Plasma enhanced atomic layer deposition and atomic layer etching of gallium oxide using trimethylgallium

Gallium oxide (Ga₂O₃) is an ultrawide bandgap material that has recently gained considerable interest due to its large direct bandgap (∼4.9 eV), high thermal and chemical stability, large dielectric constant (>10), and unique transparency in the visible and ultraviolet (UV) spectral regions.^1–5 With multiple crystal polymorphs, the bandgap of Ga₂O₃ may vary between the reported values of 4.5–4.9 eV. Among these, monoclinic β-Ga₂O₃ is the most stable phase with a bandgap of 4.9 eV. Ga₂O₃ thin films have been formed using various deposition techniques, including metalorganic chemical vapor deposition,⁶ molecular beam epitaxy,⁷ metalorganic vapor phase epitaxy,⁸ atomic layer deposition (ALD),^3,9–13 pulsed laser deposition,² and electron beam evaporation.¹⁴ Reported applications of Ga₂O₃ include gas sensors,^15,16 field-effect-transistors,^17,18 solar cells,^19,20 and deep-UV photon detectors.^21–23 

Atomic layer deposition of Ga₂O₃ using trimethylgallium [TMG, Ga(CH₃)₃] and O₂ plasma has previously been reported.^3,9,10,13,24 Other reported Ga₂O₃ ALD precursors include gallium acetylacetonate,¹² dimethylgallium isopropoxide,¹¹ gallium tri-isopropoxide,²⁵ hexakis(dimethylamido) gallium,²⁶ and pentamethylcyclopentadienyl gallium.²⁷ The films described in these reports were typically amorphous, although polycrystalline β-Ga₂O₃ may be produced by annealing at temperatures between 700 and 1000 °C. Phase selectivity of crystalline α-, β-, and ɛ-Ga₂O₃ was demonstrated using a TMG and O₂ plasma enhanced ALD process by adjusting chamber pressure and plasma composition.¹³ TMG with O₂ plasma also demonstrated a large Ga₂O₃ ALD window (RT—400 °C).^9,10

Dry etching, required for semiconductor device fabrication, involves the use of plasma and vapor phase reactants rather than wet chemicals for the removal of surface material, enabling the patterning of small features. “Conventional” dry etching methods employ plasma ions and involve a combination of chemical and physical mechanisms to remove the material. Several dry etching techniques for Ga₂O₃ have previously been investigated using a variety of F- and Cl-based species, including SF₆, NF₃, CHF₃, BCl₃, and Cl₂.^18,28–33 These reactants modify the surface, often producing a nonvolatile material, which requires removal by physical bombardment from the ions, typically Ar.^18,28,29 Due to the strong chemical bonding of Ga₂O₃, the ion energies required for material removal are typically high.^29,32–34 Therefore, these plasma-based etching techniques often result in ion-induced damage to the material, including electrically active damage which may affect device performance.^29,32,33 Development of a dry etching method for Ga₂O₃ with low surface damage is, therefore, necessary for the advancement of Ga₂O₃ processing.

Atomic layer etching (ALE) is a dry etching technique using cycles of sequential reactions for controlled material removal for atomic scale processing.^35–39 One cycle of ALE typically consists of a surface modification process followed by removal of the modified surface by either volatile release or assisted desorption. These reactions may occur through either thermochemical mechanisms or energy-assisted processes through the use of ions and plasmas.

Plasma-based ALE has been in development for decades, beginning with ALE of Si being reported in 1996.⁴⁰ These processes have been further expanded through implementation in fabrication processing as well as commercial availability of the ALE systems.^35,36,38,41 In plasma-based ALE, the modification step serves to lower the ion energy threshold to remove material from the surface relative to the unmodified material. The ions are then required to remove the modified surface through momentum transfer to surface atoms.^35,36,38,41 Typical plasma ALE processes require Ar ion energies of 30–70 eV.⁴¹ Plasma ALE is highly anisotropic, enabling directional etching for high aspect ratio nanoscale patterning. Due to the use of energy-assisted reactions in plasma-based ALE, ion-induced damage to the material may occur. Ion-free ALE processes could have the advantage of preventing ion-induced damage and producing conformal etching.

