SM Model for void formation rate as a function of temperature. SM is caused by the interaction between the thermomechanical stress in the interconnect systems and the diffusion of vacancies (Ref. 1). The graph shows that the SM is expected to occur at use condition, however at the EM test condition the SM effect is negligible.
Schematic diagrams of the (a, c) top view and (b, d) side view of via-chain test structures used for the SM-EM interaction study in (a, b) lower and (c, d) upper metal leads, respectively.
The temperature rise of lower metal multilink test structure as a function of lower metal current density, heated at . It shows that the temperature rise at EM test condition is only about 0.27 K and thus the possibility of temperature-induced failures during EM testing is very minimal.
Failure distribution of samples under test compared to EM-test only samples for the study of SM/EM interaction in the lower metal interconnects. ALTs show that failure distribution of failed samples under test has trimodal distribution. The inset shows the resistance trend for each group population.
Cross section of degraded sample after test from (a) FM-I population, where the void formed directly below the via of the cathode end causing open circuit failure. (b) FM-III population, in which the trench void formed in the Cu metal line (Ref. 16).
Failure distribution of samples under test compared to EM-test only samples for the study of SM/EM interaction in the upper metal interconnects. The lifetime degradation due to SM effect is . The inset shows the resistance trend for each group population.
TIVA analysis of one of the degraded samples from test in the top metal of the M2 test structure. The rectangle sign shows the stressed test structure. The bright spots indicate the void location where many voids are detected at the end of each single chain metal line.
FEA simulation of hydrostatic stress in the lower metal test structure (Ref. 16). The nonuniform tensile stress near the via bottom leads to a prominent stress gradient that drives vacancies toward the via.
FEA simulation of hydrostatic stress in the upper test structure. The nonuniform tensile stress near the upper metal lead (SM-I) and inside the via (SM-II). The resulting stress gradient may favor vacancies to migrate toward these two different areas during SM test.
Mechanism of SM and EM interaction–an approach using stress evolution diagram. (a) Assuming symmetrical condition where the amount of vacancies near both ends after SM test are the same. (b) As a result, the metal line become more tensile by almost the same increment ( and ) and the time to reach void nucleation is shorter, , during EM test.
(a) Assuming the amount of vacancies at the cathode region is limited. (b) As a result, there is no tensile stress developed at the cathode side of metal line after SM test. The time to reach void nucleation is the same, , during EM test.
Statistical test results for test structure. is the measured median time to failure and is the standard deviation of the natural log of the failure times for each population of samples. The error estimates for and were calculated with 90% two-sided confidence bounds.
Statistical test results for test structure.
Material properties used in our simulation model.
Article metrics loading...
Full text loading...