Effects of electrical, thermal and thermal gradient stress on reliability of metal interconnects

This thesis focuses on the reliability modeling of metal interconnects under timedependent stress. Whereas most existing reliability models are based upon the assumption that stress is constant throughout the useful life of a system, this thesis considers the more general and more realistic situation where the stress is time-dependent. In this work the stress is defined by temperature and current density variables. It is assumed that the Cumulative Density Function (CDF) is characterized by a single stress parameter that incorporates all stress-dependent variables. A closed-form expression that can be used to calculate the CDF under time-varying stress is presented and this can be used to determine the corresponding Median Time to Failure (MTF). A single parameter which can be represented as a real number is used to incorporate the total effects of the stress history making this approach applicable for dynamic power/thermal management algorithms. A reliability model that includes the effects of thermal gradient stress in the presence of temperature and current stress is also introduced. With these models, temperature measurement accuracy requirements are developed that are necessary if power/thermal management circuits are to be successful in achieving 10% accuracy in the MTF. Incorporation of a time-dependent stress model that incorporates the user-dependent electrical and thermal stress history in the power/thermal management module of a large integrated circuit offers potential for significantly improving system performance while maintaining a target reliability throughout the operating life of the integrated circuit or for improving the reliability when operated at a user-determined stress level.