Dry and wet etching for β-Ga₂O₃ Schottky barrier diodes with mesa termination

Ga₂O₃ is one of the promising materials for high-power applications due to its high critical electric field of 8 MV cm⁻¹ and a large band-gap energy above 4.5 eV.¹⁾ Bulk Ga₂O₃ with a β-gallia phase, which is the most thermally stable, is grown using melt-based methods,²⁾ potentially providing scalable wafers at a low cost. Recently, β-Ga₂O₃ fin field-effect transistors and trench metal-oxide-semiconductor (MOS) diodes have shown an excellent performance of high breakdown voltages over 1 kV.^3–5) Still, the specific on-resistance R_on is over 5 mΩcm² at 1 kV, which is much higher than the theoretical limits. These vertical devices use a deep etching to prevent the high electric field near the surface region, suffering from a plasma damage by an inductively-coupled plasma (ICP) reactive-ion etching (RIE).⁶⁾ Single crystal Ga₂O₃ is etched by chlorine-based RIE.^7–9) The plasma damage can substantially reduce a carrier mobility,¹⁰⁾ increasing R_on. The high-power β-Ga₂O₃ devices with low R_on require less plasma damage.

The plasma damage can be efficiently removed by etching with a chemical reaction. In fact, β-Ga₂O₃ Schottky barrier diodes (SBDs) with nearly unity ideality factor n are achieved by immersing in hot phosphoric-acid (H₃PO₄) solution after ICP-RIE.¹¹⁾ A Ga₂O₃ single crystal poses an excellent chemical stability. β-Ga₂O₃ is slightly etched by heating in nitric acid (HNO₃), sulfuric acid (H₂SO₄), H₃PO₄, and potassium hydroxide (KOH) solutions. The etch rate of β-Ga₂O₃ (100) at 120 °C is 1.4 nm min⁻¹ for HNO₃ (61 wt%),¹²⁾ ∼2 nm min⁻¹ for H₂SO₄ (97 wt%),¹³⁾ and ∼10 nm min⁻¹ for H₃PO₄ (85 wt%).¹⁴⁾ In hot KOH solution, photo-enhanced chemical etching with ultraviolet illumination increases the etch rate to 30 nm min⁻¹ for (010) plane and 150 nm min⁻¹ for ($\bar{2}$01) plane.¹⁵⁾ The effect of these chemical solutions on Ga₂O₃ devices has not been compared. In this study, we etched β-Ga₂O₃ under various ICP-RIE conditions and chemical solutions, and investigated the effect of various chemical solutions on the vertical β-Ga₂O₃ SBDs with mesa termination.

For dry and wet etching, we used tin-doped β-Ga₂O₃ (010) substrates (Novel Crystal Technology, Inc.). The substrate surfaces were treated by chemical mechanical polishing (CMP). The tin concentration was 4 × 10¹⁸ cm⁻³. After solvent cleaning with acetone and isopropanol, a 50-nm-thick nickel-metal mask was deposited on the substrate using an electron beam (EB) evaporation. The ridge structures with ∼300-nm heights were formed using ICP-RIE (Samco RIE-400iPS) with BCl₃ and Cl₂ mixture gases. An etch rate of the Ga₂O₃ layers was determined by scanning electron microscopy (SEM). After removing the nickel-metal mask using piranha (H₂SO₄ (95 wt%): H₂O₂ (30–35wt%) = 3:1) for 1 min, the substrates were dipped in various chemical solutions without starring on a heater for 10–30 min to remove the plasma damage. As the chemical solutions, we used piranha, H₃PO₄ (85 wt%), hydrofluoric acid (HF, 46–48 wt%), tetramethylammonium hydroxide (TMAH, 2 wt%), and KOH (3 wt%). The solution temperature was monitored by thermocouple in the heater. The wet-etched surfaces were observed by SEM.

