Process-variation- and random-dopant-induced static noise margin fluctuation in nanoscale CMOS and FinFET SRAM cells

Abstract

In this study, a three-dimensional "atomistic" coupled device-circuit simulation approach is advanced to investigate the process-variation-effect (PVE) and random dopant fluctuation (RDF) induced characteristic fluctuations in planar metal-oxide-semiconductor field-effect-transistor (MOSFET) static random access memory (SRAM) from 65nm to 16-nm gate length. Our preliminary results show that the RDF dominates the fluctuation of static noise margin (SNM). As the gate length of the planar MOSFETs scales from 65 nm to 16 nm, the normalized RDF-induced SNM fluctuation increases from 4% to 80%. To reduce the device variability induced fluctuation in circuit, a device with vertical-doping-profile and raised V th is employed. The SNM is 3 times larger than the original 16-nm-gate SRAM. Moreover, the normalized RDF-induced SNM fluctuation is reduced by a factor of 2.67. Additionally, a 16-nm-gate silicon-on-insulator fin-type field-effect-transistor is used to further improve the SNM of SRAM. Due to the superior electrostatic integrity and larger effective device width than planar MOSFETs, the SNM of 16-nm-gate FinFET SRAM is six times larger than the original 16 nm SRAM with five times smaller SNM fluctuation. The study investigates the roll-off characteristics of SNM and provides an insight into design of fluctuation resistant nanoscale SRAM.

Statistical analysis of process variation induced SRAM electromigration degradation

“…Hence, for a given bit-line width, by using both (8) and (16), the existence of EM and functional failure of a generated SRAM array sample can be verified.…”

“…Previous works [8][9] [10] only focused on relationships between process variation and functional failures in SRAM. Some papers considered reliability effects such as NBTI and concluded that slight performance degradation can be expected after 10 years of operation [11].…”