ALE via thermally driven reactions is a more recent development as a highly conformal and ion-free technique for low damage atomic scale material removal using sequential exposures of vapor phase reactants.^42–44 Thermal ALE requires self-limiting behavior of the reactant exposures, requiring the use of one reactant to modify or passivate the surface and another reactant to remove only the modified surface layer, enabling etching without ion-induced damage to the underlying material. Choice of the precursor is, therefore, critical to utilize thermally favorable reactions while avoiding spontaneous etching.

Thermal ALE occurs through surface modification followed by the formation of stable and volatile complexes. Surface modification of previously reported thermal ALE processes typically occurs through fluorination via exposure to HF, SF₄, or XeF₂.⁴⁵ The reported mechanism for volatile product formation is a ligand exchange between the modified surface species with metal organic precursors, typically involving chlorine, methyl, or ethyl groups. A report of thermal ALE of Al₂O₃ involved the use of HF and tin acetylacetonate [Sn(acac)₂, Sn(CH₃COCHCOCH₃)₂].⁴² 

HF and trimethylaluminum [TMA, Al(CH₃)₃] have previously been reported as effective reactants for the thermal ALE of Al₂O₃ and HfO₂.^43,46 Aqueous HF has also been used for wet chemical etching of Ga₂O₃ at room temperature.⁴⁷ Thermal ALE of Ga₂O₃ was recently reported by Lee et al. using HF with BCl₃, AlCl(CH₃)₂, Al(CH₃)₃, TiCl₄, and Ga[N(CH₃)₂]₃ as ligand exchange precursors. In this study, we report that HF and trimethylgallium [TMG, Ga(CH₃)₃] used in discrete subsequent exposures are effective reactants for Ga₂O₃ ALE. Figure 1 shows the reaction steps during the ALE process. First, exposure of the Ga₂O₃ film to HF fluorinates the surface. Second, the proposed mechanism during TMG exposure is a ligand exchange process between fluorine and methyl groups, resulting in volatile GaF_x(CH₃)_y species, which desorb from the surface,

where “s” denotes the substrate surface, “mod” indicates modified surface species, and “g” indicates gas or vapor phase species.

Ga₂O₃ thin films were deposited by plasma enhanced atomic layer deposition (PEALD) using TMG and O₂ plasma onto 25.4 mm diameter boron-doped Si (100) wafers supplied by Virginia Semiconductor, and then, etched by ALE using TMG and HF. The PEALD and ALE reactions were both performed in a custom stainless-steel hot wall ALD reactor with a multiwavelength ellipsometer (Film Sense, FS-1) mounted at ∼45° to the wafer surface. A two-stage dry processing pump (Ebara, A70W) with N₂ (99.999%, 14.2 SLM) dilution flow was used to reach a base pressure of 4 × 10⁻⁵ Torr during processing. Quadrupole mass spectrometry (QMS) (SRS, RGA 200) was used to monitor gas chemistries during reactions. Remote plasma generation occurred in a fused quartz tube using an inductively coupled 13.56 MHz RF generator (MKS, Elite 300), which was located approximately 25 cm above the wafer surface. This distance between plasma formation and the sample surface enables a sufficient flux of oxygen radicals while limiting the density of ions at the surface. Due to the operating pressure (100 mTorr), O radicals are expected to be completely thermalized. A bias was not applied to the grounded substrate during the deposition.

HF vapor was derived from HF-pyridine [(C₅H₅N)⋅(HF)_x] (70% HF by weight, Alfa-Aesar). HF-pyridine was transferred into the precursor vessel in a dry Ar-filled polytetrafluoroethylene glove bag. The HF-pyridine and TMG (99.998 wt. %+, STREM Chemicals, Inc.) precursors were held at 30 °C. Precursor vapor delivery occurred using ultrahigh purity Ar (99.999%) carrier gas with flow rates of 5.0 standard cubic centimeters per minute (SCCM) for TMG and 15.0 SCCM for HF. TMG was delivered to the chamber through a 100 μm orifice, which allowed for continuous reactant flow into the chamber during precursor exposure. Precursor exposure times of 30 s were used resulting in pressure transients of 600–800 and 200–400 mTorr for HF and TMG, respectively. A 30.0 SCCM flow of N₂ was used between reactant exposures for 30 s to purge the system of unreacted precursors and byproducts. Precursor pulse times were varied to achieve saturation of reactions.

