Abstract
Though wide bandgap semiconductors offer superior performance to its Si based counterpart, the current state of the art manufacturing technology produces several defects preventing devices from performing optimally. Particularly in SiC, the methods for detecting extended defects such as threading edge dislocations (TED), threading screw dislocations (TSD), basel plane dislocations (BPD), stacking faults, and polytype inclusions are well established; however, automated quantitative analysis is challenging due to the variable size, shape, and intensity of these numerous defects. This study focuses on developing machine learning models using multiple measurements with different techniques including x-ray topography (XRT) and ultraviolet photoluminescence (UVPL) to locate and quantify the microscopic defects on a macroscopic scale.
R. E. Stahlbush
U.S. Naval Research Laboratory
-
10B.1 – Mapping Defects in SiC Wafers Using a Multi-Channel Convolutional Neural Network
James Gallagher, U.S. Naval Research LaboratoryN. Mahadik, U.S. Naval Research LaboratoryR. E. Stahlbush, U.S. Naval Research LaboratoryKarl D. Hobart, U.S. Naval Research LaboratoryM.A. Mastro, U.S. Naval Research Laboratory