Fri, Feb 17, 2023 @ 02:00 PM - 03:30 PM
Ming Hsieh Department of Electrical and Computer Engineering
Conferences, Lectures, & Seminars
Speaker: Sourav Dutta, Components Research at Intel
Talk Title: Oxide Semiconductor Nanoelectronics for Energy-Efficient Computing
Series: Nano Science & Technology
Abstract: Advancements in deep learning and artificial intelligence (AI) have created a growing demand for domain-specific hardware architecture for accelerating data-intensive workloads. A key source of acceleration as well as energy-efficiency comes from the availability of dense on-chip embedded memory to overcome the memory bottleneck. While silicon transistor scaling continues to make steady progress, resulting in exponential increase in aerial transistor density to over 100 million per square millimeter, static random-access memory (SRAM), the primary workhorse for on-chip embedded memory, is approaching its scaling brick wall. In this talk, I will discuss how amorphous oxide semiconductor can allow us to escape the confines of two-dimensional scaling and embrace the third dimension for continued scaling of logic and embedded memory and enable merged logic-memory fabric. I will highlight the oustanding challenges in oxide semiconductors and discuss strategies towards obtaining back-end-of-line (BEOL) compatible high mobility n- and p-type oxide semiconductor transistor through disorder insensitive material design, band structure engineering and defect suppression. Through low temperature in situ synthesis and integration with functional materials such as ferroelectrics, I will discuss enabling BEOL compatible oxide semiconductor-based high-density embedded memory. Finally, I will discuss how oxide semiconductor-based monolithic 3D logic and memory can enable building specialized hardware tailored for accelerating key mathematical operations pertinent for AI workloads such as performing linear matrix operations or solving nonlinear coupled differential equations.
Biography: Dr. Sourav Dutta is currently a memory device research engineer at Components Research at Intel. His current research focuses on novel high-density embedded memory for Intel\'s next generation technology nodes. He received his Ph.D. in 2018 from the School of Electrical and Computer Engineering at Georgia Institute of Technology under Prof. Azad Naeemi. From 2018 to 2021, he was a postdoctoral research scholar with Prof. Suman Datta at University of Notre Dame. His research focused on emerging nanoelectronics devices with functional materials such as oxide semiconductors and ferroelectrics. He has worked on back-end-of-line (BEOL) compatible high performance oxide semiconductor transistor, high density embedded memory using oxide semiconductor and ferroelectric material, and monolithic 3D integration of logic and memory for compute-in-memory and collective computing applications. Dr. Dutta is the recipient of the 2018 Sigma Xi Best Doctoral Dissertation Award from Georgia Institute of Technology.
Host: J Yang, H Wang, C Zhou, S Cronin, W Wu
Audiences: Everyone Is Invited
Contact: Marilyn Poplawski