Analysis of the Reservoir Length and its Effect on Electromigration Lifetime
- PDF / 104,353 Bytes
- 5 Pages / 612 x 792 pts (letter) Page_size
- 1 Downloads / 228 Views
Larry Ting Dallas Reliability & ESD Lab, Texas Instruments, 12500 TI boulevard, MS 8640, Dallas, Texas 75243
N.C. Tso Intel Corporation, M.S. RA2-413, 5200 NE Elam Young Parkway, Hillsboro, Oregon 97124-6497
C-U. Kim Materials Science and Engineering Department, The University of Texas at Arlington, P.O. Box 19016, Arlington, Texas 76006 (Received 1 March 2001; accepted 29 October 2001)
This report studies the electromigration performance of W-plug via structures under the reservoir effect. The lifetime improvement factor M was observed to be a weak function of the stressing current and approximately equal to 2. A Simple model is included in the report to explain this observation. The model also predicts the most effective reservoir length for electromigration lifetime improvement.
I. INTRODUCTION
The reservoir effect has been known to improve the electromigration lifetime of via structures.1,2 A reservoir is a protrusion of an Al–Cu metal conductor beyond the via. The metal tail does not carry current in an electromigration test. However, host atoms, namely Cu and possibly Al, in a reservoir, migrate without an electric field to replenish vacant sites under the via. The motion of host atoms toward the via is in general driven by a concentraton gradient and a stress gradient.2,3 Electromigration data shown by Fujii, Koyama, and Aoyama4 clearly show that as stressing conditions, namely, current and ambient temperature, increase, the lifetime enhancement factor decreases. In addition, via electromigration lifetime is shown to be a strong function of the reservior length under fixed testing conditions. However, the median time-to-failure does not increase indefinitely with reservoir length but rather saturates beyond a critical value. This observation implies that only a certain amount of mass of metal in the reservoir is actively involved in the healing process. However, a quantitative model that predicts what the maximum reservoir length is lacking. In this report, we characterize the lifetime enhancement factor M due to the reservoir effect with respect to the stressing current. A quantitative model was developed
a)
II. EXPERIMENT
The metal stack is Ti/TiN/Al–0.5% Cu/TiN, which is referred to as barrier/Al alloy/capping layer. The nominal via size is 0.4 × 0.4 m and is W deposited by chemical vapor deposition. Further details on the manufacturing process of multilevel metallization in this experiment are discussed elsewhere.5 The metal line conductors are 3 m wide (two times the average grain size) for easy grain boundary diffusion and 250 m long to prevent the short length effect.6–8 The test structure has a 250-mlong reservoir terminated by a bond pad. A sample of at least 25 devices was tested for a good statistical analysis. Ambient temperature was fixed at 200 °C. An external electric field was applied on the test structure in such a way that the electron flow direction was from the top conductor to the bottom conductor that rendered shorter test time.9 III. RESULTS AND OBSERVATIONS
Via structures
Data Loading...
