Comparison of the Material Quality of AlxIn1−xN (x—0–0.50) Films Deposited on Si(100) and Si(111) at Low Temperature by Reactive RF Sputtering

AlxIn1−xN ternary semiconductors have attracted much interest for application in photovoltaic devices. Here, we compare the material quality of AlxIn1−xN layers deposited on Si with different crystallographic orientations, (100) and (111), via radio-frequency (RF) sputtering. To modulate their Al content, the Al RF power was varied from 0 to 225 W, whereas the In RF power and deposition temperature were fixed at 30 W and 300 °C, respectively. X-ray diffraction measurements reveal a c-axis-oriented wurtzite structure with no phase separation regardless of the Al content (x = 0–0.50), which increases with the Al power supply. The surface morphology of the AlxIn1−xN layers improves with increasing Al content (the root-mean-square roughness decreases from ≈12 to 2.5 nm), and it is similar for samples grown on both Si substrates. The amorphous layer (~2.5 nm thick) found at the interface with the substrates explains the weak influence of their orientation on the properties of the AlxIn1−xN films. Simultaneously grown AlxIn1−xN-on-sapphire samples point to a residual n-type carrier concentration in the 1020–1021 cm−3 range. The optical band gap energy of these layers evolves from 1.75 to 2.56 eV with the increase in the Al. PL measurements of AlxIn1−xN show a blue shift in the peak emission when adding the Al, as expected. We also observe an increase in the FWHM of the main peak and a decrease in the integrated emission with the Al content in room-temperature PL measurements. In general, the material quality of the AlxIn1-xN films on Si is similar for both crystallographic orientations.


Introduction
Al x In 1−x N ternary semiconductor alloys have attracted huge interest for their application in solar cells, particularly after the revision of the indium nitride (InN) band gap energy in 2001 [1]. The direct band gap (i.e., high absorption coefficient) of Al x In 1−x N, tunable from the near-infrared (0.7 eV for InN [1]) to the ultraviolet (6.2 eV for AlN [2]) range, makes it an excellent candidate for developing photovoltaic devices in combination with silicon. In addition, this material shows high resistance to thermal and mechanical stress and irradiation with high-energy particles [3], which makes it suitable for space applications.
There are several studies about the growth of Al x In 1-x N films on either Si(111) or Si(100) and its comparison with Al x In 1-x N on sapphire substrates. Bashir et al. [35] deposited InN on Si(111) by RF sputtering and obtained large crystallite size, low microstrain, and low dislocation density. Afzal et al. [22] grew Al x In 1−x N films on Si(111) at 300 • C using a magnetron cosputtering system and obtained polycrystalline films with the preferred orientation along the (101) direction, with higher crystallite size and lower surface roughness compared with other substrates such as GaAs and glass. However, the comparison of Al x In 1−x N layers simultaneously grown on both Si(111) and Si(100) substrates by reactive RF sputtering has never been reported so far.
This work presents the study of the properties of Al x In 1−x N layers with Al content ranging from 0% to 50% simultaneously deposited on silicon (100) and (111) via reactive RF sputtering at a relatively low substrate temperature, 300 • C. The layer characteristics in terms of structural, morphological, electrical, and optical properties are studied and compared considering both substrate orientations. Finally, we demonstrate the possibility of using silicon (100) as a feasible substrate for developing Al x In 1−x N layers for device applications by taking advantage of its compatibility with today's silicon-based nanotechnology.

