Using TCAD, Response Surface Model and Monte Carlo Methods to Model Processes and Reduce Device Variation

Reduction of electrical parameter variation is essential to achieve high yield and reliability in semiconductor devices. However, variation depends on a large number of process factors, which are often interdependent. In this work, well-calibrated Technology Computer-Aided-Design process and device simulations were performed in a designed experiment to develop an efficient, surrogate response surface model (RSM) of the device parameters as a function of key process factors. Monte Carlo simulations were performed with the RSM to estimate variation and design systems to reduce variation. The approach, illustrated here specifically for peripheral n-type field-effect transistors in a dynamic random-access-memory process flow, is general, easy-to-implement, and a cost-effective way to systematically identify, model, and analyze process variation.