Novel Interlayer Boosting the Performance of Evaporated Cu2O Hole‐Selective Contacts in Si Solar Cells

Passivating contacts based on transition metal oxides are of great interest for applications in crystalline silicon (c‐Si) solar cells due to their improved optical transparency and potential cost reduction. In this work, the contact resistivity and passivation for thermally evaporated Cu2O are investigated and optimized, with and without an Al2O3 interlayer, as a hole‐selective contact to c‐Si. Additionally, we implement an Al y TiO x /TiO2 stack as a novel passivating tunnel interlayer for hole‐selective contacts, achieving an implied open‐circuit voltage iV oc of 630 mV and a record‐low J 0 of 212 fA cm−2 while maintaining a contact resistivity ρ c of 62 mΩ cm2. A record‐low ρ c of 8 mΩ cm2 for Cu2O‐based contacts is also demonstrated at the expense of passivation. The addition of the interlayer resulted in a 2% absolute improvement in the efficiency of proof‐of‐concept c‐Si cells with full‐area rear Cu2O contacts, reaching 19.1%.The demonstration of this novel interlayer stack provides new avenues to improve the performance also of other hole‐selective passivating contacts.


Introduction
Crystalline silicon solar cells incorporating passivating contacts have emerged as a high-efficiency alternative to conventional passivated emitter and rear contact devices.Passivating contacts provide both carrier-selectivity and passivation to the crystalline silicon (c-Si) interface, enabling efficient extraction of one type of charge carrier while suppressing recombination associated with direct metal contacts, and thus allowing higher efficiency. [1][3] For surface passivation, controlling the carrier concentration and passivating dangling bonds at the interface are important approaches. [3]erefore, each material can induce passivation and/or selectivity by a combination of mechanisms.
In recent years, interdigitated back contact (IBC) and silicon heterojunction (SHJ) solar cells incorporating passivating contacts based on stacks of intrinsic and doped-hydrogenated amorphous silicon (a-Si:H) have achieved world-record efficiencies of over 26%. [4,5]Similar results have been obtained for cells incorporating passivating contacts based on stacks of thin SiO x and doped polycrystalline silicon (poly-Si). [6]Despite enabling excellent passivation and selectivity, these contacts require high processing temperatures (for poly-Si) or possess limited thermal stability (for a-Si:H), and involve toxic/complex processes, as well as presenting high parasitic absorption, which limits the device efficiency. [2,7]The need to find more stable, and less toxic materials/processes with improved optical transparency have led researchers to explore novel materials as passivating contacts.Transition metal oxides (TMOs) are potential candidates to replace Si-based contacts due to their high bandgap and manufacturability.In this context, hole-selective TMO materials have so far exhibited poorer performance than electron-selective contacts, [8] and thus require further attention.
