No abstract

microstructure containing lots of triple points, and harsh operating conditions may make power IC's more vulnemble than standard devices. In this paper we present strongly reduced EM-lifetimes upon ABSTRACTThe combined effects of electromigration and thennomigration are studied. Significantly shorter electromigration lifetimes are observed in the presence of a temperature gradient. This cannot be explained hy thennomigation only, but is attxibuted to the effect of temperature grament on electromigration-induced failures.

No abstract

INTRODUCTIONReliability of copper interconnect systems is very sensitive to the maximum current used in the product. Shrinking the metal dimensions from technology node to technology node yields a need for higher current density to fulfill the product needs. Since the electromigration performance is predicted to decrease for 45nm and beyond [1], higher electromigration robustness in smaller dimensions is necessary. A main driving force for electromigration improvement is engineering the interface quality/composition between Cu and Cap layer [2]. Two ways to achieve this are discussed so far. Producing a layer between Cu and Cap by silicidation of the Cu surface [3] or introducing a metal cap below the dielectric cap [1].The latter has the advantage of better resistance control but is very challenging with respect to cleaning processes and leakage/breakdown behavior.In this paper we present the results of advanced process improvements leading to very reliable full-build interconnect systems ready for high volume production. Besides the electromigration investigation, we studied stress migration performance as well as BTS and leakage measurements. We found behavior changes in almost all reliability tests suggesting differences in physical behavior, while exhibiting very strong lifetime performance. EXPERIMENTAL SET UP AND RESULTSWe evaluated a wide set of reliability measurements on 65nm and 45nm technology node material. Electromigration testing was performed in commercially available oven systems on package level. The conventional test conditions did not yield any electromigration failures. Even higher acceleration with four times higher current density did only yield a small resistance increase after 1000h of testing. For comparison non-CoWP sample on the same technology node have been tested at the same conditions. Fig. 1 (right) provides an example of the electromigration performance gain due to CoWP capping.To evaluate the electromigration kinetics of the metal cap, we accelerated the electromigration testing extremely by using the maximum available oven temperature (350°C) and very high current densities. For this investigation we used current densities of 4MA/cm2 up to 20MA/cm2 (Tab. 1). The tested structure is a single link structure connected with one via on each end to a wide feeder line. In this investigation we tested via 1 in up stream and down stream direction.The tests showed significant resistance increases above 12MA/cm2. At higher current densities the typical electromigration breakdown behavior (Step-like resistance increase) was eventually observed. Knowing that this very high current density causes significant joule heating the temperatures have been corrected by either calculating the average temperature using the TCR [4] as well as by a simulation of the exact temperature profile using 3D-FEM modeling (Fig. 2) [5]. These results give fairly good activation energy values (Tab. 2) and n-values of ~2.2 (TCR temperature corrected). Applying advanced failure analysis techniques, such as OBIRCH ...