This study was performed on Ga₂O₃ deposited on Si (100) substrates by PEALD using TMG and O₂ plasma and a substrate temperature of 200 °C. Prior to the deposition, Si wafers were cleaned using a 10 min UV O₃ exposure at atmospheric pressure. The PEALD process was monitored by in situ ellipsometry and characterized by x-ray photoelectron spectroscopy (XPS). Ga₂O₃ films were deposited using a TMG pulse of 0.1 s followed by 30 s of TMG exposure and a 100 W O₂ plasma exposure for 10s. A 30s N₂ exposure was used between reactant exposures to purge the reaction chamber. Ga₂O₃ films were confirmed to be amorphous using transmission electron microscopy.

Ellipsometry measurements were made in situ during the processing using a multiwavelength ellipsometer (Film Sense, FS-1) mounted at ∼45° to the wafer surface. The ellipsometer used four light emitting diodes (LEDs) to produce light of wavelengths centered at 465, 525, 580, and 635 nm. A Cauchy model was used for thickness measurements of the Ga₂O₃ films. Optical parameters of the Ga₂O₃ films were determined by ex situ spectroscopic ellipsometry (M2000, J.A. Woollam) from a 20 nm PEALD Ga₂O₃ film grown at 200 °C. Thickness measurements from this device were confirmed using XPS of 5 and 10 nm of Ga₂O₃ grown by PEALD on Si (100).

For in situ measurements, samples were transferred in ultrahigh vacuum for XPS (VG Scienta, R3000). Measurements were made using an Al K_α x-ray source producing monochromatic light at 1486.7 eV and a system base pressure of ∼4 × 10⁻¹⁰ Torr. Using an analyzer pass energy of 100 eV and slit size of 0.4 mm, the system may resolve binding energies to ±0.15 eV. XPS scans were fit using Gaussian–Lorentzian line shapes after a standard Shirley background was subtracted. Empirically derived atomic sensitivity factors were used to determine surface stoichiometry.⁴⁸ The Ga 3d, O 1s, C 1s, and F 1s XPS core level spectra were used for chemical state analysis at the wafer surface.

Ex situ atomic force microscopy (AFM) (Asylum Research, MFP-3D) measurements were used to determine the surface roughness of Ga₂O₃ films and were performed using scan sizes of 25.0 and 400 μm².

TMG and O₂ plasma were used to deposit Ga₂O₃ films by PEALD. Alternating exposures of HF and TMG were then used to etch Ga₂O₃ films by thermal ALE. The deposition and etching processes were monitored in situ using a multiwavelength ellipsometer. In situ XPS was employed to identify chemical states and determine surface composition.

Thickness measurements were recorded via ellipsometry using a Cauchy model throughout the depositions. The change in thickness throughout 15 cycles of a typical deposition is shown in Fig. 2(a). An average linear increase in thickness of 0.9 ± 0.1 Å was noted for the 15-cycle deposition. The ellipsometric response for three ALD cycles is shown in Fig. 2(b) with individual TMG and O₂ plasma half-cycles of one ALD cycle identified. We note that the ellipsometric thickness change could be due to adsorbed species or changes of the surface optical properties. A large thickness increase is observed as TMG is adsorbed onto the surface. Then, there is a decrease in thickness after the O₂ plasma exposure, which is interpreted as the combustion of methyl groups and oxidation of the surface. A slight decrease is observed in the ellipsometry data prior to ignition of the O₂ plasma, which is attributed to noise in the instrument data due to a change in birefringence of the incident light caused by pressure changes in the reaction chamber.

XPS was used to examine the surface composition after deposition and etching, as well as to calculate film thickness to verify the deposition and etch rates. Figure 3(a) shows the XPS spectrum of Ga 3d core level of a 20 nm Ga₂O₃ film after PEALD. A spectral deconvolution is indicated by dashed lines. An O 1s core level at 23.0 ± 0.2 eV is also visible within this spectrum. Based on the ratio of the integrated Ga 3d and O 1s areas and photoionization cross sections, an O:Ga ratio of 3:2 was obtained. The oxygen concentration was measured to be 59 ± 10%.⁴⁸ Spectral deconvolution of Ga 3d indicated two peaks with energies 18.9 ± 0.1 and 20.2 ± 0.1 eV, which are consistent with Ga–OH and Ga–O, respectively.^10,49 Similarly, O 1s primary peak position (530.8 eV) of the PEALD Ga₂O₃ shown in Fig. 3(a) is characteristic of bonding with Ga in its highest oxidation state (Ga³⁺),^49–51 while broadening on the higher binding energy side (531.8 eV) is characteristic of O–H bonding.^50,52,53 The presence of hydroxyl species is attributed to a surface termination species. Additionally, carbon impurities were below the XPS detection limit for the PEALD Ga₂O₃ films.

