Selective and nonselective plasma etching of (Al_0.18Ga_0.82)₂O₃/ Ga₂O₃ heterostructures

The ternary alloy (Al_xGa_1−x)₂O₃ demonstrates utility in modulation-doped transistors featuring high-electron-density channels formed at the interface with Ga₂O_3.^1–12 Additionally, in its polycrystalline form, it serves as a transparent conducting oxide.^13,14 The alloy has also gained considerable prominence in the domains of optoelectronics and transparent electronics owing to its noteworthy electrical conductivity and optical transparency. Its potential application extends to serving as a deep ultraviolet (UV) transparent film for optoelectronic devices.

In terms of fabricating High-Electron-Mobility Transistors (HEMTs) based on the (Al_xGa_1−x)₂O₃/ Ga₂O₃ heterostructure, one of the key steps is selective removal of a Ga₂O₃-doped contact layer from the underlying (Al_0.18Ga_0.82)₂O₃ donor layer, allowing deposition of rectifying contact.⁹ This step needs high selectivity removal of Ga₂O₃ and low damage to the (Al_xGa_1−x)₂O₃ layer. Complicating this is the fact that the Al composition in the ternary layer is generally relatively low, in the range of 0.17–0.22.^1,3,5–7,12 For manufacturing purposes, a minimum selectivity of at least 10 is desirable.⁹ The predominant methodology for the (Al_xGa_1−x)₂O₃ /Ga₂O₃ patterning involves dry etching due to the inherent absence of wet etching capabilities at room temperature, as documented in various studies.^15–24 The extent of ion-induced damage under practical etching conditions has been reported previously, with alterations in the electrical properties of the etched surface observed to extend beyond the spatial range of incident ions.^25–27 This phenomenon is commonly attributed to the rapid diffusion of primary defects, such as vacancies or interstitials.^25–28 

Differential etching rates or selectivity between two distinct layers arise from variations in the volatility of their respective etch byproducts or the development of an etch-stop material on the surface.²⁶ A representative instance is the utilization of Cl₂/SF₆ for the selective etching of GaAs in preference to AlGaAs^26,29,30 and also GaN over AlGaN.³¹ In this scenario, selectivity is established through the generation of nonvolatile AlF₃ on the AlGaAs or AlGaN surfaces subsequent to the etching of GaAs or GaN, respectively. Analogously, a comparable outcome can be achieved by introducing oxygen into a chlorine plasma to generate AlOx species on the AlGaAs or AlGaN surfaces.^26,30,31 The selectivity mechanism is predominantly governed by chemical processes, demonstrating a pronounced sensitivity to incident ion energy and flux. High-density plasmas typically exhibit lower selectivities than their reactive ion etching (RIE) counterparts.²⁶ Conventional etch-stop reactions, such as the formation of AlF₃, may be less effective, potentially resulting in premature passivation. In the case of achieving selective etching of Ga₂O₃ over (Al_xGa_1−x)₂O₃ at low Al contents, the strategy is to add F₂ or O₂ to the plasma chemistry to form nonvolatile etch products. In this work, we have chosen a source of F to avoid deleterious oxidation effects on the (Al_xGa_1−x)₂O₃ donor layer. Fluorine incorporation into Si-doped Al-containing semiconductors causes reduction in carrier density, and this effect has been used to control the threshold voltage in transistors through compensating Si donors with the negatively charged F ions.^32–36 The fluorine can be incorporated during exposure to wet chemical solutions such as HF, by direct implantation of low-energy fluorine ions or by immersion in a fluorine-containing plasma, including in Ga₂O₃.³⁵ Avoiding this F-induced donor compensation is important for maintaining the electrical characteristics of Ga₂O₃.

0.3 μm thick, silicon-doped, n-type (4.1 × 10¹⁹ cm⁻³) (Al_0.18Ga_0.82)₂O₃ layers were employed in the conducted experiments. These specimens were grown via Metal Organic Vapor Phase Epitaxy within an Agnitron Technology Agilis 100 MOVPE reactor, utilizing insulating (010) β-Ga₂O₃: Fe substrates. The layers were synthesized at approximately 600 °C on an unintentionally doped Ga₂O₃ buffer layer with a thickness of approximately 150 nm. The precursors employed in the process included triethylgallium, trimethylaluminum, O₂, with SiH₄ serving as the dopant gas and Ar as the carrier gas. A detailed description of the growth methodology has been provided previously.^2,37 The epitaxial structure is depicted in the schematic representation of Fig. 1. The β-Ga₂O₃ samples for etch rate measurements were (100) bulk, Sn-doped substrates, grown by the Edge-Fed Defined Growth method and purchased from Novel Crystal Technology (Saitama, Japan).

