A FeFET based processing-in-memory architecture for solving distributed least-square optimizations

Insik Yoon, Muya Chang, Kai Ni, Matthew Jerry, Samantak Gangopadhyay, Gus Smith, Tomer Hamam, Vijayakrishan Narayanan, Justin Romberg, Shih Lien Lu, Suman Datta, Arijit Raychowdhury

Research output: Chapter in Book/Report/Conference proceedingConference contribution

3 Scopus citations


Hf02 based ferroelectric FET (FeFET) has recently received great interest for its application in nonvolatile memory (NVM) [1]. Unlike conventional perovskite based ferroelectric materials, Hf02 is CMOS compatible and retains ferroelectricity for thin film with thickness around 10 nm. Therefore, successful integration of ferroelectric Hf02 into advanced CMOS technology makes this technology highly promising for NVM [1]. Moreover, by tuning the portion of switched ferroelectric domain, a FeFET can exhibit multiple intermediate states, which enables its application as an analog conductance in mixed-signal in-memory computing. Currently, such architectures have been applied to neuromorphic computing [2], [3]. In this paper, we present a processing-in-memory (PIM) architecture with FeFETs and demonstrate how this can be used to solve a new class of optimization problems, in particular, distributed least square minimization.

Original languageEnglish (US)
Title of host publication2018 76th Device Research Conference, DRC 2018
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Print)9781538630280
StatePublished - Aug 20 2018
Event76th Device Research Conference, DRC 2018 - Santa Barbara, United States
Duration: Jun 24 2018Jun 27 2018

Publication series

NameDevice Research Conference - Conference Digest, DRC
ISSN (Print)1548-3770


Other76th Device Research Conference, DRC 2018
Country/TerritoryUnited States
CitySanta Barbara

All Science Journal Classification (ASJC) codes

  • Electrical and Electronic Engineering


Dive into the research topics of 'A FeFET based processing-in-memory architecture for solving distributed least-square optimizations'. Together they form a unique fingerprint.

Cite this