Event
MSE Seminar: Dr. Wendy L. Sarney, Army Research Laboratory
Wednesday, April 3, 2024
3:30 p.m.
Room 2110 Chemical and Nuclear Engineering Building
Sherri Tatum
301 405 5240
statum12@umd.edu
Ferroelectric Hafnia Materials Development for Neuromorphic Applications
Abstract: We explore monolithic processing and the integration of hafnia-based ferroelectrics into neuromorphic devices. The demonstration of complementary metal organic semiconductor (CMOS)-compatible ferroelectric hafnia (HfO2) in 2011 prompted a surge of research into integrating ferroelectric-based devices with advanced semiconductor technology nodes. Current research explores and designs devices for new computing and embedded memory applications. Neuromorphic computing is a brain-inspired computing paradigm to emulate neurological activities with novel in-memory computing devices. Computers excel at analytical math and the processing of information through model development. This power-hungry computation mode does not meet the demands of edge applications for networks deprived of energy and communications. Alternatively, cognitive computing operates similarly to the human brain. Trainable networks process patterns of information and are sensitive to the details of input patterns. Edge computing allows local data processing, eliminating the need for centralized processing in contested and limited network paradigms. This reduces the latency of the network and improves resiliency, power efficiency, and speed for time-sensitive tasks.
In our lab, we study the underlying physics of switching in ferroelectric hafnia-based materials and optimize their properties for these types of applications. I will emphasize the underlying materials characterization of the phase landscape in hafnium zirconium oxide (HZO) ferroelectrics. We use rapid thermal annealing processes to achieve monolithic integration of HZO into two-dimensional field effect transistors (FeFETs). Direct imaging and spectroscopy reveal the mixed phase, multi-domain nature of HZO with small grain sizes between 10 and 20 nm. Small grains are potentially helpful for scaling analog properties into advanced CMOS nodes. Thermal processing at CMOS-compatible temperatures increases crystallinity and further aligns the phases. Ordered ferroelectric orthorhombic phases are present in the material with CMOS-compatible thermal processing temperatures.
