High field effects of GaN HEMTs.

Abstract

This report represents the completion of a Laboratory-Directed Research and Development (LDRD) program to develop and fabricate geometric test structures for the measurement of transport properties in bulk GaN and AlGaN/GaN heterostructures. A large part of this study was spent examining fabrication issues related to the test structures used in these measurements, due to the fact that GaN processing is still in its infancy. One such issue had to do with surface passivation. Test samples without a surface passivation, often failed at electric fields below 50 kV/cm, due to surface breakdown. A silicon nitride passivation layer of approximately 200 nm was used to reduce the effects of surface states and premature surface breakdown. Another issue was finding quality contacts for the material, especially in the case of the AlGaN/GaN heterostructure samples. Poor contact performance in the heterostructures plagued the test structures with lower than expected velocities due to carrier injection from the contacts themselves. Using a titanium-rich ohmic contact reduced the contact resistance and stopped the carrier injection. The final test structures had an etch constriction with varying lengths and widths (8x2, 10x3, 12x3, 12x4, 15x5, and 16x4 {micro}m) and massive contacts. A pulsed voltage input and a four-point measurementmore » in a 50 {Omega} environment was used to determine the current through and the voltage dropped across the constriction. From these measurements, the drift velocity as a function of the applied electric field was calculated and thus, the velocity-field characteristics in n-type bulk GaN and AlGaN/GaN test structures were determined. These measurements show an apparent saturation velocity near to 2.5x10{sup 7} cm/s at 180 kV/cm and 3.1x10{sup 7} cm/s, at a field of 140 kV/cm, for the bulk GaN and AlGaN heterostructure samples, respectively. These experimental drift velocities mark the highest velocities measured in these materials to date and confirm the predictions of previous theoretical models using ensemble Monte Carlo simulations.« less