(AlxGa1-x )2O3 is an ultra-wide bandgap semiconductor that is attracting attention for use in high power transistors and deep ultraviolet (DUV) photodetectors (1)(2)(3)(4) . To this point, most of the work has focused on the β-polytype with a bandgap tunability ranging from 4.8 eV (Ga2O3) to 8.7 eV (Al2O3) (1,2) . The advantages of β-(AlxGa1-x)2O3 grown on β-Ga2O3 include the high channel mobility at the interface and ability to use the heterostructure in field effect transistor structures with improved mobility, as well as Schottky diodes and photodetectors (1,2) . A drawback of this materials system during growth with techniques such as Metal Organic Chemical Vapor Deposition is the difficulty in maintain phase stability at high Al contents (1) .
The α-polytype is a potential solution to extending the stability to higher Al contents and higher temperatures (2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12) . This metastable, corundum crystal structure α-polytype, transforms to the βpolymorph at temperatures above 750-900°C under equilibrium conditions (5) . Of all reported polymorphs, this phase has the highest optical bandgap with energies ranging between 5.0 eV and 5.3 eV for the binary. A number of different growth methods have recently been shown to produce high quality rhombohedral α-AlxGa1-x)2O3 over a broad composition range and in particular, the use of Al2O3 cap layers during annealing is able to preserve the α-phase (16) . The α-polytype has an even larger bandgap than the β-polytype and can be grown on low-cost c-plane sapphire substrates which have the same lattice structure, albeit with relatively high dislocation densities (2,19,22,43) . Figure 1 shows the crystal structure of rhombohedral α-Ga2O3. This rhombohedral phase can also occur for Al2O3 and In2O3, which allows a potential bandgap engineering over the entire composition range (34,45,46) , rendering this polymorph interesting for deep UV photodetectors or quantum-well infrared photodetectors (1,2) , in addition to the power device applications (35)(36)(37) .
It is clearly of interest to establish the band alignments for commonly employed dielectrics on this materials system. SiO2 is one of the most promising choices, due to its large bandgap and maturity as a dielectric in semiconductor devices. We have previously measured the band alignment and its thermal stability for SiO2 and Al2O on β-AlxGa1-x)2O3 for 0.2< x <0.65 (51,52) . In this paper we report on the measurement of valence band offsets in the SiO2/α-AlxGa1x)2O3 system using alloys grown by Pulsed Laser Deposition (PLD) (49,50) over a wide composition range. This was done using circular half-segmented (Ga2O3/Al2O3) ceramic targets to synthesize large diameter large thin films with a lateral varying cation composition (1,7,8,45,46) .
Since the bandgap of the α-polytype alloys is even larger than those of the β-polytype alloys, this system represents a test in terms of finding dielectrics that exhibit acceptable band offsets (> 1eV is a common rule of thumb) in both the conduction and valence bands.
The α-AlxGa1-x)2O3 alloys were grown by combinatorial PLD using a KrF excimer laser (Coherent LPX Pro 305, 248 nm, 2.6 J.cm -2 ) directed through a quartz window onto a segmented ceramic target (Ga2O3/Al2O3) under an O2 partial pressure (1,49,50) . The substrate and target were mounted rotatable opposite each other at a distance of 10 cm and with a small lateral offset to each other (46) . PLD has been utilized previously to obtain AGO alloys over a broad composition range for both αand β-polytypes (8,20,40) The resulting cation ratio was determined by energy dispersive X-ray spectroscopy (EDX) measurements performed using a FEI Nova Nanolab 200 equipped with an Ametek EDAX detector and the crystal structure by X-ray diffraction (XRD). measured by a PANalytical X'pert PRO MRD diffractometer equipped with a PIXcel3D detector operating in 1D scanning line mode with 255 channels (for 2Ɵ-ω scans), receiving slit mode (for Φ-scans) and fast 2D frame based mode (for reciprocal space maps). The cation gradient ranged between x= 0.13 and x=0.84 as measured by EDX and marked as black dots in Figure 2(a). The composition between these data points was interpolated and the resulting cation ratio represented as false color map. Along the cation gradient direction, marked with a black arrow, additional EDX measurements were performed in 1mm steps to obtain the spatial Al dependence with high lateral resolution as shown in Figure 2(b) (20) . The Al incorporation covers a range of 0.13 ≤ x ≤ 0.84. The rhombohedral crystal structure of the thin films was confirmed by the 2Ɵ-ω XRD scans. The resulting false color map is presented in Figure 3 and reveals the crystallization in the rhombohedral phase in the entire investigated Al range. We examined four compositions, x=0.26, 0.42, 0.58 and 0.74. More details and discussion on the characterization of these samples can be found in Hassa et al. (20) . The SiO2 layers were deposited by Atomic Layer Deposition (ALD) at 200°C using the thermal mode in a Cambridge Nano Fiji 200. The precursors were Tris (dimethylamino) silane and an inductively coupled plasma (ICP) of O2 at 300 W (51,52) . After solvent cleaning, the substrates were exposed to Ozone for 15 min to form a protective oxide that is easily thermally desorbed during the ALD step. Both thick (200 nm) and thin (1.5 nm) layers of SiO2 were deposited for measuring bandgaps and core levels on the α-(AlxGa1-x)2O3.
