Characterization of self-heating leads to universal scaling of HCI degradation of multi-fin SOI FinFETs

Hai Jiang,SangHoon Shin,Xiaohui Liu,Xing Zhang,Muhammad A. Alam
DOI: https://doi.org/10.1109/IRPS.2016.7574506
2016-01-01
Abstract:SOI FinFETs and other Gate-all-around (GAA) transistors topologies have excellent 3-D electrostatic control and therefore, have been suggested as potential technology options for sub-14 nm technology nodes. Unfortunately, the narrow gate geometry and reduced gate pitch suppress heat dissipation and increase thermal cross-talk, leading to severe self-heating of these transistors. Self-heating degrades performance and makes the classical reliability theories based on T <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">L</sub> ~T <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">sub</sub> irrelevant. In this paper, first, we propose a physics-based thermal circuit compact model for multi-fin SOI FinFETs to characterize self-heating and validate the results by AC conductance method. Next, we analyze HCI degradation varying with the number of fin (N <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">fin</sub> ), chuck temperature (T <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">sub</sub> ) and AC frequency (f). The results show that HCI degradation dependent variables (N <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">FIN</sub> , T <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">sub</sub> , f) can be correlated to the lattice temperature (T <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">L</sub> = g(N <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">FIN</sub> , T <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">sub</sub> , f)) and obey the universal degradation curve (ΔV <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">th</sub> (T <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">L</sub> ) = f(S(T <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">L</sub> ) × t)). Si-O bond-dispersion model explains the universal curve; therefore, the model can be used for a long term reliability projection with arbitrary combination N <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">fin</sub> , T <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">sub</sub> , f, etc.
What problem does this paper attempt to address?