$α$-(Al_xGa_1−x)₂O₃ single-layer and heterostructure buffers for the growth of conductive Sn-doped $α$-Ga₂O₃ thin films via mist chemical vapor deposition

Recently, the growth of (Al_xGa_1−x)₂O₃/Ga₂O₃ heterostructures has attracted large attention of the research community for the next-generation ultra-wide bandgap high power devices.^1–4 Very high-quality β-(Al_xGa_1−x)₂O₃/Ga₂O₃ heterostructures have been grown by plasma-assisted molecular beam epitaxy (PAMBE).¹ Based on this work, β-(Al_xGa_1−x)₂O₃/Ga₂O₃ modulation-doped field-effect transistors have also been demonstrated.^5–7 On the other hand, a high Al content x can be realized in the α-phase (Al_xGa_1−x)₂O₃ because α-(Al_xGa_1−x)₂O₃ crystalizes in the same corundum structure as the sapphire substrate.^8–10 Previously, mist chemical vapor deposition (mist CVD) has demonstrated its capability in growing complicated multi-layer corundum-type heterostructures.^11–13 It has also been shown that the dislocation density, which was typically high in undoped α-Ga₂O₃,^14–16 could be reduced by the use of a complicated α-(Al_xGa_1−x)₂O₃ heterostructure buffer grown via mist CVD,¹¹ epitaxial lateral growth via halide vapor phase epitaxy (HVPE),^14,15 or the growth of thick films.¹⁶ However, the effect of (Al_xGa_1−x)₂O₃ buffers with different Al-contents x on the properties of subsequently grown conductive Sn-doped α-Ga₂O₃ (Sn:α-Ga₂O₃), in particular, on the dislocation density and conductivity, has not been systematically studied.

In this work, we grew six single-layer and two heterostructure buffers using a third generation mist CVD system. The properties of these buffers as well as the Sn:α-Ga₂O₃ thin films grown on them were systematically studied by x-ray diffraction (XRD) and transmission electron microscopy (TEM). PtO_x and AgO_x Schottky diodes (SDs) were fabricated on the conductive thin films, and their electrical characteristics were examined. The Sn:α-Ga₂O₃ thin film grown on a thin strained heterostructure buffer showed the best characteristics overall.

Six single-layer buffers were grown on c-plane sapphire substrates using a so-called third generation mist CVD system (3rd G mist CVD system), which consists of several solution chambers, a mist mixing chamber, and a fine-channel reactor.^8,17,18 The Al/Ga flow rates were 0/5, 1/4, 2/3, 3/2, 4/1, and 4.5 l/min/0.5 l/min. The 3rd G mist CVD system enables the growth of complicated multi-layer heterostructures in a single run. In this work, two heterostructures were grown as the buffers that consisted of six layers, whose thicknesses were aimed at 20 nm and 100 nm. The Al content of each layer was decreased step-by-step by changing the Al/Ga flow rates from 4.5/0.5 to 0/5, which were the same flow rates used during the growth of single-layer buffers. The Sn:α-Ga₂O₃ films were subsequently deposited for 15 min on the buffers using a one-chamber system.¹⁹ The schematic layouts of the samples are shown in Fig. 1. The growth of the (Al_xGa_1−x)₂O₃ buffers and conductive Sn:α-Ga₂O₃ thin films is detailed in Table I. The samples are hereafter referred to as the 00/50, 10/40, 20/30, 30/20, 40/10, and 45/05 single-layers and 20-nm and 100-nm heterostructures. Out-of-plane (0006) and in-plane (10$1¯$4) XRD measurements were performed with Rigaku ATX and SmartLab systems, respectively, on all samples, whereas reciprocal space maps were performed on the 20-nm and 100-nm heterostructures after the deposition of Sn:α-Ga₂O₃ with the Rigaku SmartLab system. TEM samples were prepared using a focused ion beam (FIB, FEI QUANTA 3D 200 i) system on the 00/50 and 40/10 single-layer and 100-nm heterostructure samples. TEM images were taken with a JEOL JEM-2100F system. PtO_x and AgO_x SDs were fabricated on the conductive Sn:α-Ga₂O₃ films using reactive RF sputtering.^20–25 The SDs were characterized by a Keysight B1506A Power Device Analyzer.

