AlxIn1−xN on Si (100) Solar Cells (x = 0–0.56) Deposited by RF Sputtering

We investigate the photovoltaic performance of solar cells based on n-AlxIn1−xN (x = 0–0.56) on p-Si (100) hetero-junctions deposited by radio frequency sputtering. The AlxIn1−xN layers own an optical bandgap absorption edge tuneable from 1.73 eV to 2.56 eV within the Al content range. This increase of Al content results in more resistive layers (≈10−4–1 Ω·cm) while the residual carrier concentration drops from ~1021 to ~1019 cm−3. As a result, the top n-contact resistance varies from ≈10−1 to 1 MΩ for InN to Al0.56In0.44N-based devices, respectively. Best results are obtained for devices with 28% Al that exhibit a broad external quantum efficiency covering the full solar spectrum with a maximum of 80% at 750 nm, an open-circuit voltage of 0.39 V, a short-circuit current density of 17.1 mA/cm2 and a conversion efficiency of 2.12% under air mass 1.5 global (AM1.5G) illumination (1 sun), rendering them promising for novel low-cost III-nitride on Si photovoltaic devices. For Al contents above 28%, the electrical performance of the structures lessens due to the high top-contact resistivity.


Introduction
Nowadays, the main goal of the photovoltaic industry is to develop novel technologies that manage to simultaneously improve device efficiency and reduce fabrication costs. In that sense, materials such as perovskites [1] and III-V compounds [2] have obtained great results. However, new low-cost materials are needed to complement the already established Si technology. III-nitride semiconductors and alloys are very promising for application in solar cells because of their tunable wide direct bandgap energy from the near-infrared (0.7 eV, InN) to the ultraviolet (6.2 eV, AlN) and their particular material properties, such as thermal and chemical stability, make them quite promising to be used in space applications [3].
The difficulties of growing high-quality and single-phase Al x In 1−x N relies on the big differences between its binary constituents, InN and AlN (bonding energies, lattice mismatch, and growth temperature) being responsible for the immiscibility gap, and therefore for the phase separation and composition inhomogeneities commonly present in the alloy [4].
The growth of Al x In 1−x N alloys has been reported by different techniques, but mainly by metal-organic chemical vapour deposition [5][6][7], molecular beam epitaxy [8][9][10], and sputtering [11][12][13]. The first two growth techniques achieve the fabrication of high crystal quality layers at high growth In this paper, we report the electrical and optical properties of n-Al x In 1-x N on p-Si (100) hetero-junctions deposited at 550 • C by RF magnetron sputtering as a function of the Al mole fraction (x = 0-0.56) with the propose of developing operative low-cost solar cells. The photovoltaic performance of these junctions is also studied and deeply analysed.

Al x In 1−x N Deposition by RF Sputtering
Al x In 1−x N samples were deposited on p-doped 300-µm-thick Si (100) with a resistivity of 1-10 Ω·cm using a reactive RF magnetron sputtering system with 2 inch confocal magnetron cathodes of pure indium (4N5) and pure aluminium (5N), and pure nitrogen (6N) as the reactive gas. The substrates were chemically cleaned in organic solvents before being loaded into the chamber where they were outgassed at 600 • C for 30 min. Then, substrates were cooled down to the growth temperature. Before deposition, targets and substrate were cleaned using a plasma etching with Ar (5N) in the growth chamber. Optimized Al x In 1−x N layers present in this study were deposited with a nitrogen flow of 14 sccm at a pressure of 0.47 Pa. The RF power applied to Al target, P Al , was varied from 0 to 225 W (samples D1 to D7 in Table 1) to modulate the Al mole fraction, while the RF power applied to In target and the temperature were fixed to 30 W and 550 • C, respectively. Under these deposition conditions the deposition rate increases linearly with P Al from 0.59 nm/min for P Al = 0 W to 1.66 nm/min for P Al = 225 W. Taking this in mind, we have adjusted the deposition time to target a nominal thickness of the Al x In 1−x N layer of 90 nm. X-ray reflection (XRR) measurements performed in these films point to an average thickness of the Al x In 1−x N of 88 ± 8 nm. More details about the deposition procedure and characterization of the samples can be found in ref. [28].

