[PhD] VLSI Interconnect Reliability

Shaoyi Peng

Modeling and Simulation Methods for VLSI Interconnect Reliability Focusing 

on Time Dependent Dielectric Breakdown

PhD Dissertation in Electrical Engineering

University of California Riverside

https://escholarship.org/uc/item/966241xk (March 2021)

Abstract: Time dependent dielectric breakdown (TDDB) is one of the important failure mechanisms for Copper (Cu) interconnects that are used in VLSI circuits. This reliability effect becomes more severe as the space between wires is shrinking and low-k dielectric materials (low electrical and mechanical strength) are used. There are many studies and theories focusing on the physics of it. However, there is limited research from the electronics design automation (EDA) perspective on this topic, aiming to evaluate, or alleviate it from the perspective of designing a VLSI chip. This thesis compiles several studies into evaluating TDDB on the circuit level, and engineering methods that help the evaluation. The first work extends the study of a published physics model on simplified yet practical cases. It simplifies the calculation of lifetime by deriving an analytic solution and applying fitting methods. The second study proposes a new way to evaluate lifetime of a chip by extending the models of simple interconnect structures to the complete chip. This method is more robust as it focuses more on a complete chip. However, heavy dependence of finite element method (FEM) makes the flow very slow. The third study adopts machine learning methods to accelerate this slow evaluation process. The proposed method is also applicable to other similar electrostatics applications. Last but not least, the fourth study focuses on a GPU based LU factorization algorithm, which, on a broader aspect, is a universal numerical algorithm used in many different simulation applications, which can be helpful to TDDB evaluations as it can be used in FEM.

Fig: Structure of two copper interconnect wires and the IMD in the cross-section SEM image after TDDB failure [sem]

REF: 

[sem] N. Suzumura, S. Yamamoto, D. Kodama, K. Makabe, J. Komori, E. Murakami, S. Maegawa, and K Kubota. A new TDDB degradation model based on Cu ion drift in Cu interconnect dielectrics. In IEEE Int. Reliability Physics Symposium (IRPS), pages 26–30, 2006.

[paper] Memory Technology – A Primer for Material Scientists.

Schenk, Tony, Milan Pesic, Stefan Slesazeck, Uwe Schroeder, and Thomas Mikolajick

Memory Technology–A Primer for Material Scientists

Reports on Progress in Physics (2020)

Abstract - From our own experience in the group, we know that there is quite a gap to bridge between scientists focused on basic material research and their counterparts in a close-to-application community focused on identifying and solving final technological and engineering challenges. In this review, we try to provide an easy-to-grasp introduction to the field of memory technology for materials scientists. As an understanding of the big picture is vital, we first provide an overview about the development and architecture of memories as part of a computer and point out some basic limitations that all memories are subject to. As any new technology has to compete with mature existing solutions on the market, today's mainstream memories are explained and the need for future solutions is highlighted. The most prominent contenders in the field of emerging memories are introduced and major challenges on their way to commercialization are elucidated. Based on these discussions, we derive some predictions for the memory market to conclude the paper.

TABLE OF CONTENTS

1. INTRODUCTION
2. OVERVIEW AND BASIC LIMITATIONS
3. COMMERCIALLY AVAILABLE MAINSTREAM MEMORIES

3.1. Static and Dynamic Random Access Memory (SRAM/DRAM)
3.2. Flash Memory and Solid-State Drive (SSD)
3.3. Magnetic Hard Disk Drives (HDD) and Magnetic Tapes
3.4. Outlook: Market Trends and Drivers

4. EMERGING MEMORIES

4.1. Resistance-based Read-out: Memory Concepts and Basic Considerations
4.2. Anion migration or valence change memory (VCM)
4.3. Cation migration or electrochemical metallization memory (ECM)
4.4. Phase change memory (PCM)
4.5. Magnetoresistive memory (MRM)
4.6. Ferroelectric Memory (FEM)
4.7. Miscellaneous

5. SUMMARY AND CONCLUSION

FIG: Evolution of the mainstream solutions for the respective memories classes. The introduction of Flash memory partially bridged a technology gap around the year 2009. Today, two types of so-called storage-class memories – a memory-type SCM (SCM 1) and a storage-type SCM (SCM 2) – were proposed to overcome the memory gap. NAND flash already fulfills the role of a mainstream SCM 2. For SCM 1, 3D XPoint could be a promising candidate, but is not a dominant mainstream memory. In future, we will likely see different types of SCMs and NV-RAM with different specifications as required by the respective application – because in the end, the overall system cost decides about the choice of the memory.