Towards Understanding Early Failures Behavior during Device Burn-In: Broadband RF Monitoring of Atomistic Changes in Materials

Yaw S Obeng,Papa K Amoah,Victor H Vartanian,Chukwudi A Okoro,Johnny Dai

doi:10.1149/2.0411609jss

Abstract

In this paper, we attempt to understand the physico-chemical changes that occur in devices during device “burn-in”. We discuss the use of low frequency dielectric spectroscopy to detect, characterize and monitor changes in electrical defects present in the dielectrics of through silicon vias (TSV) for three dimensional (3D) interconnected integrated circuit devices, as the devices are subjected to fluctuating thermal loads. The observed changes in the electrical characteristics of the interconnects were traceable to changes in the chemistry of the isolation dielectric used in the TSV construction. The observed changes provide phenomenological insights into the practice of burn-in. The data also suggest that these “chemical defects” inherent in the ‘as-manufactured’ products may be responsible for some of the unexplained early reliability failures observed in TSV enabled 3D devices.

Highlights

In this paper, we attempt to understand the physico-chemical changes that occur in devices during device “burn-in”
We discuss the use of low frequency dielectric spectroscopy to detect, characterize and monitor changes in electrical defects present in the dielectrics of through silicon vias (TSV) for three dimensional (3D) interconnected integrated circuit devices, as the devices are subjected to fluctuating thermal loads
The observed changes in the electrical characteristics of the interconnects were traceable to changes in the chemistry of the isolation dielectric used in the TSV construction

Summary

Introduction

We attempt to understand the physico-chemical changes that occur in devices during device “burn-in”. The concept is frequently used in justifying burn-in strategies for improving system reliability.[3,6,7] the value of the bathtub curve in characterizing infant mortalities has been questioned.[8] The initial decreasing failure rates in the infant mortality region of the bathtub curve assumes some design or manufacturing defects cannot be completely eliminated, resulting in a subpopulation of “weak devices”; the “weak devices” could either die off completely or get stronger under stress While this explanation of infant mortality is not analytically self-consistent, the bathtub model has historically been used successfully in the semiconductor industry.[8,9,10] We need to develop physical evidence-based insights to help bridge the gap between theory and practice of burn-in, and to understand, at least phenomenologically, what happens during post device fabrication burn-in.

Results

Discussion

Conclusion