High performance tunnel junctions were made from sputtered and annealed p-type CuAlO2 and n-type ZnSnO3 with suitable band alignment for both low resistance and alignment to typical inorganic materials needed for a tandem solar cell. The devices not only exhibit low resistance, they are also thermally stable, capable of sustaining postdeposition temperatures up to 600 °C. This is a key requirement for many high performance multijunction thin film inorganic solar cells. The CuAlO2 top-layer remains amorphous, providing a diffusion barrier for top cell stack processing. The materials’ stack gives a negligible voltage drop, and the visible-spectrum transparency is near 100%. XPS measurements show that unannealed Cu in the Cu-Al-O films is in the +2 oxidation state, while in the films annealed at 500 °C and above, Cu is in the +1 oxidation state. This suggests that annealing is necessary to form CuAlO2. A near-broken gap alignment provides a low resistance contact with band alignment that is nearly ideal for a tandem device.

The efficiency of single-junction photovoltaic devices is approaching a fundamental bound to power conversion efficiency known as the Shockley–Queisser limit.1 The only proven route to circumventing this limit is through the use of multiple junctions in series. In such a device, a wide bandgap semiconductor on top absorbs the high energy photons from the solar spectrum, while allowing lower energy photons to be transmitted to be absorbed by a narrower gap semiconductor. This approach minimizes thermalization losses and so boosts power conversion efficiency. Since all high performance solar cells consist of forward biased pn junctions, a multijunction device (p/n/p/n) necessarily creates a reverse biased np junction at the center. Therefore, some type of structure that can transform the upward flow of electrons from the bottom cell to a downward flow of holes from the top cell is required. Such a structure must produce no more than a small voltage drop, must be optically transparent over the range of wavelengths absorbed by the bottom cell, and must be thermally stable since the upper cell must be built on top of it.2 

Existing multijunction photovoltaics (PV) devices, made from epitaxially grown single crystal films, use heavily doped n+/p+ tunnel junctions to overcome current blocking by the reverse biased diode. These heavily doped junctions have depletion widths smaller than the effective tunneling distance and so provide a low resistance, optically transparent conversion. Furthermore, the diffusion coefficient of the dopants is low at the growth temperature. While this works well, their high cost makes epitaxial devices prohibitively expensive for terrestrial use. In low-cost (i.e., polycrystalline) photovoltaic materials, the presence of grain boundaries leads to a significant diffusion during the temperature steps needed to fabricate the upper solar cell.3 This interdiffusion counter dopes both sides of a tunnel junction leading to poor performance. The lack of such a connective structure has been a key barrier to low-cost multijunction PV.

The primary figure of merit for such a structure is the specific contact resistance (RC) calculated as the inverse of the slope of the J-V curve at zero voltage. While values as low as 10−7 Ω cm2 have been demonstrated in single crystal tunnel junctions,4 solar applications have relatively low current densities and so could accept RC’s as large as 0.1 Ω cm2. We have shown that a p-type Cu2O/n-type ZnSnO3 stack provides ohmic behavior with an RC of 0.37 W cm2.5 However, the p-type Cu2O layer in our previous devices was not thermally stable, decomposing into CuO and Cu inclusions at 450 °C. Such temperatures are required to fabricate the upper cell in a multijunction device.

Figure 1 shows how such a device operates. Even if the bulk of the two films is lightly doped, the type III band alignment creates extremely heavily doped interfacial layers allowing it to behave as a tunnel junction. Thus, conduction band electrons flowing to the broken gap junction (BGJ) from the right can readily tunnel into the valence band of the p-type material, creating a hole current from the left. The converse, a hole current from the left creating an electron current from the right, is equally valid. Based on literature values of the electron affinities,6 we suggested that the Cu2O/ZnSnO3 structure may have been operating as a BGJ even through the electron affinity of Cu2O is known to depend on process conditions.7 Here we report a stack consisting of n-type ZnSnO3 and a new p-type material which shows not only much lower resistivity than our first paper, but more importantly, greatly improved thermal stability. We then measure the band positions and show that the device has a near-broken gap band alignment.

Fig. 1.

Band diagram of a broken gap tunnel junction.

Fig. 1.

Band diagram of a broken gap tunnel junction.

Close modal

Since Cu2O proved thermally unstable, we looked for an alternative p-type material with similar values of electron affinity (χ) and bandgap (EG). CuAlO2 appeared to be a likely candidate. There are relatively few papers on the electrical properties of CuAlO2. One of the first, using laser ablation to deposit thin CuAlO2 films, reported a bandgap of 3.5 eV and a room temperature resistivity of 10.5 Ω cm.8 A report of samples made by reactive dc magnetron sputtering found similar results, a bandgap of 3.54 eV and a room temperature resistivity of 3.1 Ω cm, but required additional oxygen during deposition and a high temperature postdeposition anneal.9 The latter suggests that the material may meet the requirement for thermal stability.

