Surface Electromigration, Void Dynamics, and the Reliability of IC Interconnects

Z.-S. Choi, R. Monig, T. Chookajorn, T.J. Park, C.V. Thompson
Sponsorship: Intel, AMD, SRC

Electromigration is atomic diffusion due to a momentum transfer from conducting electrons. Electromigration of metallic IC interconnects is and will remain a major reliability concern as future technologies demand increasing device and wire densities as well as higher current densities. In Cu-based metallization, electromigration occurs by diffusion of Cu at the interface between polycrystalline Cu and the dielectric overlayer, and it leads to formation of voids that cause an increase in resistance and to failure. The rate of failure is therefore highly dependent on the Cu atomic diffusivity, which is affected by the grain structure of the Cu, as well as the stress conditions. In situ scanning electron microscope observations have been performed on passivated Cu interconnects of different widths during accelerated electromigration tests. In some cases, voids form and grow at the cathode. However, an alternative failure mode is also observed, during which voids form distant from the cathode and drift toward the cathode, where they eventually lead to failure. The number of observations of this failure mode increased with increasing line width. During void motion, the shape and the velocity of the drifting voids varied significantly. Postmortem electron backscattered diffraction (EBSD) analysis was performed (see Figure 1), and a correlation of EBSD data with the in situ observations reveals that locations of voids, their shape evolution, and their motion all		strongly depend on the locations of grain boundaries and the crystallographic orientations of neighboring grains. [1] A separate experiment determined surface electromigration rates on oxide-free surfaces of unpassivated damascene Cu interconnect segments through electromigration testing under vacuum. Electromigration-induced voids grew at the cathode end of the segments due to a flux divergence at refractory-metal-lined vias to the lead lines below the test segment. Diffusivity on a clean Cu surface was determined by measuring the size of the voids as a function of time and test temperature at a fixed current. An activation energy of 0.45±0.11 eV and a pre-factor of 3.35 x 10−12 m2/s were found for the product of the effective charge z* and the surface diffusivity Ds [2]. Through correlations of void growth rates with the crystallographic textures of adjacent grains (Figure 2), relative surface diffusivities for grains with different crystallographic orientations have been determined [3]. Data acquired in the experiments described above are being used in simulations of electromigration-induced failure, to develop improvedmethods for reliability projections based on accelerated electromigration tests.
Figure 1: In situ SEM images of the cathode of a test structure, showing void drift toward cathode end. The test line is surrounded by a Cu-extrusion monitor. Bottom image is a texture mapping by EBSD obtained after EM test.		Figure 2: EBSD has been used to determine the crystallographic orientations of grains adjacent to voids. Correlation with void growth rates allows determination of relative values of the surface diffusivity.

References
[1] Z.-S. Choi, R. Mönig, and C.V. Thompson, “Dependence of the electromigration flux on the crystallographic orientations of different grains in polycrystalline copper interconnects,” Applied Physics Letters, vol. 90, p. 241913, 2007.
[2] Z.-S. Choi, R. Mönig, and C.V. Thompson, “Activation energy and pre-factor for surface electromigration and void drift in Cu interconnects,” Journal of Applied Physics, vol. 102, p. 083509, 2007.
[3] Z.-S. Choi, R. Mönig, and C.V. Thompson, “Effects of microstructure on the formation, shape, and motion of voids during electromigration in passivated copper interconnects,” Journal of Materials Research, vol. 23, p. 383-391, 2008.