A design study on the scaling limit of ultra-thin silicon-on-insulator MOSFETs.

摘要

As bulk CMOS is approaching its scaling limit, SOI CMOS is gaining more and more attentions and is considered as a potential candidate for achieving 10-nm CMOS. Fully-depleted SOI MOSFETs have several inherent advantages over bulk MOSFETs-low junction capacitance, no body effect and no need for body doping to confine gate depletion. This dissertation presents a comprehensive, 2-D simulation-based design study on the scaling limit of ultra-thin silicon-on-insulator MOSFETs.; Starting with the lateral-field analysis of fully-depleted (FD) SOI MOSFETs, it is shown that the general scale-length model is inapplicable for predicting the minimum scalable channel length Lmin when the buried-oxide is very thick. The scaling of FDSOI MOSFETs is independent of the buried-oxide thickness. An empirical Lmin prediction equation is developed by approximating the constant L min contours in a design plane of silicon-film and gate-dielectric thickness. Ultimately, Lmin ∼5tSi with a high-k gate dielectric. Other factors such as body doping, substrate biasing, and buried-insulator permittivity &egr;BOX and bandgap affecting short-channel scaling of FDSOI are also investigated. Empirical Lmin prediction equations are developed for FDSOI devices with body doping and low-k buried-insulators. In principle, the Lmin can be improved from ∼5t Si to ∼2tSi by body doping. The Lmin can also be reduced 15% shorter from &egr;BOX = 3.9&egr;0 to &egr; BOX = &egr;0.; Finally, the scaling limit of FDSOI MOSFETs is discussed. From the electrostatic perspective 10-nm FDSOI CMOS requires scaling both high-k gate-dielectric and silicon-film thickness to their limits of ∼2 nm. However, silicon-film thickness cannot below ∼3 nm to avoid severe mobility degradation. The scaling limit of FDSOI MOSFETs with a feasible HfO2 gate dielectric is then projected to be ∼17 nm. 10-nm FDSOI CMOS can be achieved only if there is a breakthrough on thin silicon-film mobility.