As show in Figure 1.1, the copper line is completely embedded in a barrier layer and in a capping layer in the dual-damascene technology. EM tests have shown that the quality of these interfaces is of crucial importance for the EM lifetimes [26,27,28,29]. Therefore, since the introduction of copper interconnects, their surface and interface EM properties have been extensively studied.
The adhesion between the copper and the surrounding layers is significantly influenced by the choice of processes and materials, so that several combinations have been investigated [30]. Good adhesion is characterized by a tightly bonded interface, which reduces the diffusion along this path and, consequently, increases EM lifetimes. In contrast, the weak bonding between the copper and surrounding layers forms a poorly adhering interface, which leads to higher diffusivities. This increases material transport prior to void nucleation and also speeds-up void evolution and growth, resulting in shorter EM lifetimes. Moreover, the critical stress for void nucleation can be significantly lower at sites of weak adhesion.
The interface between copper and the capping layer is considered to be the dominant diffusivity path in dual-damascene interconnects. This has been confirmed by in situ SEM experiments which have shown that void nucleation, migration, and growth occur mainly along the copper/capping layer interface [31,32,33]. The properties of this particular interface play, therefore, a key role for the electromigration failure.
SiN and SiC based films are widely used as capping materials for copper interconnects.
One of the major tasks during process development is to find capping layer materials, as well as pre-clean and deposition techniques that yield good interface adhesion to obtain adequate EM performance.
Recently, Zschech et al. showed that the bonding strength of the copper/capping layer interface can be increased by using a metallic coating, such as CoWP, before the SiN capping layer deposition [28].
For such a structure they measured an EM activation energy of 1.9 - 2.4 eV, which is much higher than the 0.9 - 1.2 eV typically obtained for the standard SiN capping layers.
This significant reduction in material transport, and also reduced void growth rates, is explained by a Cu/CoWP interface presenting an epitaxy-like transition (i.e. non-interrupted array of parallel lattice planes) between the copper and the capping layer, formed by metallic Cu-Co bonding and a highly ordered interface.
Similar results were obtained by Yan et al. using Cu
Sn coatings [27].
Strengthening the copper/capping layer interface is especially valuable in narrow lines, where due to the bamboo copper grain structure the activation energy for failure is close to the bulk value. This results in a very large EM lifetime. However, as the copper grain size decreases with the line width beyond the 65 nm node, and the lines present again polycrystalline structure, grain boundary diffusion becomes more important and is likely to dominate the failure mechanism [18].
Regarding the barrier layer, tantalum is the most commonly used material. A good adhesion has been observed between copper and tantalum, so that a high activation energy, about 2.1 eV has been reported [19]. Nevertheless, some works have found relatively low activation energies, approximately 1.0 eV, for this interface [13,34]. This can be explained by the oxidation of the copper surface, so that the adhesion between the oxidized copper and tantalum layer is reduced, becoming a path for rapid diffusion.
From these observations, one can see that the quality of interfaces in copper interconnects is very sensitive to process variations. Any oxygen which is allowed to diffuse to the copper surface will reduce the adhesion and lead to increased diffusion, consequently, reducing the EM lifetime. Poor cleaning practice prior to the application of the capping layer has also been found to degrade the interconnect lifetime [35].