ZnSnO3 was sputtered from a target pressed from a 50/50 mixture of ZnO and SnO2 sold by the American Elements company (Los Angeles, CA). CuAlO2 was deposited from a custom-made ceramic sputter target. Copper oxide powder (99.99% Cu2O, >10 μm particle size, Sigma Aldrich) was mechanically ground down to an average diameter of under 1 μm using a ball mill. Aluminum oxide powder (99.99% alpha phase Al2O3, ∼100 nm avg. particle size, Inframat Advanced Materials) was mixed with the copper oxide powder in equal atomic portions. The mixture was heated at 1100 °C overnight and reground the next day. This process was repeated three times until there were very few visually identifiable white and red particles remained. The resultant powder had a light bluish-white appearance. We used Energy Dispersive Spectroscopy (EDS) to confirm that the vast majority of the powder was stoichiometric and XRD to verify the existence of the delafossite CuAlO2 phase. The powder was then pressed at 2000 psi for 1 h to form a 3 in. diameter by 0.125 in. thick disc, which was then sintered at 1160 °C for 16 h. The final target was bonded to a copper plate, and a 5 h burn-in of the target at 100 W rf power was performed. The target was subsequently stored in a desiccator when not in use.

Films were sputtered on oxidized silicon wafers from this target at an rf power of 100 W, the maximum power we could use due to the poor thermal stability of the electrically conductive bonding adhesive. The argon flow was set to 50 sccm while the oxygen flow was set to 5 sccm, producing a chamber pressure of 5 mTorr. The substrate to target distance was 12.5 cm. No intentional heating of substrate was applied. The deposition rate was found to be 0.042 nm/min, meaning that a deposition time of 5 h was needed to get 10 nm films. Films thicknesses as measured by both ellipsometry and profilometry were found to be in very good agreement. The measured deposition rate is nearly an order of magnitude lower than a previous report where the film was reactively sputtered in a magnetron plasma.10 This may be explained by the use of rf rather than dc sputtering and the low power density used (2.2 W/cm2).

Atomic forces microscope measurements reveal 0.3–0.5 nm rms roughness for as-deposited films, with no statistically significant change after annealing at temperatures up to 600 °C. This is unsurprising given the reported lack of grain growth in CuAlO2 up until 800 °C.11 X-ray diffraction measurements suggested amorphous films for all sputter conditions, however, the very thin films measured may have led to weak signals and so the lack of peaks cannot be considered definitive. Attempts to use 4-point dc electrical measurements of 10 nm films were unsuccessful, showing an unmeasurably large resistivity. With films this thin, one must consider the effect of surface states. Reddy found carrier concentrations of 1.5–8.9 × 1017 cm−3 depending on process conditions in their reactively sputtered films. For a 10 nm thick film, this corresponds to a sheet carrier concentration of 1.5–8.9 × 1011 cm−2, a value that could easily be trapped or depleted by surface states in an unpassivated film.

As-deposited films demonstrated transmission close to 95% for photon energies below the bandgap of 2.7 eV [Figs. 2(a) and 2(b)]. However, this is substantially below the theoretical direct gap value of 3.5 eV.12 A 5 min RTA anneal at 500 °C in N2 increased the transmission to nearly 100% and shifts the bandgap to 3.5 eV in agreement with similar work by Lan et al.11 Once annealed at 500 °C, films were found to be stable, both optically and chemically. Further annealing at 600 °C for more than 1 h had no measurable impact. The bandgaps in the as-deposited films vary by about 0.25 eV from sample to sample. However, this drops to about 0.10 eV after annealing at 500 °C or above. Others have also seen a significant variation in the as-deposited films.13 

Fig. 2.

(a) Optical bandgap and (b) transmission plots of CuAlO2 films sputtered using the same sputtering conditions with various postdeposition anneals. The margin of error was ∼±3%.

Fig. 2.

(a) Optical bandgap and (b) transmission plots of CuAlO2 films sputtered using the same sputtering conditions with various postdeposition anneals. The margin of error was ∼±3%.

