第27卷第4期 半导体学报 Vo1.27 No.4 2006年4月 CHINESE JOURNAL OF SEMICONDUCTORS Apr.,2006 Ph0t0lith0graphy Process Simulation for Integrated Circuits and Microelectromechanical System Fabrication Zhou Zaifa ,Huang Qing’an,and Li Weihua (Key Laboratory ofMEMS oftheMinistry ofEducation,Southeast University,Nanifng 210096,China) Abstract:Simulations of photoresist etching,aerial image,exposure,and post・bake processes are integrated to ob。 tain a photoIjthography process simulation for microelectromechanical system(MEMS)and integrated circuit(IC) fabrication based on three.dimensional(3D)cellular automata(CA).The simulation results agree well with availa- ble experimental results.This indicates that the 3D dynamic CA model for the photoresist etching simulation and the 3D CA model for the post.bake simulation could be useful for the monolithic simulation of various lithography Processes.This is determined to be useful for the device-sized fabrication process simulation of IC and MEMS. Key words:cellular automata;process simulation;photoIjthography simulation;model;TCAD EEACC:2570;2560;2550E CLC number:TN4 Document code;A Article ID:0253・4177(2006)04-0705・07 the simplest models for implementation in MEMS 1 IntrOductiOn and IC fabrication process simulation,the cellular automata model,which was first presented by von As the dimensions of integrated circuits(IC) Neumann following Ulam’s suggestionsL .has are scaling down to the nanometer regime and the been successfully applied to the simulation of vari- complexity of micro..electro--mechanical system OUS fabrication processes such as photoresist etch- (MEMS)design and fabrication is increasing, ingEs~7l,depositionE8,9],and silicon anisotropic three.dimensional(3D)photolithography simula- etchingE10].The 3D dynamical CA model for pho. tion has become necessary for accurate analysis of toresist etching simulation[11 J.which is extended complex structures such as contacts,corners, from the 2D dynamical CA modelL5J.has been islands,and 3D defectsE ] Simulation aids in un— presented and tested using wel1.known etching derstanding the effects of parameters during pho— rate distribution functions and is demonstrated to tolithography so that masks can be optimized and be accurate,fast and stable. the accuracy of critical structures and the reliabil- In this paper,simulations of the photoresist itv of devices can therefore be improvedL .The etching process using the 3D dynamic CA model simulation of photolithography is a complex task and the post.bake process using the 3D CA model that usually includes precise modeling in aerial im- are integrated together for the first time with an age simulation,exposure simulation,post-bake aerial image simulation and an exposure process simulation,and photoresist etching simulation simulation to accurately describe the effects of pa- (development simulation)E3].The etching simula- rameters during the phOtOlithOgraphy process.The tion is usually the most time,consuming step and simulation results agree well with available exper- greatly affects the accuracy of the whole photoli— imental results.This indicates that the 3D dynamic th0graphy simulation.So a fast and accurate 3D CA model for photoresist etching simulation and model for photoresist etching is greatly desired 3D CA model for post.bake simulation are accu- for effective photolithography process simulation. rate,fast,and can be integrated with other photo- Due to its advantages such as ease in handling lithography simulation steps.This will be useful topological changes,adaptive mesh methods,and for the simulation of the device.sized fabrication *Project supported by the National Outstanding Young Scientists Foundation of China(No 50325519) 十Corresponding author.Email:zhouzaifa@yahoo.com.cn Received 17 January 2006,revised manuscript received 15 February 2006 @2006 Chinese Institute of Electronics 维普资讯 http://www.cqvip.com