Materials and Methods
Al x In 1-x N layers were simultaneously deposited in a reactive RF magnetron sputtering system (AJA International, ATC ORION-3-HV, Scituate, MA, USA) on three substrates: p-doped 375 µm thick Si(100), p-doped 500 µm thick Si(111) (both with a resistivity of 1-10 Ωcm), and on 500 µm thick (0001)-oriented sapphire. This system was equipped with a 2 inch confocal magnetron cathodes of pure In (4N5) and pure Al (5N). The base pressure of the system was in the order of 10 −7 mbar. The substrate-target distance was fixed at 10.5 cm, and the temperature during the deposition was monitored with a thermocouple placed in direct contact with the substrate holder. The substrates were chemically cleaned in organic solvents before being loaded in the chamber, where they were outgassed for 30 min at 550 • C and then cooled down to the growth temperature. Prior to the deposition, the surface of the targets and the substrates were cleaned using a soft plasma etching with Ar (2 sccm and 20 W), causing no damage to the surface. Al x In 1−x N layers were deposited in a pure N 2 atmosphere with a nitrogen flow of 14 sccm and a pressure of 0.47 Pa. The RF power applied to the Al target, P Al , was set to 0, 100, 150, 175, and 225 W (samples M1-M5, respectively), while the RF power applied to the In target and the temperature were fixed to 30 W and 300 • C, respectively. A sputtering time of 3 h was used for the InN sample, 5 h The alloy mole fraction, crystalline orientation, and mosaicity of the films were evaluated by high-resolution X-ray diffraction (HRXRD) measurements using a PANalytical X'Pert Pro MRD system (Malvern, UK). In addition, the thicknesses of the layers were obtained from field-emission scanning electron microscopy (FESEM) images. Atomic force microscopy (AFM) was employed to study the surface morphology and estimate the root-mean-square (rms) surface roughness using a Bruker multimode Nanoscope IIIA microscope in tapping mode (Billerica, MA, USA). Additionally, transmission electron microscopy (TEM) provided a deeper understanding of the structural properties of the interface between the deposited material and the substrate. The electrical properties of the films were analyzed using room-temperature Hall-effect measurements in a conventional Van der Paw geometry.
Finally, photoluminescence measurements were carried out at room temperature by exciting the samples with~20 mW of a continuous-wave laser diode emitting at λ = 405 nm focused on a 1 mm diameter spot. The emission was collected with a 193 mm focal-length Andor spectrograph equipped with a UV-extended silicon-based charge-coupled-device (CCD) camera operating at −65 • C between 200 and 1100 nm.

Structural Characterization
To study the structural quality of the layers, HRXRD 2θ/ω scans were carried out on the Al x In 1−x N layers grown on Si(100) and Si(111), with the results shown in Figure 1a,b, respectively. All layers presented a wurtzite crystalline structure highly oriented along the c-axis, and no other crystallographic phases were detected. The increase in P Al shifted the (0002) and (0004) reflection peaks assigned to Al x In 1−x N toward higher diffraction angles, which confirmed the reduction in the c lattice parameter. The Al mole fraction of the alloy was estimated by applying Vegard's Law [39] to the AlN-InN system, using the c lattice parameter obtained from HRXRD and assuming fully relaxed layers. The calculated Al mole fraction, x, scales linearly with P Al between x = 0 and x = 0.49 or 0.48, for Si(100) and Si(111) substrates, respectively, as summarized in Table 1.
The FWHM of the ω-scan (rocking curve) of the (0002) Al x In 1−x N diffraction peak provides information about the mosaicity of the material. In this study, Al x In 1−x N layers grown on both silicon substrates showed similar values, in the 3-6 • range, without a clear trend (Table 1). This indicated that the mosaicity is independent of the crystal orientation of the silicon substrate.
The FWHM of the ω-scan (rocking curve) of the (0002) AlxIn1−xN diffraction peak provides information about the mosaicity of the material. In this study, AlxIn1−xN layers grown on both silicon substrates showed similar values, in the 3-6° range, without a clear trend (Table 1). This indicated that the mosaicity is independent of the crystal orientation of the silicon substrate.

Morphological Characterization
In order to investigate the morphology of the layers, they were studied by FESEM and AFM techniques. Figure 2a-c show the FESEM images of samples grown on Si(100) and Si(111). The morphology of the layers evolved from nanocolumnar for pure InN (sample M1) toward grain-like compact when increasing the Al content (samples M3 and M5) for both substrate orientations. Such a trend was already observed in similar AlxIn1−xN samples deposited on Si(111) by RF sputtering (40 W In, 300 °C) with similar Al compositions [31]. The observed phenomena could be attributed to changes in the surface diffusion of adatoms due to the increased kinetic energy of the incoming Al species, which can determine the layer morphology for both substrate orientations. The observed morphological transition was accompanied by a modification of the sample surface roughness. The rms roughness was measured by AFM images scanned in a 2  2 μm area ( Figure 3). The results showed a surface roughness evolution from 11.5 (Al content x = 0) to 2.5 nm (x ≈ 0.36) for Si(100) and from 13.0 (Al content x = 0) to 2.5 nm (x ≈ 0.36) for Si(111), as summarized in Table 1, and in agreement with previously published results [31]. The roughness remained almost constant for samples with an Al content in the range within x ≈ 0.36-0.42 (see Table 1), and it finally dropped up to ≈2.5 nm

Morphological Characterization
In order to investigate the morphology of the layers, they were studied by FESEM and AFM techniques. trend (Table 1). This indicated that the mosaicity is independent of the crystal orie of the silicon substrate.

