适用于超深亚微米集成电路制造与验证流程的光学邻近修正方法研究

《适用于超深亚微米集成电路制造与验证流程的光学邻近修正方法研究》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、浙江大学电气工程学院博士学位论文适用于超深亚微米集成电路制造与验证流程的光学邻近修正方法研究姓名：陈晔申请学位级别：博士专业：电路与系统指导教师：严晓浪;史峥20080501浙江大学博士学位论文的模型，这通常是困难而且是耗时的。本文介绍了一种新的光刻建模环境。除了实现了传统的建模流程之外，新工具还提供一组易于定制以及扩展的模块。使用这些模块，用户可以通过编写简明的Tcl脚本来实现的新的光刻模型。这能够大幅提高建模的效率。关键词：光学邻近修正、光刻仿真、光刻友好的版图设计、可制造性设计、亚波长光刻IIAbstractInthelastseveraldecades,thef

2、astdevelopingofmicroelectronictechnologyandgrow-ingrequirementsfromthevariouskindofapplicationsmakethemodemintegratedcircuit(IC)morecomplicatedinscaleandfunctionally．ThesewillinevitablyleadtothehigherdemandsformanufacturingtechniquesofICsystems．AsfarasthestateoftheartofICmanufacturingsys

3、tems，thetechniqueofSOcalledResolutionEnhance-mentTechniques(RETs)makeitpossibletoproductfeatures2oreven3timessmallerthanthewavelengthofilluminationsystem．Thelimitsofopticshavebeenpushedwellbeyondwhatwasexpectedjustafewyearsago．Asfornow,illuminationof193nmwavelengthisusedin45nmICproductio

4、n．AmongallkindsofRETs，OpticalProximityCorrection(OPC)isthemostsophisticatedandwidelyusedtechnique．Itachievestheaimofcancelingoutdistortionandimprovingtheprint-abilitybymodifyingmaskpat-terns．Sophisticatedcorrectionalgorithm，goodlayoutdesignstyleandaccuratemodeloflithographysystemarethree

5、keysofsuccessfulOPC．Inthisthesiswewillintroduceoureffortsonallthesethreeaspects．Model-basedOPC(MBOPC)isaniterativeprocess．Itsconvergencedependsonboththeinitialvaluesandtheselecteddirectionofiteration．Inthisthesis，anewkindofMBOPCalgorithmisintroducedtoimprovetheconvergentspeedandconvergen

6、trange,whichisbasedongeneralizedintensitydistributionfunctionsandtakesthein-teractionsbetweenedgesintoaccount．Tomakethenewalgorithmfaster,afewmeth-odstoacceleratetheJacobianmatrixcalculationandlinearsystemsolvingprocessisalsointroduced．Anedgebiasmodelingmethodisdevelopedinthisthesis,whic

7、hhasveryfastcorrectionspeedandmoderatecorrectionaccuracy．Betterinitialvaluesfortheiterativeprocessofconventionalmodel-basedOPCalgorithmcanbeprovidedbythismethod，withalittlerun-timepenalty．Thenumberofiterationsisreducedbecauseofthebetterinitialvalues．Thusthetotalrun-timeis