7O6 半导体学报 第27卷 process of IC,and MEMS 2 3D dynamical CA model The photoresist is divided into f×m×凡identi— cal cubic cells with side length a,and the 3D Moore neighborhood is adopted in the 3D dynamic CA model,as shown in Fig.1.There are 6 adjacent ca— bic cells,12 diagonal cubic cells,and 8 point cubic cells in the neighborhood of cell(i,f,k).A boundary cell between the internal photoresist and etchant will be etched by etchant flowing Fig.1 Moore neighborhood of the cellular automata from its neighbors.Theoretically,the effect of etchant from all neighbors should be taken into account when the state of cell(i, ,k)is upda— ted.But the 8 point diagonaI cells do not signifi. cantly affect the change of the state of cell (i,J,k).so the effect of the point diagonal cells is neglected in this mode1.The Iocal state Cf.i,k (t)of each cell at time t is defined as the ratio of the etched volume V。(t)to the total volume V : CiJ. (t)= V。(t) .(1) The update rule is then given by Ref.[11]: Cl,』. (tl+丁)=Cf,J.k( t1)+ V (tl+丁) dIac t+V。(tl+丁) 口3 V (tl+丁) dj 。。 =(a—df^一dif)× (dj,d七f+dild kh+djhd kl+djhd )+ (a—d yh—d y1)(dlfd七f+dlfd +di^d +df^d )+ (a—d —d七f)(dlfdjf+dlfdih+di^dff+di^dih)+ (口一di^一d f)(a—d f—d )(d,f+djh)+ (盘一dih—dff)(a—d“一d蛐)(djf+df^)+ (盘一d蛐~d f)(a—dff—djh)(d{f+df^)+ ditdjtd^f+dftd id舳+ditdj^d +d.dj^d + df^djid七i+di^dffd舳+di^d『^d +df^d『^d (3) n V (tl+T)di4go l 百/./R;,,, (T+Tcf+1,,, +l(t1)) + D 2 百J、. , (丁+Tci ̄i,j,k-i(f1))。+ D^2i.j k(丁+TcH.,. +l(f1)) + D^2f,,. (丁+Tci-1,j,k-1(f1))。+ D^2f.,,^(丁+Tci.j十1, +1(f1))。+ D 百n 2 (丁+Tc l, 1(f1)) + D2百』、flJ’ (丁+Tci,t-i,k+i(f1)) + D 百n 2 (丁+Tcf,卜l 七1(f1)) + D2百n f,,, (71+Tci ̄l,i ̄l,k(f1))。+ D 2百J、l小 (7]+Tci ̄l,j-l,k(f1)) + D^2百j,J. (丁+Tcl-i4 ̄l,k(f1))。+ D^2百f,,,k(丁+Tci-l,i-l,k(f1)) (4) where V (tl+丁) dj 。 t and V (tl+丁)d。 go l de・ scribe the effects from the adjacent and the diago— nal cells,respectively.Tc(t1)is the time compen— sation value.The parameter D is adopted to de— scribe the effect of the diagonal cells,and differ— ent values will be used accordi ng to the state of the diagonal cells and the state of the shared adja— cent cells of these diagonal cells and cell(i, ,足). The variables in Eq.(3)are defined as fd“=R (T+Tci l.』, (t1)) l df^=R¨, (T+Tc『+1.J. (t1)) {] =d Rjl (R i,Ⅲj,k ( +丁T 小 T fc ’t./j-+1,. (t)1)) … ㈣ l d kJ=R J k(T+Tc1.j1 (ti)) 【d =R , (T+Tcf}J l(t1)) In the 3D dynamic CA model,only the boundary cells are processed,and an interior cell that has at least one fully etched adjacent or diag- onal neighbor becomes a boundary cel1.At the be- ginning of the simulation,the photoresist is divid- 维普资讯 http://www.cqvip.com
第4期 Zhou Zaifa et a1.: Photolithography Process Simulation for Integrated Circuits and… 707 ed into f×m×凡cubic cells,and all the cells are duce the simulation sDeed——the time step value in the“0”state(unetched),and the time compen. in the 3D dynamic CA modeI of T=a/IOR…is sation value for all cells is“0”.Initial conditions adopted,where R is the maximum etch rate of are imposed by changing the states of some cells, the cells. and the etching process starts along the etch The model has been tested using the well— boundary cells.Some boundary cells will gradually known etching rate distribution functions and has become fully etched,and those fully etched cells been found to be stable,fast,and accurateL .The will be excluded from the boundary cell array dur. 1.0 m×1.0 m x 1.0 m photoresist was divided ing the simulation.At the same time,some interior into 100×100×100 identical cubic cells.and the cells will become boundary cells according to the test function is given bv[12] CA rules.These new boundary cells will be insert— R( )=4x m/s (6) ed into the boundary eell array。and pointers cor— where x and Y are the distances on the wafer,and responding to these new boundary cells are also z is the depth of the photoresist.The 3D simula. created.Considering that there is always a trade. tion profiles corresponding to different etching off between simulation accuracy and speed—— time steps are shown in Figs.2(a)and(b).The namely.a large time step will reduce the simula— simulation time was found to be 35.9s on a Sun tion accuracy.whereas a small time step will re一 Ultra work station. 