From the hole-selective layers reported, the best results have so far been obtained with contacts based on V 2Àx O x , WO x, and MoO x .[11][12][13] Efficiencies as high as 23.8% have been achieved using MoO 3 to substitute the p-type a-Si:H layer in SHJ solar cells. [14]Although they usually present sufficiently low contact resistivity (<100 mΩ cm 2 ), these hole-selective materials underperform in terms of passivation when used on their own, requiring additional interlayers to improve them.Best results have been obtained with intrinsic a-Si:H as interlayer, as in the aforementioned SHJ cells, however, this compromises some of the advantages of using these materials.Some alternative interlayers include SiO 2 , Al 2 O 3 , and HfO 2 . [15,16][18] Al 2 O 3 appears particularly promising in this role because of its negative fixed charge that could enhance its performance when coupled with hole-selective contacts. [15]nother potential hole-selective material is cuprous oxide (Cu 2 O).In contrast to the other hole-selective materials mentioned above, Cu 2 O promotes selectivity through a small valence band offset and large conduction band offset with silicon (%0.2 and %0.9 eV respectively). [19]In this respect, it is more closely analogous to common electron-selective TMOs like TiO 2 .Additionally, its intrinsic p-type conductivity, high hole mobility (%80-256 cm 2 V À1 s À1 ), [20] relatively high work function (%4.6-5 eV), [19,21] electronic bandgap of %2.1 eV, and reasonably wide optical direct bandgap (%2.6 eV) [22,23] make Cu 2 O a promising candidate for such an application.
Although Cu 2 O is a well-known semiconductor material, research integrating Cu 2 O as a hole-selective layer for siliconbased solar cells is scarce in the literature.Early work, using sputtering, reported efficiencies of 3.39% and 5.4% for pure Cu 2 O and B-doped Cu 2 O, respectively. [19,24]Subsequently, 6.03% was achieved with thermally evaporated Cu 2 O. [25] Using nitrogen as a dopant, sputtered CuO x :N resulted in contact resistivity values as low as 11 mΩ cm 2 , on the same order of magnitude as other state-of-the-art hole-selective contacts, although neither passivation nor cell results were reported. [26]Using spin-coating, Liu et al. pushed the efficiency to 9.54%. [27]Nguyen et al. employed Cu 2 O, deposited by atomic layer deposition, on top of a-Si:H, obtaining an efficiency of 13.7%. [28]Most recently, record efficiencies of 19.71% and 20.35% have been demonstrated by Li et al. who deposited Cu 2 O by e-beam evaporation on top of a 1 nm Al 2 O 3 interlayer, which promoted some passivation by preventing the formation of interfacial defects as a result of copper diffusion. [29,30]Despite some promising results on the device level, the literature lacks extensive information on contact resistivity and passivation of Cu 2 O-based hole-selective contacts; two crucial factors when applying such contacts in c-Si solar cells.
In this work, we deposit Cu 2 O by thermal evaporation and evaluate the contact resistivity and passivation of Cu 2 O-based contacts.We additionally investigate both well-known and new interlayers as a means to improve the passivation of Cu 2 O and (potentially) other hole-selective materials.