The etching was conducted using a PlasmaTherm 790 Inductively Coupled Plasma reactor. Etching was performed under the following conditions: discharges of 15 SCCM of Cl₂ and 5 SCCM of Ar, sometimes with the addition of 5 SCCM CHF₃ at a fixed pressure of 5 mTorr. The ICP source power and rf chuck power were systematically varied to manipulate plasma density and ion energy, respectively, thereby influencing the etch rates. The etching was nominally performed at room temperature with the samples connected to the water-cooled platen with thermal grease. Given the short etch times, we do not expect a significant temperature rise and there was no evidence of thermal degradation of the resist masks. Etch rates were determined by assessing the etch depth using a Tencor profilometer after the removal of the photoresist. Notably, brief plasma exposures lasting approximately 1–2 min did not result in significant surface roughening. The samples were lithographically pattered using AZ 4620 photoresist with a thickness of ∼1 μm. All of the etch processes were done at room temperature. Etch time was 2 min. Etch depth and surface root mean square (RMS) roughness of the processed samples were measured with a surface profilometer (Tencor alpha-step IQ).

The x-ray photoelectron spectroscopy (XPS) system utilized in this study was a Physical Instruments ULVAC PHI instrument, equipped with an aluminum (Al) x-ray source characterized by an energy of 1486.6 electron volts (eV) and a source power of 300 W. The analysis parameters included a spot size of 100 μm in diameter, a take-off angle of 50°, and an acceptance angle of ±7°. For high-resolution scans, the electron pass energy was set at 23.5 eV, while for survey scans, it was set at 93.5 eV. The overall energy resolution of the XPS system was approximately 0.5 eV, and the observed binding energy exhibited an accuracy within 0.03 Ev.

Figure 2(a) shows the etch rates of (Al_0.18Ga_0.82)₂O₃ and Ga₂O₃ in 15 SCCM Cl₂/5 SCCM Ar discharges as a function of ICP power at a fixed RF source power of 50 W. Figure 2(b) shows the etch rates as a function of RF source power at a fixed ICP power of 400 W. The DC self-bias is indicated in both cases. This parameter increases with rf power due to increased ionization of the gas in the plasma, leading to a higher density of charged particles.²⁶ This increased charge density can contribute to a higher self-bias voltage. By contrast, increasing ICP source power typically leads to a higher electron density in the plasma. Higher electron density means more efficient ionization of the gas in the plasma. Increased ionization can result in a higher density of positive ions in the plasma, which can reduce the sheath potential and, consequently, the DC self-bias voltage.²⁶ The self-bias voltage refers to the DC voltage that develops across the sheath of a plasma discharge. The sheath is a thin layer of plasma near the electrode where the electric field is concentrated. The resultant average ion energy is the sum of the dc self-bias and the plasma sheath potential.²⁵ The latter is ∼23 V under the conditions used in the experiments.²⁵ Therefore, the ion energies range from ∼73 to 523 eV over the range of conditions we examined. There is little selectivity between the etch rates of Ga₂O₃ and (Al_0.18Ga_0.82)₂O₃ under any conditions in the Cl₂/Ar discharges.

Figure 3 shows the resultant selectivities for dry etching of Ga₂O₃ over (Al_0.18Ga_0.82)₂O₃ in Cl₂/ Ar discharges as a function of either RF power or ICP source power. These are in the range 0.5–1.3 and are well short of practical values.

The effect of addition of CHF₃ to promote an etch stop reaction is shown in Fig. 4, where the etch rates are given of (a) (Al_0.18Ga_0.82)₂O₃ and (b) Ga₂O₃ in 15 SCCM Cl₂/ 5 SCCM Ar/ 5 SCCM CHF₃ discharges as a function of ICP power at a fixed RF source power of 50 W. The rates of the alloy decrease significantly with the addition of trifluoromethane for this low rf power condition. By contrast, the etch rate of the Ga₂O₃ increases with this addition, possibly because methyl-based products are formed that add to the removal rate of GaCl_x products, particularly at high ICP source powers.