Close modal

Composition measurements of the CuAlO2 films were attempted using three methods, however, determining an accurate composition of such thin layers proved difficult. RBS measurements of a thin film deposited on a Ta/Si substrate showed Cu:Al ratios of 5:4, i.e., slightly copper rich while Auger depth profiling by sputtering with a 4 keV Ar+ beam gave Cu:Al ratios between 0.5:1 and 0.25:1. Considerable work to resolve this inconsistency led us to believe that the substantial mass difference between Cu and Al creates a preferential knock-on of the copper atoms during sputter-depth profiling.14 As will be discussed in Sec. IV, XPS measurements of the rf sputtered films done under more carefully controlled conditions confirmed the film stoichiometry as 1:1 and the dependence of the indicated composition on the sputtering parameters.

To make full tunnel junctions, zinc stannate was deposited by rf magnetron sputtering of a ceramic target in a pure Ar environment in an AJA ATC 2000 sputtering system using a target to substrate distance of 12.5 cm, as described previously.5 The chamber pressure was 5 mTorr, while the sputtering power was maintained at 125 W. The substrate was unbiased and the stage was unheated. These conditions were found to produce stoichiometric n-type films with a carrier concentration of 1018 cm−3 at a deposition rate of 1.33 nm/min. The bandgap was found to be 3.5 eV, which is in good agreement with the literature.

Devices were constructed on ITO-coated glass. A 110 nm thick aluminum doped ZnO layer was deposited on the ITO-coated glass substrate using a Savannah atomic layer deposition system from Cambridge NanoTech. This improved lateral conduction from the device stack to the bottom electrodes, better approximating the window-layer of a photovoltaic cell.15 In general, the samples were heated above 100 °C in situ to remove hydroxides and water adsorbed during air exposure, and then allowed to cool to below 45 °C prior to thin film deposition. The exception was CuAlO2 since it was deposited in the same chamber as ZnSnO3 and so did not experience air exposure. Devices were fabricated using a two-step photolithography/liftoff process in which the ZnSnO3 and CuAlO2 layers are patterned in the first liftoff and the metal contacts are patterned in the second liftoff. After each photoresist develop step, samples were subjected to 1 min of oxygen plasma ashing to remove any remaining resist residue in the patterned region. After the CuAlO2 was deposited, the samples were diced to produce multiple devices for subsequent anneals. Next, rapid thermal annealing was done at temperatures from 300 to 500 °C for 5 min and at 600 °C for over 1 h. Each of these anneals was done prior to top contact deposition to avoid problems associated with metal contact diffusion,16 since in the projected application metal contacts would not be present during top cell deposition in a multijunction solar cell. Finally, the top contact was deposited through a stencil mask. After each process step, a piece of the sample was stored for future reference and characterization. For each film deposition, additional witness substrates (soda lime glass and silicon) were included for characterization of the film properties.

Electrical measurements were conducted using two geometries: 3-electrode and 4-electrode. Both geometries used a four-contact measurement method, where a current is driven between an outer pair of contacts and a voltage drop is measured between an inner pair of contacts. The 3-electrode geometry used an inverted “T” shape configuration with a single top electrode centered on the device that had both current and voltage contacts, and two lower electrodes offset to either side of the device. The two lower electrode configuration was chosen to eliminate the voltage drop associated with the lateral conduction in the electrode beneath the stack. Device sizes ranged from 3 × 3 to 5 × 5 mm2. Due to the need for a mechanical sample alignment from the previous RTA steps, the mask for metal electrode 1 was smaller than the junction oxide stack by 200 μm. This gives an alignment tolerance of ±100 μm. Gold was used for the top metal electrode because it provided both a low-resistance path to the device and a reliable low-resistance contact to CuAlO2.

The initial 3-electrode measurements were done using an Agilent 4156 Parameter Analyzer. The voltage sweeps were limited to ±2 V to prevent damage known to occur at the metal-oxide interface.17 

Measurements done in the 3-electrode configuration on CuAlO2/ZnSnO3 devices showed BGJ-like behavior with a specific contact resistance of 2–4 Ω cm2 in the as-deposited devices. The measurements were nearly ohmic, albeit with a slight curvature (Fig. 3). This was virtually eliminated after annealing at 300 °C, but the measured resistance is found to increase from 3 to 8 Ω cm2 after a 5 min 300 °C RTA and from 3 to 9 Ω cm2 after a 5 min 500 °C RTA. The increased variation in contact resistance was a consistent feature of devices using CuAlO2 films.

Fig. 3.

3-Electrode I-V results of 80 nm ZTO/10 nm CuAlO2 devices on ITO/glass.

Fig. 3.

3-Electrode I-V results of 80 nm ZTO/10 nm CuAlO2 devices on ITO/glass.