Materials Characterization
As copper has up to three possible oxidation states, it is important to verify whether the thermal evaporation successfully produced Cu 2 O, the target material of this study.We used X-ray photoelectron spectroscopy (XPS) to confirm this hypothesis.The Cu 2p 3/2 spectra (Figure S1, Supporting Information) show a sharp peak at 932.5 eV, as well as a broader, weaker one at 934.3 eV, which are likely associated with Cu 2 O and CuO respectively. [31]The assignment of the weaker peak to CuO is supported by the presence of its typical satellite peak in the 940-945 eV range, with the expected intensity ratio.We interpret the presence of some CuO as being due to surface oxidation of the Cu 2 O film in ambient, as previously reported. [32]It is worth mentioning that the samples stayed exposed to air much longer prior to XPS measurement than for the lifetime and contact measurements.It is also important to note that the Cu metal and Cu 2 O 2p 3/2 binding energies differ by only 0.1 eV.Therefore, one needs to also evaluate the Cu Auger electron (LMM) spectra in order to distinguish the different oxidation states.As shown in Figure 1a, we observe peaks at 913 eV and 916.7 eV, which are associated with the Cu þ state, [33] showing no presence of Cu 0 and Cu 2þ , confirming the deposition of Cu 2 O.
The ultraviolet photoelectron spectroscopy (UPS) spectrum, in Figure 1b, reveals a work function of 4.6 eV, in the range of reported values in the literature. [19]To confirm the work function value from UPS and to check the homogeneity of the film, we performed Kelvin Probe mapping in the dark.The small work function variation observed across an area of %0.17 cm 2 (Figure 1c) suggests that a 7 nm thick film is uniform and homogenous.The average work function derived from this area is 4.63 eV, identical to that determined via UPS (within the experimental uncertainty, especially of the latter technique).
Linear fits in Figure 2a,b show the extraction of the valence band maximum (VBM) and bandgap determined respectively from the photoemission yield and optical absorption coefficient obtained from ellipsometry.A VBM of 5.17 eV implies no valence band offset with silicon, which is ideal for a hole-selective contact.We extracted the bandgap energy based on the Tauc equation with n = 3/2 (direct forbidden bandgap), as both theoretical and experimental studies support the fact that the lowest energy optical transition for Cu 2 O is of this type. [21,23]A bandgap energy of 2.1 eV is consistent with previously reported values in the literature and results in a large conduction band offset with silicon (%1 eV), which should block electron transport, assisting hole-selectivity.The grazing incidence X-ray diffraction (GIXRD) diffractogram of Figure 2c shows a peak at 36.4°, which corresponds to the (111) plane of Cu 2 O. Therefore, the thin-film (%7 nm) contains some level of crystallinity, which could influence the surface passivation.