Material Characterization
The electrical and optical properties were analysed in Al x In 1−x N layers deposited on sapphire substrates at the same time and under the same growth conditions. We used room-temperature Hall-effect measurements under the conventional van der Pauw geometry. The optical properties of the layers were estimated from transmittance measurements carried out at normal incidence using an optical spectrum analyser over the 350-1700 nm wavelength range at room temperature.

Device Fabrication
Al x In 1−x N on Si structures were processed into devices of~0.7 cm 2 , with both top and bottom contacts formed by 100 nm-thick Al deposited by RF sputtering. Before the deposition of the Al x In 1−x N film, the bottom contact to p-Si was sputtered and then annealed at 450 • C during 3 min under nitrogen atmosphere to provide an ohmic behaviour [25]. Afterwards, Al x In 1−x N films were fabricated and finally, the top n-contact was deposited. Figure 1a,b show a schematic description of the Al x In 1−x N on Si solar cell structure with a top-view optical microscopy image of a representative device, respectively.
The resistivity of the top and bottom contacts was obtained from transmission line measurements on Al x In 1−x N on sapphire samples and bare p-Si substrates, respectively. The ohmicity of the n-Al x In 1−x N top contact was evaluated measuring the current density-voltage curve between two Al pads (2.5 × 1 mm 2 ) spaced from 0.5 to 3 mm in the dark; while the ohmicity of the p-Si bottom contact was evaluated in bare p-Si wafers measuring the current density-voltage curve between two Al pads (1 × 0.5 cm 2 ) spaced 1-6 mm in the dark [25]. Figure 1c presents the evolution of the n-contact resistance as a function of P Al used for the Al x In 1−x N growth, which increases from 0.07 Ω to 2.2 MΩ in the studied range. The bottom p-contact resistivity is~6 Ω·cm 2 .

Device Characterization Techniques
Devices were characterized by current density-voltage curves performed in dark and under standard solar illumination at 25 °C. Dark measurements were carried out with a four-point probe station, whereas measurements under illumination were recorded in a solar simulator with an AM1.5G spectrum under 1 sun (Pin = 100 mW/cm 2 ). External quantum efficiency (EQE) measurements were carried out at room temperature exciting with a 250 W halogen lamp coupled to an Oriel Cornerstone 130 1/8 m monochromator in the visible range (400-1100 nm) at zero bias. A red He-Ne laser (λ = 633 nm) with an output power of 0.6 mW was used to calibrate the generated photocurrent versus the input optical power.

Results and Discussion
The structural, morphological and optical quality and properties of the AlxIn1−xN on Si (100) layers understudy was deeply analysed in reference [28]. In this report, our group demonstrated the successful deposition of high-quality and wurtzite single-phase n-type AlxIn1−xN layers with an Al mole fraction over the range of x = 0 to x = 0.56 on p-Si (100) substrates by RF magnetron sputtering. High-resolution X-ray diffraction measurements revealed that the layers own wurtzite single-phase crystalline structure oriented along the c-axis perpendicular to the sample surface. 2θ/ω scans performed in these layers show the diffraction peaks associated to the Si (100) and the AlInN (0002) reflections revealing that there is no phase separation in the AlInN alloy along the studied range of Al power supply and that no other parasitic reflections related to other phases were detected. On the other hand, the full width at half maximum of the rocking curve around the AlxIn1−xN (0002) diffraction peak decreased from ~9° to ~3° when incorporating Al to the AlxIn1−xN alloy. The root mean square surface roughness of the samples, estimated from atomic force microscopy, evolved from 20 nm for InN to 1.5 nm for Al0.56In0.44N films. For Al mole fractions equal and below 0.35, samples showed strong low-and room-temperature photoluminescence emission, which blueshifted from 1.59 eV for InN to 1.82 eV for Al0.35In0.65N samples at 11 K. Taking into account these results, the developed n-AlxIn1−xN/p-Si junctions presented promising material properties to explore their performance operating as solar cells.
