Influence of mask material on the electrical properties of selective area epitaxy GaN microstructures

In the last two decades, GaN based device technology has emerged as a promising candidate to overcome the limitations of conventional semiconductor technologies in the areas of high-frequency and high-power electronic applications. Due to the lack of native, low cost, readily available substrates, GaN is typically grown heteroepitaxially on foreign substrates such as sapphire, silicon, and silicon carbide. The inherent lattice and thermal mismatch with these substrates causes stress in the heteroepitaxially grown film which results in the formation of dislocations, defects, and cracking, all of which are known to affect the device performance.^1,2 To overcome these issues, multiple reports have been published showing successful development of selective area epitaxy (SAE) based growth techniques resulting in improved crystal quality.^3–6 The homoepitaxial and localized nature of SAE growth technique mitigates the effect of lattice and thermal mismatch stresses while reducing the whole-wafer bowing effect. SAE also enables three-dimensional stress relaxation of the overgrown crystal structure in both vertical and lateral growth directions.⁷ These advantages have led to a high interest in the SAE process for the growth of large scale light-emitting diodes and high-power electronic devices.^8–10 

Multiple reports have been published that focus mainly on the optical properties of structures grown by SAE technique.^11–14 Despite the progress made in this area, only a small body of research has reported on the electrical properties of structures grown by SAE technique.^15,16 Recently, metal-like conductivity in Si-doped GaN microwires grown selectively on N-polar GaN has been demonstrated.¹⁶ The high dopant incorporation was attributed to the polarity of the underlying N-polar GaN film. In our work, we observe high electrical conductivity in SAE grown Ga-polar structures, suggesting little influence from the polarity of the crystal. Knowledge of the factors causing such high electrical conductivity in SAE GaN is essential for controlling material properties and to enable the design of device quality structures. Although the crystal quality of the structures grown by SAE technique has been repeatedly shown to be better than conventional planar structures, the effect of mask material and a good understanding of the factors responsible for the high electrical conductivity in the SAE structures is lacking. In this study, we report on the electrical properties of GaN microstructures grown by SAE technique using both dielectric and metal masks. The effect of mask selection on the impurity incorporation and electrical behavior of the unintentionally doped (UID) structures has been systematically studied. Highly conductive behavior showing low resistivity in the range of 0.18–0.29 mΩ cm and carrier concentrations greater than metal–nonmetal (MNM) transition level was observed in SAE microstructures grown on Ga-polar GaN films on sapphire substrates.

Structures used in this study were grown using a vertical Veeco metal organic chemical vapor deposition (MOCVD) system. Approximately 3.5 μm of UID GaN film on the sapphire substrate was used as an underlying template for the SAE overgrowth. The GaN template had a background carrier concentration of 5 × 10¹⁶ cm⁻³, electron Hall mobility of 332 cm²/V s, and bulk resistivity of 0.323 Ω cm. Samples were patterned with two types of mask material: a SiO₂/SiN_x dielectric mask and a tungsten (W) metal mask. Plasma enhanced chemical vapor deposition technique was used to deposit 150 nm SiO₂/150 nm SiN_x mask layer stack, and electron-beam evaporation was used to deposit 60 nm of W mask on the GaN templates. Samples with SiO₂/SiN_x mask were patterned and dry etched with plasma assisted reactive ion etching followed by buffered hydrofluoric acid wet etching to create mask openings. For samples patterned with W mask, H₂O₂ based wet etching was performed at room temperature. After etching, the samples were cleaned with organic solvents before loading into the MOCVD reactor. Mask window openings for SAE growths were 50 × 500 μm rectangular and 450 × 450 μm square microstructures. SAE GaN (∼1 μm) microstructures were grown on both mask types. Growths were performed at a reactor pressure of 100 Torr and a growth temperature of 1040 °C. Trimethylgallium (TMGa) and ammonia were used as group III and V sources with a V/III ratio of 3000 and TMGa flow rate of 102 μmol/min. Around 15 000 sccm of hydrogen was used as the carrier gas and 8000 sccm of nitrogen was used as a dilution gas. The growth parameters were kept identical for planar and SAE GaN samples. On samples with different mask materials, growths were performed separately. The details of the samples used in this study are shown in Table I.

To compare the electrical properties of SAE and planar samples, two-terminal device structures were fabricated. Isolation mesas were created on planar samples using low power BCl₃/Cl₂/Ar based inductively coupled plasma dry etching. Ohmic contacts were fabricated on the 50 × 500 μm rectangular microstructures of both SAE and planar samples. The ohmic contacts consisted of electron-beam evaporated Ti (15 nm)/Al (60 nm)/Mo (35 nm)/Au (60 nm) metal stack which was RTA annealed at 850 °C for 30 s in nitrogen ambient. On sample C, the W mask was etched away with H₂O₂ before fabrication of ohmic contacts. Device structures were protected with a resist mask during the wet etching process. To perform Hall measurements on the SAE microstructures, 50 × 50 μm Ti (60 nm)/Al (300 nm) contact pads were deposited by electron-beam evaporation at four corners of the 450 × 450 μm squares. Surface morphology of the SAE structures was characterized by atomic force microscopy (AFM) imaging with Bruker Dimension Icon scanning probe microscope and scanning electron microscope (SEM) imaging with LEO (Zeiss) 1550. Two-terminal current–voltage (I-V) measurements were performed using a Keithley 4200 parameter analyzer. Secondary ion mass spectroscopy (SIMS) was performed with PHI 6650 to profile the concentration of C, O and Si.

