Eduard A Cartier
Analog devices for Artificial Intelligence: Device requirements, device concepts and reliability challenges
Specialized hardware for deep learning using analog memory devices has the potential to outperform conventional GPUs by a large margin. At the core of such hardware are arrays of non-volatile-memory (NVM) devices that can perform the simple matrix operations needed for deep learning in parallel and in constant time. Several implementations can be found in the literature that use different materials as memory elements, including phase-change-memory (PCM), resistiverandom-access-memory (RRAM), electrochemical-randomaccess-memory (ECRAM), and ferroelectric devices. While the current focus is to demonstrate functionality, there is an increasing concern about the reliability margins of this emerging technology. In this presentation, we will describe device requirements, practical implementations, and discuss possible reliability exposure in terms of variability, stability and drift, retention and durability.
Eduard A. Cartier earned the B.E.E degree at the catholic boarding school "Collegium Maria Hilf" in Schwyz, Switzerland and the M.S and Ph.D degrees from the Swiss Federal Institute of Technology in Zurich, Switzerland. In his master thesis he studied the electronic structure of metallic glasses using high resolution X-ray photoelectron spectroscopy and inverse photoelectron spectroscopy. His Ph.D thesis reported on the electronic structure and on atomic defects in graphite intercalation compounds and in metallic glasses as measured with positron annihilation techniques. He received the silver medal of the Swiss Federal Institute of Technology for his outstanding Ph.D thesis. From 1986 to 1988 he worked at the ABB Research Center in Baden-Dattwil, Switzerland. While at ABB, his research centered around hot carrier transport in organic materials such as saturated long chain hydrocarbons and polymers and on dielectric breakdown of polymer insulators in high power applications. Since 1989, he is a research staff member of the IBM Research Division at the T.J. Watson Research Center in Yorktown Heights, NY, USA. His research activities at the T.J. Watson Research Center concentrated around hot carrier transport in silicon and silicon dioxide and around hot carrier induced oxide degradation relevant to FET in CMOS technologies and nonvolatile memory device operation and on the understanding of basic processes relevant to dielectric breakdown. Of particular interest to him was the investigation of the role played by hydrogen in the hot-carrier-induced oxide degradation processes leading to dielectric breakdown and device failure. He obtained two IBM internal awards for outstanding contributions in the field of hot carrier transport in insulators and in the field of oxide reliability predictions for electronic devices. For the last ten years, his work concentrated on the development of CMOS devices with high-k dielectrics and metal gates in the gate stack as a replacement for traditional SiO2/ploy-Si gate stacks.