To develop efficient devices, it is of crucial importance to proper control the n-type doping. For this aim, Hall effect measurements were carried out in AlxIn1−xN films deposited on sapphire at the same time and under the same growth conditions to explore the evolution of their electrical characteristics, such as carrier concentration, layer resistivity, and carrier mobility, depending on the

Device Characterization Techniques
Devices were characterized by current density-voltage curves performed in dark and under standard solar illumination at 25 • C. Dark measurements were carried out with a four-point probe station, whereas measurements under illumination were recorded in a solar simulator with an AM1.5G spectrum under 1 sun (P in = 100 mW/cm 2 ). External quantum efficiency (EQE) measurements were carried out at room temperature exciting with a 250 W halogen lamp coupled to an Oriel Cornerstone 130 1/8 m monochromator in the visible range (400-1100 nm) at zero bias. A red He-Ne laser (λ = 633 nm) with an output power of 0.6 mW was used to calibrate the generated photocurrent versus the input optical power.

Results and Discussion
The structural, morphological and optical quality and properties of the Al x In 1−x N on Si (100) layers understudy was deeply analysed in reference [28]. In this report, our group demonstrated the successful deposition of high-quality and wurtzite single-phase n-type Al x In 1−x N layers with an Al mole fraction over the range of x = 0 to x = 0.56 on p-Si (100) substrates by RF magnetron sputtering. High-resolution X-ray diffraction measurements revealed that the layers own wurtzite single-phase crystalline structure oriented along the c-axis perpendicular to the sample surface. 2θ/ω scans performed in these layers show the diffraction peaks associated to the Si (100) and the AlInN (0002) reflections revealing that there is no phase separation in the AlInN alloy along the studied range of Al power supply and that no other parasitic reflections related to other phases were detected. On the other hand, the full width at half maximum of the rocking curve around the Al x In 1−x N (0002) diffraction peak decreased from~9 • to~3 • when incorporating Al to the Al x In 1−x N alloy. The root mean square surface roughness of the samples, estimated from atomic force microscopy, evolved from 20 nm for InN to 1.5 nm for Al 0.56 In 0.44 N films. For Al mole fractions equal and below 0.35, samples showed strong low-and room-temperature photoluminescence emission, which blue-shifted from 1.59 eV for InN to 1.82 eV for Al 0.35 In 0.65 N samples at 11 K. Taking into account these results, the developed n-Al x In 1−x N/p-Si junctions presented promising material properties to explore their performance operating as solar cells.
To develop efficient devices, it is of crucial importance to proper control the n-type doping. For this aim, Hall effect measurements were carried out in Al x In 1−x N films deposited on sapphire at the same time and under the same growth conditions to explore the evolution of their electrical characteristics, such as carrier concentration, layer resistivity, and carrier mobility, depending on the alloy mole fraction. As plotted in Figure 2, the resistivity of the III-nitride layers increases four orders of magnitude from 0.2 mΩ·cm to 3.4 Ω·cm for the InN (D1) to the Al 0.45 In 0.55 N samples (D5), respectively. We would like to point out that AlN material usually owns a resistivity higher than 10 14 Ω·cm [29]. In this sense, the residual n-type carrier concentration decreases by almost three orders of magnitude from 7.5 × 10 21 cm −3 to 1.6 × 10 19 cm −3 , and the carrier mobility decreases from 4.2 cm 2 /V·s to 0.1 cm 2 /V·s within the same Al mole fraction range. The responsible for the high carrier concentration of these layers are the unintentional doping coming from hydrogen and/or oxygen impurities incorporated during the sputtering deposition, and point defects like nitrogen vacancies, as previously reported [30][31][32].
Materials 2020, 12, x FOR PEER REVIEW 5 of 11 magnitude from 7.5 × 10 21 cm −3 to 1.6 × 10 19 cm −3 , and the carrier mobility decreases from 4.2 cm 2 /V·s to 0.1 cm 2 /V·s within the same Al mole fraction range. The responsible for the high carrier concentration of these layers are the unintentional doping coming from hydrogen and/or oxygen impurities incorporated during the sputtering deposition, and point defects like nitrogen vacancies, as previously reported [30][31][32].    Figure 3 shows the optical transmittance spectra measured in Al x In 1−x N samples deposited simultaneously on sapphire under the same deposition conditions vs the Al mole fraction x. The optical absorption band-edge blue shifts as expected with the Al content. However, transmittance measurements of layers with thickness below 100 nm never reach zero values at short wavelengths. This effect can lead to an overestimation of the estimated bandgap energy.