Morphological Characterization
In order to investigate the morphology of the layers, they were studied by and AFM techniques. Figure 2a-c show the FESEM images of samples grown on and Si(111). The morphology of the layers evolved from nanocolumnar for pure InN ple M1) toward grain-like compact when increasing the Al content (samples M3 a for both substrate orientations. Such a trend was already observed in similar A samples deposited on Si(111) by RF sputtering (40 W In, 300 °C) with similar Al co tions [31]. The observed phenomena could be attributed to changes in the surfac sion of adatoms due to the increased kinetic energy of the incoming Al species, wh determine the layer morphology for both substrate orientations. The observed morphological transition was accompanied by a modification sample surface roughness. The rms roughness was measured by AFM images sca a 2  2 μm area ( Figure 3). The results showed a surface roughness evolution fro (Al content x = 0) to 2.5 nm (x ≈ 0.36) for Si(100) and from 13.0 (Al content x = 0) to (x ≈ 0.36) for Si(111), as summarized in Table 1, and in agreement with previous lished results [31]. The roughness remained almost constant for samples with an tent in the range within x ≈ 0.36-0.42 (see Table 1), and it finally dropped up to ≈ The observed morphological transition was accompanied by a modification of the sample surface roughness. The rms roughness was measured by AFM images scanned in a 2 × 2 µm area ( Figure 3). The results showed a surface roughness evolution from 11.5 (Al content x = 0) to 2.5 nm (x ≈ 0.36) for Si(100) and from 13.0 (Al content x = 0) to 2.5 nm (x ≈ 0.36) for Si(111), as summarized in Table 1, and in agreement with previously published results [31]. The roughness remained almost constant for samples with an Al content in the range within x ≈ 0.36-0.42 (see Table 1), and it finally dropped up to ≈2.5 nm for an Al content of ≈50%. This surface roughness reduction was attributed to an increase in the adatom energy and mobility when increasing P Al , in agreement with results obtained in similar Al x In 1-x N-on-Si(100) samples deposited at a higher temperature (550 • C) [27], where the surface roughness was 2.0 and 1.5 nm for x ≈ 0.35 and x ≈ 0.56, respectively.
for an Al content of ≈50%. This surface roughness reduction was attribut in the adatom energy and mobility when increasing PAl, in agreement tained in similar AlxIn1-xN-on-Si(100) samples deposited at a higher tem [27], where the surface roughness was 2.0 and 1.5 nm for x ≈ 0.35 and x ≈ 0 The interface between the AlxIn1−xN and the silicon substrate was s mission electron microscopy (TEM) measurements. Figure 4 shows th TEM images of an AlxIn1−xN (x ≈ 0.36) layer deposited on Si(100) and Si the epitaxial growth along the c-axis for the two silicon orientations. I images reveal the formation of an amorphous layer of ~2.5 nm at the laye face (see the inset of both figures), which may have weakened the inte phases and reduced the influence of the silicon orientation on the qual layer deposited on top. The similar structural quality obtained growing o was also confirmed by the comparable grain size estimated from STEM 5a,b). Thus, the structural quality is conserved even when grown on a although a clearer boundary between the amorphous interfacial layer an was observed in this case.  The interface between the Al x In 1−x N and the silicon substrate was studied by transmission electron microscopy (TEM) measurements. Figure 4 shows the cross-sectional TEM images of an Al x In 1−x N (x ≈ 0.36) layer deposited on Si(100) and Si(111), evidencing the epitaxial growth along the c-axis for the two silicon orientations. In both cases, the images reveal the formation of an amorphous layer of~2.5 nm at the layer/substrate interface (see the inset of both figures), which may have weakened the interactions between phases and reduced the influence of the silicon orientation on the quality of the nitride layer deposited on top. The similar structural quality obtained growing on both substrates was also confirmed by the comparable grain size estimated from STEM images (Figure 5a,b). Thus, the structural quality is conserved even when grown on a cubic substrate, although a clearer boundary between the amorphous interfacial layer and the nitride one was observed in this case.
for an Al content of ≈50%. This surface roughness reduction was attributed to an in in the adatom energy and mobility when increasing PAl, in agreement with resul tained in similar AlxIn1-xN-on-Si(100) samples deposited at a higher temperature (5 [27], where the surface roughness was 2.0 and 1.5 nm for x ≈ 0.35 and x ≈ 0.56, respect The interface between the AlxIn1−xN and the silicon substrate was studied by mission electron microscopy (TEM) measurements. Figure 4 shows the cross-sec TEM images of an AlxIn1−xN (x ≈ 0.36) layer deposited on Si(100) and Si(111), evide the epitaxial growth along the c-axis for the two silicon orientations. In both case images reveal the formation of an amorphous layer of ~2.5 nm at the layer/substrate face (see the inset of both figures), which may have weakened the interactions be phases and reduced the influence of the silicon orientation on the quality of the n layer deposited on top. The similar structural quality obtained growing on both subs was also confirmed by the comparable grain size estimated from STEM images (F 5a,b). Thus, the structural quality is conserved even when grown on a cubic sub although a clearer boundary between the amorphous interfacial layer and the nitrid was observed in this case.

