Author: Darryl A. Gleason

Publisher:

Published: 2019

Total Pages: 178

ISBN-13:

In this work, scanning probe microscopy (SPM) methods are developed and extended to spatially resolve performance-hampering electrically-active defects, known as traps, present in AlGaN/GaN Schottky barrier diodes (SBDs) and high electron mobility transistors (HEMTs). Commercial devices used in these studies were cross-sectioned to expose electrically-active regions which are traditionally inaccessible to SPM techniques. Surface potential transients (SPTs) are collected over the cross-sectioned faces of devices using nanometer-scale scanning probe deep-level transient spectroscopy (SP-DLTS), a millisecond time-resolved derivative technique of scanning Kelvin probe microscopy (SKPM) that was implemented with a custom system designed to study SBDs and HEMTs in cross-section. Detected SPTs are indicative of carrier emission from bulk defect-related trap states. In conjunction with similar measurements of these trap states using macroscopic techniques, finite-element simulations provide strong, corroborating evidence that observable SPTs are produced by traps located in the bulk of these samples and are therefore not a result of surface states or surface-related phenomena. GaN-based materials offer advantages over many alternatives in high-frequency and high-voltage applications. Features including a wide bandgap and a large breakdown voltage often translate to improved efficiency, performance, and cost in many electronic systems. However, GaN-based material research is still maturing, and charge trapping may be a limiting factor in GaN electrical performance and therefore hinder its widespread application and adoption. Determining the signatures and spatial distributions of active traps in GaN devices is critical for understanding trap-related mechanisms of device failure as well as the growth or fabrication steps which may be responsible for introducing these defect states. Powerful techniques like deep-level transient spectroscopy (DLTS) exist for identifying specific traps in GaN, but the macroscopic variants of DLTS measure averaged trapping characteristics and are unable to precisely spatially locate the traps they measure. SP-DLTS is an extension of atomic force microscopy (AFM) and was developed approximately seven years prior to this writing. The technique uses SKPM to measure the local surface potential which is sensitive to modulations in the local trapped charge. Probing and analyzing the temperature-dependent SPTs using the same approach applied in the aforementioned conventional techniques reveals the signatures of traps which dominate the local SP-DLTS signal. Performing this measurement over a grid of locations (i.e. a map) provides nanometer-scale resolution of transients and therefore active trap modulation. However, device geometry is one primary limitation of plan-view or "top-down" SP-DLTS due to the sensitivity of the technique only to near-surface charge. Device features like electrodes can mask or electrically screen traps located in active device regions. Furthermore, in commercial devices like those studied here, metallic and passivation layers bury, screen, and/or mask traps in many device regions and completely prevent SP-DLTS probe access. Here, commercial AlGaN/GaN SBDs and HEMTs are cross-sectioned to expose their length and depth with sufficiently low surface damage to permit electrical access to traps beneath the cross-sectioned surface. SP-DLTS is used to detect and identify two distinct trap species with energies near EC − 0.6 eV and EC − 0.9 eV. Unlike macroscopic techniques, SP-DLTS affords trap studies under arbitrary bias conditions; the measurements indicate that trap occupancy modulation is observable during both the device on- and off-state, the latter of which is generally unreported in the literature since macroscopic techniques typically measure trap emission during the device on-state. In addition to qualitatively reproducing these experimental results, finite-element HEMT simulations reveal that current leakage mechanisms and the dopant-to-trap ratio in the GaN buffer likely strongly influence the signatures of detected traps by DLTS-based techniques. Collectively, this experimental and computational approach makes a significant advancement in the study and characterization of traps in AlGaN/GaN devices.