Abstract: With the rapid development of microelectronics technology, the failure of interconnects in the integrated circuit raises wide attention. The interconnects inevitably exist some drawbacks, such as voids and cracks. If the drawbacks nucleate, grow and change their shape to form crack-like slits oriented perpendicular to an interconnect line, an open circuit could result. This is a common form of interconnects failure. And interface migration is one of the main mechanisms leading to the evolution of microstructure. Based on the classic theory and weak statement of interface migration, a finite-element method is developed to simulate the evolution of intragranular voids in copper interconnects caused by interface migration induced by stress migration. The validity of the method is confirmed by the agreement of the numerically simulated the undulating surface with that predicted theoretically. Through a large number of numerical simulations, we find that the evolution of the intragranular voids has two trends, namely, void growth and void shrinkage. And the shape of the void is governed by the stress, $\beta$ , the linewidth, ${\stackrel{\text{?}}{\sigma }}_c$ , and the initial aspect ratio of the intragranular void, ${\stackrel{\text{?}}{h}}_c$, and there exist critical values for these parameters ( ${\beta }_c$, $\stackrel{\text{?}}{h}\mathrm{?}{\stackrel{\text{?}}{h}}_c$ and $\stackrel{\text{?}}{\sigma }$ ). When $\stackrel{\text{?}}{h}$, $\beta$ or $\stackrel{\text{?}}{\sigma }$, the intragranular void will grow along the major axis; otherwise, the intragranular void will shrink into a cylinder. The increase of the stress, or the aspect ratio, or the decrease of the linewidth is beneficial to void growth. And the area of void growth will increase faster with bigger $\stackrel{\text{?}}{h}$ , smaller $\beta$ or bigger $\mathrm{~}$ . But, the decrease of the stress or the aspect ratio, or increase the linewidth accelerates void shrinkage and the shrinkage area will decrease faster with smaller $\mathrm{~}$ , bigger $\beta$ or smaller $\sigma$ .