暑0 i 毫0 s 占 。O O 7 1 Fig.2 Simulation profiles with Equation(6) (a)and(b)correspond to different etching times The 3D dynamic CA model for the simulation resist—coated wafer.The exposure process gener— of photoresist etching has been found to be accu— ates acid in the resist.During the subsequent post— rate,fast,and stable,but it is necessary to in. bake process.the resist undergoes an acid.cata. tegrate simulations of aerial xmage'exposure,post— 1yzed cross—linking reaction which determines the bake,and photoresist etching in order to accurate— etch rate distribution in the photoresist.In the last ly describe the effects of parameters during the step,the photoresist is etched according to the photolithography process.In the following sec— etch rate distribution obtained from the two for. tion,the photoresist etching simulation using the met simulation steps.The typical simulation steps 3D dynamic CA model will be integrated with for the photolithography process of chemical am- simulations of the aerial image,exposure process, plification photoresists are shown in Fig.3. and post—bake process. : 3 Simulation and discussion Fig.3 Typical simulation steps for photolithography process of chemical amplification photoresists The general steps for phOtOlithOgraphy simu- lation are aerial image simulation,exposure simu- The new SPLAT[ 。]can efficiently express lation,post—bake simulation,and photoresist etch— the influence of various parameters of the image ing simulation.The aerial image exposure process tool for high numerical aperture(NA)lithogra— involves calculating the image distribution on the phy resists(NA ̄0.6).With the parameters listed 维普资讯 http://www.cqvip.com
708 半导体学报 第27卷 in Table 1,such as partial coherence ,focus,and exposure dose I,the new SPLAT is successfully employed to simulate the aerial image on the re. sist.coated wafer.and bottom anti.reflective coat— ing(BARC)technology is implemented using AR3 anti.reflectant.Then the well—known Dill’s Table 1 Parameters for hard—contact exposure process of UVTMl l3 chemical amplification photore- sist Simulation step Parameters Reference n‘ill∞n=1.577,n it=1.786, k ili =一3.588,rtAR3:1.46, Aerial image kA1t3:一0.47,focus 0, [13,15] J=13rrd/cm ,KrF. NA=0.63, =0.6 Dill A=0,0746 ̄m~・ Exposure DiDill Cll B=0.563%m。。。. =0.041lcm2/mJ. [15,16] [PAG]o=i Time=90s, Temperature=125 ̄C。 k1=0.7068s~, Post.bake k2=0.0001s。。。。 [15~173 P=1,r=1,q 1.34, [口]=0.264,Mo=1 N 1.345,Rmax 0.8944Fm/s, Photoresist etch R i =5.8×10 Fm/s, nsimulation —notch=22.367, [18] M【n。t h=0.586,Time=45s, 一 Temperature=20"C ABC model[ 。which was first presented to model the exposure process of diazo—type resists,is em— ployed to model the exposure process of chemical— ly amplified photoresists.Dill’s ABC model is used to describe the reaction during the exposure process as a[P AG———a —t 一]:一CDill I[-PAG] (7) ot=A 1[PAG]+B Di1l (8) where A Dill(/am ),B Di1l(/am一 ),and C Dill(cm / mJ)are the parameters of Dill’s ABC mode1. [PAG]is a normalized concentration of photoac. id generator,and I is the exposure illumination power density obtained from the aerial image sim. ulation.Equations(7)and(8)are used to calcu. 1ate parameter L J and the normalized acid gen. crated.After the exposure is calculated,the nor— malized acid generated is[A](O)=1一[PAG], which will be used as the initial condition for the following post—bake simulation. The exact changes and mechanism during the post—bake process are complex,but a relatively simple theory is presented to explain the reaction mechanismE .According to this theory。the reac tion occurring in the photoresist cells can be de. scribed by a cellular automata model given by mi,i,k(t)=mi,i(t一 )一k1 EA]I。