Close modal

Measurement of the contact resistance in such a device is difficult, especially if the contact resistance is low. The total BGJ film thickness is ∼20 nm, while the width is typically 2–5 mm, an aspect ratio of at least 105 to 1. Any lateral current would create a series resistance that could easily mask the true device contact resistance. This is particularly problematic for the relatively lightly doped p-type CuAlO2 layer. To get a more accurate determination of the device performance, 4-point electrode measurements were done using the split-ring test structure shown in Fig. 4. For these measurements, both top and bottom electrodes were patterned to ensure a more vertical potential gradient. The top of the device was contacted through an outer split-ring electrode 1 to drive the current to a lower split-ring electrode 2, while the potential drop across the junction was measured between the two inner pins (electrodes 3 and 4) for the bottom and top electrodes.

Fig. 4.

Top and side views of the geometry for measuring the resistance of the tunnel junction stack using a 4-electrode method. Electrodes 1 and 4 are contacted using a thin gold wire spring to avoid damage to the device, while electrodes 2 and 3 are contacted using a tungsten tip probe.

Fig. 4.

Top and side views of the geometry for measuring the resistance of the tunnel junction stack using a 4-electrode method. Electrodes 1 and 4 are contacted using a thin gold wire spring to avoid damage to the device, while electrodes 2 and 3 are contacted using a tungsten tip probe.

Close modal

This 4-electrode configuration had extremely small voltage drops across the device, making accurate dc measurements extremely difficult. A Stanford Research Systems SR830 DSP lock-in amplifier was used to detect these small signals. The internal function generator with a 1 V rms signal at 190 Hz was used as a source. The signal output was applied to both the oscilloscope and the BNC central conductor, which was also connected to electrode 1, while the ground shield was connected to electrode 2. The specific contact resistance was calculated by dividing the measured resistance by the pn junction area in the device.

The data (not shown) show that less than 20 μV is actually dropped across the bilayer stack. This corresponds to an RC less than 1 mΩ cm2. For a typical tandem PV short circuit current, ∼20 mA/cm2, this translates to a voltage drop of less than 20 μV. This value of RC proved to be thermally stable, showing no measurable change, even for 60 min anneals at 600 °C. This suggests that the CuAlO2 layer is far more stable than the Cu2O layer used in our previous work. This conclusion is confirmed by Auger depth profiling which showed that film stacks annealed at temperatures up to 600 °C had no discernable interdiffusion. Nor were Cu inclusions seen in the annealed films.

To verify the optical performance of the full stack (without electrodes), we characterized it on a UV-vis test apparatus (Fig. 5). As expected from the individual film results, all devices showed optical transmission well above 90% for 90 nm thick as-deposited films. The transmission was near 100% (within experimental uncertainty) for 20 nm thick films at higher post deposition anneal temperatures (Fig. 5). The improved transmission occurs primarily for photons with energies of 2.2 eV and above. This is almost certainly associated with the formation of CuAlO2, partially during the 300 °C anneal and more fully during the 500 °C anneal as detailed in Fig. 2.

Fig. 5.

(a) Optical transmission for 80 nm ZnSnO3/10 nm CuAlO2 devices on ITO/glass and (b) 10 nm ZnSnO3/10 nm CuAlO2 devices. The estimated uncertainty in the transmission data was ±4%.

Fig. 5.

(a) Optical transmission for 80 nm ZnSnO3/10 nm CuAlO2 devices on ITO/glass and (b) 10 nm ZnSnO3/10 nm CuAlO2 devices. The estimated uncertainty in the transmission data was ±4%.

Close modal

While these experimental results are very encouraging, they do not prove that a broken gap junction occurs across the n-type ZnSnO3/p-type CuAlO2 heterointerface. Conduction may involve conventional band to band tunneling, tunneling due to a broken gap structure,18 thermionic emission, or trap assisted tunneling if the stack interface has a high concentration of defect states as in multijunction amorphous silicon cells.19 The contact resistance showed virtually no dependence on measurement temperature, eliminating the thermionic process. The position of the bands is of critical importance to distinguishing between the remaining alternatives. Furthermore, if one wants to use such a stack to create a low-resistance tunnel junction, they must be aligned to the semiconductors on both sides of the stack. To determine the band positions samples of CuAlO2, ZnSnO3, and stacks of the two materials were sent to National Renewable Energy Laboratory (NREL) for detailed thin film analysis using x-ray photoelectron spectroscopy (XPS) and ultraviolet photoelectron spectroscopy (UPS). For CuAlO2, both as-deposited and annealed samples were examined.