Cu 2 O Hole-Selectivity
To assess the hole-selectivity, we began by studying the contact resistivity behaviour for Cu 2 O single layers as a function of thickness, as shown in Figure 3.The data shows an initial decrease in contact resistivity followed by a sharp increase; which is a similar trend to other hole-selective contacts, including CuO x :N. [26,34]he high contact resistivity values for the thinnest films may relate to non-uniformity, or even not-completely-closed films due to limitations in the ability of thermal evaporation to deposit ultra-thin films.ρ c decreases to <100 mΩ cm 2 for films thicker than 4 nm, reaching a minimum average value of 72 mΩ cm 2 for 7 nm and remaining approximately constant up to 10 nm before again rising above 1000 mΩ cm 2 at 20 nm.The low contact resistivity reported here for 4-10 nm thick films could be linked to the high hole-mobility of Cu 2 O as well as the close valence band alignment of Cu 2 O with silicon.The minimum value of 72 mΩ cm 2 reported here is lower than values reported previously for single, non-doped Cu 2 O layers. [30]Nevertheless, the inset figure reveals that the IV curves measured for the transfer length measurement (TLM) structure show a slight S-shape curve, which is not ideal for a contact application.Based on these results, we fixed the thickness of Cu 2 O at 7 nm for subsequent experiments.
In terms of passivation, Cu 2 O has previously demonstrated very poor passivation in the literature with recombination current density prefactor J 0 on the order of %10 4 fA cm À2 . [30]e observed a similar behaviour for the thermally evaporated Cu 2 O single-layer films.As-deposited Cu 2 O by itself was unable to effectively passivate the surface, and we were unable to appropriately extract J 0 and lifetime.This is unsurprising as the relatively low work function of 4.6 eV is expected to be associated with a rather low hole concentration (given the 0.4-0.6 eV difference between the Cu 2 O Fermi level and VBM), and these would be insufficient to induce significant band bending at the Si surface.Furthermore, it has been shown before that when in contact with silicon, Cu 2 O loses oxygen, which could result in a more metallic-like state at the interface. [30,35]Metallic Cu can easily diffuse into Si and form defects at the silicon surface, [29] degrading the surface passivation and likely causing Fermi-level pinning. [36]herefore, adding an interlayer between silicon and Cu 2 O could mitigate these effects and improve surface passivation as reported for both Cu 2 O and other hole-selective contacts. [17,18,30]owever, there is generally a trade-off between contact resistivity and passivation when it comes to adding interlayers, which has not been extensively investigated for Cu 2 O.In the next section, we report the use of single layers of thermal and plasma atomic layer deposition (ALD) Al 2 O 3 as interlayers and their effects on the passivation and contact resistivity.

