IRPS

Session 7A – Dielectric Reliability (Front-end and Back-end)

Session 7A – Dielectric Reliability (Front-end and Back-end)

Session Co-Chairs: Ernest Wu, IBM, Bonnie Weir, Avago
Section D

1:25 p.m. – Session Introduction

1:30 p.m.

7A.1 The Physical Mechanism Investigation between HK/IL Gate Stack Breakdown and Time-dependent Oxygen Vacancy Trap Generation in FinFET Devices

C.-H. Yang, S.-C. Chen, Y.-S. Tsai, R. Lu, Y.-H. Lee, Taiwan Semiconductor Manufacturing Company

In this paper, the detailed TDDB models of HK/IL gate stack were well established through the analysis of the oxide trap generation in FinFET technology. We systematically characterized gate oxide traps of HK and IL layers by AC admittance and SILC spectrum methodologies. We found that NMOS TDDB is sensitive to HK traps, while PMOS TDDB is mostly attributed by IL traps and HK deep traps. In addition, the gate oxide deep traps were found to be highly responsible for the permanent damage during stress. Such that, the physical mechanisms of HK/IL gate stack breakdown in FinFET devices can be successfully explained through the scenario of time-dependent HK and IL trap generations. Through the trap studies, TDDB reliability of FinFET technology can be successfully improved.

1:55 p.m.

7A.2 Multiphysics based 3D Percolation Framework Model for Multi-Stage Degradation and Breakdown in High-κ / Interfacial Layer Stacks

S. Mei, N. Raghavan, K. Shubhakar, M. Bosman*, K.-L. Pey, Singapore University of Technology and Design, *A*STAR

Softer stages of dielectric degradation and breakdown deserve in-depth studies given their relevance and finite probability of occurrence in the 14 nm and sub-10 nm CMOS technology nodes of the future. This study presents a multi-physics based percolation framework model to simulate the dynamics of the sequential and competitive evolution of soft breakdown (SBD) spots during degradation of a dual-layer high-κ interfacial layer (HK-IL) dielectric stack. The presented model leverages on the combined use of Kinetic Monte Carlo (KMC) routine to describe the microstructural variations and stochastics of defect nucleation and growth as well as the finite element model (FEM) to quantify the spatio-temporal evolution of the potential, field and temperature distributions in the stack. The results point to IL being the first to breakdown and confirm physical and atomistic evidences that suggest the preferential nucleation of SBD spots beneath the grain boundary regions in the HK layer. Furthermore, our model enables visualization of multiple SBD paths in the IL layer before HK eventually percolates, which supports earlier statistical studies.

2:20 p.m.

7A.3 Layout Dependence of Gate Dielectric TDDB in HKMG FinFET Technology

W. Liu, E. Wu*, F. Guarin, C. Griffin, R. Dufresne, D. Badami, M. Shinosky, D. Brochu, GLOBALFOUNDRIES, *IBM TJ Watson Research Center

In this work, we report for the first time the experimental evidence of layout dependence on gate dielectric time-dependent-dielectric-breakdown TDDB in a leading edge HKMG FinFET technology. Structures with identical total effective gate area but various Fin and finger configurations per unit cell exhibit more than 10X difference in Tbd and Qbd. Fin number per unit cell is the major impact factor based on leakage / stress current comparisons and Fin number scaling conversion. The implication of these findings on technology qualification methodology is that one needs to evaluate layout sensitivity for a given process under development to confirm that reliability assessment are valid for structures with various layout configurations. Processes with strong layout dependence should be optimized before technology qualification.

2:45 p.m.

7A.4 CAFM Based Spectroscopy of Stress-Induced Defects in HfO2 with Experimental Evidence of the Clustering Model and Metastable Vacancy Defect State

A. Ranjan, N. Raghavan, K. Shubhakar, R. Thamankar, J. Molina*, S.J. O’Shea*, M. Bosman*, K.L. Pey, Singapore University of Technology and Design, *NIAOE, **A*STAR

In this study, we perform random telegraph noise (RTN) spectroscopy on ultra-thin HfO2 dielectric films using a conductive atomic force microscope (CAFM), enabling accurate assessment of single or few defect kinetics in very small area regions spanning 40 × 40 nm2. Our characterization results show that bias-dependent RTN trends can be clearly detected at high spatial resolution using the CAFM system. Experimental evidence of the metastable nature of oxygen vacancy defects is presented and the nanoscale BD results provide further support to the time-dependent defect clustering model that is recently proposed for oxide breakdown [1, 2]. Statistics of CAFM breakdown voltage also show a trimodal distribution that corresponds to percolation nucleation at grain (G), grain boundary/triple point (GB/TP) sites and G-GB interfaces.