Figure 6 shows typical UPS results for ZnSnO3. It was found that the surface cleaning process plays an important role in the results. Sputtering sufficient to remove all adsorbed carbon led to erroneous indications of composition. This was also found for cleaning by UHV baking. UHV anneals at 200 °C or low energy (∼1 keV) Ar+ sputtering were found to remove most of the surface carbon without affecting the underlying film. The work function varied a bit depending on surface preparation details, but typical values were on the order of 4.2 ± 0.1 eV. Figure 6 illustrates the method for extracting the work function from UPS spectra. In this example, a linear extrapolation shows that the secondary electron cutoff (SEC) falls at 16.98 eV on the binding-energy scale. From this value, the work function, ϕ, is calculated according to ϕ = hν – SEC = 4.24 eV, where hν = 21.22 eV is the He I photon energy. The valence-band maximum (VBM = EF – EV) was similarly found via linear extrapolation of the valence-band onset (not shown) to be 3.38 ± 0.03 eV. From the measured bandgap (3.5 eV), this suggests that the film is very heavily doped n-type (of order 1019 cm−3), with the Fermi level located ∼0.12 eV below the conduction-band minimum (CBM), indicating that the electron affinity is 4.12 ± 0.1 eV.

Fig. 6.

UPS measurements of ZnSnO3 sample after a 200 °C UHV anneal shows that the SEC is 16.98 eV. This corresponds to a work function of 4.24 eV.

Fig. 6.

UPS measurements of ZnSnO3 sample after a 200 °C UHV anneal shows that the SEC is 16.98 eV. This corresponds to a work function of 4.24 eV.

Close modal

The measured carrier concentration is consistent with the results inferred by UPS by Minami et al.,20 but differs somewhat from the work of Choi who found carrier concentrations in the 1016 cm−3 range in as-deposited samples and in the 1019 cm−3 range after a 600 °C anneal.3 Choi found that the affinity also varied with annealing, ranging from ∼3.7 eV in as-deposited films to ∼4.7 eV for samples annealed at 600 °C. The difference in our results and Choi’s may reflect the effects of traps, experimental differences such as the effects of surface contamination, or the presence of different phases with deposition and postdeposition processes.

XPS measurements were repeated for as-deposited CuAlO2 samples. A similar sensitivity to surface preparation was observed, particularly for the Cu 2p3/2 signal. Even modest surface sputtering was found to shift the peak by ∼1 eV to lower binding energies. UHV baking at 150 °C was found to provide an acceptable carbon reduction without changing the film properties. X-ray valence-band data were found to be more reliable than UPS, showing almost no change with surface preparation. We attribute this to the difference in excitation depth: UV photons are more strongly absorbed near the surface than are the Al k-alpha x rays used in the XPS measurements. The XPS-derived valence-band maximum depended on the sections of the spectra that were fit, as the CuAlO2 valence-band onsets were not as linear as the ZnSnO3 data, making the VBM less certain.

Figure 7 shows XPS results for an as-received sample, a sample that underwent a low energy sputter clean, and samples that had UHV bakes at 150 and 175 °C. The value (Ef – Ev) varied with both surface cleaning procedure and data fitting window. For unannealed films, we find affinity values ranging from 3.3 to 4.1 eV assuming the ∼2.7 eV bandgap measured by absorption measurements for the as-received films. The Fermi level is clearly near the center of the gap, however, the CuAlO2 literature that exists uniformly identify the material as p-type. We take these results to indicate a high concentration of deep levels in unannealed CuAlO2 films, leading to very low carrier densities. We cannot definitively determine whether these states are on the surface or in the bulk. The presence of these traps may be related to the observed nonlinear behavior in the XPS results.

Fig. 7.

XPS data for as-received CuAlO2 with various surface preparations showed substantial nonlinearity as the probed energy approached the valence band.

Fig. 7.

XPS data for as-received CuAlO2 with various surface preparations showed substantial nonlinearity as the probed energy approached the valence band.

Close modal

High-resolution scans of the Cu 2p3/2 line and the Cu LMM Auger line were consistent with Cu+2. This is unexpected for the CuAlO2 delafossite structure where Cu should be in the +1 oxidation state. This is supported by the presence of “shake-up” peaks in the XPS spectrum commonly associated with Cu+2.21 One possibility is the existence of the spinel compound CuAl2O4 rather than the delafossite structure.22 However, XPS measurements found the film to be stoichiometric with a Cu/Al ratio of 0.98, a value well within the uncertainty of the technique. All of the p-type films were both extremely thin and amorphous so x-ray diffraction could not be used to distinguish the phase. Grazing incidence (13°) XPS was done to further examine the surface of the unannealed sample. Under this condition, the Cu/Al ratio was found to be 0.39, a value much closer to the spinel compound. This may be due to preferential surface segregation of the aluminum oxide, known to occur in multimetal oxides where one metal has a much higher oxygen affinity23 and in doped metal oxides.24 