Al 2 O 3 Interlayer
We investigated the use of 2-10 cycles of plasma and thermal Al 2 O 3 followed by 7 nm of Cu 2 O. Figure S2 and S3, Supporting Information, show, respectively, the passivation and contact resistivity results with Al 2 O 3 interlayers deposited   In order to push the performance further, we investigated a novel Al y TiO x /TiO 2 stack as an alternative passivating tunnel interlayer.This highly transparent stack has previously been reported by us as an effective electronselective contact. [37]However, in this work we keep the total thickness of the Al y TiO x /TiO 2 stack low to allow the transport of holes via tunneling, and show for the first time that this stack can also be employed effectively as a passivating tunnel interlayer in a hole-selective contact structure.
The stack consists of a first layer of Al y TiO x formed by alternating cycles of Al 2 O 3 and TiO 2 (each super-cycle consisting of one cycle of TMA þ H 2 O followed by one cycle of TiCl 4 þ H 2 O).This initial layer is capped by a layer of pure TiO 2 (TiCl 4 þ H 2 O) to form the interlayer stack.Both layers of the interlayer stack are amorphous as previously reported by our group. [37]Passivation results with only the Al y TiO x /TiO 2 interlayer (Figure 4a) demonstrated excellent performance even when decreasing the total number of cycles to low values.Note that in the following we count one super-cycle of Al y TiO x as one cycle when counting total cycles.We obtained iV oc above 700 mV for interlayer thicknesses >=20 cycles, reaching 723 mV for the thickest interlayer (total cycle count = 39), without any post-deposition treatment.
We next proceeded to evaporate 7 nm of Cu 2 O on top of the Al y TiO x /TiO 2 interlayer to investigate the contact resistivity and passivation.Figure 4a,b respectively show the iV oc and contact resistivity values found as a function of the total number of ALD cycles for the interlayer.From the trend in Figure 4a, the improvement in passivation increases with the interlayer thickness.In contrast to the case of the single Al 2 O 3 interlayers, the deposition of Cu 2 O decreases iV oc , with the magnitude of the reduction decreasing as the number of cycles increases.It seems that the thicker the interlayer becomes, the more it can protect the surface passivation from the Cu 2 O deposition.As previously mentioned, possible causes for poor passivation by Cu 2 O films are potential Cu diffusion and/or the formation of a poor-quality interface with silicon.Since the film is crystalline, there is a lattice mismatch at the Si interface with Cu 2 O, which is generally associated with poor passivation.Therefore, having a sufficient thickness for the interlayer could mitigate these problems to a certain extent.
iV oc reaches a maximum of 673 mV at 39 total cycles (11:28 Al y TiO x :TiO 2 cycles), much higher than the values obtained for a single Al 2 O 3 interlayer.On the other hand, such a relatively thick interlayer leads to a very high contact resistivity, as shown in Figure 4b.When varying the number of cycles for the interlayer, one sees an interesting trend in contact resistivity.Despite the presence of a dielectric layer, there is a range of thicknesses where the contact resistivity drops considerably compared to the case without any interlayer, reaching values as low as 12 mΩ cm 2 for 14 cycles of the interlayer stack.The contact resistivity remains <100 mΩ cm 2 up to 21 cycles before it steeply rises.The reason behind the improvement in contact resistivity could be Fermilevel depinning at the silicon interface.We reduce the density of interface states, which leads to depinning of the interface after a certain interlayer thickness (%10 cycles), as shown in Figure 4.
Compared to the case with Al 2 O 3 interlayers (Figure S3, Supporting Information), the contact resistivity with the Al y TiO x /TiO 2 stack remains within an acceptable range up to a much higher cycle count (5 vs 21 cycles) before it begins to increase significantly.The explanation likely involves the lower ALD deposition temperature (75 °C).At this temperature, the density of both Al 2 O 3 and TiO 2 decreases, and a more amorphous film with weaker dielectric behaviour is formed. [38,39]t such low temperatures, SiO 2 formation is also inhibited or limited according to transmission electron microscopy images reported in our previous work. [37]These low-temperature characteristics give considerably more room to expand the interlayer thickness.It is important to note that the sheet resistance derived from the slopes of the resistance versus spacing plots, measured using TLM, have similar values with and without the interlayer, implying that the flow of current between contacts continues to occur via hole conduction through the p-type substrate rather than via electron conduction in a potential n-type inversion layer (the obtained sheet resistance values are also far too low to be attributable to inversion layer conduction).This confirms that the contacts remain hole-selective despite the use of interlayer materials previously employed in electron-selective contacts.
