Current Status of LSI Micro-fabrication and Future Prospect for 3D System Integration

  • PDF / 1,082,330 Bytes
  • 10 Pages / 612 x 792 pts (letter) Page_size
  • 88 Downloads / 193 Views

DOWNLOAD

REPORT


0970-Y05-02

Current Status of LSI Micro-Fabrication and Future Prospect for 3D System Integration Kazuya Okamoto Osaka University, Osaka, 565-0871, Japan ABSTRACT Micro-fabrication technology for LSI devices has been progressing technically over time. However, the scenario of continuing to produce ever-finer feature geometries has a low probability. Resolution capabilities will reach a critical limit due to conventional CMOS performance threshold and chip economy. Therefore, to assure continued performance improvements for future devices, front-end fabricators should consider a new 3 dimensional structure (3D-LSI) using a "Through-Si-Via" process. At the same time, the definition of the semiconductor device should be updated to "System & Design Integration (S&DI)." S&DI will provide the needed feedback to launch a new field of clear applications, based on a total system solution with innovative design, fabrication, inspection and evaluation equipment. 3D-LSI and S&DI will have a tremendous impact on the future electronics industries. 1. INTRODUCTION In the past quarter century, the semiconductor industry had seen unprecedented progress with compound annual growth rate (CAGR) at more than 13% compared to 5% of the gross world product as shown in figure 1. However, the rate is currently almost reaching a plateau. This indicates that the current status of mankind is changing and is different from the past of the 70s and 80s, when the semiconductor itself was applied to various kinds of equipment. In the future the semiconductor (LSI) will work in the heart of the BRICs market, but apparently it is the time for a change. In this paper, the author will discuss the current status of LSI miniaturization and bring up necessity of 3D monolithic integration as FEOL (Front-End-of-Line) and BEOL (BackEnd-of-Line). Since the invention of solid-state transistors in 1947, LSI technology has been progressing using Dennard’s (1/k) rule [1] based on the guiding principle of “Moore’s miniaturization rule.” LSI forms the basis of highly-networked information society, and everyone agrees its demand is assured. But we have to consider new emerging rules such as Amdahl’s rule, which predicts transformation of MPU architecture. In fact, in 2005 Intel quit the development of the conventional MPU with a single-core and extremely high frequency, and announced dual-core MPU and the maximum frequency at 4GHz to reduce the power consumption. Currently Intel is planning to extend the number of cores. At the IDF2006 fall meeting [2], the development status of a new MPU stacked with 80 cores and 20MB-SRAM was officially presented along with the excellent performance of 1TFLOPS at 3.1GHz. Figure 2 shows the transition of the value of semiconductors and communications. The two items show similar trends, and shows degradation in value at about 1/10 per decade. At 30 years in the future, the degradation value is 1/1000, which means the value itself is free of charge and the scheme of mankind is dramatically

changing. In actual fact, the semiconducto