Figure 2 plots the (0006) reflection XRD spectra measured on the samples before (dashed lines) and after (solid lines) the deposition of the Sn:α-Ga₂O₃ films. In the single-layer buffers, the XRD (0006) reflection shifted to the high angle direction with the increase in the Al flow rate indicating higher Al incorporation. Using Vegard’s law,⁸ the Al content x was calculated to be 0, 0.09, 0.21, 0.36, 0.57, and 0.66 for the 00/50, 10/40, 20/30, 30/20, 40/10, and 45/05 buffers, whose thicknesses were 95 nm, 118 nm, 127 nm, 117 nm, 106 nm, and 111 nm, respectively, as calculated from the Laue fringes. In the 100-nm heterostructure buffer, the (0006) reflections from the 45/05, 40/10, 30/20, and 20/30 layers can be seen, whereas those from the 10/40 and 00/50 layers seemed to merge into each other. In the 20-nm heterostructure buffer, in addition to the merging of the (0006) reflections from the 10/40 and 00/50 layers, those from the 40/10 and 45/05 layers were also not clearly discernible.

Figures 3(a)–3(d) summarize Al-content-dependent full widths at half maximum (FWHMs) of the rocking curves (RCs) of the single-layer [(a) and (c)] and heterostructure [(b) and (d)] samples measured in the tilt [(0006) reflection, (a) and (b)] and twist [(10$1¯$4) reflection, (c) and (d)] configurations. The (10$1¯$4) reflection 2θ/θ, Δω, and ϕ scans of the Sn:α-Ga₂O₃ thin films are shown in Figs. S1(a)–S1(c) of the supplementary material. In Fig. 3(a), black squares and red circles represent the FWHMs of the Sn:α-Ga₂O₃ films and the single-layer buffers only (without Sn:α-Ga₂O₃), respectively. In the tilt configuration, as x of the single-layer buffers exceeded 0.4, the FWHMs of the Sn:α-Ga₂O₃ films abruptly increased, even though those of the corresponding buffers without Sn:α-Ga₂O₃ were rather similar. This indicates that the tilting of the mosaic blocks in the Sn:α-Ga₂O₃ films became significantly more severe at x > 0.4. The 100-nm heterostructure buffer caused more tilting to the mosaic blocks in the Sn:α-Ga₂O₃ film in comparison with the 20-nm one, as evident in the FWHMs shown in Fig. 3(b). The inferior tilting property of the mosaic blocks may be attributed to the (i) large lattice mismatches between Sn:α-Ga₂O₃ and high Al content α-(Al_xGa_1−x)₂O₃ single-layer buffers and (ii) higher roughness typically observed in thick layers grown via mist CVD (the 100-nm heterostructure in this case). Unevenness in rough buffers might induce more tilting. In the twist configuration, while the FWHMs gradually increased with x as a general trend, the FWHM of the 100-nm heterostructure sample was less than that of the 20-nm one, in sharp contrast with the tilting behavior. Apparently, while the lattice mismatch might still induce more twisting in the Sn:α-Ga₂O₃ films grown on the single-layer buffers, the unevenness, in contrast, might reduce the twisting. With low FWHMs in both tilt and twist configurations, the 20-nm heterostructure buffer produced a Sn:α-Ga₂O₃ film of best crystallinity overall.

Figures 4(a) and 4(b) display the reciprocal space maps (RSMs) of the 20-nm and 100-nm heterostructures with the Sn:α-Ga₂O₃ layer on top. The RSMs were measured around the (10$1¯$(10)) reflections of the α-(Al_xGa_1−x)₂O₃ layers (0 ≤ x ≤ 1), which appear as peaks and halos in the maps. Although five intermediate α-(Al_xGa_1−x)₂O₃ (0 < x < 1) layers were deposited, only three and four halos appeared in the RSMs of the 20-nm and 100-nm heterostructure samples, respectively. This was in agreement with the 2θ/θ plots shown in Fig. 2. In the 20-nm heterostructure, the halos originating from the α-(Al_xGa_1−x)₂O₃ (0 < x < 1) films lied below the straight line that connects peaks originating from the Sn:α-Ga₂O₃ film and α-Al₂O₃ sapphire substrates [see Fig. 4(a)]. Clearly, strain remained in thin layers of the 20-nm heterostructure. This may be caused by the out-of-plane tensile strain along the c axis²⁶ and/or the in-plane compressive strain along a and b axes.²⁷ On the contrary, all the peaks and halos lied almost on a straight line in the 100-nm heterostructure sample except for the halo originating from the 45/05 layer. The straight line indicates the linearity of the dependence of a and c lattice constants on the Al content x, which again confirms the validity of Vegard’s law in the α-(Al_xGa_1−x)₂O₃ system reported previously⁸ but in the condition that the layers were stacked on each other. The slight deviation of the 45/05 halo from the straight line suggests that the film did not fully relax; possibly, partial strain²⁶ remained in the layer.

