In situ etching of β-Ga₂O₃ using tert-butyl chloride in an MOCVD system

The ultra-wide bandgap semiconductor β-Ga₂O₃ (∼4.8 eV) has garnered attention recently as a platform for power electronics and radio frequency devices.¹ The ultra-wide bandgap results in a high critical breakdown field strength yielding a superior Baliga's figure of merit relative to semiconductors like SiC and GaN.² Progress in β-Ga₂O₃ research has been spurred by the availability of large-area (up to 4 in.) melt-grown substrates³ and the ease of n-type doping.⁴ Metalorganic chemical vapor deposition (MOCVD) has emerged as a technique capable of producing high-quality β-Ga₂O₃ thin films with room-temperature electron mobilities approaching the polar optical phonon limit.^5–8 A low-damage in situ etch to minimize contamination or plasma-induced damage before subsequent deposition of n+ material⁹ or dielectrics¹⁰ will be key for enabling higher-performance devices. In this study, we investigate the use of tert-butyl chloride (TBCl) as a precursor for in situ etching of β-Ga₂O₃.

In situ etching of β-Ga₂O₃ has been demonstrated using a flux of elemental gallium in molecular beam epitaxy (MBE) and using triethylgallium in MOCVD.^11,12 The etch mechanism for both leverages the formation of volatile gallium suboxides.¹³ The use of elemental Ga can, however, potentially leave gallium metal droplets on the surface necessitating an ex situ HCl wet etch. Agnitron Technologies has demonstrated that these Ga droplets can be removed in situ with TBCl etching; however, their use of TBCl for etching of β-Ga₂O₃ itself was not promising, requiring much higher TBCl molar flow compared to those used in this work to achieve appreciable etch rates.¹⁴  In situ etching of β-Ga₂O₃ has also been demonstrated in halide vapor phase epitaxy (HVPE) systems using HCl gas.^15,16

UV-vis optical reflectometry was used to measure heteroepitaxial film thickness, while x-ray diffraction (XRD) (PANalytical Empyrean) was used to determine homoepitaxial film thickness. Homoepitaxial films grown for determining the etch rate included a thin (∼10 nm) β-(Al_0.07Ga_0.93)₂O₃ interface followed by 200–300 nm of β-Ga₂O₃ which provided an index contrast resulting in Laue oscillations.²⁵ Atomic force microscopy (AFM) was used to evaluate the surface morphology.

We used heteroepitaxial β-Ga₂O₃ to map out the etch rate as a function of TBCl molar flow at temperatures between 700 and 900 °C (Fig. 1). At a fixed reactor pressure of 15 Torr, we found that etch rate increases linearly with TBCl molar flows between ∼20 and 61 μmol/min, which enables fine control of the etch rate.

In general, the etch rate increases with increasing temperature; however, the slope of the etch rate vs TBCl molar flow jumps sharply from 800 to 850 °C, indicative of a sudden change in the etch-limiting step. At a fixed TBCl molar flow of ∼61 μmol/min, the etch rate follows an Arrhenius relationship (Fig. 2). There are two distinct activation energy regimes that are commonly observed in CVD growth²⁶ and etching²⁷ processes. In our work, the low-temperature (LT) regime below ∼800 °C has a higher apparent activation energy of ∼1.59 and ∼1.75 eV for 15 and 30 Torr, respectively. The high temperature (HT) regime above ∼800 °C exhibits a much lower apparent activation energy of ∼0.11 and ∼0.04 eV for 15 and 30 Torr, respectively. These values are much lower than the Ga-O bond dissociation energy (∼3.88 eV),²⁸ and we confirmed that even at the highest etch temperature employed, there is negligible thermal decomposition of β-Ga₂O₃.²⁹ 

During the etching process, the etchant adsorbs on the surface and then reacts to form an etch product, followed by the desorption of the etch product from the surface. Based on the data from Tsang,²⁴ at temperatures between 700 and 1000 °C used in our work and estimated residence times³⁰ for precursors in the heated boundary layer above the susceptor, TBCl pyrolyzes into isobutene and hydrogen chloride (HCl) (Fig. 3). This temperature range, however, is not high enough to enable further gas-phase pyrolysis of HCl (Fig. 3).³¹ Therefore, it is reasonable to assume that HCl is the etchant.

