The emission microscopy （EMMI） test is proposed as an effective method to control the polysilicon over-etching time of advanced CMOS processing combined with a novel test structure, named a poly-edge structure. From the values of the breakdown voltage （Vbd） of MOS capacitors （poly-edge structure） ,it was observed that,with for the initial polysilicon etching-time, almost all capacitors in one wafer failed under the initial failure model. With the increase of polysilicon over-etching time, the number of the initial failure capacitors decreased. Finally, no initial failure capacitors were observed after the polysilicon over-etching time was increased by 30s. The breakdown samples with the initial failure model and intrinsic failure model underwent EMMI tests. The EMMI test results show that the initial failure of capacitors with poly-edge structures was due to the bridging effect between the silicon substrate and the polysilicon gate caused by the residual polysilicon in the ditch between the shallow-trench isolation region and the active area, which will short the polysilicon gate with silicon substrate after the silicide process.
Breakdown voltage （Vbd） and charge to breakdown （Qbd） are two parameters often used to evaluate gate oxide reliability. In this paper,we investigate the effects of measurement methods on Vbd and Qbd of the gate oxide of a 0.18μm dual gate CMOS process. Voltage ramps （V-ramp） and current ramps （J-ramp） are used to evaluate gate oxide reliability. The thin and thick gate oxides are all evaluated in the accumulation condition. Our experimental results show that the measurement methods affect Vbd only slightly but affect Qbd seriously,as do the measurement conditions.This affects the I-t curves obtained with the J-ramp and V-ramp methods. From the I-t curve,it can be seen that Qbd obtained using a J-ramp is much bigger than that with a V-ramp. At the same time, the Weibull slopes of Qbd are definitely smaller than those of Vbd. This means that Vbd is more reliable than Qbd, Thus we should be careful to use Qbd to evaluate the reliability of 0.18μm or beyond CMOS process gate oxide.