So far, we have described the interlayer without distinguishing the relative influence of the Al y TiO x and TiO 2 layers.Figure 5a,b display the contact resistivity and passivation resulting from varying the Al y TiO x layer thickness while fixing the TiO 2 thickness at 15 cycles.We can observe a sharp increase in iV oc with increasing Al y TiO x cycles, going from around 470 mV for 2 Al y TiO x cycles to up to 657 mV for 8 cycles.Interestingly, there seems to be some saturation of the passivation, which is very sensitive to the deposition of Cu 2 O as demonstrated above.In terms of contact resistivity, Figure 5b shows a similar trend to the one reported in Figure 4b.An initial decrease in contact resistivity occurs between 2 and 5 Al y TiO x cycles, reaching values around 14-28 mΩ cm 2 before again increasing with further increases in thickness.As solar cell efficiency is more sensitive to passivation quality than contact resistivity (for contact resistivity values <100 mΩ cm 2 ), we aimed at achieving the highest number of Al y TiO x cycles while maintaining a <100 mΩ cm 2 ohmic contact.Thus, the optimum combination was 6 Al y TiO x and 15 TiO 2 cycles, yielding a contact resistivity of 62 mΩ cm 2 and maximum iV oc of 630 mV.The corresponding recombination current density prefactor J 0 was 212 fA cm À2 per side, which is the lowest ever reported for Cu 2 O.   Above 6 cycles, the contact resistivity becomes non-ohmic as indicated in Figure S4, Supporting Information.
We also tried varying TiO 2 thickness with 6 Al y TiO x cycles, but this resulted in undesired contact resistivity and Schottky behaviour (Figure S5a, Supporting Information).Therefore, we instead investigated the behaviour of contact resistivity and passivation resulting from varying the TiO 2 capping layer thickness while fixing the number of Al y TiO x cycles at 4. Figure 7 shows the result of varying the number of TiO 2 cycles from 2 to 35 in this case.Figure 7a reveals that iV oc is less sensitive to the variation of TiO 2 thickness, reaching only 578 mV for a of 4 Al y TiO x super-cycles and 35 TiO 2 cycles.Interestingly, contact resistivity showed an ohmic behaviour (Figure S4b, Supporting Information) even with 35 cycles of TiO 2 .Notably, we achieved a minimum contact resistivity of 8 mΩ cm 2 , the lowest ever reported for Cu 2 O-based contacts, with a combination of 4 Al y TiO x and 10 TiO 2 cycles as shown in Figure 7b.Indeed, TiO 2 has been previously reported as a potential hole-selective layer for c-Si due to its negative fixed charge. [40]Nevertheless, this stack presents a considerably poorer passivation when compared to the interlayer with 6 super-cycles of Al y TiO x and 15 cycles of TiO 2 reported in Figure 4, the best overall combination obtained for this interlayer stack.
The excellent performance of the Al y TiO x /TiO 2 tunnel interlayer benefited the surface passivation and contact resistivity of the hole-selective contact.One additional explanation for the improved performance is the potential change in Cu  The Al y TiO x /TiO 2 interlayer greatly improves the overall performance of Cu 2 O contacts through different mechanisms as demonstrated above.A useful way of visualizing these mechanisms is through the energy-level diagram for the structures.Figure 9 shows the energy levels for the materials studied in this work.The values for Al y TiO x and TiO 2 are taken from our previous work. [37]roof-of-concept solar cells (Figure 10a) were fabricated to evaluate the effect of adding the optimized interlayer stack at a device level.Compared to the control device, without the interlayer, the efficiency of the champion device increased by 2%  absolute, from 17.1 to 19.1%, when adding the interlayer stack, as shown by the current density-voltage ( J-V) curves in Figure 10b.The largest part of this improvement is due to a 45 mV gain in V oc when adding the interlayer, confirming the passivation improvement observed in lifetime structures.Additionally, a higher fill factor (75.58 vs 79.67%), attributable to reduced series resistance (0.18 vs 0.54 Ω), contributes to the improved performance.We note that the absolute efficiency of both sets of devices is limited by a relatively low value of J sc , which can be correlated to the non-ideal optical properties of the front-side anti-reflection coating, as indicated by the relatively low external quantum efficiency (EQE) through the visible region (Figure 10d).Non-uniformity of the front-side antireflection coating may also account for the significant variation of J sc and fill factor observed for both sets of devices (Figure 10c).By contrast, V oc shows relatively little variation within each batch, indicating good uniformity of the surface passivation.EQE measurements also reveal a slightly improved response in the infrared region  (1000-1200 nm) for the device with the interlayer, which indicates reduced surface recombination at the rear side.
Therefore, the novel interlayer stack significantly improved the performance of solar cells with Cu 2 O contacts by boosting passivation and reducing the contact resistivity of the contact structure.

