Origin of etch delay time in Cl2 dry etching of AlGaN/GaN structures

Abstract

The etch delay time commonly found during dry etching of AlGaN and GaN has been experimentally proven to be due to the presence of hard–to–etch surface oxides. A ${\mathrm{BCl}}_3$ deoxidizing plasma, followed by a ${\mathrm{Cl}}_2$ etching plasma, was found to give dead-time-free aluminum-mole-fraction-independent etch rates. No selectivity between GaN and AlGaN has been observed up to an aluminum mole fraction of 35%. The aluminum-mole-fraction-dependent etch rates commonly reported in literature have been related to the different dead-times associated with dissimilar surface oxides, disproving the more common explanations in terms of the higher binding energy of AlN compared to GaN and/or the lower volatility of ${\mathrm{AlCl}}_x$ compared to ${\mathrm{GaCl}}_x.$