Etching of elemental layers in oxide molecular beam epitaxy by O₂-assisted formation and evaporation of their volatile (sub)oxide: The examples of Ga and Ge

Transparent semiconducting oxides like Ga₂O₃, In₂O₃, SnO₂, and GeO₂ have been rediscovered as promising (ultra-)wide bandgap semiconductors for applications in power electronics.^1–5 Their growth as epitaxial thin films by molecular beam epitaxy (MBE) is beneficial for materials exploration and device applications, both requiring a high-degree of purity and crystallinity.

In the MBE growth of oxides possessing volatile suboxide such as Ga₂O₃, In₂O₃, SnO₂, and GeO₂ from elemental sources (Ga, In, Sn, and Ge), the provided elemental flux is oxidized via a first oxidization step to form suboxides (Ga₂O, In₂O, SnO, and GeO) on the substrate. The suboxide is further oxidized via a second oxidization step to form solid oxide thin films.^6,7 The competing desorption of the intermediately formed suboxide (typically having a higher vapor pressure than its cation element) can decrease the thin film growth rate. While in an oxide MBE-growth chamber, suboxides were also found to form (and evaporate) readily from the elemental sources at a typical molecular O₂ background pressure present during growth,^8,9 their oxidation into the stable oxide (e.g., Ga₂O₃, In₂O₃, SnO₂, GeO₂, and SiO₂) required more reactive oxygen species, e.g., provided by an oxygen plasma.^7,8,10–13 Solely, the growth of In₂O₃ at a low growth rate of 0.6 nm/min has been demonstrated using molecular O₂.¹⁴ 

Despite these beneficial applications, the elemental fluxes are prone to leave unwanted elemental adsorbates, layers, or droplets on the surface. The removal of these elemental layers requires heating to the desorption temperature of the element, high-energy sputtering, or ex situ wet-chemical etching—all of which may create unacceptable degradation of the surface.

For this study, Ge and Ga cation layers were grown in high vacuum (background pressure 10⁻⁸ mbar) on 2-in. c-plane sapphire [Al₂O₃(0001)] wafers at a temperature of 400 °C by MBE. The rough backside of the single-side polished substrate was sputter-coated with titanium to allow for noncontact substrate heating by radiation from the substrate heater. The substrate temperature (T_sub) was measured with a thermocouple placed behind the substrate heater. Standard shuttered effusion cells were used to evaporate Ge (7 N purity) and Ga (7 N purity) from pyrolytic BN crucibles. The beam equivalent pressure (BEP) of the cations and O₂, proportional to the particle flux, were measured by a nude filament ion gauge positioned at the substrate location. A schematically drawn MBE growth chamber used in this study can be seen in Fig. 2. Table I illustrates the geometries of all the used MBE components. The BEPs are given in units of mbar and are converted into the equivalent particle flux (atoms cm⁻² s⁻¹) by multiplying the measured growth rate of the GeO₂ and Ga₂O₃ layer under conditions of full Ge and Ga incorporation by the cation number density of rutile phase GeO₂ (2.4 × 10²² cm⁻³)⁷ and monoclinic phase Ga₂O₃ (3.6 × 10²² cm⁻³)²⁸ and by kinetic gas theory in the case of O₂. For layer deposition, the used Ge and Ga -cell temperatures of 1300 °C (Note that this high temperature leads to a relatively fast degradation of the used standard effusion cell.) and 900 °C resulted in Ge and Ga-fluxes (Φ) of Φ_Ge = 2.2 × 10¹⁴ cm⁻² s⁻¹ and Φ_Ga = 2.2 × 10¹⁴ cm⁻² s⁻¹ impinging on the substrate, respectively.

Next, we provided molecular O₂ to etch the deposited Ga and Ge layers at elevated substrate temperatures (650 °C for Ga and 700 °C for Ge) that allow suboxides to desorb. For this purpose, a mass flow controller supplied molecular O₂ from the research-grade O₂ gas (6 N purity) and the O₂ flow was set as standard cubic centimeters per minute (SCCM). The flux Φ of desorbing species from the layer surface was measured in situ by line-of-sight quadrupole mass spectrometry (QMS, Hiden Analytical “HAL 511 3F”). The QMS ionizer was run at an electron energy of 50 eV to obtain optimal sensitivity. Therefore, some of the measured signals might be affected by fragmentation of suboxide molecules into cation and oxygen atoms.²⁹ To assess the surface coverage, the process was additionally in situ monitored by reflection high-energy electron diffraction (RHEED). The electron gun supplies electrons with kinetic energy of E_kin = 20 keV impinging on the growth surface at grazing incidence <2^o. Scattered electrons are hitting a phosphor screen positioned in the diffracted beam path and the resulting image is recorded by a CCD camera.