These results suggest that the as-deposited CuAlO2 films may be a highly defective mixture of CuAl2O4 and CuO, both of which contain Cu+2. If the two materials are present in equal amounts, the film would have the same stoichiometry as CuAlO2, the observed “bulk” composition of the film. As previously mentioned, as-deposited films showed a smaller apparent bandgap in UV-vis measurements (Fig. 1). The spinel compound has a bandgap of 4 eV in the crystalline phase25 but only about 2.2 eV in small-grain films.26 Thus, it is possible that post deposition annealing at 500 °C is required to form the delafossite structure in our samples. This has some support in the very limited literature that exists for these materials.27 Hsieh et al. found that rf sputtered film must be annealed at 800 °C to be sufficiently crystalline to observe CuAlO2 structure by x-ray diffraction,28 although the degree of crystallinity in as-deposited films depended on rf power.29 Given the low power used in our films, the apparently amorphous nature of our films is unsurprising.

CuAlO2 films annealed at 500 °C for 5 min were scanned by XPS [Fig. 8(a)]. Such a temperature cycle is consistent with top cell formation in a tandem solar cell. It was found that the results were less dependent on surface preparation, with EF − EV = 0.09 eV ± 0.04 eV. Also, the linear part of the date is extended, with deviation from linearity beginning at about ∼20% of the peak value of count rate rather than the ∼40% seen in the unannealed sample. This makes accurate extrapolations much more certain. The results would suggest heavy p-type doping, a result that is inconsistent with transport measurements. However, a closer look at Fig. 8(a) shows a high concentration of defect (i.e., band tail) states. It is possible that these states localize a high concentration of holes, leaving few for conduction in spite of the small value of EF − EV. Again, this may be either a surface or a bulk effect. The work function of CuAlO2 was measured by UPS and found to be 21.22 – 16.74 eV = 4.48 eV [Fig. 8(b)]. Using this value, we arrive at a valence-band maximum (EV) of 4.57 eV. Using the optically measured bandgap (3.5 eV, Fig. 1), this gives the conduction band minimum of 1.07 eV below the vacuum level. These values are close to other reports in the literature.8,30–33 Near-valence-band states have generally been attributed to copper vacancies in Cu2O.34 Given our model that the material is transitioning from Al2O3/Cu2O to CuAlO2 during anneals at or above 500 °C, the observed defect states may be related to such metal atom vacancies.

Fig. 8.

(a) XPS measurement of EF − EV for two CuAlO2 samples shows significant band tail states near the valence band and (b) the work function for these samples as extracted from UPS measurements.

Fig. 8.

(a) XPS measurement of EF − EV for two CuAlO2 samples shows significant band tail states near the valence band and (b) the work function for these samples as extracted from UPS measurements.

Close modal

Finally, we combine the results in Fig. 9 to compare the band alignment of the individual materials. It is found that CuAlO2 does not form a true broken gap structure with ZnSnO3. Instead, the results suggest a band overlap of 0.45 eV. The linear behavior and low specific contact resistance that was observed can be explained in terms of the proposed high concentration of near-valence-band defects in CuAlO2 as discussed above. As shown in Fig. 8(a), these extend about 0.5 eV from the valence-band maximum. These states may provide an effective way to surmount the small EC,n/EV,p barrier enabling a low-resistance contact and creating a near-broken gap junction.18 In agreement with Hsieh,28 higher temperature anneals improve crystallinity, but in this model improved crystallinity may actually degrade performance by removing these states.

Fig. 9.

Band alignment referenced to the vacuum level for an annealed ZnSnO3-CuAlO2 device. The alignment shows a 0.45 eV barrier, however, the band tail states [Fig. 8(a)] shown in gray cover this energy range.

Fig. 9.

Band alignment referenced to the vacuum level for an annealed ZnSnO3-CuAlO2 device. The alignment shows a 0.45 eV barrier, however, the band tail states [Fig. 8(a)] shown in gray cover this energy range.

Close modal

A p-type material has been reactively sputtered from a custom CuAlO2 ceramic target with properties consistent with the literature and composition confirmed by XPS after a 500 °C anneal. The material deposition rate is very slow and morphology is amorphous. The film’s electrical properties vary considerably between depositions and area analyzed. The results for films annealed at 500 °C are much more consistent. XPS measurements indicate that unannealed films show Cu in the +2 oxidation state, while 500 °C annealed films have Cu in the +1 oxidation state. The latter is consistent with CuAlO2. This interpretation is supported by optical measurements that show the apparent bandgap shifting from 2.7 to 3.5 eV upon annealing. The small value of the difference in energy between the Fermi energy and the top of the valence band, 0.09 ± 0.04 eV, is surprising given the low conductivity of the films. This was attributed to a very high concentration of band tail states extending from the valance band edge into the gap. The best measurements indicate a work function of 4.48 eV. Combined with our optically measured bandgap, we arrive at values of 1.07 and 4.57 eV for the conduction and valence-band edges, respectively. The CuAlO2 layer was found to be stable up to the highest temperature tested (600 °C).

