Multi-Cycle High Temperature Rapid Thermal Annealing for Dislocation Elimination in 4H-SiC Epitaxy

Abstract

High power devices fabricated in 4H-silicon carbide (SiC) epilayers are known to suffer from forward voltage degradation caused by basal plane dislocation (BPD) induced stacking faults (SF) [[1], [2]]. Over the past decade, several efforts have been successful in removal of more than 90% of the BPDs in the substrate by converting them to benign threading edge dislocations (TED) during epigrowth [[3], [4]]. However, BPDs in the many areas of the epilayers are still > 100 cm-2, which adversely affects the device yield and reliability. We have developed a technique to remove the existing BPDs from the epilayers by using a novel multi-cycle rapid thermal annealing (MRTA) process [[5]], where the BPDs in the epilayers were eliminated, while preserving the epitaxial surface.We used 4 offcut, 15 mm, n-type 6E15 4H-SiC epilayers, commercially obtained. The as-grown samples were imaged using ultraviolet photoluminescence (UVPL) imaging. Carriers were excited with the 334 nm line from an Ar-ion laser, and images were collected in the emission range of 600-1000 nm. The samples were graphite capped and annealed at 1750 oC-1875 oC for a total time of ~5 minutes in 20 cycles. The annealing was performed in a custom designed chamber with inductive sample heating and ultrapure N2 overpressure at 0.41MPa. The base temperature was maintained near 1500 oC, from where the temperature was ramped up and down to the desired temperature in rapid cycles to get an accumulated time of 5 mins at the desired temperature. The combination of high N2 overpressure and rapid cycles enabled us to perform the anneals at a higher temperature than conventional methods, while preserving the graphite cap, and thus maintaining a pristine epitaxial surface. Post-annealed samples were also imaged to observe the BPDs in the epilayers. The UVPL images of the as grown samples showed an average BPD density ~1000 /cm2. We wanted to achieve two goals with our annealing efforts, which were to eliminate BPDs in the epitaxial layers and to preserve the epitaxial surface morphology. We conducted several previous attempts to anneal similar samples using high temperature microwave annealing as well as annealing under high N2 pressure. From our work with microwave annealing, we were able to eliminate BPDs by annealing at above 1750 oC. However, the surface morphology was poor. The graphite capping technique was tweaked but the microwave annealing performed at ambient pressure conditions was not able to achieve a good sample surface. Annealing under high pressure preserved the surface, however, could not go to temperatures higher than 1750 oC without damaging the cap. For complete BPD elimination much higher temperature annealing was required. Thus, the novel MRTA process was adopted, which went to higher temperatures and preserved the graphite cap. Results from the post annealed samples using the different techniques would also be presented.The UVPL images of the samples after the MRTA process indicated a surface without any significant pits or generation of any new BPDs during the annealing process, which has been shown to happen using conventional annealing at these high temperatures. Almost all the BPDs that were previously present disappeared in the epilayer. It is proposed that during annealing BPDs that had not yet converted to threading edge dislocations (TED) during epigrowth forms a TED near the surface. Then the TEDs at the end of all the BPDs glide in a prismatic plane causing the BPD to shorten. Since this glide doesn't occur in the lower energy basal plane, the required energy for this glide is provided by the very high annealing temperature. This way a BPD segment converts to a TED with the TED glide and thus shortens the BPD. The UVPL images of the annealed samples also show a dot where the original BPD had propagated from the substrate, which represents a TED.[[1]] J. P. Bergman, Mater. Sci. Forum 353, 299 (2001)[[2]] R. E. Stahlbush, Mater. Sci. Forum 389, 427 (2002)[[3]] S. Ha, Phys. Rev. Lett. 92, 175504 (2004)[[4]] R. L. Myers-Ward, Mat. Sci. Forum 615-617, 105 (2009)[[5]] T. J. Anderson, Elect. Lett. 50, 197 (2014)Figure: Typical UVPL image of as grown sample with BPD density ~1000 /cm2. The short horizontal white lines are BPDs. Lower image is sample after MRTA at 1875 oC for 5mins, with no BPDs. The long scratch line is scribe mark on backside.