Coupling-Constrained Dummy Fill for Density Gradient Minimization

Dummy fill is typically performed by foundries to achieve layout pattern uniformity for chip yield enhancement. However, filling dummies may greatly increase interconnect coupling capacitance and lead to explosion of mask data. Therefore, it is desirable to simultaneously consider the coupling constraints, the number of inserted dummies, and the density gradient during dummy fill. In this paper, we present dummy fill algorithm that adopts a three-stage technique of coupling constraint regulation, gradient-driven multilevel dummy density analysis, and ILP-based fill synthesis for dummy feature minimization. Based on the Gaussian smoothing, the density gradient can be greatly minimized. Experimental results show that our algorithm can reduce the standard deviation of the metal density by up to 37%, the average neighboring density difference by up to 49%, and dummy counts by 41% with a reasonable runtime overhead, which preserves more flexibility for subsequent manufacturability enhancement.