Conclusion
We performed an optimization of thermally evaporated Cu 2 O as a hole-selective contact for c-Si solar cells.An optically transparent Al y TiO x /TiO 2 stack was successfully implemented as a novel passivating tunnel interlayer to improve passivation and contact resistivity.We studied different combinations for the interlayer stack, with best results for 6 cycles of Al y TiO x and 15 cycles of TiO 2 , reaching an iV oc of 630 mV, record-low J 0 of 212 fA cm À2 and ρ c of 62 mΩ cm 2 .On the other hand, a stack of 4 cycles of Al y TiO x and 10 cycles of TiO 2 led to a record-low ρ c of 8 mΩ cm 2 .The improvement in both passivation and contact resistivity is due to the improved chemical passivation and increased work function of Cu 2 O provided by the interlayer.Proof-of-concept solar cell results demonstrated a 2% absolute gain in efficiency compared to devices without the interlayer, with V oc and fill factor increasing by around 45 mV and 4% absolute, respectively, confirming the results from the test structures.This work opens new avenues to improve the performance of existing holeselective contacts, increasing their efficiency potential while avoiding optically absorbing materials like a-Si:H, and bringing costs down with low-temperature processes.

Experimental Section
Cu 2 O films were deposited by thermal evaporation using an Angstrom Engineering thermal evaporation system.The deposition rate was kept constant at 0.2 Å s À1 .Cu 2 O powder (99.9%) was acquired from Alfa Aesar.Boron-doped, as-cut Cz p-type silicon wafers (2.1-2.9Ω cm, 150 μm) were used for contact resistivity measurements.These wafers were etched by tetramethylammonium hydroxide to remove saw damage.For lifetime measurements, we used boron-doped, double-side-polished, float-zone p-type silicon wafers (1-5 Ω cm, 285 μm).Substrates were cleaned by standard Radio Corporation of America (RCA) cleans [41] followed by a dip in 1% hydrofluoric acid to remove the RCA-grown oxide layer.We initially studied the behaviour of Cu 2 O single layers, and subsequently investigated the insertion of interlayers to improve passivation.All interlayers were deposited by atomic layer deposition in a Beneq TFS 200 system.They included: 1) single Al 2 O 3 layers; and 2) an Al y TiO x /TiO 2 stack, which has been previously reported by our group as an electron-selective contact. [37]The former was deposited at 200 °C using trimethylaluminum (TMA) and either H 2 O or O 2 plasma as reactants.The Al y TiO x /TiO 2 stack was deposited at 75 °C using TMA, titanium tetrachloride (TiCl 4 ), and H 2 O, following the same procedure reported in our previous work. [37]or cell fabrication, we employed Cz p-type silicon wafers (1.8 Ω cm, 250 μm).A high-temperature tabula rasa treatment was performed prior to cell processing to improve the bulk lifetime. [42]2 Â 2 cm 2 solar cells were defined on the front surface using photolithography.The front side of the cells was then textured, and a full-area front-side phosphorus diffusion was performed at 780 °C for 30 min.Anti-reflection and surface passivation of the front side were provided by a stack of thermal SiO x and plasma-enhanced chemical-vapor-deposited silicon nitride (SiN x :H) layers.Contact openings were formed by photolithography, and an evaporated stack of Cr/Pd/Ag, followed by silver electroplating, formed the front metal grid.Finally, a full-area rear contact was formed by depositing Al y TiO x / TiO 2 and Cu 2 O/Pd/Ag stacks by ALD and thermal evaporation, respectively.
Contact resistivity was extracted using TLM measurements.Currentvoltage (I-V ) measurements were performed at room temperature using a Keithley 2425 SourceMeter.TLM contact patterns were formed by thermally evaporating stacks of 10 nm of palladium and 150 nm of silver through a shadow mask.Surface passivation quality was evaluated via excess carrier lifetime measurements using a Sinton Instruments WCT-120 photoconductance lifetime tester.For lifetime measurements, the films were deposited on both sides to form symmetrically passivated samples.
GIXRD patterns were performed on a PANalytical X'Pert PRO MRD system with Cu Kα radiation and an incident angle of 0.6°.Film thickness and absorption coefficient were determined using an ex situ J. A. Woollam M2000D spectroscopic ellipsometer, using a B-spline model with the Cu 2 O optical constants determined in our previous work as a starting point. [43]XPS was performed to determine the oxidation state of the film, using a Thermo Scientific Nexsa Surface Analysis System equipped with a hemispherical analyser at a base pressure of 5.0 Â 10 À9 mbar.The incident radiation was monochromatic Al Kα x-rays (1486.6 eV) at 72 W (6 mA and 12 kV, 400 μm 2 spot).Survey and high-resolution scans were recorded at analyser pass energies of 150 and 50 eV and step sizes of 1.0 and 0.1 eV, respectively.A low-energy dual-beam (ion and electron) flood gun was used to compensate for surface charging.UPS was carried out in-situ alongside the XPS measurement to determine the work function and valence band energy of the material.UV radiation of 21.22 eV was used together with a À 10 V bias to distinguish secondary electron cut-off from the low kinetic energy electron scattering.A KP Technology Φ 4 ultrahigh-vacuum Kelvin probe system was used for additional measurements of work function and valence band energy.A mapping mode was employed to determine the homogeneity of the work function.The VBM energy was extracted using photoemission yield spectroscopy with a monochromated UV source (200-360 nm) in the same KP Technology system.The illuminated current density-voltage ( J-V ) was performed in-house using an FCT-450 tool from Sinton Instruments.EQE measurements were performed using a QE-R system from Enlitech.