f。 (t一 )× mf_j. (t—T)×T (9) EA]i,,. (t)=[ ]lIj, (t一 )一 k 2[ ] . (t一 )×T (10) where T is the time step in the post.bake simula. tion,[ ] . (t)is the normalized acid generated in cell(i,j,k)at time t,and the initial value of [A]llj。 (t)is[A] . (0).Also,mi,j,k(t)is the normalized inhibitor concentration in cell(i,f, k)at time t,k】and k 2 are the reaction rate coef. ficients,and q is a parameter related to the type of photoresist.The final normalized inhibitor con— centration m f.“ after exposure simulation and post—bake simulation can be calculated using Eqs. (9)and(10).Thus,according to the notch mod. elc 。the etch rate for all photoresist cells can be calculated by R(mj。i,k)=R 4x(1一ml,J。 ) × (a+1)(1一mf。j。 )n_notch a+(1一mi,j,k) 曲 +R i (11) (n notch+1) a= =_1——(1一Mth Dt h)n_D。 (12) (n notch一1) where R is the maximum etch rate of the pho— toresist,R I is the minimum etch rate of the pho— toresist, is the sensitivity of the developer,n notch is the sensitivity of the notch,Mth aoteh iS the threshold inhibitor concentration where the notch occurs,and m i.j. is the final normalized inhibitor concentration in cell(i,f,k). According to the available parameter values of UV M 113 chemical amplification photore. sistE ,as listed in Table 1。the etch rate for all photoresist cells can be determined so that the 3D dynamic CA model can be employed to accom. plish photoresist etching simulation.Thus the whole photolithography of UV 1 13 chemical amplification photoresist can be successfully simu— lated. Figures 4 and 5 show the simulation profiles 维普资讯 http://www.cqvip.com
第4期 Zhou Zaifa et a1.: Ph0t0lith0graphy Process Simulation for Integrated Circuits and… 709 of photolithography of UV M 1 13 photoresist cor— ing experimental resultsL ].Figures 5(a)and(b) show cross-sectional views of the simulation and the corresponding experimental results[ 8]for responding to two different mask shapes.Figures 4(a)and(b)are 3D views of the simulation re— suits corresponding to different time steps.The l10nm/220nm line/space with a different focus value of一0.2.The etching time in Fig.5 iS 45s. etching times are 22s and 45s for Figs.4(a)and (b),respectively.Figures 4(e)and(d)are cross— The above simulation results agree well with a— sectional views of the simulation and correspond vailable experimental results. 88 88 (c) Fig.4 Simulation profiles of ph0t0lith0graphy process for UV 113 chemical amplification photoresist (a), (b)3D view of the simulation results corresponding to 22s and 45s;(c),(d)Cross,section view of simulation and corresponding experimental results[ 。] Fig.5 Cross.section view of simulation(a)and corresponding experimental(b)results 。 for llOnm/220nm line/space with a different focus value of一0.2 Figure 6 shows the photolithography simula— Although only the photolithography process tion results for capacitor fabrication,which also a- of UVTM 113 chemical amplification photoresist iS grees with available experimental resultsL3’ .The simulated in this paper,since the aerial image,ex. results indicate that the photoresist etching simu- posure,post—bake,and photoresist etching simula— lation using the 3D dynamic CA model can be SUC- tions are relatively independent,the 3D dynamic cessfully integrated with other photolithography CA model for the simulation of photoresist etch— simulation steps,SO that the effect of parameters ing and 3D CA model for the simulation of post- during the photolithography process can be accu— bake can also efficiently simulate the photolithog- rately described. raphy process of other photoresists. 维普资讯 http://www.cqvip.com