As an application example, we demonstrate the in situ removal of Ga droplets from a SiO₂ mask directly after in situ patterning of Ga₂O₃ by Ga-etching. For this purpose, an MBE-grown, ≈500 nm-thick Ga₂O₃ layer was covered by a ≈75 nm-thick SiO₂ hard mask (deposited using sputtering and structured by contact lithography and CHF₃-based reactive ion etching) and subsequently loaded into the MBE growth chamber. Ga-etching was performed by exposure to a Ga flux of Φ_Ga = 6.6 × 10¹⁴ cm⁻² s⁻¹ at T_sub = 650°C in the absence of O₂ for a total etching time of 40 min, resulting in an etch depth of ≈140 nm (determined by profilometry measurement). Subsequently, we in situ removed the Ga droplets that remained on the SiO₂ mask by exposure to 1 SCCM O₂ for 90 min at the same T_sub. The untouched structured-Ga₂O₃ sample, the structured-Ga₂O₃ sample after Ga-etching, as well as a Ga-etched Ga₂O₃ sample after molecular O₂ exposure were observed by top-view scanning electron microscopy (SEM).

Figure 3 presents the QMS signal of Ga and Ga₂O (proportional to the desorbing flux of Φ_Ga and $Φ Ga 2 O$ and the corresponding RHEED images at different stages during the elemental layer deposition and its subsequent etching by O₂. When the Ga shutter was opened, the elemental desorption slightly increased and then rapidly faded, corresponding to an almost full adsorption of the provided flux. After closing the Ga shutter, T_sub was immediately increased to 650 °C at 0.5 °C/s to facilitate Ga₂O desorption in the following etching process. The elevated T_sub did not result in detectable desorption of the already grown cation layer (the QMS signal before supplying O₂ is negligible), while the disappeared streaky RHEED pattern (middle) clearly indicates a substrate coverage by this layer. A dramatic $Φ Ga 2 O$ signal increase can be observed when an O₂ flow (1 SCCM) started impinging on the surface. This observation confirms that O₂ reacted with Ga to form Ga₂O via Eq. (3) at a temperature that allows suboxides to desorb. $Φ Ga 2 O$ fades gradually from the maximum value, likely due to the gradual decrease of surface fraction covered by the elemental layer. The Ga signal during etching is related to the fragmentation of Ga₂O molecules by the electrons of the ionizer in the quadrupole mass spectrometer.²⁹ The complete removal of the Ga layer is evidenced by the disappearance of the $Φ Ga 2 O$ signal and by the reappearance of the streaky RHEED pattern of the substrate.

To determine the efficiency of the etching process, we established a quantitative relation of impinging O₂-flux and desorbing Ga₂O flux at varying flow rates of O₂. Figure 4(a) illustrates the QMS signal of $Φ Ga 2 O$ during the deposition of 6 equal layers of metallic Ga and their in situ etching by O₂ at a decreasing flow, which were 2.00, 1.50, 1.00, 0.80, 0.50, 0.25 SCCM, respectively. These experiments were carried out in sequence using Ga deposition and etching temperatures of 400 and 650 °C, respectively. Similar to Fig. 3, a sharp increase of $Φ Ga 2 O$ was detected when O₂ was supplied, and different O₂-fluxes were able to fully convert the Ga layers into evaporated Ga₂O, leaving behind a clean surface. Apparently, the maximum $Φ Ga 2 O$ decreases with reduced impinging O₂-flux and the required time to completely remove the same amount of Ga increases simultaneously.