XPS survey scans were performed to measure the chemical state of the SiO2 and (AlxGa1x)2O3 layers in a Physical Instruments ULVAC PHI, with an Al x-ray source (energy 1486.6 eV, source power 300W) (51)(52)(53) . The analysis size was 100 µm diameter, with a take-off angle of 50° and acceptance angle of ±7 degrees. The electron pass energy was 23.5 eV for high-resolution scans and 93.5 eV for survey scans. The total energy resolution of this XPS system is 0.5 eV, and the accuracy of the observed binding energy is ~ 0.03 eV. Charge compensation was performed using an electron flood gun and simultaneous ion beam. The C 1s core levels of the surface adsorbate (284.8 eV) were used to calibrate the binding energy (53) . Only the relative energy position is needed to determine the valence band offsets, so the absolute energy calibration has no effect on that offset. The samples were electrically insulated from the chuck to avoid uneven charge dispersion along the sample. All electron analyzers and equipment were grounded. Differential charging was not observed in any of the samples with the use of the electron gun (51,52) . The SiO2 bandgap was obtained from Reflection Electron Energy Loss Spectroscopy (REELS) using a 1 kV electron beam and hemispherical electron analyzer. The bandgaps of the (AlxGa1-x)2O3 for each composition were obtained from XPS energy loss measurements of the O1S peak. This is done simultaneously with the band alignment measurements (52,53) .
Figure 4 shows the high resolution XPS spectra for the α-(AlxGa1-x)2O3 to SiO2 vacuumcore delta regions of the four compositions, while the data for the SiO2 is shown in Figure 5. The valence band offsets were extracted from the shift of the core levels for the heterostructure samples with the thin dielectric on top of the four different compositions of the alloy. A compilation of the valence band maxima (VBM) are collected in Table 1, with values of (2.7±0.15) eV for x=0.26, (3.3±0.15) eV for x=0.42, (4.4±0.15)eV for x=0.58 and (4.5±0.15) eV for x=0.74. These were obtained by fitting of the leading edge of the valence band. The error bars in the different binding energies were combined in a root sum square relationship to determine the overall error bars in the valence band offsets (ΔEV) (51) . These band offsets are then obtained by measuring the shift of the core levels in the α-(AlxGa1-x)2O3 when SiO2 was deposited. These are also tabulated in Table 1, with values of -1.2 eV for x=0.26, -0.2 eV for x=0.42, 0.2 eV for x=0.58 and 0.4 eV for x=0.74.
Once the valence band offsets are established, to determine the conduction band offset, it is necessary to measure the band gap of each composition. We measured the bandgaps of the four (AlxGa1-x)2O3 compositions, as shown in Figure 6 (a), from the separation between the core level peak energy and the onset of inelastic (plasmon) losses in each O 1s photoemission spectra (36) . To find the band-gap energy, a linear fit is made to the measured loss spectra curve near the approximate location of onset of inelastic losses. Next, by subtracting the background, the "zero" level is determined. The energy corresponding to the onset of inelastic losses is found by extrapolating the linear-fit line and calculating its intersection with the "zero" level (33,40) . The bandgap energy is equal to the difference between the core-level peak energy and the onset of inelastic losses.
The bandgap energies were determined as 5.8 eV for x=0.26, 6.1 eV for x=0.42, 6.4 eV for x=0.58 and 7 eV for x= 0.74, respectively. Figure 7 shows the relationship between bandgap of the alloys and Al composition for our samples. As previously reported, the bandgap as a function of composition, x, is given by (28,54) 𝐸𝐸 𝐺𝐺 (𝑥𝑥) = (1 -𝑥𝑥)𝐸𝐸 𝐺𝐺 𝐺𝐺𝐺𝐺𝐺𝐺 + 𝑥𝑥𝐸𝐸 𝐺𝐺 𝐴𝐴𝐴𝐴𝐺𝐺 -𝑏𝑏𝑥𝑥 (1 -𝑥𝑥)
where EG GaO and EG AlO are the bandgaps of the α-polytype binary endpoints and b is the bowing parameter. This has been previously reported in the range 0.84-2.13 (9,34) . We get a value closer to 3 for our samples in the composition range measured. It was noted by Varley (53) that the b values are strongly influenced by the selective occupation of different cation sublattice by Al and that a single bowing parameter is not accurate over the entire composition range
The SiO2 bandgap was obtained from Reflection Electron Energy Loss Spectroscopy (REELS) and showed a value of 8.7 eV, similar to previous reports (51,52) .
The conduction band offsets ΔEC were obtained from the relation
The corresponding conduction band offsets were 4.1 eV (x=0.26), 2.8 eV (x=0.46), 2.1 eV (x=0.58) and 1.30 eV (x=0.74).
Figure 7 shows the band diagrams for the SiO2/α(AlxGa1-x)2O3 heterostructure. The SiO2 has adequate conduction band offsets, but the valence band offsets are small and actually negative at
The valence band offsets of SiO2/ α-(