0005329, the Center for the Next Generation of Materials by Design, an Energy Frontier Research Center funded by the U.S. Department of Energy, Office of Science. This work was performed in part at the CNS, a member of the National Nanotechnology Coordinated Infrastructure Network (NNCI), which is supported by the National Science Foundation under NSF Award No. 1541959, at Harvard’s X-ray laboratory, and at the Buonassisi group at the Massachusetts Institute of Technology (MIT). CNS is part of Harvard University.