Figures 5(a)–5(f) show weak beam dark field (WBDF) [(a)–(c)] and selected area electron diffraction (SAED) [(d)–(f)] images of the 00/50 [(a) and (d)], 40/10 [(b) and (e)] single-layer, and 100-nm heterostructure [(c) and (f)] buffers with Sn:α-Ga₂O₃ thin films on top. As shown in Figs. 5(d)–5(f), the reflections g used in the two-beam condition were (0006), (10$1¯$(10)), and (1$1¯$08) for the 00/50, 40/10 single-layer, and 100-nm heterostructure samples, respectively. The corresponding tilting axes were close to the ⟨1$3¯$20⟩, ⟨(⁠$10¯$⁠)461⟩, and ⟨$1¯1¯$20⟩ zone axes. Threading dislocations were clearly visible in the WBDF images, and they appeared as long streaks. The densities of the dislocations in the Sn:α-Ga₂O₃ layers on the 00/50, 40/10 single-layer, and 100-nm heterostructure buffers were estimated^11,28 to be ∼7 × 10¹⁰, ∼7 × 10¹⁰, and 6 × 10¹⁰ cm⁻², respectively. Clearly, the use of thick α-(Al_xGa_1−x)₂O₃ buffers did not significantly reduce the dislocation densities. Thicknesses of the layers in the 100-nm heterostructure buffer were 73 nm, 121 nm, 116 nm, 134 nm, 116 nm, and 112 nm in the order from the top to the bottom. The last two layers at the bottom had significantly fewer dislocations than the others. A large number of dislocations started to appear at the 30/20 layer and propagated up to the Sn:α-Ga₂O₃ films. Interestingly, the top ∼10 nm to 30 nm layer of the Sn:α-Ga₂O₃ films was riddled with dot-like defects instead of streaks as observed in the 00/50 single-layer and 100-nm heterostructure samples [see the insets of Figs. 5(a) and 5(c)]. Apparently, not all threading dislocations propagated up to the surface of the Sn:α-Ga₂O₃ films.

Figures 6(a) and 6(b) plot the J–V characteristics of PtO_x/and AgO_x/Sn:α-Ga₂O₃ SDs fabricated on the samples, while Figs. 6(c) and 6(d) present the breakdown characteristics of the diodes under reverse bias. Representative fits for the J–V curves are shown in Figs. S2(a)–S2(f) of the supplementary material. Note that the diodes were fabricated on 6 × 6 mm² pieces cut from the same 12 × 12 mm² substrates for each sample. The diodes properly rectified with low reverse bias currents and yielded the highest rectification ratios of 5 × 10⁷ and 2 × 10⁷ for PtO_x and AgO_x SDs fabricated on the strained 20-nm heterostructure sample, respectively. The PtO_x SDs had a much smaller hysteresis than the AgO_x ones. While the Sn:α-Ga₂O₃ films grown on the 00/50 buffers (from the 00/50 single-layer, the 20-nm, or 100-nm heterostructure buffers) produced consistently high forward currents, some non-uniformity was observed in the Sn:α-Ga₂O₃ films grown on other buffers. This was also evident in the conductivity measured from transmission line model (TLM) structures fabricated on the PtO_x and AgO_x chips, which is shown in Figs. 7(a) and 7(b). In addition, low forward currents and conductivities were observed in the Sn:α-Ga₂O₃ films grown on buffers with the Al content in the middle range (x = 0.21, 0.36, and 0.57). There were no clear correlations between the forward currents of the diodes (or the conductivities) and the FWHMs obtained from the XRD RC FWHMs (Fig. 3). Apparently, the overall crystallinity of the whole Sn:α-Ga₂O₃ thin films did not strongly influence the conductivity, which might be associated only with the dot-like defect region observed in the TEM images. As shown in Figs. 7(c) and 7(d), the dependences of the breakdown voltages across the samples are rather similar in both PtO_x and AgO_x. Both dependences show a dip at the 40/10 sample and high breakdown voltages at the 30/20 and 45/05 samples. This indicates that the breakdown voltages depended mainly on the Sn:α-Ga₂O₃ layer, not on the nature of the Schottky contacts, i.e., PtO_x or AgO_x. The breakdown voltages did not depend on the conductivity either.