To overcome this, we estimated the bandgap energy from transmittance spectra measured in thick Al x In 1−x N on sapphire samples (thicknesses above 300 nm) with an Al mole fraction similar to those of the samples of this paper grown at 300 • C by RF sputtering [25]. The method to estimate the bandgap energy from transmittance measurements is described in Ref. [25]. Table 1 shows the comparison between the optical bandgap energy estimated from both set of samples, the ones of this paper (thin ones) and the ones of Ref. [25] (thick ones). From these data, we can see how the bandgap energy E g increases from 1.73 eV to 2.56 eV according to the rise in Al mole fraction from x = 0 to 0.56, as summarized in Table 1.
It has to be noted that the high values of E g are related to the so-called Burstein-Moss effect, i.e., the band filling of the conduction band and the blue shift of the bandgap energy of the material due to an increase of the unintentional doping (residual carrier concentrations close to 10 21 cm −3 ) caused by impurities like oxygen and defects like nitrogen vacancies. Figure 3 shows the optical transmittance spectra measured in AlxIn1−xN samples deposited simultaneously on sapphire under the same deposition conditions vs the Al mole fraction x. The optical absorption band-edge blue shifts as expected with the Al content. However, transmittance measurements of layers with thickness below 100 nm never reach zero values at short wavelengths. This effect can lead to an overestimation of the estimated bandgap energy. To overcome this, we estimated the bandgap energy from transmittance spectra measured in thick AlxIn1−xN on sapphire samples (thicknesses above 300 nm) with an Al mole fraction similar to those of the samples of this paper grown at 300 °C by RF sputtering [25]. The method to estimate the Current density vs voltage (J-V) curves of the processed devices were measured in the dark to analyse the electrical behaviour of the diodes. Figure 4 illustrates the J-V curves of the fabricated n-p junctions with structures D1-D7 under dark conditions pointing to a rectifying behaviour. The reverse saturation current density (J 0 ), series and shunt resistances (R s , R sh ) and the diode ideality factor (η) were estimated fitting experimental data to the expression: , and V T is the thermal voltage given by V T = k B · T/q (26 meV at room temperature).
Materials 2020, 12, x FOR PEER REVIEW 6 of 11 bandgap energy from transmittance measurements is described in Ref. [25]. Table 1 shows the comparison between the optical bandgap energy estimated from both set of samples, the ones of this paper (thin ones) and the ones of Ref. [25] (thick ones). From these data, we can see how the bandgap energy Eg increases from 1.73 eV to 2.56 eV according to the rise in Al mole fraction from x = 0 to 0.56, as summarized in Table 1. It has to be noted that the high values of Eg are related to the so-called Burstein-Moss effect, i.e., the band filling of the conduction band and the blue shift of the bandgap energy of the material due to an increase of the unintentional doping (residual carrier concentrations close to 10 21 cm −3 ) caused by impurities like oxygen and defects like nitrogen vacancies. Current density vs voltage (J-V) curves of the processed devices were measured in the dark to analyse the electrical behaviour of the diodes. Figure 4 illustrates the J-V curves of the fabricated n-p junctions with structures D1-D7 under dark conditions pointing to a rectifying behaviour. The reverse saturation current density (J0), series and shunt resistances (Rs, Rsh) and the diode ideality factor (η) were estimated fitting experimental data to the expression: J = J0 · [exp (Vd / (η · VT)) − 1] + Vd / Rsh, where Vd is the diode voltage (Vd = V -J · Rs), and VT is the thermal voltage given by VT = kB · T / q (26 meV at room temperature). From the results summed up in Table II, we deduce that Rs increases from ~4.7 Ω·cm 2 for InNuntil ~60.3 Ω·cm 2 for AlxIn1−xN-based devices (x = 0.56), probably due to the high-top n-contact resistivity present at high Al mole fractions (see Figure 1c). The gradual increase of Rsh of more than two orders of magnitude (Rsh ~ 12-3100 kΩ·cm 2 ) and the drop of the reverse saturation current density (J0 ~ 3.6−0.15 mA/cm 2 ) with the Al mole fraction are assigned to the improved material quality of the AlxIn1−xN layers compared to their InN counterparts, as previously demonstrated by our group [28]. At