The etch-rate dependence of β-Ga₂O₃ (010) on the ratio of Cl₂ and BCl₃ gases using ICP-RIE is shown in Fig. 1(a). The ICP power, bias power, cooling-water temperature, and chamber pressure were 150 W, 30 W, 20 °C, and 0.6 Pa, respectively. For the BCl₃/Cl₂ mixture gas, the etch rate of Ga₂O₃ increased with increasing the BCl₃ gas composition, corresponding to the other report.¹⁶⁾ This indicates that not only the chlorine radicals in the plasma, but also unsaturated radicals produced from the BCl₃ additive contribute to the Ga₂O₃ etching.^8,17,18) As shown in Fig. 1(b), the etch rate of Ga₂O₃ increased with increasing the ICP power. The high ICP power increases the plasma density, or the number of chlorine radicals, enhancing the chemical reaction on surfaces. The chemical etching provides a lower plasma damage in comparison with sputtering. The etch rate of 97 nm min⁻¹ was achieved at the ICP power of 400 W with the Cl₂/BCl₃ mixing gas of 10/10 sccm.

Zoom In Zoom Out Reset image size

The bird-view SEM images of the Ga₂O₃-ridge structure after ICP-RIE are shown in Fig. 2. The Ga₂O₃-ridge structures etched with only BCl₃ gas had the rough sidewall, as shown in Fig. 2(a), while those etched with the BCl₃/Cl₂ mixture gas had the smooth sidewall, as shown in Fig. 2(b). We suppose that the nickel-metal mask reacts with oxide in Ga₂O₃ for the ICP-RIE with BCl₃ gas, causing the redeposition of Ni-O impurities around the sidewall. Photoresist and SiO₂ may be used as a mask for ICP-RIE using only BCl₃ gas. The roughness of the sidewall was independent on the ICP power, as shown in Fig. 2(c). The 50-nm-thick nickel mask was removed for 3 min at the ICP power of 400 W, as shown in Fig. 2(d), indicating the selectivity of ∼6, which is higher than the SiN_x mask.¹⁸⁾

Zoom In Zoom Out Reset image size

After etching β-Ga₂O₃ (010) at the ICP power of 300 W with the Cl₂/BCl₃ mixture gas of 10/10 sccm and removing the nickel mask, the Ga₂O₃-ridge structures were dipped in various acid and alkali solutions. Although both solutions of HF at room temperature and piranha at 120 °C hardly etch β-Ga₂O₃ (010), the H₃PO₄ solution at 80 °C shows slightly anisotropic etching, as shown in Fig. 3(b). These solution temperatures were too small precisely to determinate the etch rate. Zhang et al. reported that hot H₃PO₄ solution of β-Ga₂O₃ (010) has a higher etch rate of the vertical plane to near [101] and [001] directions than that of the vertical plane to near [−101] and [103] directions.¹⁴⁾ The appeared plane in hot H₃PO₄ solution may be the stable plane, or the vertical plane to near [−101] and [103] directions.

Zoom In Zoom Out Reset image size

The TMAH solutions hardly etch β-Ga₂O₃ (010) at 80 °C, as shown in Fig. 3(c), while the KOH solutions at 80 °C show effective etching, as shown in Fig. 3(d). The etch-rate difference between the TMAH and KOH solutions is attributed to the potassium ions, which have the large diffusion length in oxide layers,¹⁹⁾ and a higher density of OH⁻ ions in the KOH solution than that of the TMAH solution. We suggest that Ga₂O₃ is dissolved in the base solutions after Ga–O compounds are formed by reacting Ga atoms with OH⁻ ions.^15,20) Thus, the etching behavior in alkali solutions significantly depends on the crystal orientation, which defines the different Ga-to-O ratio, dangling bond densities, and surface energies. The strong Ga–O bonds in some crystal orientations including the (010) plane may suppress the reaction of Ga atoms with OH⁻ ions,¹⁵⁾ resulting in chemically stable surfaces, or anisotropic etching.

