The impact of pressure and abrasives on the subsurface damage for chemical mechanical polished (010) β-Ga₂O₃ substrates was assessed. A combination of 1 kPa of applied pressure and colloidal silica was found to be necessary for achieving both smooth surfaces (< 0.5 nm rms) and subsurface-damage-free material. The as-received grounded surfaces of (010) Tamura substrates were lapped and chemical mechanical polished to a damage-free state. Symmetric (020) triple-axis X-ray diffraction rocking curves were employed to assess the subsurface lattice damage by measuring diffuse scatter intensity (i.e. lattice damage from lapping/polishing) quantified by rocking curve widths below the full width at half maximum (FWXM where X < 0.5) [1,2]. The rocking curve FWHM and FW(0.001)M for the as-received ground surface were ~180” and ~8300”, respectively. These broad widths correspond to cracks, voids, and dislocations induced by the grinding process. The substrates were lapped with 5 μm alumina particles, then lapped with 0.3 μm alumina particles in water. Colloidal alumina in NaOCl was then used to polish, which was followed by colloidal silica in NaOH. A final cleaning step used abrasive-free diluted bleach and citric acid to remove residual silica on from the surface [3]. Compared to our previous work that used diluted bleach and citric acid to polish various III-V materials [1,2], we show this chemistry combination is inert for β-Ga₂O₃ but is effective in cleaning off colloidal silica particles and other forms of silicon on the substrate surface. The final rocking curve FWHM and FW(0.001)M were ~16” and ~150”, respectively after using the colloidal silica, which matches the widths of the as-polished commercial wafers. Material removal rates were measured for each lapping and polishing steps: 5 μm alumina particles yielded a removal rate of ~20 μm/hr, and colloidal silica yielded the slowest ~0.4 μm/hr. The depths of the subsurface damage induced by: lapping with 5 μm alumina particles was ~20 μm and lapping with 0.3 μm alumina was ~6 μm. Determining both the removal rates and depth of subsurface damage is crucial in optimizing lapping and polishing recipes.

The authors would like to acknowledge the support from the Office of Naval Research through a MURI program, grant No. N00014-18-1-2429.

References

1. S. Hayashi, et al., ECS Trans., 16(8), 295 (2008).

2. S. Hayashi, et al., J. Electrochem. Soc., 155(2), H113 (2008).

The formation of insulating layers by acceptor dopants is important for β-Ga₂O₃ device fabrication. Commonly used acceptors in Ga₂O₃ are Mg, Cr, Zn and Cu. Typically the acceptors are introduced during growth, or by ion implantation. Acceptor incorporation by indiffusion is also viable but controlling the depth and concentration can be challenging. In this study, we utilize a combination of experimental techniques and first-principles calculations to identify the mechanism for Zn indiffusion, and to assess the prevailing defect configurations and their thermal stability.

Zn has been introduced into (001) and (-201) oriented β-Ga₂O₃ through vapor phase in sealed evacuated quartz ampules heated to temperatures in the range 900-1100°C, for 1 h. The concentration of Zn as a function of depth was measured using Secondary Ion Mass Spectrometry. The Zn concentration was found to reach a value of approximately 10²⁰ cm^-3, before it drops off abruptly, i.e., the concentration versus depth profile resembles that of a box profile. The profiles were successfully fitted by a trap-limited diffusion model, assuming that interstitial Zn (Zn_i) is the mobile specie, which then is trapped and released through dissociation from a hereto unknown defect. From this model, Zn_i migration barriers of 2.0±0.1 eV and 2.2±0.1 eV were extracted for the (001) and (-201) orientations, respectively, with corresponding dissociation energies of 3.2±0.8 eV and 3.0±0.5 eV.

Through first-principles calculations it was found that the Zn_i donor favors a split-interstitial configuration in which it shares a tetrahedral lattice site with Ga. The calculated migration barrier was found to be 2.3 eV in both [001] and [-201] directions, which is in line with the extracted values from the trap-limited model. Ga vacancies are known to trap donor impurities in β-Ga₂O₃ and is a candidate for the trap. However, the Zn_Ga acceptor is predicted to have a dissociation energy of 7.0 eV, making it highly stable and unlikely to dissociate once formed. Interestingly, the Zn_Ga acceptor can trap a second Zn_i on a Ga site, forming the donor complex Zn_iZn_Ga. The dissociation energy of this complex was found to be 3.7 eV. After Zn indiffusion, the samples where highly conductive, suggesting that the dominating defect is a donor, e.g., Zn_iZn_Ga. After a second heat treatment at 1000°C under O₂ flow, the samples turned insulating. This is consistent with Zn_iZn_Ga donors being the dominant defect after indiffusion, as such complexes would be expected to dissociate during the post-diffusion anneal, leaving behind compensating Zn_Ga acceptors.

β -Ga₂O₃ has received widespread attention due to its ultrawide bandgap, which can enable applications under extreme conditions. Ultrafast laser irradiation of β -Ga₂O₃ provides a means to explore the response of the material under these extreme conditions, which can generate point defects as well as modify structural features with electronic properties that differ from the pristine surface. However, an understanding of defects generated by femtosecond laser irradiation in the vicinity of laser induced periodic surface structures (LIPSS) remains to be explored. We correlate topographic features with the presence of defects and relative crystallinity using depth- and spatially resolved cathodoluminescence spectroscopy (DRCLS).We also explore how these defects change work function and associated band bending using Kelvin Probe Force Microscopy (KPFM). Defects are found to correlate with crystalline order and near-surface morphology, even in morphologically flat areas, thus a factor in building many applications such as ultrashallow ohmic contacts and quantum dot patterning.

Scanning electron microscopy, AFM, KPFM and DRCLS provide near-nanometer scale depth and spatially resolved information that reveals the interplay between topography, lattice disorder, and defect formation in the vicinity of LIPSS formed by femtosecond laser irradiation. Reduction in crystalline order can be correlated with the concentration of a defect with an ~2.4 eV optical emission but is also suggested to be affected by defects not observed by DRCLS, including a continuum of states increasing towards the band edges, typical of amorphous semiconductors. Thermodynamic calculations currently available suggest oxygen interstitials and divacancy complexes as likely candidates for the ~2.4 eV emission. Furthermore, laterally resolved CLS reveals that defects can be generated even in the region surrounding the LIPSS. KPFM measurements along with spatially resolved CLS reveal differences in the work function and crystalline order between crests and troughs, lending support to the hypothesis that the LIPSS are formed via the diffusion of point defects because of a relaxation of the surface instability caused by ultrafast laser irradiation. These results suggest that defect formation is an intrinsic feature of ultrafast laser irradiation and LIPSS formation in Ga₂O₃, which has extensive implications for both exploring the robustness of this material under extreme conditions as well as for device fabrication involving LIPSS. This work supported by AFOSR grants FA9550-18-1-0066, FA9550-20-1-0278, and FA9550-16-1-0069 and NSF grant DMR-18-00130.