The QMS signal of the desorbing Φ_Ge and Φ_GeO, as well as the surface development monitored by RHEED during Ge layer deposition and O₂-etching experiment, shown in Fig. 6, exhibit qualitatively similar behavior to that observed for Ga. To enhance GeO desorption in the etching process, O₂ was supplied at T_sub = 700 °C. A significant Φ_GeO signal increase can be observed when the O₂ (1 SCCM) approached the surface, confirming that O₂ reacted with Ge to form GeO via Eq. (4). The disappearance of Φ_GeO and reappearance of the streaky RHEED pattern of the substrate proved a complete removal of the Ge layer. Similarly, we determined η for Ge etching using the same methodology employed in our Ga experiment. A surface coverage of D_Ge = 7.85 × 10¹⁷ cm⁻², a desorbing flux of Φ_GeO = 1.58 × 10¹⁴ cm⁻² s⁻¹ was obtained based on the using experiment parameters, while 1 SCCM of O₂ corresponding to a P_O2= 1.55 × 10⁻⁵ mbar, which can be translated into O₂-flux of Φ_O2= 4.2 × 10¹⁵ cm⁻² s⁻¹ by Eq. (6). Consequently, a $η Ge= Φ GeO 2 ∗ Φ O 2$ $×100%=1.8%$ representing removal efficiency of O₂ was obtained according to Eq. (4).

To better illustrate the application scope of our studies, we conducted experimental tests on the device structuring process by in situ Ga-etching a SiO₂-masked Ga₂O₃ layer. Figure 7 showcases the top-view SEM images of the masked and structured Ga₂O₃ samples. By comparing Fig. 7(a) with Fig. 7(b), as anticipated, we observed Ga droplets remaining on the SiO₂ mask after structuring Ga₂O₃ by exposing it to the Ga flux. However, these droplets can be completely removed in situ by providing O₂ following the oxide etching process, as evidenced by a clean mask surface depicted in Fig. 7(c).

In conclusion, we have demonstrated a process in which molecular O₂ is utilized to remove elemental Ga and Ge layers in an MBE growth chamber through formation and desorption of their volatile suboxides at temperatures lower than those required for elemental desorption. Under the investigated (nonoptimized) conditions about 1.8%–3.5% of the provided O₂ contributed to cation removal.

We further showcased the application of this O₂-assisted cation removal process for the in situ removal of the residual Ga droplets from the SiO₂ mask directly after structuring a Ga₂O₃ layer by in situ etching using a Ga atomic flux.

We predict the O₂-assisted cation removal process to be generally applicable in situ within an oxide MBE or MOCVD system to remove residual elemental layers (that may occur after exposure to the cation fluxes, e.g., during in situ oxide etching or substrate-temperature calibration) if their volatile suboxides or oxides exhibits a higher vapor pressure than the respective elements, such as B, In, La, Si, Sn, Sb, Mo, Nb, Ru, Ta, V, and W.^9,31

The fraction of O₂ contributing to cation removal may be increased by using different substrate temperatures from those employed in our experiments to affect activation of suboxide formation or parasitic O₂ desorption from the substrate. An increase of the total cation layer removal rate, however, necessitates at some point higher O₂-fluxes feeding the suboxide formation and higher substrate temperatures to ensure correspondingly faster suboxide desorption rates. These efficiency and speed aspects provide room for future optimization of the O₂-assisted cation removal process, particularly by studying its substrate-temperature dependence.

The authors thank Hans-Peter Schönherr, Claudia Hermann, Sander Rauwerdink, and Walid Anders for technical support, Steffen Breuer for discussion, as well as Jingxuan Kang for critically reading the manuscript. This work was performed in the framework of GraFOx, a Leibniz-ScienceCampus partially funded by the Leibniz association. W.C. gratefully acknowledges financial support from the Leibniz association under Grant No. K417/2021.

The authors have no conflicts to disclose.

Wenshan Chen: Formal analysis (equal); Investigation (equal); Visualization (equal); Writing – original draft (equal); Writing – review & editing (equal). Kingsley Egbo: Formal analysis (supporting); Investigation (supporting); Writing – review & editing (supporting). Huaide Zhang: Investigation (supporting); Writing – review & editing (supporting). Andrea Ardenghi: Investigation (supporting); Writing – review & editing (supporting). Oliver Bierwagen: Conceptualization (equal); Funding acquisition (equal); Project administration (equal); Supervision (equal); Writing – review & editing (equal).

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