The (010) orientation was used for the above dry and wet etchings of β-Ga₂O₃. Currently, β-Ga₂O₃ (010) epilayers are not commercially available. Using the (001) orientation, we investigated the effect of various chemical solutions on the vertical β-Ga₂O₃ SBDs with mesa termination. We used 10-μm-thick silicon-doped Ga₂O₃ films grown on 650-μm-thick tin-doped β-Ga₂O₃ (001) substrates by halide vapor-phase epitaxy. The substrates were thinned to ∼300 μm thickness from (00$\bar{1}$) backside by grinder, followed by CMP treatment with colloidal silica for 6 min at 50 rpm. After a 100-nm-thick nickel-metal mask was deposited on (001) surfaces by the EB evaporation, a 770-nm-deep mesa termination was fabricated by RIE at the ICP power of 300 W with the Cl₂/BCl₃ mixing gas of 10/10 sccm. The nickel metal was removed by the piranha solution in order to remove both plasma damages on the mesa surface and sidewall, which are potentially caused by the 300 W ICP power due to the large-size mesa patterns. Then, the samples were dipped in various chemical solutions at 80 °C for 10 min. For a cathode contact, Ti (20 nm)/Au (50 nm) metal stacks were deposited on the (00$\bar{1}$) backside, followed by a thermal annealing at 550 °C for 1 min in a nitrogen ambient. Circular anode contacts with various diameters of 100–500 μm were prepared on the (001) surfaces using Ni (20 nm)/Au (50 nm) metal stacks. The current density–voltage (J–V) and capacitance–voltage (C–V) characteristics of the β-Ga₂O₃ (001) SBDs with mesa termination were performed using an Agilent B1505 semiconductor analyzer. The breakdown voltages of the Ga₂O₃ SBDs were determined in Fluorinert FC-72.

The C–V measurements of the Ga₂O₃ SBDs were performed at 1 MHz with a DC bias sweeping from −3 V to 0 V using 500-μm-diameter Ni contacts. The effective donor concentrations N_d–N_a and built-in potential V_d are determined from the equation of $\tfrac{1}{{C}^{2}}=\tfrac{2\left({V}_{d}-V\right)}{{A}^{2}e{\varepsilon }_{s}{\varepsilon }_{0}\left({N}_{d}-{N}_{a}\right)},$ where e is the electron charge, ε_s (∼10) is the relative permittivity of β-Ga₂O_3,²¹⁾ and A is the area of the anode electrode. N_d—N_a and V_d of the n-type Ga₂O₃ epilayer are calculated to be 2.5 × 10¹⁶ cm⁻³ and 0.88 eV, respectively. The Schottky barrier height Φ_{B_CV} is determined from the equation of $e{{\rm{\Phi }}}_{B}={{\rm{eV}}}_{d}+{E}_{c}-{E}_{f}+kT,$ where k is the Boltzmann constant, T is the absolute temperature, and E_c and E_f are the bottom of the conduction band and the Fermi level in β-Ga₂O₃, respectively. By assuming silicon donors fully ionized at room temperature, E_c-E_f is equal to $kT\,\mathrm{ln}\,\tfrac{{N}_{c}}{{N}_{d}-{N}_{a}},$ where N_c ($=2{\left(\tfrac{2\pi {m}_{n}kT}{{h}^{2}}\right)}^{\tfrac{3}{2}}$) is the effective density of states in the conduction band. m_n (∼0.28m₀ for Ga₂O₃) is the density-of-state effective mass for electrons²²⁾ and h is the Plank constant. By ignoring the image-force-induced lowering of the barrier height, Φ_{B_CV} of Ni for the Ga₂O₃ SBDs is calculated to be 1.03 eV. This value is slightly lower than those of the other reports (1.2 ± 0.2 eV)^4,23–26) and the ideal Schottky barrier height, which is estimated to be 1.2 eV by assuming electron affinities of 4.0 eV for UID β-Ga₂O₃ and a work function of 5.2 eV6 for Ni.²⁷⁾ The low Φ_B may result from the lateral distribution of barrier inhomogeneity due to the large-sized electrode.^24,28)