Near-broken gap junction devices have been developed by sputtering using thin films of ZnSnO3 and CuAlO2. The band alignment of the 500 °C annealed n-type and p-type materials may be appropriate for solar cells with an affinity of 3.7–4.0 eV, although with a band overlap of 0.32–0.54 eV. The absence of a barrier in electrical measurements was attributed to band tail states in the CuAlO2. The optimized deposition and postdeposition RTA processing gives nearly ideal behavior: ohmic I-V profiles with a specific resistivity below 1 mΩ cm2 and optical transmission over 95%. BGJ devices were stable up to 600 °C for over 1 h, which makes them highly compatible with multijunction solar cells.

The authors acknowledge funding from the Department of Energy SunShot grant (No. DEEE0005319). Part of this work was authored in part by Alliance for Sustainable Energy, LLC, the manager and operator of the National Renewable Energy Laboratory for the U.S. Department of Energy (DOE) under Contract No. DE-AC36-08GO28308. Funding provided by the U.S. DOE Office of Energy Efficiency and Renewable Energy, Solar Energy Technologies Office. Portions of this work were conducted in the Minnesota Nano Center, which is supported by the National Science Foundation (NSF) through the National Nano Coordinated Infrastructure Network (NNCI) under Award No. ECCS-1542202. Part of this work was carried out in the University of Minnesota Characterization Facility, a member of the NSF-funded Materials Research Facilities Network via the MRSEC program. B.B. received support from the University of Minnesota MRSEC under Award No. DMR-1420013 from NSF.

The views expressed in the article do not necessarily represent the views of the DOE or the U.S. Government. The U.S. Government retains and the publisher, by accepting the article for publication, acknowledges that the U.S. Government retains a nonexclusive, paid-up, irrevocable, worldwide license to publish or reproduce the published form of this work, or allow others to do so, for U.S. Government purposes.