the same time, the ideality factor η drops from ~6.0 for devices with an Al mole fraction of 0.28 to ~3.0 for higher ones. From the results summed up in Table 2, we deduce that R s increases from~4.7 Ω·cm 2 for InN-until~60.3 Ω·cm 2 for Al x In 1−x N-based devices (x = 0.56), probably due to the high-top n-contact resistivity present at high Al mole fractions (see Figure 1c). The gradual increase of R sh of more than two orders of magnitude (R sh~1 2-3100 kΩ·cm 2 ) and the drop of the reverse saturation current density (J 0~3 .6−0.15 mA/cm 2 ) with the Al mole fraction are assigned to the improved material quality of the Al x In 1−x N layers compared to their InN counterparts, as previously demonstrated by our group [28]. At the same time, the ideality factor η drops from~6.0 for devices with an Al mole fraction of 0.28 tõ 3.0 for higher ones.  Figure 5 presents the J-V measurements of all devices under AM1.5G illumination (1 sun) carried out to evaluate their photovoltaic performance in terms of open-circuit voltage (V oc ), short-circuit current density (J sc ), fill factor (FF), and power conversion efficiency (Eff.). J sc was calculated taking into account the net area of the devices without the shadow of the top contact fingers of~0.13 cm 2 .    The analysis of the photovoltaic parameters Voc, Jsc, FF and efficiency extracted from the J-V curves of Figure 5 is summarized in Table 2 as a function of the Al mole fraction x. This analysis reveals that the open-circuit voltage, the short-circuit current density and the FF increase from 0.36 V to 0.39 V, from 13.4 mA/cm 2 to 17.1 mA/cm 2 and from 30.7% to 31.4% for InN and Al0.28In0.72N-based devices, respectively. This improvement leads to a maximum conversion efficiency of the AlxIn1−xN on Si devices of 2.2% with 28% of Al content.
However, the high serial resistance obtained for devices growth under an Al power supply of 100 W (D2) leads to a low open-circuit voltage and short-circuit current values. This effect may be attributed to some alloy disorder on the nitride/Si interface that can be found at low Al power supplies and degrades the electrical behaviour of the junction [33].
When increasing the Al mole fraction above 28% the Voc, Jsc, and efficiency drop down to 0.29 V, 0.5 mA/cm 2 , and 0.05% for the Al0.56In044N-based device, respectively. This trend is contrary as expected from theory when assuming the positive influence of increasing the Al mole fraction of the alloy, hence the bandgap energy, on the open-circuit voltage and the short-circuit current of the device, as previously reported [27]. So, the drop of the photovoltaic performance of the junction for Al contents above ∼30% is attributed to the deterioration of the electrical properties of the AlxIn1−xN The analysis of the photovoltaic parameters V oc , J sc , FF and efficiency extracted from the J-V curves of Figure 5 is summarized in Table 2 as a function of the Al mole fraction x. This analysis reveals that the open-circuit voltage, the short-circuit current density and the FF increase from 0.36 V to 0.39 V, from 13.4 mA/cm 2 to 17.1 mA/cm 2 and from 30.7% to 31.4% for InN and Al 0.28 In 0.72 N-based devices, respectively. This improvement leads to a maximum conversion efficiency of the Al x In 1−x N on Si devices of 2.2% with 28% of Al content.
However, the high serial resistance obtained for devices growth under an Al power supply of 100 W (D2) leads to a low open-circuit voltage and short-circuit current values. This effect may be attributed to some alloy disorder on the nitride/Si interface that can be found at low Al power supplies and degrades the electrical behaviour of the junction [33].
When increasing the Al mole fraction above 28% the V oc , J sc , and efficiency drop down to 0.29 V, 0.5 mA/cm 2 , and 0.05% for the Al 0.56 In 044 N-based device, respectively. This trend is contrary as expected from theory when assuming the positive influence of increasing the Al mole fraction of the alloy, hence the bandgap energy, on the open-circuit voltage and the short-circuit current of the device, as previously reported [27]. So, the drop of the photovoltaic performance of the junction for Al contents above~30% is attributed to the deterioration of the electrical properties of the Al x In 1−x N layer (resistivity increases and carrier concentration and mobility decrease with the Al content) that increases the serial resistance of the junction and the top contact resistance of the device.
