TY - JOUR
T1 - Void dynamics in copper-based interconnects
AU - Choi, Zung Sun
AU - Lee, Junghoon
AU - Lim, Meng Keong
AU - Gan, Chee Lip
AU - Thompson, Carl V.
PY - 2011/8/1
Y1 - 2011/8/1
N2 - The electromigration lifetimes of Cu-based interconnects are strongly influenced by whether voids are present before electromigration, and by where fatal voids initially form and grow. Modeling, simulations, and comparisons with in situ experiments are used to establish criteria for void formation away from the cathode end of a copper interconnect. It is shown that observation of voids at locations other than the cathode strongly suggests that the voids grew from pre-existing voids. When pre-existing voids are within a current-density- dependent critical length from the cathode, new voids are unlikely to nucleate at the cathode and failure occurs only when the pre-existing voids grow. As these voids grow, they will either lead directly to open-circuit failure or, once they reach a critical size, they will de-pin from grain boundaries and drift toward the cathode. In the latter case, multiple voids might accumulate and coalesce to cause failure. This mechanism has been observed in both side-view and top-down in situ accelerated life-time testing. It is shown that the relative importance of these various void-induced failure mechanisms depends on the current density, and is different under typical accelerated test conditions from what is expected at service conditions.
AB - The electromigration lifetimes of Cu-based interconnects are strongly influenced by whether voids are present before electromigration, and by where fatal voids initially form and grow. Modeling, simulations, and comparisons with in situ experiments are used to establish criteria for void formation away from the cathode end of a copper interconnect. It is shown that observation of voids at locations other than the cathode strongly suggests that the voids grew from pre-existing voids. When pre-existing voids are within a current-density- dependent critical length from the cathode, new voids are unlikely to nucleate at the cathode and failure occurs only when the pre-existing voids grow. As these voids grow, they will either lead directly to open-circuit failure or, once they reach a critical size, they will de-pin from grain boundaries and drift toward the cathode. In the latter case, multiple voids might accumulate and coalesce to cause failure. This mechanism has been observed in both side-view and top-down in situ accelerated life-time testing. It is shown that the relative importance of these various void-induced failure mechanisms depends on the current density, and is different under typical accelerated test conditions from what is expected at service conditions.
UR - http://www.scopus.com/inward/record.url?scp=80051937927&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=80051937927&partnerID=8YFLogxK
U2 - 10.1063/1.3611408
DO - 10.1063/1.3611408
M3 - Article
AN - SCOPUS:80051937927
SN - 0021-8979
VL - 110
JO - Journal of Applied Physics
JF - Journal of Applied Physics
IS - 3
M1 - 033505
ER -