Figure 3(b) shows the XPS scans of the Ga 3d core level of a Ga₂O₃ film after 15 cycles of ALE, ending with TMG exposure, with spectral deconvolution indicated by dashed lines. An upward shift of 1.2 eV in binding energy may be observed in all peaks. Despite a binding energy shift, energy splitting between the Ga–OH and Ga–O states remains consistent. Surface impurities of C and F were detected after ALE via XPS with similar concentrations of C and F, each accounting for up to 3 at. %. Surface atomic concentration of O was reduced to 53.0 ± 0.5% at the surface after ALE, while Ga content remained constant, indicating favorable substitution of oxygen by carbon and fluorine.

Thickness change per cycle remained approximately linear throughout the etching process. The linearity of the ALE process at 300 °C is shown in Fig. 4(a) throughout 15 cycles of HF and TMG exposures on a Ga₂O₃ film. The ellipsometric response for three ALE cycles is shown in Fig. 4(b) with the process steps annotated for one cycle. One cycle of the process shown consisted of a 0.1 s pulse of HF, followed by 30 s of exposure prior to a 30 s N₂ purge, and then an 8 s pulse of TMG, followed by 30 s of TMG exposure prior to another 30 s N₂ purge. The HF and TMG pulses and exposures have been highlighted in the plot. A slight increase in measured thickness is observed at each HF exposure. A decrease in thickness is observed during TMG exposure as the ligand exchange occurs and the resulting products desorb from the surface. No discernible plateau was observed in the ellipsometry data during individual reactions, which is attributed to noise in the instrument detection caused by birefringence of the incident light through pressure changes in the system during pulses of precursor and processing gases. Precursor exposure times and N₂ purge times were adjusted to ensure complete reactions.

Self-limiting behavior is an important aspect of ALE, indicating that a process is restricted to affect only the surface atoms. The self-limiting behavior of the ALE process was determined by incrementing precursor exposure times prior to the N₂ chamber purge. Forty cycles of ALE were performed for each set of pulse times. The etch rates for each set of exposure times were taken as the average changes in thickness per cycle. As shown in Fig. 5, there is an increase in etch rate as the TMG pulse time is increased from 0.1 to 12.0 s before the etch rate saturates at approximately 1.0 Å/cycle. HF pulse and N₂ purge times were also adjusted. An HF pulse time of 0.1 s was determined to be sufficient for complete reactions as no increase in etch rate was observed at higher pulse times.

AFM measurements were taken after PEALD of 10 nm of Ga₂O₃ on Si and ALE of 5 nm. A scan size of 25.0 μm² was used to investigate the surface structure. An RMS surface roughness of 0.55 ± 0.05 nm was determined for 10 nm of PEALD Ga₂O₃ deposited at 200 °C on Si (100). After the removal of 5.0 nm by ALE at 300 °C, RMS roughness was 0.56 ± 0.05 nm, indicating uniform and conformal etching of the film.

Thin films of Ga₂O₃ were deposited by PEALD using O₂ plasma as the oxidizing agent and TMG serving as the Ga source and then etched by thermal ALE using exposures of HF and TMG within the same reaction chamber.

During ALD, the TMG reaction step results in adsorbed precursor molecules on the substrate surface. The O₂ plasma exposure serves to both combust unreacted methyl groups at the surface and oxidize the adsorbed Ga.

The proposed overall ALE reaction can be written as^43,54

which results in the production and desorption of H₂O and volatile GaF(CH₃)₂ species. This occurs through two discrete reactions at the surface, including fluorination with HF,

and ligand exchange with TMG,

QMS measurements indicated the presence of GaF(CH₃)₂ monomers among volatile products. Clancey et al. have reported that ALE of Al₂O₃ resulted in volatile products formed in dimers including Al₂F₂(CH₃)₃⁺ and Al₂F(CH₃)₄⁺ after the loss of a methyl group.⁵⁵ However, our system is not capable of detecting sufficient mass to confirm the formation of higher order Ga polymers.