Figure 1 .
Figure 1.a) XPS spectrum of Cu LMM for Cu 2 O, b) Cu 2 O UPS spectrum showing extracted work function and valence band level.c) Kelvin probe workfunction map of Cu 2 O thin film.

Figure 2 .
Figure 2. a) Cube root of photoemission yield showing extraction of VBM, b) Tauc plot for n = 3/2 (direct forbidden transition) showing extraction of optical bandgap, and c) GIXRD diffractogram of Cu 2 O thin film deposited by thermal evaporation.All films in the analysis were 7 nm thick.

Figure 3 .
Figure 3. Contact resistivity versus Cu 2 O thickness.Inset shows IV curves measured for the TLM structure with a 7 nm thick film.Error bars show standard deviation across 2 depositions.
by thermal and plasma ALD.Additionally, the passivation was further investigated by annealing the stacks at 250 °C in O 2 .iV oc improved following Cu 2 O evaporation and annealing for both thermal and plasma Al 2 O 3 interlayers, reaching 575-580 mV.On the other hand, the contact resistivity generally worsened as the number of Al 2 O 3 cycles increased.Exceptions to this were the samples with 5 cycles of thermal Al 2 O 3 or 2 cycles of plasma Al 2 O 3 , which showed similar ρ c (62 mΩ cm 2 ) to the sample without an interlayer.2.4.Al y TiO x /TiO 2 Stack Interlayer Despite the overall performance of the Cu 2 O contacts having improved with the Al 2 O 3 interlayer, the resulting passivation remains relatively poor, in part due to the fact that the starting point, without Cu 2 O, was quite low.

Figure 4 .
Figure 4. a) Implied open-circuit voltage and b) contact resistivity for Cu 2 O with ALD Al y TiO x /TiO 2 interlayer as a function of the total number of ALD cycles.Inset figures show the sample structures used for each measurement.Labels at each data point indicate the number of Al y TiO x super-cycles and TiO 2 cycles, respectively.Error bars show the minimum/maximum across 2 depositions.For data points without error bars, only one sample was measured.
Figure 6a,b show the J 0 extraction and IV curves, respectively.

Figure 5 .
Figure 5. Implied open-circuit voltage a) and contact resistivity b) for Cu 2 O with ALD Al y TiO x /TiO 2 interlayer as a function of number of Al y TiO x supercycles.The number of TiO 2 cycles was fixed at 15. Inset figures show the sample structures used for each measurement.Error bars show the minimum/ maximum across 2 depositions.

Figure 6 .
Figure 6.Auger-corrected inverse lifetime a), with linear fit used to extract J 0 , and TLM IV curves b) for the best interlayer stack (6:15 Al y TiO x /TiO 2 cycles).
2 O work function once an interlayer is inserted in the structure.Similarly to the case for MoO x , the chemical reduction of Cu 2 O at the silicon interface could lead to a lower work function, which could be prevented by adding a chemically stable interlayer.To test this hypothesis, we measured the work function of the Cu 2 O film with the best interlayer combination (6 Al y TiO x and 15 TiO 2 cycles).

Figure 8
compares the spatial variation of the work function of Cu 2 O with and without the interlayer.When inserting an interlayer, the average work function increased from 4.63 to 4.74 eV, which is consistent with the idea that the interlayer provides protection from Cu 2 O reduction.As discussed above, copper (Cu) easily diffuses and forms defects when in contact with the silicon surface.Therefore, by having an oxide interlayer, we can inhibit the diffusion, avoid the reduction of Cu 2 O at the interface, and thus increase the work function.It is also worth noting the decrease in the standard deviation of the work function, indicating an improvement in work function uniformity.

Figure 7 .
Figure 7. a) Implied open-circuit voltage and b) contact resistivity for Cu 2 O with ALD Al y TiO x /TiO 2 interlayer as a function of the number of TiO 2 cycles.The number of Al y TiO x super-cycles was fixed at 4. Inset figures show the sample structures used for each measurement.Error bars show the minimum/ maximum across 2 depositions.For data points without error bars, only one sample was measured.

Figure 8 .
Figure 8. Work-function mapping of the as-deposited Cu 2 O without (red) and with (yellow) an interlayer stack of 6 Al y TiO x and 15 TiO 2 cycles.

Figure 9 .
Figure 9. Energy levels of layers employed in this work.

Figure 10 .
Figure 10.a) Schematic diagram of the solar cell structure with the Al y TiO x /TiO 2 interlayer stack.b) J-V curves for the champion devices with and without the interlayer stack.c) Box plot of cell parameters for cells with and without the interlayer.d) EQE measurement of the champion cells showing the infrared region in the inset.