The J–V characteristics of the β-Ga₂O₃ (001) SBDs treated by various chemical solutions are shown in Fig. 4. R_on of Ga₂O₃ SBDs treated by the H₃PO₄, piranha, and KOH solutions was 0.91, 41.0, and 850 mΩ-cm², respectively. The low R_on of the Ga₂O₃ SBDs treated by the H₃PO₄ solution is attributed to the substrate thinning, low ohmic-contact resistance, and low plasma damage. The maximum current density and on/off ratio was 340 A cm⁻² at +2 V and ∼10¹⁰, respectively. The forward bias J–V characteristics were analyzed using the thermionic emission (TE) model expressed as $J={J}_{0}\left[\exp \left(\tfrac{{\rm{eV}}}{nkT}\right)-1\right],$ where J₀ ( $={A}^{* }{T}^{2}\,\exp \left(-\tfrac{e{{\rm{\Phi }}}_{B}}{kT}\right)$) is the saturation current density. The Schottky barrier height Φ_{B_JV} at the metal/semiconductor interface and the Richardson constant A* ($=\tfrac{4\pi e{m}_{n}{k}^{2}}{{h}^{3}}$) are required to fit the exponential portion of the J–V characteristics. Using A* = 33.5 Acm⁻² K⁻², Φ_{B_JV} for the Ga₂O₃ SBDs treated by the H₃PO₄, piranha, and KOH solutions are calculated to be 0.97, 0.99, and 1.00 eV, respectively, close to Φ_{B_CV}. n of the Ga₂O₃ SBDs treated by the H₃PO₄, piranha, and KOH solutions are 1.03, 1.21, and 1.35, respectively. The Ga₂O₃ SBDs treated by the H₃PO₄ solution have n close to unity, implying that the nearly ideal J–V characteristic is realized by removing the plasma damages. In these devices, we could not clarify the relation between the electrical properties and the plasma-damage area, i.e. the mesa top surface or sidewall. Further investigation using smaller-size electrodes is necessary.

Zoom In Zoom Out Reset image size

For the reverse bias J–V characteristics, we use the thermionic field emission (TFE) model expressed as ${J}_{TFE}=\tfrac{{A}^{* }Te\hslash E}{k}\sqrt{\tfrac{\pi }{2{m}_{n}kT}}$ $\exp \left[-\tfrac{1}{{kT}}\left(e{{\rm{\Phi }}}_{B}-\tfrac{\alpha {\left(e\hslash E\right)}^{2}}{{m}_{n}{\left({kT}\right)}^{2}}\right)\right],$^29,30) where E is the electric field at the metal/semiconductor interface ( $E=\sqrt{\tfrac{2e({N}_{d}-{N}_{a})\left({V}_{d}-V\right)}{{\varepsilon }_{s}{\varepsilon }_{0}}}$). V_d and N_d − N_a are obtained from C–V measurements, and Φ_B is obtained from the forward bias J–V measurements. The calculated curves using the TFE model are shown as a solid line at the reverse-bias side in Fig. 4. The TFE model is very close to the experimental data with the H₃PO₄ treatment, indicating that the reverse current is the intrinsic characteristics of the Ni/Ga₂O₃ Schottky contact. The SBDs treated by piranha and KOH solutions had the large leakage current, implying that the plasma damage remains. The current density of Ga₂O₃ SBDs increased with increasing reverse bias and reaches 1 A cm⁻² at the reverse bias around 300 V, corresponding to the TFE model. We consider that the leakage current derived from the plasma damage, defects and surface charges³¹⁾ are negligibly small for β-Ga₂O₃ (001) SBDs treated by the H₃PO₄ solution. Further reduction of the reverse current density could be achieved using an anode electrode with larger Φ_B and a Ga₂O₃ epilayer with a lower donor concentration.³²⁾

To summarize this work, we investigated dry and wet etchings of β-Ga₂O₃ and fabricated vertical Schottky barrier diodes (SBDs) with mesa termination using an optimal etching condition. Using an inductively-coupled plasma reactive-ion etching with a nickel-hard mask, a β-Ga₂O₃ (010) mesa structure is obtained at an etch rate of 77 nm min⁻¹ in BCl₃/Cl₂ mixture gas, showing a smooth sidewall. We compare various acid and alkali solutions to remove the plasma damage. By immersing the SBDs in hot phosphoric-acid solution, the differential on resistance and ideality factor of β-Ga₂O₃ (001) vertical SBDs with mesa termination is reduced to 0.91 mΩcm² and 1.03, respectively. Current density at the reverse bias is in good agreement with thermionic field emission model.

Acknowledgments