Electrical Characterization
The electrical properties of the AlxIn1-xN layers could only be addressed for s deposited on sapphire substrates, because the silicon conduction masked the layer nal whenever a silicon substrate was used. The values of resistivity, carrier concen and mobility were obtained for simultaneously grown layers with an Al conten 0.32. Samples with higher Al content showed a resistivity above 10 mΩ·cm, mak Hall effect measurement unreliable.
The layer resistivity increased from 0.38 mΩ·cm for InN to 8 mΩ·cm for Al0.32 while the carrier concentration decreased from 1.73 × 10 21 cm −3 for InN to 2.48 × 1 for Al0.32In0.68N. On the other hand, the values of mobility showed no clear trend, s with a value of 9.5 cm 2 /V.s for InN and decreasing to 3.2 cm 2 /V.s for Al0.32In0.68N peak of 11.5 cm 2 /V.s for Al0.14In0.86N. The values of resistivity and mobility obtained Al0.32In0.68N sample are similar to those reported by Liu et al. [38] (1.2 mΩ·cm an cm 2 /V·s, respectively, for a ~90 nm Al0.28In0.72N layer deposited by RF sputtering at 6 The high carrier concentration of the layers is related to the unintentional dopin impurities such as hydrogen or oxygen during growth [32], and it was also obser Nuñez-Cascajero et al. [26], where similar AlxIn1−xN on sapphire with homogeneo tribution of oxygen were obtained.

Optical Characterization
The apparent optical band gap energy of the samples deposited on sapphire timated through room-temperature optical transmittance measurements follow procedure described in Ref. [26] (See Table 2 for all optical data). Figure 6 sho squared absorption used for this estimation, obtained from the transmittance spec picted in the inset of the figure for each sample.

Electrical Characterization
The electrical properties of the Al x In 1-x N layers could only be addressed for samples deposited on sapphire substrates, because the silicon conduction masked the layered signal whenever a silicon substrate was used. The values of resistivity, carrier concentration, and mobility were obtained for simultaneously grown layers with an Al content up to 0.32. Samples with higher Al content showed a resistivity above 10 mΩ·cm, making the Hall effect measurement unreliable.
The layer resistivity increased from 0.38 mΩ·cm for InN to 8 mΩ·cm for Al 0.32 In 0.68 N, while the carrier concentration decreased from 1.73 × 10 21 cm −3 for InN to 2.48 × 10 20 cm −3 for Al 0.32 In 0.68 N. On the other hand, the values of mobility showed no clear trend, starting with a value of 9.5 cm 2 /V.s for InN and decreasing to 3.2 cm 2 /V.s for Al 0.32 In 0.68 N with a peak of 11.5 cm 2 /V.s for Al 0.14 In 0.86 N. The values of resistivity and mobility obtained for the Al 0.32 In 0.68 N sample are similar to those reported by Liu et al. [38] (1.2 mΩ·cm and 11.4 cm 2 /V·s, respectively, for a~90 nm Al 0.28 In 0.72 N layer deposited by RF sputtering at 600 • C). The high carrier concentration of the layers is related to the unintentional doping from impurities such as hydrogen or oxygen during growth [32], and it was also observed by Nuñez-Cascajero et al. [26], where similar Al x In 1−x N on sapphire with homogeneous distribution of oxygen were obtained.