The resultant etch selectivities for dry etching of Ga₂O₃ over (Al_0.18Ga_0.82)₂O₃ in Cl₂/Ar/CHF₃ discharges are shown in Fig. 5 as a function of ICP power at a fixed rf power of 50W. Very high selectivities >30 are obtained at moderate ICP source powers, with a maximum value of 55. The incident ion energies are ∼100 eV under these conditions, making them ideal for low damage removal of a Ga₂O₃ contact layer to expose the underlying ternary for deposition of the gate contact. We only did limited experiments with varying the CHF₃ ratio at a fixed rf or ICP power, but the selectivity was already increased to values >30 even at low CHF₃ additions. This is not surprising given the low total flow rate of all the gases. We did not increase the CHF_3­ flow rate above 5 SCCM to avoid excessive polymerization within the chamber.

It is also worth noting that additional OES data to understand the chemistry of the plasma would be useful, although these chemistries are common in selective dielectric etching in semiconductor technology, and balancing the etch and deposition is crucial.

Figure 6 shows XPS data from the ternary alloy before and after exposure to either Cl₂/Ar or Cl₂/5 Ar/CHF₃ discharges. There are clearly F residues remaining on the etched surface in the latter case, which promotes the etch stop reaction. The formation of AlF₃ is always the dominant factor in obtaining selective etching of the Ga-dominant alloy relative to the Al-containing alloy.^26,29–31 F can be removed by standard RCA-type wet cleaning. There is no residual polymer on the surface from the XPS data, with the carbon bonded to itself or oxygen. In addition, the formation of GaF₃ would retard the etch rate of Ga₂O₃ by essentially the same amount as for the AGO. Our experience with producing selective etching in the GaAs/AlGaAs and GaN/AlGaN systems by addition of F-containing gases (usually SF₆) suggests that it is reasonable to infer the same mechanism as proven in those systems.^29–31 Note also that under these plasma conditions, there is significant dissociation of atomic fluorine from CHF₃.³⁸ This tool shows the usual series of F atomic lines from 623.9 to 775.4 nm, far more prominent than the CF_X lines in the 203–388.3 range. The XPS is not consistent with polymer deposition. C is only bonded to itself and oxygen.

Finally, we note there was no change in the sheet carrier density of the films after the selective dry etch process, indicating no significant incorporation of F ions into the material.

High dry etch selectivities are achieved for Ga₂O₃ over (Al_0.18Ga_0.82)₂O₃ in Cl₂/Ar/CHF₃ discharges at low rf electrode powers and moderate ICP source powers. The CHF₃ addition is the key to enhancing the selectivity. Under most conditions with Cl₂/Ar plasma chemistries, the etching is nonselective. These results provide a simple ICP chemistry for selective and nonselective patterning of Ga₂O₃/(Al_0.18Ga_0.82)₂O₃ heterostructures. Achieving high selectivity would become easier at higher Al contents. In addition, another route that could be explored is the use of other C-free, fluorine sources such as SF₆, NF₃, or other gases. These provide F concentrations without the addition of carbon. There was no additional roughening of the surface for the selective etch condition. Similarly, under our pressure and gas flow rate conditions, polymer deposition within the chamber was negligible for the duration of the experiments.

This research was performed as part of the Interaction of Ionizing Radiation with Matter University Research Alliance (IIRM-URA), sponsored by the Department of the Defense, Defense Threat Reduction Agency under Award No. HDTRA1-20-2-0002. The content of the information does not necessarily reflect the position or the policy of the federal government, and no official endorsement should be inferred. This work at UF was also supported by NSF DMR No. 1856662.

The authors have no conflicts to disclose.

Hsiao-Hsuan Wan: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Validation (equal); Visualization (equal); Writing – original draft (equal). Chao-Ching Chiang: Data curation (equal); Investigation (equal); Methodology (equal); Validation (equal). Jian-Sian Li: Conceptualization (equal); Data curation (equal). Fan Ren: Conceptualization (equal); Funding acquisition (equal); Project administration (equal); Resources (equal); Supervision (equal); Writing – review & editing (equal). Fikadu Alema: Conceptualization (equal); Formal analysis (equal); Funding acquisition (equal); Resources (equal); Writing – review & editing (equal). Andrei Osinsky: Conceptualization (equal); Formal analysis (equal); Funding acquisition (equal); Resources (equal); Writing – review & editing (equal). Stephen J. Pearton: Conceptualization (equal); Funding acquisition (equal); Methodology (equal); Project administration (equal); Supervision (equal); Writing – review & editing (equal).

The data that support the findings of this study are available within the article.