Scan chain diagnosis has become a critical issue to yield loss in modern technology. In this thesis, we present a scan chain partition algorithm to improve scan chain fault diagnosis resolution. In our scan chain partition algorithm, we rst construct a controllability graph by netlist ependency, then perform the partition algorithm to decide which scan chain a scan cell belongs to. The experimental results show that our method can reduce the number of suspect scan cells from 378-4 to at most 2-3 for most cases of ITC'99 benchmarks.