Optical Characterization
The apparent optical band gap energy of the samples deposited on sapphire was estimated through room-temperature optical transmittance measurements following the procedure described in Ref. [26] (See Table 2 for all optical data). Figure 6 shows the squared absorption used for this estimation, obtained from the transmittance spectra depicted in the inset of the figure for each sample. timated through room-temperature optical transmittance measu procedure described in Ref. [26] (See Table 2 for all optical dat squared absorption used for this estimation, obtained from the tra picted in the inset of the figure for each sample.   Figure 7 shows the low-(11 K) and room-temperature (300K) PL emission of samples M1 (InN) and M2 (Al x In 1−x N, x-0.12, 0.16), grown on Si(100) and Si(111). No PL emission was observed for Al x In 1−x N layers with higher Al content than 16%, independent of the crystal orientation of the substrate. The results obtained from the analysis of the PL measurements in terms of the main peak emission energy, FWHM, and integrated intensity are summarized in Table 3.   Figure 7 shows the low-(11 K) and room-temperature (300K) PL emission of s M1 (InN) and M2 (AlxIn1−xN, x-0.12, 0.16), grown on Si(100) and Si(111). No PL e was observed for AlxIn1−xN layers with higher Al content than 16%, independen crystal orientation of the substrate. The results obtained from the analysis of the P urements in terms of the main peak emission energy, FWHM, and integrated inten summarized in Table 3.

Temperature
Main Peak Emission FWHM 2 Integrated Int  The dominant room-temperature emission energy centered at ≈1.60 and ≈1.80 eV for the M1 (InN) and M2 (Al x In 1−x N, x-0.12, 0.16) samples deposited on both silicon substrates, respectively. The position of the emission energy practically stayed the same, while the intensity decreased when increasing the temperature from 11 to 300 K, as expected. However, the presence of an emission at room temperature was a clear indication of the good crystalline quality of the samples. The FWHM of the PL emission of the samples was similar for both types of substrates, being slightly higher for sample M2, probably due to the alloy disorder present in the Al x In 1−x N layer.
Then, assuming that the band gap energy was similar for the samples grown on Si and sapphire, we could extract an approximate value for the Stokes shift as the difference between the band gap energy obtained from transmission measurements (Eg Abs ) and the PL emission energy (E PL ) at 300 K. The obtained Stokes shift was around~130 and~60 meV for InN (M1) and Al x In 1−x N (M2), respectively. These values pointed to a reduced band tail for the Al x In 1−x N samples compared with the InN ones, which could be related to the change in layer morphology (and probably the surrounding of the involved emission centers) when introducing aluminum into the InN binary.
Lastly, comparing each sample on both substrates, they showed a very similar emission shape and integrated intensity, even though the Al x In 1-x N ones had double the layer thickness compared with their InN counterparts. This result pointed to an enhancement of the nonradiative recombination channels due to Al incorporation, which could increase the lattice disorder and defects.

Conclusions
Al x In 1−x N films with low-to-mid Al content (x-0-0.50) were deposited via RF sputtering on different substrates, i.e., Si(100) and Si(111), for their comparison. The increase in the Al mole fraction improved the structural and morphological quality of the layers, achieving a minimum FWHM of the (0002) Al x In 1−x N rocking curve of~2.8 • and a minimum rms surface roughness of~2.5 nm for samples grown on both Si substrates with x-0.49. FESEM images showed a morphological transition from nanocolumnar toward a grain-like compact morphology when aluminum was introduced. Cross-sectional TEM images revealed a~2.5 nm thick amorphous layer in the interface between the nitride material and the substrate, which could be responsible for the weak coupling between the active layer and the substrate. This finding allows the development of Al x In 1−x N with similar material quality on both silicon substrate orientations.
Hall-effect measurements revealed a carrier concentration above 10 20 cm −3 for the Al x In 1-x N layers with x < 0.32, probably induced by the unintentional doping of the material during deposition. Additionally, the Al x In 1-x N layers (x ≤ 0.16) deposited on both Si substrate orientations exhibited similar PL emission in terms of shape, energy, FWHM, and integrated intensity at room temperature, showing a reduction in the PL emission efficiency when introducing the Al compared with the one obtained for InN layers.
In this work, we demonstrated the ability to produce high-quality Al x In 1−x N layers on Si with low-to-mid Al content via RF sputtering regardless of the chosen substrate orientation.