The International Technology Roadmap for Semiconductors outlines the need for high-K dielectric based gate-oxide Metal Oxide Semiconductor Field Effect Transistors for sub-45 nm technology nodes. Gate oxides of hafnium seem to be the nearest and best alternative for silicon dioxide, when material, thermal and structural properties are considered. Usage of poly-Si as a gate electrode material degrades the performance of the device and hence gate stacks based on metal gate electrodes are gaining high interest. Though a substantial improvement in the performance has been achieved with these changes, reliability issues are a cause of concern. For analog and mixed-signal applications, low-frequency (I /f~ noise is a major reliability factor. Also in recent years. low frequency noise diagnostics has become a powerful tool for device performance and reliability characterization.

This dissertation work demonstrates the necessity of gate stack engineering for achieving a low I/f noise performance. Changes in the material and process parameters of the devices, impact the 1/f noise behavior. The impact of 1/f noise on gate technology and processing parameters xvere identified and investigated. The thickness and the quality of the interfacial oxide, the nitridation effects of the layers, high-K oxide, bulk properties of the high-K layer. percentage of hafnium content in the high-K, post deposition anneal (PDA) treatments, effects of gate electrode material (poly-silicon. fully silicided or metal). Gate electrode processing are investigated in detail. The role of additional interfaces and bulk layers of the gate stack is understood. The dependence of low-frequency noise on high and low temperatures was also investigated. A systematic and a deeper understanding of these parameters on 1/f noise behavior are deduced which also forms the basis for improved physics-based 1/f noise modeling. The model considers the effect of the interfacial layer and also temperature, based on tunneling based thermally activated model. The simulation results of improved drain-current noise model agree well with the experimentally calculated values.