Professor Michael O. Thompson

Cornell University

Autonomous High-throughput Exploration and -Ga₂O₃:

Understanding Processing, Phase Formation, and Metastability

For over 40 years, my lab has explored thermal processing of materials on time scales from nanoseconds to hours, focused initially on annealing implant-induced damage in semiconductors but expanding to the understanding of phase formation in general as a function of composition and thermal processing conditions.  In this seminar, I will present results from two current projects: (i) understanding phase formation and silicon implant activation in the -Ga₂O₃ system and (ii) high-throughput automated and autonomous exploration of unary, binary, and ternary oxide systems during millisecond heat and quench. 

Ga₂O₃ is an ultra-wide bandgap semiconductor that promises to dramatically expand high voltage, high power, and high frequency applications for a broad range of energy and communication applications.  The system is also structurally rich with a wide range of known polymorphs, including the -phase for which bulk melt-grown crystals are commercially available.  Doping via silicon ion-implantation is critical in -Ga₂O₃ to form low resistance contacts and selective area doping for advanced device structures.  In contrast to elemental semiconductors, annealing of -Ga₂O₃ is extremely sensitive to point- and extended-defect populations that depend on both time, temperature and the annealing ambient.  By careful control of annealing conditions, we demonstrate the ability to achieve activation of Si to concentrations near 10²⁰ cm^-3 with mobilities above 75 cm²/V-s.  At higher concentrations, metastable -phase forms in the implant damaged regions.  To understand this phase selection, rapid heat and quench of amorphous Ga₂O₃ and related alloys was observed during millisecond Laser Spike Annealing (LSA).  The defect spinel -phase is observed to nucleate first under a wide range of conditions potentially explaining its frequent formation during growth and processing of -Ga₂O₃.

In a collaboration with Prof. van Dover and others, we have combined extensive automation of LSA, binary and ternary composition libraries, and fast structural characterization at CHESS to develop a artificial intelligence Scientific Autonomous Reasoning Agent (SARA) that promises to dramatically accelerate exploration and exploitation of new material systems.  Using lateral gradient LSA with a focused X-ray synchrotron source, we generate and characterize over a thousand unique processing conditions per minute, with upwards of 100,000 conditions on a single ternary compositional library plate.  To autonomously prioritize experiments to develop high-dimensional phase-processing maps at a commensurate rate, SARA utilizes a hierarchical set of characterization and decision algorithms based on nested active learning cycles and machine learning models incorporating underlying physics and end-to-end uncertainty quantification.  Applications of SARA to several materials systems will be shown to demonstrate the efficacy of this approach.  Finally, I will discuss very recent work on time-resolved structural formation in thin films with microsecond resolution.  In addition to conventional nucleation and growth, these studies suggest an alternative continuous transformation between phases occurring on the millisecond timescale.

 

Bio: 

Dr. Thompson received his BS in Applied Physics from CalTech in 1979 and MS/Ph.D degrees in Applied and Engineering Physics from Cornell in 1984.  After completing his Ph.D, he joined the faculty in the Department of Materials Science at Cornell continuing his work on the interaction of materials with intense laser sources.  He has co-authored over 100 journal publications, is co-inventor on 25 patents, and has founded or co-founded three startup companies.  He was the recipient of the 2009 SEMI Award for technical contributions to the semiconductor industry. 



Dr. Thompson’s research has focused primarily on the behavior of semiconductor materials under pulsed and CW laser exposure.  His group has explored limits to crystal growth, metastable impurity incorporation, point defects, interface stability, explosive crystallization, and heteroepitaxy.  He was involved in development of melt-annealing methods to fabricate thin-film transistors on glass and flexible substrates, and the use of CW non-melt laser annealing (Laser Spike Annealing) for ultra-shallow junctions in advanced VLSI.  More recently, his group has focused on implantation and annealing in ultrawide bandgap semiconductors, and on expanding use of non-melt sub-millisecond annealing in combinatorial explorations of metastable phase formation in complex oxides.  To address data explosion in these high-throughput experiments, his group is now heavily invested in AI based autonomous workflows to accelerate identification and optimization of new materials.