Automated reliability calculation of failure rate, lifetime extrapolation and prediction for embedded Metal-Insulator-Metal capacitors using an optimized Time-Dependent-Dielectric-Breakdown model

This study presents an algorithm for automated reliability analysis of embedded Metal-Insulator-Metal (MIM) capacitors with high-k dielectrics. With the proposed algorithm, Time-Dependent-Dielectric-Breakdown (TDDB) data of embedded capacitors of different physical dimensions measured at various stress conditions (temperature and electric field) can be analyzed in a uniform way based on a supervised learning approach. Instead of analyzing each influence parameter separately, the data is combined using automated linear regression, based on decision tree learning and thus defining a multi-dimensional plane. With this approach a simultaneous consideration of all affecting parameters and their interaction on the dataset can be achieved, while removing statistical outliers and extrapolating the reliability behavior of any failure percentage. To support the most common conduction mechanism related to the embedded ΜΙΜ capacitors, Poole-Frenkel emission, a new TDDB model with changed coordinate system is proposed.