Exploiting Carbon Nanotube FET and Magnetic Tunneling Junction for Near-Memory-Computing Paradigm

Abstract

The traditional von Neumann computing architecture based on metal-oxide field-effect-transistors (MOSFETs) is more and more incompetent for the increasing demand for computing speed and energy efficiency in the Internet of Things (IoT) and intelligent development. Carbon nanotube field-effect-transistors (CNTFETs) are expected to achieve significant energy efficiency benefits versus today's silicon-based FETs. In this article, we combine CNTFETs and spin-transfer torque magnetic tunneling junctions (STT-MTJs) to build near-memory computing circuits, whose high speed and low power consumption characteristics are shown through the investigation of three nonvolatile logic gates: The read speed of CNTFET/STT-MTJ nonvolatile logic circuit is approximately 40% of the MOSFET/ STT-MTJ analog circuit, and the total power consumption and read energy of the same logic circuit can be approximately saved 17%-37% compared with that of MOSFET/ STT-MTJ analog circuit. Our research shows the advantages of the integration of CNTFETs and STT-MTJs in the nonvolatile logic circuit, which has great significance for the development of beyond-complementary metal-oxide semiconductor (CMOS) electronics.