A scientific study of the problems
of digital engineering for space flight systems,
with a view to their practical solution.
Characterization and Modeling of a Highly Reliable Metal-to-Metal Antifuse for High-Performance and High-Density Field-Programmable Gate Arrays
Chih Ching Shih, Roy Lambertson, Frank Hawley, Farid Issaq, John McCollum, Esmat Ilamdy, Hiroshi Sakurai*,Hiroshi Yuasa*, Hirotsugu Honda*, Tohru Yamaoka*, Tetsuaki Wada**, and Chenming Hu***
Actel Corporation, 955 E. Arques Ave., CA 94086, USA
Matsushita Electronics Corpoeration, *Kyoto Research Laboratory, **Quality Laboratory,
Kyoto, Japan
***University of California at Berkeley, Berkeley, CA 94720, USA
Abstract
Reliability of a new amorphous silicon/dielectric antifuse is characterized and modeled. Unprogrammed antifuse leakage and time-to-breakdown are functions not only of applied voltage but also of stressing polarity and temperature. Both breakdown and leakage criteria are used to investigate their effects on time-to-fail. A thermal model incorporates the effects of programming and stress currents, ambient temperature, and variation of antifuse resistance with temperature. Measured temperature dependence of antifuse resistance is for the first time used to derive key physical parameters in the model.
Table of Contents
I. Introduction
II. Antifuse Structure & Characteristics
- Leakage and Breakdown Voltage Characteristics
- Programmed Antifuse Resistance Characteristic
III. Unprogrammed Antifuse Reliability
- Median-time-to-fall versus I/V of Unprogrammed Antifuses
- Programmed Antifuse Resistance versus Stress Time
- The Effects of Read Current, Programming Current, and Ambient Temperature on ON-state Lifetime
- AC versus DC Current Stress
IV. Conclusions
List of Figures
Figure 1. Cross-sectional view of a new metal-to-metal antifuse for triple-level metal CMOS process.
Figure 2. Typical I=V characteristic of a single unprogrammed a–Si/dielectric metal-to-metal antifuse at an ambient temperature of 25 °C.
Figure 3. Typical I=V characteristic of a single unprogrammed a–Si/dielectric metal-to-metal antifuse at an ambient temperature of 150 °C.
Figure 4. Leakage current per unprogrammed antifuse at 3.6 V as a function of ambient temperature. The activation energy for the leakage current is about 0.3 eV.
Figure 5. Breakdown voltage distribution for single antifuse as ambient temperature of 25 °C. The ramp rate is 0.1 V/ms.
Figure 6. Programmed antifuse resistance distributions for programming current IPP of 10, 15, 20, and 25 mA.
Figure 10. Unprogrammed antifuse leakage at 3.6 V versus stress time. Stressing conditions: +8.5 V on top electrode and Ta = 25 °C.
Figure 11. Unprogrammed antifuse leakage at 3.6 V versus stress time. Stressing conditions: +7.5 V on top electrode and Ta = 25 °C.
Figure 12. Unprogrammed antifuse leakage at 3.6 V versus stress time. Stressing conditions: -9 V on top electrode and Ta = 25 °C.
Figure 13. Unprogrammed antifuse leakage at 3.6 V versus stress time. Stressing conditions: -8 V on top electrode and Ta = 25 °C.
Figure 14. Median-time-to-fail of unprogrammed antifuse vs. I/V for different failure criteria with positive stress voltage on top electrode and Ta = 25 °C.
Figure 15. Median-time-to-fail of unprogrammed antifuse vs. I/V for different failure criteria with positive stress voltage on top electrode and 150 °C.
Figure 16. Median-time-to-fail of unprogrammed antifuse vs. I/V for different failure criteria with negative stress voltage on top electrode and 25 °C.
Figure 17. Median-time-to-fail of unprogrammed antifuse vs. I/V for different failure criteria with negative stress voltage on top electrode and 150 °C.
Figure 21. Programmed antifuse resistance versus stress time.
Figure 22. Lognormal plot of time-to-fail of programmed antifuse resistance versus stress time for different stress currents. The failure criterion is a 10% increase in resistance.
Figure 23. Programmed antifuse median-time-to-fail is dependent of programming current and stress current. t50 data about 107 seconds are extrapolated.
Figure 24. The lifetime of programmed antifuses is dependent on the ambient temperature. t50 data about 107 seconds are extrapolated.
Figure 25. Time-to-fail data for DC and 1 MHz AC current stress tests.
List of Tables
Table 2. Characteristics of a New Metal-to-Metal Antifuse Incorporating Amorphous Silicone and Dielectric Layers.
Conclusions
Off-state reliability of this new antifuse is shown to be suitable for V8 = 3.6 V operation using the I/V model for lifetime extrapolation. Since the saturation current at the operating voltages does not exceed the maximum allowable leakages. It is seen that the leakage increase over time behavior is not a major reliability concern.
The temperature coefficient of resistance of the programmed antifuses is found to be dependent on the programming current. A thermal model for self-heating is presented. The model compares favorably with measured resistance as a function of current for different programming currents. Ambient temperature dependence of programmed antifuse time-to-fail is reported for the first time. The arrhenius model predicts on-state lifetime lower-than-actual data for high ambient temperatures. This results in conservative maximum allowable read current in design. The effects of ambient temperature as well as frequency and stress waveform on AC lifetime is still in need of further investigation.
The new antifuse structure has a combination of electrical and reliability characteristics that make it suitable for high-performance, high-density FPGAs.
Home
Last Revised January 09, 2002
Digital Engineering Institute
Web Designer: Richard Katz