AFM scans of 5 × 5 μm on planar and SAE samples showed microscopically smooth and flat surface morphology. RMS roughness of 0.4, 2.6, and 2.09 nm was obtained in samples A, B, and C, respectively. As measured by SEM and shown in Fig. 1, growths with SiO₂/SiN_x mask showed high mask/deposition selectivity, whereas the growths with W mask showed roughening of the mask and polycrystalline deposition on the mask's surface. SEM images of the fabricated rectangular SAE microstructures are shown in Fig. 2.

Room temperature two-terminal I-V characteristics of the devices are shown in Fig. 3. Higher electrical conductivity was observed in SAE GaN devices in comparison to planar GaN device structures. In SAE devices with both mask types, the maximum current density was 40× times higher than in devices from planar GaN sample.

To extract the electrical parameters of the SAE samples, room temperature van der Pauw Hall effect measurements were performed on 450 × 450 μm square structures. The carrier concentration (η), mobility (μ), and resistivity (ρ) obtained from these measurements are listed in Table II.

The measured electrical properties indicate the presence of high electron concentrations due to unintentional n-type doping in the SAE growths. Sources of unintentional n-type conductivity in GaN are donor impurities such as O, Si, and native defects such as N-vacancies (V_N).¹⁷ Since V_N has high formation energy in n-type GaN, incorporation of Si and O donors is more likely due to lower incorporation energy which is further reduced when the Fermi level position is close to conduction band.¹⁸ To measure the concentration of unintentionally incorporated dopants, dynamic SIMS profiling was performed on the 450 × 450 μm square structures of SAE samples and the planar GaN template, the results of which are listed in Table III. The typical analysis area of SIMS was 100 × 100 μm at the center of the square microstructures.

In the planar UID GaN template, Si concentration was below the detection level. In the SAE GaN, C levels were comparable to the planar GaN; however, Si and O concentrations of 7 × 10¹⁸ cm⁻³ were measured for samples B and C, respectively. Although SiO₂/SiN_x mask shows high mask/deposition selectivity, high concentration of Si in sample B indicates incorporation of Si from the mask into the crystal structure. In case of sample C, O from the electron-beam evaporated W is one potential source. The interactions between the MOCVD precursors and the mask materials are not well established. Although there have been some reports on the detection of Si in SAE growths with SiO₂ mask,^11,19 the quantification of the unintentionally incorporated dopants from the SiO₂/SiN_x and W masks and the effect of these mask materials on the electrical behavior of SAE GaN structures has not been reported previously. From Tables II and III, it is evident that the high conductivity observed in SAE structures is due to unintentionally incorporated Si and O dopants. Comparing concentration of dopants from SIMS and Hall measurements in sample B ([Si] = 7 × 10¹⁸ cm⁻³; η = 8 × 10¹⁸ cm⁻³) and sample C ([O] = 7 × 10¹⁸ cm⁻³; η = 1.59 × 10¹⁹ cm⁻³), the dopant concentrations and the carrier concentrations are quite comparable. Theoretical findings have shown that formation energies for compensating acceptors such as Ga-vacancies and substitutional C in N-site are low in n-type GaN, making them energetically favorable to form in large concentrations.^20–22 However, our findings show that the compensation in SAE growths is quite low and nearly all dopants are activated at room temperature. The measured carrier concentrations in the SAE structures are shown to be higher than the MNM transition level for n-type GaN. The critical concentration of uncompensated donors for the MNM transition has been theoretically predicted and experimentally demonstrated to be 1.6 × 10¹⁸ cm⁻³.^23,24 Above the MNM concentrations, the energy for dopant activation/incorporation is shown to decrease drastically and the activation energy for electron transport in the GaN conduction band vanishes causing semiconducting material to transition into metallic state. The large influx of adatoms on the growing 3D structure (as opposed to a 2D growth for a planar structure) could be a key factor in influencing the high dopant incorporation in substitutional donor sites and in the reduction of the compensating acceptors. Given the complex nature of SAE growth kinetics, it is likely that the formation of compensating acceptors is supressed at growth temperatures for concentrations above the MNM transition.

The findings of this study further emphasize the need for development of a stable mask material along with development of a greater understanding of the key factors that influence compensation and high dopant incorporation in III-nitride SAE growths. Such understanding will pave the way to developing high performance and high efficiency devices.

The electrical properties of selective area epitaxially grown GaN microstructures with dielectric (SiO₂/SiN_x) and metal (W) masks have been systematically studied and reported in this work. High electrical conductivity with carrier concentrations higher than the metal-nonmetal transition levels were observed in SAE microstructures. SIMS measurements indicate the presence of high unintentional Si doping in SAE growths with SiO₂/SiN_x mask and high unintentional O doping in SAE growths with W mask. It is shown that dopant concentrations from SIMS are similar to the carrier concentrations measured from Hall, suggesting minimal compensation by acceptors in SAE structures.

This work was supported by a Grant No. W911NF-16-2-0031 from U.S. Army Research Laboratory. The authors acknowledge valuable contributions from Niaz Mahmud for assistance with preparation of the Hall samples and Michael Yakimov for electrical measurements and many helpful discussions.