HF exposures result in the formation of a stable, nonvolatile fluoride at the surface, as confirmed by XPS. Ellipsometry indicated that HF exposure does not spontaneously etch the Ga₂O₃ film at 300 °C, as repeated exposures of HF without additional reactant exposures did not produce an evident change in film thickness. Ligand exchange between TMG and the fluorinated surface produces volatile GaF_x(CH₃)_y species which desorb from the surface. Repeated TMG exposures did not etch Ga₂O₃ without prior fluorination, as ligand exchange with the fluorinated surface is necessary for volatile product formation and, therefore, vital for material removal.

XPS results indicated the presence of residual C and F (<3 at. %), resulting from precursor exposures during multiple cycles of ALE. This rate of residual surface species is typical of thermal ALE processes.^42,43,56,57 Reduction of these residuals during etching may be possible through alternative refinement of the etching parameters, although this is expected to result in diminished etch rates. Additionally, residual impurities may be removed via additional surface treatment after ALE.

Shorter N₂ purge times between precursor exposures resulted in unusually high deposition and etch rates for ALD and ALE. With an N₂ purge time 7.5 s, a deposition rate of 1.6 ± 0.2 Å/cycle and an etch rate of 2.0 ± 0.1 Å/cycle were observed, which are higher than the typically reported values for ALD and ALE processes. This enhanced rate was attributed to residual reactants near the surface from incomplete purges producing non-ALD and non-ALE behavior. N₂ purge time of 30 s was determined to be sufficient for both PEALD and ALE processes.

TMA has previously been used for ALE of several other materials, but our literature search did not identify other reports of ALE using TMG for ligand exchange. The viability for TMG to be used in the etching processes of Ga₂O₃ when used with HF, therefore, expands the selection of precursors which may be used for ALE. The use of TMG as an ALE reactant also enables simple implementation of a single system for both ALD and ALE of Ga₂O₃. This requires only the addition of an HF source to an existing ALD system equipped with TMG and an oxygen source, such as O₂ plasma, O₃, or H₂O.

Gallium nitride (GaN) is a related material that has also gained considerable interest recently due to its favorable material properties, which provide potential applications in power electronics and memory devices, in addition to its currently realized applications in LEDs, lasers, and other photoelectronic devices.⁵⁸ As the native oxide of GaN is GaO_x, the ALE of Ga₂O₃ may provide a unique pathway for a conversion etch process. As others have reported, direct fluorination of GaN using HF is not a favorable process.⁵⁹ Therefore, oxidation of GaN could provide a pathway for fluorination and a conversion etch method similar to those used for the ALE of Si, SiO₂, W, and other materials.^60–62 

PEALD of Ga₂O₃ was studied using TMG and O₂ plasma as precursors, and ALE of Ga₂O₃ was studied using TMG and HF. In situ multiwavelength ellipsometry and x-ray photoelectron spectroscopy studies demonstrated both ALD and ALE of Ga₂O₃ achieved with a growth per cycle of 0.9 ± 0.1 Å and an etch per cycle of 1.0 ± 0.1 Å. HF was effective in the surface modification of Ga₂O₃, as it has been demonstrated for ALE of Ga₂O₃ with various other precursors, including boron trichloride BCl₃, dimethyl aluminum chloride AlCl(CH₃)₂, trimethyl aluminum Al(CH₃)₃, titanium tetrachloride TiCl₄, and tris-dimethylamido-gallium Ga[N(CH₃)₂]₃.⁵⁷ Ligand exchange using TMG as the metal precursor was demonstrated in this ALE process and PEALD process. Previously reported demonstrations of ALE using TMG were not identified in our literature review, suggesting this as a new use of TMG as an effective metal precursor for ligand exchange in ALE.

This work was supported by Advanced Research Projects Agency-Energy under the PN-Diodes Program through Grant No. DE-AR0001691. We acknowledge the use of facilities within the John. M. Cowley Center for High Resolution Electron Microscopy and the NanoFab, at Arizona State University, supported in part by the NSF Program No. NNCI-ECCS-1542160. We acknowledge Dr. Zhiyu Huang for performing ex situ spectroscopic ellipsometry measurements. We would like to thank Dr. Xingye Wang and Dr. Mei Hao for useful discussions.

The authors have no conflicts to disclose.

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