Positive Bias Temperature Instability Effects in advanced High-k / Metal Gate NMOSFETs

Abstract

The recent impressive progress in the performance of High-k/Metal Gale (HKMG) transistors and the complicated nature of the multi layer gate material have brought to the forefront of High-k/Metal Gate technology development the issue of reliability and in particular the Positive Bias Temperature Instability (PBTI) [1-7]. Most of theses studies showed a significant positive threshold voltage shift for n-channel MOSFET's subjected to positive bias temperature stressing, which was attributed to the preexisting electron traps in the high-k layer. However, other authors have argued that voltage stress-induced defects generation may also contribute to the device threshold voltage instability [5-7], The effectiveness of stress-induced trap generation in these gate stacks along with the time and voltage dependence of the PBTI induced device degradation still remain controversial issues with the conclusions varying with the applied techniques, test conditions and the gate stack process details. To complicate matters more, conclusions derived from the widely employed high voltage stresses of short duration may no longer be valid for the much lower actual device operation bias conditions. In this work, we propose a novel methodology, where a fast measurement technique for threshold voltage instability characterization (to minimize inaccuracy of the measurements due to the well known relaxation phenomenon [8]) is combined with a fast switching modular level long term stress capability. This approach allowed for the assessment of PBTI for a wide range of stress voltages including those very close to the transistor operational voltage. To the best of our knowledge, this is the first time such a thorough PBTI characterization is reported in the literature. We find that although high voltage stress conditions may favor electron trapping as the predominant aging mechanism, as stress bias reduces, another mechanism "kicks in", namely trap generation. Based on these findings, a preliminary semi-empirical PBTI model is proposed which is found to agree well with the obtained experimental data.