A B S T R A C TA new method has been developed t o evaluate the stress-migration (S.M.) endurance of layered A1 interconnects stacked with refractory metals. The stressmigration endurance of layered interconnects has so far been evaluated in a manner similar to that for single layer interconnects, by monitoring the resistance change. However, in the case of layered interconnects it is hard to detect the resistance change by the conventional electrical method since the resistance change arising from void generation in the A1 portion of the layered interconnect is negligibly small. In this work, a new method for monitoring stress relaxation characteristics in layered A1 interconnects has been developed. A stress relaxation model based on creep deformation theory was applied and coupled with the Tezaki model [l] to predict S.M. lifetime. The stress in interconnects was directly measured by the X-ray diffraction method. The Stress relaxation ratio and S.M. lifetime were derived from the stress relaxation characteristics in the layered interconnect. The S.M. endurance of layered interconnects in multilevel interconnection structures was evaluated utilizing the new method. It was found for the first time that the S.M. endurance of the lower-most interconnect is the best in a multilevel interconnection structure.
I . I N T R O D U C T I O NIn future ULSI's, four or five level A1 interconnection structures are required for realizing higher density and better device performance. The stressmigration (S.M.) reliability of A1 interconnects in such multilevel structures is of serious concern. Layered A1 interconnects stacked with refractory metals such as TiN/Ti are widely used in ULSIs in order to improve the interconnection reliability. In general, S.M. failure in the layered interconnects has been determined by the same method employed for single-layer A1 interconnects viz., detection of the resistance change. However, the conventional method is not accurate in evaluating the S.M. endurance of layered interconnects because the degree of the resistance change due to void formation in A1 portion is too small t o detect electrically. Figure 1 shows an estimation of the resistance change in Al/TiN/Ti layered interconnects with a slit-void in the A1 portion. Figure 1 shows that for line lengths normally used in evaluating S.M. endurance, open failure in the A1 interconnects cannot be detected by monitoring the interconnect resistance. Figure 2 shows an example of the results obtained by performing the S.M. test on layered interconnects using the conventional method. The resistance increase of the interconnect could not be detected although the S.M. test was continued for almost 10000 hours. Therefore, a new evaluation method sensitive enough to detect the void formation resulting from stress-migration is required. . . Une LenQlh (m) 10 0.0001 -0.2 0.4 0.6 0.8 1.0 1.2 Silt Width (pm) Fig. 1 with a slit-void in the A1 portion.
Estimation of the resistance change in layered interconnects
256