1.
W.
Shockley
and
H. J.
Queisser
,
J. Appl. Phys.
32
,
510
(
1961
).
2.
M.
Yamaguchi
,
Sol. Energy Mater. Sol. Cells
75
,
261
(
2003
).
3.
Y.-Y.
Choi
,
S. J.
Kang
, and
H.-K.
Kim
,
Curr. Appl. Phys.
12
,
S104
(
2012
).
4.
V.
Kristijonas
,
T.
Marcel
,
A.
Shamsul
, and
A.
Markus-Christian
,
Semicond. Sci. Technol.
26
,
075021
(
2011
).
5.
F.
Johnson
,
S. H.
Song
,
J.
Abrahamson
,
R.
Liptak
,
E.
Aydil
, and
S. A.
Campbell
,
Sol. Energy Mater. Sol. Cells
132
,
515
(
2015
).
6.
J.
Robertson
and
S. J.
Clark
,
Phys. Rev. B Condens. Matter
83
,
075205
(
2011
).
7.
J.
Deuermeier
,
J.
Gassmann
,
J.
Brötz
, and
A.
Klein
,
J. Appl. Phys.
109
,
113704
(
2011
).
8.
Hiroshi
Kawazoe
,
Masahiro
Yasukawa
,
Hiroyuki
Hyodo
,
Masaaki
Kurita
,
Hiroshi
Yanagi
, and
Hideo
Hosono
,
Nature
389
,
939
(
1997
).
9.
A.
Sivasankar Reddy
,
P.
Sreedhara Reddy
,
S.
Uthanna
, and
G.
Mohan Rao
,
J. Mater. Sci. Mater. Electron.
17
,
615
(
2006
).
10.
Blake L.
Stevens
,
Cathleen A.
Hoel
,
Carolyn
Swanborg
,
Yang
Tang
,
Chuanle
Zhou
,
Matthew
Grayson
,
Kenneth R.
Poeppelmeier
, and
Scott A.
Barnett
,
J. Vac. Sci. Technol. A
29
,
011018
(
2011
).
11.
W.
Lan
,
W. L.
Cao
,
M.
Zhang
,
X. Q.
Liu
,
Y. Y.
Wang
,
E. Q.
Xie
, and
H.
Yan
,
J. Mater. Sci.
44
,
1594
(
2009
).
12.
F. A.
Benko
and
F. P.
Koffyberg
,
J. Phys. Chem. Solids
45
,
57
(
1984
).
13.
Yuh
Kumekawa
,
Motohiro
Hirai
,
Yuhki
Kobayashi
,
Satoshi
Endoh
,
Eri
Oikawa
, and
Takuya
Hashimoto
,
J. Therm. Anal. Calorim.
99
,
57
(
2010
).
14.
S.
Hofman
,
Proceedings of the International Conference on Quantiative Surface Analysis
,
Teddington, UK
,
16–17 October 1979
(Wiley, Hoboken, New Jersey,
1979
).
15.
F. J.
Pern
,
R.
Noufi
,
B.
To
,
C.
DeHart
,
X.
Li
, and
S. H.
Glick
,
Proc. SPIE
7412
, 74120K (
2008
).
16.
J. D.
McBrayer
,
R. M.
Swanson
, and
T. W.
Sigmon
,
J. Electrochem. Soc.
133
,
1242
(
1986
).
17.
R. E.
Hummel
,
Int. Mater. Rev.
39
,
97
(
1994
).
18.
Sang Ho
Song
,
Eray
Aydil
, and
Stephen A.
Campbell
,
Sol. Energy Mater. Sol. Cells
133C
,
133
(
2015
).
19.
Steven S.
Hegedus
,
Frank
Kampas
, and
Jianping
Xi
,
Appl. Phys. Lett.
67
,
813
(
1995
).
20.
Tadatsugu
Minami
,
Toshihiro
Miyata
, and
Takashi
Yamamoto
,
Surf. Coat. Technol.
108–109
,
583
(
1998
).
21.
D. C.
Frost
,
A.
Ishitani
, and
C. A.
McDowell
,
Mol. Phys.
24
,
861
(
1972
).
22.
C.
Baratto
,
R.
Kumar
,
G.
Faglia
,
K.
Vojisavljević
, and
B.
Malič
,
Sens. Actuators B Chem.
209
,
287
(
2015
).
23.
J.
Camra
,
E.
Bielańska
,
A.
Bernasik
,
K.
Kowalski
,
M.
Zimowska
,
A.
Białas
, and
M.
Najbar
,
Catal. Today
105
,
629
(
2005
).
24.
R. G.
Egdell
and
S. C.
Parker
,
Science of Ceramic Interfaces
, Material Science Monographs, edited by
J.
Nowotny
(Elsevier, Amsterdam,
1991
), Vol. 41.
25.
L. C.
Leu
,
D. P.
Norton
,
G. E.
Jellison
, Jr.
,
V.
Selvamanickam
, and
X.
Xiong
,
Thin Solid Films
515
,
6938
(
2007
).
26.
Masoud
Salavati-Niasari
,
Fatemeh
Davar
, and
Masoud
Farhadi
,
J. Solgel Sci. Technol.
51
,
48
(
2009
).
27.
Yang
Ren
,
Gaoyang
Zhao
,
Chenxi
Zhang
, and
Yuanqing
Chen
,
J. Solgel Sci. Technol.
61
,
565
(
2012
).
28.
Ping-Hung
Hsieh
,
Yang-Ming
Lu
,
Weng-Sing
Hwang
,
Wei-Luen
Jang
,
Chung-Li
Dong
, and
Ting-Shan
Chan
,
Mater. Chem. Phys.
144
,
547
(
2014
).
29.
Ping-Hung
Hsieh
,
Yang-Ming
Lu
, and
Weng-Sing
Hwang
,
Ceram. Int.
40
,
9361
(
2014
).
30.
W.
Lan
,
M.
Zhang
,
G.
Dong
,
P.
Dong
,
Y.
Wang
, and
H.
Yan
,
Mater. Sci. Eng.
139
,
155
(
2007
).
31.
N.
Tsuboi
,
Y.
Takahashi
,
S.
Kobayashi
,
H.
Shimizu
,
K.
Kato
, and
F.
Kaneko
,
J. Phys. Chem. Solids
64
,
1671
(
2003
).
32.
N.
Tsuboi
,
Y.
Itoh
,
J.
Ogata
,
S.
Kobayashi
,
H.
Shimizu
,
K.
Kato
, and
F.
Kaneko
,
Jpn. J. Appl. Phys.
46
,
351
(
2007
).
33.
N.
Tsuboi
,
T.
Moriya
,
S.
Kobayashi
,
H.
Shimizu
,
K.
Kato
, and
F.
Kaneko
,
Jpn. J. Appl. Phys.
47
,
592
(
2008
).
34.
L. Y.
Isseroff
and
E. A.
Carter
,
Chem. Mater.
25
,
253
(
2013
).