Electromigration study of Al and Cu metallization using WLR isothermal method

Abstract

would not be achieved. As Cu-based technology becomes the mainstream metallization for the semiconductor industry, WL EM Wafer level electromigration behavior of copper and aluminum using technique will Play a much more important role for all Practical isothermal stress was investigated in this paper. Lifetime, lognormal Purposes. In this Paper, we rePo* and compare the WL EM testing standard deviation, and activation energy were evaluated as a results of both CU and AI using the WL isothermal (ISOT) technique. function of stress temperature as well as line width. Temperature The testing method keeps the test line at a Constant temperature dependence of the embedded 2D thermal behavior was modeled via by monitoring line resistance during test. The advantage, of this the initial stress current versus the initial resistance correlations. he method is that variations in the thermal conductance of the aanspofi mechanisms in the highly accelerated wafer level environment around the line due to variations in oxide thickness and elecfromigration were observed to be the same those in line cross sectional area are automatically accounted for, thus making moderately accelerated conventional package level electromirration ' control very accurate.-. . . .for both Cu-based and AI-based systems. EXPERIMENTAL PROCEDURES : INTRODUCTION Electromigration (EM) performance of interconnect conductors inULSl microelectronic chips for the purpose of projecting lifetimes at use conditions is typically determined by strcssing.chips mounted in ceramic packages at high temperatures in an oven. In this arrangement, temperature and current density can .he varied independently. Limitations to this approach are 1) small sample size, 2) high cost, 3) package build time, and 4) extended test duration. Wafer level (WL) testing, in contrast, is performed on,a wafer probe station using current to supply both temperature and current acceleration. Stresses take only seconds or minutes instead of several hundred hours and sample sizes are generally larger. The major disadvantage to WL testing is that the temperature and current density cannot be controlled independently, and for this reason, little trust is placed in lifetime projections produced by this method. In addition to coupled parameters, several different types of WL tests are possible, and the results are different for each. These issues were addressed for AI-based technology [ 1,2].In today's advanced high performance technologies, the need for reduced resistance and capacitance interconnects has led to an evolution kom AI-based to Cu-based interconnect processes. Since the resistance of Cu wiring-is about 35% lower than AI and the melting point of Cu (1083T) is 423°C higher than AI (660"C), the EM performance of Cu is about 100 times bener than AI fiom the package level test reported by IBM [3]. It has become-an increasing challenge to obtain the necessary data for determining lifetime predictions and use conditions for each successive generation of Cubased technology by means of traditio...

Electromigration of Flat and Bent Copper Lines Patterned with a Plasma-Based Etch Process

“…31 For the single-level structure, the failure is sensitive to the film's microstructure and interface characteristics. 32 Previous studies showed that lattice diffusion occurred at a temperature higher than 230°C. Because EM tests in this study were carried out at 320-380°C, it is possible that both the lattice diffusion and the grain-boundary diffusion coexist, which results in a large value.…”

“…31 Furthermore, the difference in the thermal expansion coefficient ͑CTE͒ between Cu and the dielectric film weakens the interface in the EM test at a high temperature. 32 An E a value of 0.65 eV was reported for Cu with the PECVD SiN x cap layer. 31 However, EM damage during the lifetime test is a complex and dynamic process.…”