To explore the theoretical limits of performance of these n-p junctions without the above-mentioned deterioration, we have performed simulations of the J-V curves under AM1.5G illumination (1 sun) of our devices with the PC1D program [34] following the procedure described in the paper [27]. First of all, we have simulated the InN-based device to have a starting point for the next simulations of the devices with increasingly Al mole fraction. As plotted in Figure 6, we have fitted the simulated J-V curve of the InN-based device to the experimental one, taking into account the thickness of the InN layer obtained by X-ray reflection measurements [28] and the n-type carrier concentration of the InN layer of Figure 2. the thickness of the InN layer obtained by X-ray reflection measurements [28] and the n-type carrier concentration of the InN layer of Figure 2.
Afterwards, to study the evolution of the photovoltaic performance of the devices through their J-V curves with the Al mole fraction we have kept the thickness, doping and top and bottom contact resistances fixed to the values used for the simulation of the InN (see legend of Figure 6), while we have changed the following simulation parameters with the Al mole fraction: bandgap energy, electron affinity, refractive index, dielectric constant, intrinsic concentration, and Nc/Nv ratio. We note that the PC1D program offers the possibility to take into account the experimental absorption of the layers, so in our case, we have used the transmittance data measured in AlxIn1−xN on sapphire samples grown at 300 °C with thicknesses above 300 nm and with Al mole fraction similar to those of this study [25]. From these simulations and under these particular conditions of nitride layer thickness and n-doping concentration, we can conclude that theoretically the photovoltaic performance of the AlxIn1−xN on Si (100) devices would improve with the Al content as theoretically expected (higher Voc, Jsc and conversion efficiency), achieving a maximum of Voc = 0.58 V, Jsc = 23.9 mA/cm 2 and efficiency of 4.2% for n-AlxIn1−xN on p-Si devices with 50% Al.  Afterwards, to study the evolution of the photovoltaic performance of the devices through their J-V curves with the Al mole fraction we have kept the thickness, doping and top and bottom contact resistances fixed to the values used for the simulation of the InN (see legend of Figure 6), while we have changed the following simulation parameters with the Al mole fraction: bandgap energy, electron affinity, refractive index, dielectric constant, intrinsic concentration, and Nc/Nv ratio. We note that the PC1D program offers the possibility to take into account the experimental absorption of the layers, so in our case, we have used the transmittance data measured in Al x In 1−x N on sapphire samples grown at 300 • C with thicknesses above 300 nm and with Al mole fraction similar to those of this study [25]. From these simulations and under these particular conditions of nitride layer thickness and n-doping concentration, we can conclude that theoretically the photovoltaic performance of the Al x In 1−x N on Si (100) devices would improve with the Al content as theoretically expected (higher V oc , J sc and conversion efficiency), achieving a maximum of V oc = 0.58 V, J sc = 23.9 mA/cm 2 and efficiency of 4.2% for n-Al x In 1−x N on p-Si devices with 50% Al.
To investigate the influence of the impinging wavelength on the photocurrent of the devices, responsivity measurements were carried out in the 375-1100 nm wavelength range. The external quantum efficiency (EQE) stands for the fraction of collected electron-hole pairs per incident photon and was estimated from these measurements as EQE = (J op /P op )·(hc/qλ), where J op is the photocurrent density, P op the optical power density impinging the device, q the electron charge, h is Planck's constant, c is the speed of light and λ the wavelength of the incident light. The spectral evolution of the EQE curves of the devices is displayed in Figure 7. For Al mole fractions below 0.35, the device EQE presents a broad spectral response covering the full solar spectrum from the visible to the near-infrared range with a peak in the visible above the AlxIn1−xN bandgap energy followed by a photoresponse in the near-infrared with a cut-off ~1100 nm related to the Si band edge. Moreover, the band edge at short wavelengths follows the AlxIn1−xN absorption shift with the Al mole fraction as plotted the normalized EQE in the inset of Figure 7. However, for Al mole fractions above 0.35, the EQE drops accordingly with J-V results of Figure  7 without observing any contribution coming from the Si layer. This lessening of the photovoltaic performance of the devices is attributed to the increase of the AlxIn1−xN resistivity of three orders of magnitude that lead to deterioration of the top contact performance. These results are very promising for AlxIn1−xN on Si structures working as solar cells. Nevertheless, the degradation of the photovoltaic performance for devices with Al contents above 35% need to be overcome. For this aim, new strategies should be incorporated, such as doping the AlxIn1−xN layer with Si and/or introducing a thin highly n-doped InN layer on top of the junction to improve the contact resistance.