In the HT regime, the dominant GaCl_n species is presumably GaCl based on thermodynamic calculations of gas-phase HVPE growth of GaN.⁴⁴ The low apparent activation energy in the HT regime agrees with the apparent activation energy (∼0.08 eV) extracted above 800 °C from atmospheric pressure HCl etching of β-Ga₂O₃.¹⁵ The etch rate in the HT regime is determined by the surface concentration of HCl as evidenced by the linear dependence of the etch rate on reactor pressure in Fig. 1(b).⁴⁵ 

At high substrate temperatures, MOCVD growth typically occurs in a mass-transport-limited regime. In the absence of significant gas-phase parasitic reactions, and when all the flows are held constant, the growth rate is independent of total reactor pressure within this mass-transport-limited regime.^30,46 To determine whether etching with TBCl occurs in a mass-transport-regime, the pressure dependence of the etch rate was explored at a fixed TBCl flow rate at two temperatures: 750 °C (LT regime) and 900 °C (HT regime). The reactor pressure was controlled, independent of the total gas flow, using a computer-controlled butterfly valve in the exhaust manifold. Figures 1(b) and 1(c) show that the etch rate is increasing, and not saturating, with increasing reactor pressure, indicating that etching is not occurring in a mass-transport-limited regime even at the highest substrate temperatures investigated in this study. An increase in etch rate with reactor pressure is likely due to the increased surface coverage of the etchant, HCl, with increasing HCl partial pressure.⁴⁵ 

In order to investigate the effect of oxygen on the etch rate, we introduced a 50 sccm flow of O₂ into the reactor during etching. In the LT regime, the etch rate was largely not affected; however, in the HT regime, the etch rate was suppressed by a factor of ∼2 as seen in Fig. 4. We believe that the HT etch rate is suppressed due to the competing HVPE growth back reaction.

Next, we etched co-loaded heteroepitaxial $2¯01$ and homoepitaxial (010) β-Ga₂O₃ samples to investigate the etch rate anisotropy. The etch rate for homoepitaxial samples with ∼61 μmol/min TBCl molar flow was determined by XRD (Fig. S1) for four distinct conditions (15 Torr at 750 and 875 °C, and 30 Torr at 750 and 875 °C). Figure 5 summarizes these etch rates, revealing the anisotropy. Also plotted in Fig. 5 are the ratios between the calculated (010) and $2¯01$ dangling bond densities (ρ, green) and surface energies (E, blue), which agrees well with the experimental data.⁴⁷ Etch rate anisotropy has also been observed during etching of β-Ga₂O₃ using HCl in an HVPE reactor.¹⁵ 

To investigate the surface morphology resulting from TBCl etching, ∼400 nm thick homoepitaxial (010) unintentionally doped (UID) β-Ga₂O₃ samples were grown and then immediately in situ etched, without cooling down, to a depth of ∼100 nm for each of the four conditions shown in Figs. 6(b)–6(e). The resulting surface morphology resembles that resulting from hot phosphoric wet etching⁴⁸ and does not exhibit characteristic faceting of the (110) plane along the [001] direction typically seen post-growth⁴⁹ [as shown for an unetched film in Fig. S1(b)] or after elemental Ga etching.^11,12 In general, we observe that etching under higher pressures and low temperatures results in smoother surfaces. Currently, the exact mechanism for surface roughening is unclear, but we note that conditions for smoother etch morphologies also result in longer surface residence time of gas-phase species. We note that heteroepitaxial $2¯01$ oriented β-Ga₂O₃ films grown on c-plane sapphire also exhibited increased surface roughness after etching using TBCl (Fig. S2).

The electrical properties are not compromised for films grown after a 30 min ex situ 48% hydrofluoric acid etch plus a ∼50 nm in situ, ∼61 μmol TBCl etch at 750 °C. (010) homoepitaxial films doped⁵⁰ to ∼1 × 10¹⁷ and ∼2 × 10¹⁸ cm⁻³ exhibit mobilities of ∼115 and ∼102 cm²/V s, respectively. We demonstrate that subsequent regrowth after in situ etching of the substrate results in sub-nanometer RMS roughness (Fig. S3). The surface morphology of regrowth after in situ etching homoepitaxial films is comparable (∼1 nm RMS) and is the subject of future work.

In summary, this study investigated the in situ etching of both heteroepitaxial $2¯01$ and homoepitaxial (010) β-Ga₂O₃ films by TBCl in an MOCVD system over a temperature range of 700–1000 °C, pressure of 10–60 Torr, and TBCl molar flow of ∼20 to 61 μmol/min. Two distinct regimes for TBCl etching of β-Ga₂O₃ were observed. The LT regime, below ∼800 °C, exhibits an apparent activation energy of ∼1.59 and ∼1.75 eV for 15 and 30 Torr, respectively. In the LT regime, we hypothesize that the etch rate is limited by the desorption of GaCl_n or likely H₂O. In the HT regime, we hypothesize that the thermodynamically favored etch product is GaCl and the apparent activation energy is low. The relationship between the etch rate of $2¯01$ and (010) β-Ga₂O₃ scales by the ratio of surface energies. Finally, the surface morphology of in situ etched homoepitaxial films was evaluated, and it was determined that the lower temperature, higher pressure etch resulted in smoother surfaces. This work lays the groundwork for utilizing in situ TBCl etching and regrowth to obtain low-resistance ohmic contacts and improve the performance of β-Ga₂O₃ based devices.