Characterization and Performance Enhancement of Electrically Programmable Fuse (eFuse)

博士 === 國立成功大學 === 微電子工程研究所 === 106 === Under the guidance of Moore's Law, the manufacturing semiconductor process of integrated circuits has been going forward. And the complexity of IC chips has been also greatly increased. How to ensure each IC workable and prevent it from failure is a big to...

Full description

Bibliographic Details
Main Authors: Chang-ChienWong, 翁彰鍵
Other Authors: Shoou-Jinn Chang
Format: Others
Language:en_US
Published: 2018
Online Access:http://ndltd.ncl.edu.tw/handle/kmb2r9
Description
Summary:博士 === 國立成功大學 === 微電子工程研究所 === 106 === Under the guidance of Moore's Law, the manufacturing semiconductor process of integrated circuits has been going forward. And the complexity of IC chips has been also greatly increased. How to ensure each IC workable and prevent it from failure is a big topic because the cost of IC design and semiconductor manufacturing has been increased radpidly. In 2002, the Electrically Programmable Fuse (eFuse) technology was originally published in an IBM's technical paper at embedded DRAM conference. It has successfully adopted electromigration into semiconductor manufacturing even electromigration technology has traditionally had an adverse effect on wafer performance and has been avoided in design. And IBM announced that e-Fuse is successfully implemented into semiconductor circuit of its micro-processors on 300mm fabrication in 2004. Now, this eFuse has been a popular one of current one-time electrically programmable (OTP) components and widely adopted into complementary metal-oxide-semiconductors (CMOS) related semiconductor chips from 0.25um and beyond. In recent years, Many studies have discussed the programming conditions and physical characteristics of different eFuse structures. This dissertation is divided into two main parts to study different eFuse structures. The first part is the related research about polysilicon eFuse, and the second one is about active eFuse. In the part of polysilicon eFuse studies, the characterization and performance of an electrically programmable fuse (eFuse) prepared with compatible fabrication processes for complementary metal-oxide-semiconductors (CMOS) manufacturing combined with dual-stress-liner (DSL) or single-stress-liner (SSL) technique are discussed. According to the 1st experimental result, the preapred polysilicon fuse capped with compressive-stress or tensile-stress film show different electrical and physical behaviors. The programming mechanism included electromigration mode (EM) and rupture mode (RM) is well performed by referring to the analysis of transmission electron microscopy (TEM) and energy dispersive X-ray spectrometry (EDS). And the use of stress buffer layer in 2nd experiment is also studied to mitigate the intensity of the tensile-stress film. Based on the theories of electromigration, stress-migration and Blech effect, the compressive-stress film is found to contribute to void nucleation which not only increases the programmed fuse resistance but also acts as a silicide-layer refill inhibitor to provide more reliable programmed eFuse functionality. Finally, 1st experiment shows that the minimum and maximum programming currents are 7 mA and 13 mA with the ratio of post-programmed resistance (Rf) to the unprogrammed resistance (Ri), 〉10^3. And 2nd experiment has the minimum and maximum programming currents as 5 mA and 10 mA with the Rf/Ri ratio, 〉10^2. In the last research of active eFuse, an active fuse is successfully implemented in an SOI process fully compatible with the current CMOS technology. The structure of SOI inherently provides an isolation environment for active eFuse and is helpful to the thermal effect. The programming performance is studied with regard to the doping type of the diffusion layer. The P-type active fuse was observed to have better programming performance than the N-type at both room and high temperatures. The formation of diffusion break after programming assists the performance of the P-type active fuse. Finally, this experiment shows that the minimum and maximum programming currents are 16 mA and 22 mA with the Rf/Ri ratio, 〉10^3.