Conclusions
Photovoltaic devices based on n-AlxIn1−xN on p-Si (100) hetero-junction solar cells with Al content ranging from 0% to 56% were fabricated by RF sputtering. The variation on the Al content leads to an evolution of the optical bandgap edge from 1.73 to 2.56 eV. Increasing the Al leads also to a 7-fold increase of the top n-contact resistance due to the higher AlxIn1-xN resistivity and the drop of the n-type carrier concentration. Best results achieved for solar cells with x = 0.28 are Voc = 0.39 V, Jsc = 17.1 mA/cm 2 , and FF = 31.4% under AM1.5G equivalent illumination (1 sun), pointing to a conversion efficiency of 2.12% without any top/back coating or surface treatment. For Al contents above 28 %, the photovoltaic performance is gradually reduced due to the strong top contact resistance, agreeing with the EQE of the devices, which drastically drops. Theoretical simulations of these structures point to a maximum of Voc = 0.58 V, Jsc = 23.9 mA/cm 2 , and efficiency of 4.2% for n-AlxIn1−xN on p-Si devices with 50% Al and a III-nitride resistivity in the ≈10 −4 Ω·cm range.  For Al mole fractions below 0.35, the device EQE presents a broad spectral response covering the full solar spectrum from the visible to the near-infrared range with a peak in the visible above the Al x In 1−x N bandgap energy followed by a photoresponse in the near-infrared with a cut-off~1100 nm related to the Si band edge. Moreover, the band edge at short wavelengths follows the Al x In 1−x N absorption shift with the Al mole fraction as plotted the normalized EQE in the inset of Figure 7.
However, for Al mole fractions above 0.35, the EQE drops accordingly with J-V results of Figure 7 without observing any contribution coming from the Si layer. This lessening of the photovoltaic performance of the devices is attributed to the increase of the Al x In 1−x N resistivity of three orders of magnitude that lead to deterioration of the top contact performance. These results are very promising for Al x In 1−x N on Si structures working as solar cells. Nevertheless, the degradation of the photovoltaic performance for devices with Al contents above 35% need to be overcome. For this aim, new strategies should be incorporated, such as doping the Al x In 1−x N layer with Si and/or introducing a thin highly n-doped InN layer on top of the junction to improve the contact resistance.

Conclusions
Photovoltaic devices based on n-Al x In 1−x N on p-Si (100) hetero-junction solar cells with Al content ranging from 0% to 56% were fabricated by RF sputtering. The variation on the Al content leads to an evolution of the optical bandgap edge from 1.73 to 2.56 eV. Increasing the Al leads also to a 7-fold increase of the top n-contact resistance due to the higher Al x In 1-x N resistivity and the drop of the n-type carrier concentration. Best results achieved for solar cells with x = 0.28 are V oc = 0.39 V, J sc = 17.1 mA/cm 2 , and FF = 31.4% under AM1.5G equivalent illumination (1 sun), pointing to a conversion efficiency of 2.12% without any top/back coating or surface treatment. For Al contents above 28 %, the photovoltaic performance is gradually reduced due to the strong top contact resistance, agreeing with the EQE of the devices, which drastically drops. Theoretical simulations of these structures point to a maximum of V oc = 0.58 V, J sc = 23.9 mA/cm 2 , and efficiency of 4.2% for n-Al x In 1−x N on p-Si devices with 50% Al and a III-nitride resistivity in the ≈10 −4 Ω·cm range.