Research On Test Pattern Generation Methods For The Optical Proximity Correct

The rapidly development of semiconductor technology has entered the nanometer-scale fabrication era nowadays.Integrated circuit manufacture approaching to the resolution limit under the condition of low k1 lithography,especially to those technology nodes below 65 nm,even in the case of using the most advanced immersion optical techniques.Complex two-dimensional layout patterns became increasingly sensitive to photo-resist bridging and necking defect.The rapidly development of resolution enhancement technology(RET)and design for manufacture to be more inevitable and essential when applied in the semiconductor technology,it helped the technology update to 16 nm even 10 nm.Optical proximity correct technique was the most frequently-used and the primary method for RET.The model-based OPC favored by every foundry with higher correction accuracy and more proven technology.The OPC model was generated by the CD data of the test patterns in the same process,then based on the model and combined with several OPC correcting algorithms to handle the customer's GDS data.The GDS data were much more complicated than before while the OPC model was requiring higher correction accuracy.This paper aimed to introduce one kind of one and two dimensional OPC test patterns generate methods with highly precision.For one dimensional OPC test patterns,firstly selecting several optical signal sampling points on the full pitch patterns,comparing the actual CD data with the target and then combining with the assist features to decide the test patterns pitch setting rule.After the contrast test,the OPC test pattern number was decreased 14% and the time for model establishment shrink 7.65% compared with the traditional way,moreover the real GDS patterns edge placement error within spec increased from95.88% to 96.72%.The two dimensional OPC test patterns generate method was applying the Monte Carlo method,which based on four groups of typical unit patterns with a fixed filling probability randomly filled in the frame cell.It also took the mask rule checker into account to present patterns without any design rule violation,unless the frame was being filling-in.Through the contrast test,it easier to discovered that the proposed method was improved by 13% and 15% individually for pattern bridging and necking which significantly efficient in hotspot patterns simulation and forecast.And the fixed filling probability made the final OPC test patterns proximate to the real design pattern density which ensures the OPC model simulation accuracy.