In-Situ Observations of Electromigration-Induced Void Dynamics
- PDF / 1,919,869 Bytes
- 7 Pages / 414.72 x 648 pts Page_size
- 80 Downloads / 210 Views
(1)
where A is a constant of proportionality, A.H is the activation energy, and k is Boltzmann's constant. The value of the current density exponent, n , has been measured to be close to 2[1]. This value of n = 2 has been predicted as a result of vacancy diffusion mass transport mechanism modeled by Shatzkes and Lloyd[2]. In order for a void to nucleate, a critical vacancy concentration is required. They found that the time to attain the critical EM-induced vacancy concentration required for void nucleation at the cathode end of a conductor stripe is proportional to]j -2. Korhonen et. al.[3] determined the stress build-up transient in a conductor stripe upon application of current stress. The time to attain the critical stress required for void formation was found to be proportional to j -2. Although the void nucleation stage of EM damage has been modeled successfully, deviations from n = 2 in Black's equation have been observed experimentally[4] and can range between 1 and 7. This discrepancy may be due to the void growth stage following void nucleation. It is thought that the typical interconnect failure mechanism is void nucleation and growth at a fixed point, extending across the width of the conductor. However, the actual void growth dynamics are not fully understood. In their well-known in-situ EM study, Thomas and 163
Mat. Res. Soc. Symp. Proc. Vol. 404 0 1996 Materials Research Society
Calabrese[5] observed that some voids are very mobile and move along the length of the conductor toward the cathode in the opposite direction of the electron flow. Other voids were immobile and grew at their cathode side where the current density at the perimeter of the void was at a minimum. Hillocks were observed to grow in the vicinity of voids, appearing to grow from the substrate-metal interface. Levine and Kitcher[6] observed microscopic voids forming as early as 1% of the failure time with a maximum void density achieved between 10% and 20% of the failure time. They concluded that voids must grow, move and coalesce in order to achieve open circuit failure. Besser et. al.[7] tried to relate void movement and growth with void size but found no clear relationships. In this study, electron-beam lithography (EBL) was used to fabricate gold interconnect test lines with pre-patterned voids in order to observe movement and growth in a field-emission scanning electron microscope (FESEM). Various void shapes and sizes were patterned into the Au conductors in order to determine the relationships between dynamic void behavior and initial void geometry. Au was used as the interconnect material because aluminum is known to form a self-passivating oxide skin on its surface, thus passivating the interior surfaces of Al voids[8]. Test lines were not passivated because of the obvious difficulty in leaving a cavity under a passivation layer following lithographic patterning of a void. The presence of prepatterned voids was found to force the line failure in only about 40% of the cases. The EM forces within the line tended to decrease
Data Loading...
