Real-Time In-Situ Chemical Sensing in AlGaN/GaN Metal-Organic Chemical Vapor Deposition Processes for Advanced Process Control

Abstract

Gallium nitride and its alloys promise to be key materials for future semiconductor devices aimed at high frequency, high power electronic applications. However, manufacturing for such high performance products is challenged by reproducibility and material quality constraints that are notably more stringent than those required for optoelectronic applications. To meet this challenge, in-situ mass spectrometry was implemented as a real-time process- and wafer-state metrology tool in AlGaN/GaN/AlN metal-organic chemical vapor deposition processes on semi-insulating SiC substrate wafers. Dynamic chemical sensing through the process cycle, carried out downstream from the wafer, revealed generation of methane and ethane reaction byproducts, as well as other residual gas species. Real-time metrics were derived based on the chemical signals to predict/control material quality and thickness of critical layers within the heterostructure in real time during growth, and corresponding metrologies were used for real-time advanced process control. Using the methane/ethane ratio, GaN epilayer crystal quality was predicted in real time to 2 5% precision, which was verified by post-process x-ray diffraction. Moreover, the same real-time metric predicted material quality as indicated by post-process photoluminescence band-edge intensities to ~5% precision. The methane/ethane ratio has a fundamental significance in terms of the intrinsic chemistry in that the two byproducts are believed to reflect two parallel reaction pathways leading to GaN-based material growth, namely the gas phase adduct formation route and the surface route for direct precursor decomposition, respectively. The fact that lower methane/ethane ratios consistently yield better material quality suggests that the surface pathway is preferred for high quality GaN growth. In addition, a metric based on methane and ethane signals integrated through the AlGaN growth period (~1 min or less) enabled prediction of the cap layer thickness (~20 nm) to within ~1% precision, which was verified by post-process x-ray reflectance. These types of real-time advanced process control activities in terms of fault detection and management, course correction, and pre-growth contamination control have made significant contributions to the GaN-based semiconductor development and manufacturing at Northrop Grumman Electronics Systems in terms of improved material quality, yield, and consequent cost reduction, and they are now in routine use.