Statistical analysis of reliability and variability effects in CMOS technologies based on single-defect spectroscopy

Abstract

The metal-oxide-semiconductor field effect transistor (MOSFET) is the integral building block of modern electronic devices. Its influence has triggered substantial cultural, societal, and economic shifts over the past few decades. Recognizing the sustained relevance of MOSFET technology, there is a considerable emphasis on amplifying their performance through ongoing research and development. Nevertheless, it is crucial to acknowledge that MOSFETs contain defects at the atomic scale due to their internal structure, fabrication procedures, and operating conditions. These defects represent obstacles to the reliability and functionality of MOSFETs as they trap electric charges and degrade the performance of these transistors. The impact of these defects is increasingly prominent as the trend toward transistor miniaturization continues. This situation raises issues such as bias-temperature instabilities (BTI) and random telegraph noise (RTN).This work aims to contribute to understanding defects’ impact on the performanceof MOSFETs through experimental characterization and statistical modeling. The main focus is to analyze the impact of single defects on the device performance. By leveraging the extended measure-stress-measure (eMSM) on nanoscale devices, it is possible to extract the impact of single defects on the threshold voltage shift ∆Vth , one of the most core parameters of a transistor. Single-defect spectroscopy, used in conjunction with the defect-centric model (DCM) to evaluate the data, enables the accurate extraction of defect characteristics. This is essential for understanding the statistically distributed nature of active defects, particularly in terms of their number and impact on ∆Vth . This approach is especially crucial for identifying high-impact defects, known as ”killer” defects, which can lead to immediate device and circuit failures in nanoscale nodes. From the results of the statistical analysis, the impact of device geometry and body bias on average threshold shift that can be induced by a single defect can be obtained, enhancing the understanding of time-dependent variations across different technologies. Additionally, reliability simulations are employed to calculate the cumulative response of many defects on large devices and enable a comparison of theoretical trap parameters with the measurement data. For the experiments, three distinct technologies are investigated, enabling conclusions to be drawn about their distribution andphysical properties. Combining experimental measurements, statistical modeling, and compact-physics simulations can lead to understanding defects in MOSFETs and device and circuit performance and reliability.