7lO 半导体学报 第27卷 gi、I_暑u口 6 6 Fig.6 Photoresist etching profiles for capacitor fab— rication(a)and(b)correspond to different etching times. 4 Conclusion The photoresist etching simulation using the 3D dynamic CA model and post.bake simulation using the 3D CA model have been successfully in. tegrated with aerial image simulation and expo. sure process simulation for UV 1 13 chemical am. plification photoresist.The simulation results a. gree well with available experimental results.indi. cating that the 3D dynamic CA model for photo— resist etching simulation and 3D CA model for post-bake simulation are accurate,fast,and can be integrated with other photolithography simulation steps.This will be useful for the device—sized fab rication process simulation of IC and MEMS. References [1]Strojwas A J,Zhu Zhengrong,Ciplickas D.et a1.Layout manufacturab订ity analysis using rigorous 3-D topography simulation.IEEE International Semiconductor Manufacturing Symposium,2001:263 [2]Cole D C,Barouch E,Conrad E D,et a1.Using advanced simulation to aid microlithography development.Proceedings of the IEEE,2001,89(8):1194 [3]Karafyllidis I.A three-dimensional photoresist etching simu- lator for TCAD.Modeling and Simulation in Materials Sci- ence and Engineering,1999,7(2):157 [4一 Neumann J.Theory of self-reproducing automata.Urbana: University of Illinois Press,1966 [5]Zhou Zaifa,Huang Qing’an,Li Weihua,et a1.A novel 2-D dynamic cellular automata model for photoresist etching process simulation.Journal of Micromechanies and Micro engineering,2005,15(3):652 [6]Zhou Zaifa,Huang Qing’an,Li Weihua,et a1.Negative chemical amplification process simulation in integrated cir- cults and microelectromechenical systems fabrication.ICO- 20,2005 [7]Strasser E,Selberherr S.Algorithms and models for cellular based topography simulation.IEEE Trans Computer-Aided Design Integrated Circuit Systems,1995,14(5):1104 [8]Karafyllidis I,Georgoulas N,Hagouel P I,et a1.Simulation of deposition-topography granular distortion for TCAD.Model- ing and Simulation in Materials Science and Engineering, 1998,6(2):199 [9]Huang Qing’an,Zhou Zaifa,Li Weihua,et a1.A modified cellular automata algorithm for boundary movement of dep。 osition topography simulation.Journal of Micromechanics and Microengineering,2006,16(1):1 [1O]Jiang Yanfeng,Huang Qing’an.Simulation of silicon aniso- tropic etching using cellular automata method.Chinese Jour- nal of Semiconductors,2005,26(3):618(in Chinese)[姜岩 峰,黄庆安.硅各向异性腐蚀的原子级模拟.半导体学报, 2005,26(3):618] Fill Zhou Zaifa,Huang Qing’an,Li Weihua。et a1.A novel 3-D dynamic cellular automata model for photoresist etching process simulation.IEEE Trans Computer-Aided Design of Integrated Circuits and Systems,accepted [12- 1Ferguson R A.Modeling and simulation of reaction kinetics in advanced resist processes for optical lithography.PhD Dissertation,University of California,Berkeley,1991 [13]Yeung S,Derek L,Robert L,et a1.Extension of the Hopkins theory of partially coherent imaging to include thin-film in- terference effects.Proc SPIE,1993,1927:452 [14]Dill F H,Neureuther A R,Tuttle J A.et a1.Modeling projee. tion printing of positive photoresists.IEEE Trans Electron Devices,1975,22(2):456 L15] Erdmann A,Henke W,Robertson S。et a1.Comparison of simulation approaches for chemically amplified resists.Proc SPIE,2001,4404:99 L16j Roberson S,Mack C A,Maslow M.Towards a universal re。 sist dissolution model for lithography simulation.Proc SPIE, 2001,4004:1l1 [17]Jakatdar N,Bao J W,Spanos C J.et a1.A parameter extrac. tion framework for DUV lithography simulation.Proc SHE, 1999,3677:447 [18-1 Baluswamy P,Weatherly A,Kewley D,et a1.Practical resist model calibration.Proc SPIE,2003,5040:1556 维普资讯 http://www.cqvip.com
第4期 Zhou Zaifa et a1.: Ph0t0lith0graphy Process Simulation for Integrated Circuits and… 71 1 集成电路和微电子机械系统加工过程中的光刻工艺模拟 周再发 黄庆安 李伟华 (东南大学MEMS教育部重点实验室,南京210096) 摘要:基于3D元胞自动机方法实现了影像成形、曝光、后烘和光刻胶刻蚀过程等集成电路和微电子机械系统加工 过程中的光刻过程模拟模块的集成.模拟结果与已有实验结果一致,表明基于3D元胞自动机方法的后烘和光刻胶 划蚀模拟模块的有效性,这对于实现集成电路和微电子机械系统的器件级的工艺模拟具有一定的实用性. 关键词:元胞自动机;工艺模拟;光刻模拟;模型;计算机辅助设计 EEACC:2570;2560;2550E 中图分类号:TN4 文献标识码:A 文章编号:0253—4177《2006)04-0705-07 *国家杰出青年科学基金资助项目(批准号:50325519) 十通信作者.Email;zhouzaifa@yahoo.tom.Cll 2006.O1.17收到,2006.02-15定稿 @2006中国电子学会
因篇幅问题不能全部显示,请点此查看更多更全内容