Figure 8 plots the dependence of the effective barrier height Φ_eff on the ideality factor n derived from the J–V characteristics.^21,29 As can be seen, the effective barrier height vs ideality factor data could be fitted by linear dependences. Previously, the linear dependence was explained by the inhomogeneity of Schottky barrier heights.³⁰ Most likely, the inhomogeneity was also responsible for the large ideality factors typically observed in the diodes fabricated on α-Ga₂O₃ thin films.^23,31,32 The extrapolation of the linear fits to the unity ideality factor yielded the average (or sometimes called “homogeneous”) barrier heights of 1.60 eV and 1.62 eV for PtO_x and AgO_x SDs, respectively. The fact that these values were almost equal regardless of the contacting materials suggests that the Fermi level was strongly pinned³⁰ at the level of ∼1.6 eV below the conduction band in mist-CVD-grown Sn:α-Ga₂O₃. This level might be linked to the 3.6 eV photoluminescence peak observed in heavily Sn-doped α-Ga₂O₃ reported previously.³¹ 

Overall, the Sn:α-Ga₂O₃ thin film deposited on the 20-nm heterostructure buffer produced best characteristics, such as the highest on–off ratios and barrier heights, lowest ideality factors (in both PtO_x and AgO_x SDs), a high breakdown voltage (in the PtO_x SD), consistently high conductivity, and low RC FWHMs.

In conclusion, a third generation mist CVD system was used to grow six single-layer and two heterostructure α-(Al_xGa_1−x)₂O₃ buffers for the subsequent deposition of conductive Sn-doped α-Ga₂O₃ thin films. The Al contents x increased from 0 to 0.66 in the six single-layer buffers. The heterostructure buffers consisted of six layers laying on top of each other, which were grown using the same recipe as the six single-layer buffers. The thicknesses of each layer in the two heterostructures were ∼20 nm and 100 nm. For Sn:α-Ga₂O₃ films grown on single-layer buffers, the FWHM of the (0006) XRD RCs abruptly increased as x exceeded 0.4, indicating the severe tilting of the mosaic blocks. Nevertheless, mono-crystallinity was maintained in all grown layers including those in complicated 20-nm and 100-nm heterostructure buffers. Strain was observed in the 20-nm heterostructure, while the layers in the 100-nm heterostructure almost fully relaxed and Vegard’s law was followed even when the α-(Al_xGa_1−x)₂O₃ thin films were stacked on each other. Transmission electron microscopy analyses show that the threading dislocation densities remained high in the order of 10¹⁰ cm⁻² despite the employment of the buffers. PtO_x and AgO_x SDs were fabricated on the Sn:α-Ga₂O₃ films. The Schottky barrier height vs ideality factor plots could be fitted by linear dependences, indicating that the large ideality factors observed in α-Ga₂O₃ SDs were due to the large inhomogeneity of the Schottky contacts. The extrapolation of the dependences of the PtO_x and AgO_x SDs yielded homogeneous Schottky barrier heights of ∼1.60 eV and 1.62 eV, respectively, suggesting that the Fermi level was strongly pinned at the ∼E_c − 1.6 eV level. Overall, the Sn:α-Ga₂O₃ film grown on the strained 20-nm heterostructure showed best characteristics, such as a low RC FWHM, consistently high conductivity, high Schottky barrier heights and on–off ratios, low ideality factors, and high breakdown voltage in the PtO_x SD.

See the supplementary material for 2θ/θ, Δω, and ϕ scans of the (10$1¯$4) reflection measured in the samples (Fig. S1) and representative J–V characteristics with the corresponding fits of PtO_x and AgO_x Schottky contacts (Fig. S2).

The data that support the findings of this study are available from the corresponding author upon reasonable request.

We would like to thank Professor H. Kohno and Professor N. Nitta for technical advices during TEM observation. G.T.D. would like to thank Professor M. W. Allen and Dr. A. M. Hyland for the technical support during the deposition of PtO_x contacts. This work was supported by the JSPS KAKENHI, Grant Nos. JP16F16373, JP15H05421, and JP18H01873.