User view of MEB Maskless Lithography

Transcription

1 User view of MEB Mskless Lithogrphy Jck J.H. Chen 2010 Interntionl Symposium on Lithogrphy Extensions Kobe, Oct 22, 2010

2 Outlines ITRS rodmp Performnce of MAPPER Pre-lph TSMC Chllenges nd solutions of MEB ML2 for HVM Conclusions 2

3 ITRS Rodmp 2009 ML2 How fr cn we go? 3

4 Drivers of MEB ML2 Cost The lowest CoO by the ssumption of >100WPH cluster tool Sve on ll criticl msks! Short cycle time for new tpe-out Remove msk mking nd verifiction cycles Esy for smll volume prototyping or whtever modifiction of lyout Avoid double ptterning less lyers, less D0 Extensibility For production: Resolution hs no limits but trdes off Throughput more prllelism!!! For R&D: possibly do N+1 or N+2 node development t existing tool Lrge DoF ( µm insted of nm ) Less design rule restrictions Mix&mtch with msk-bsed opticl scnners 4

5 Cost Advntge of ML2 Cost MPW = Cost PWP = Costtool + Cost MPW High cost due to expensive msks Cost per wfer pss Costtool = Mskprice($) Wfers Toolprice($) Productivity (Wfers) Msk-bsed opticl Lithogrphy ML2 (high cost) ML2 (low cost) If CostML_tool > CostMB_tool If CostML_tool ~< CostMB_tool ML2 more economicl MB more economicl Brekeven # W1fers 5 Then ML2 is lwys cheper! # Wfers per msk set

6 MAPPER Pre-Alph TSMC Shipped into Gig-Fb12 since Jul 09 Strted exposure since Dec bem@ 5keV, rster scn exposure Stndlone with 300mm wfer stge Only using CAR Successfully exposed more thn 210 wfers Resolution ws stedily t ~30nm HP by 35nm spot size. Hs just upgrded to 25-nm spot size. Asterix06 Tool Utiliztion Summry in %, 4.63%, 53hrs 99hrs 14.62%, 312hrs 38.86%, 830hrs 39.43%, 842hrs Dily Strt 6 Down Engineering Exposure Wit

7 Exposure Overview 10 bems K1 K2 J1 I1 J2 I2 11 bems G1 F1 D1 D2 B1 A1 D3 B2 A2 D4 B3 A3 D5 B4 A4 D6 B5 A5 B6 A6 F8 D7 A7 E9 C8 B8 F10 E10 D10 C10 B9 A9 H10 G10 D9 C9 A8 I10 F9 D8 B7 J10 H9 G9 E8 C7 J9 H8 F7 K10 I9 G8 E7 C6 J8 H7 F6 K9 I8 G7 E6 C5 J7 H6 F5 K8 I7 G6 E5 C4 J6 H5 F4 K7 I6 G5 E4 C3 J5 H4 F3 K6 I5 G4 E3 C2 J4 H3 F2 K5 I4 G3 E2 C1 J3 H2 G2 E1 K4 I3 H1 300mm wfer K3 B10 A10 Exposure re of 110 bems: 1.4 mm x 1.5 mm With the blnker rry module nd opticl dt pth, ech bem writes individul lyout. Stge position controlled by CAP sensors limited movement rnge Not yet scnning exposure not yet stitching 7

8 Exposure Overview 10 bems K1 K2 J1 K3 J2 In cse of full field 13k bems I1 I2 I3 1x Field 11 bems H1 G1 H2 G2 F1 E1 B1 EO slit with 13,000 bems 300mm wfer A1 B2 A2 D4 B3 A3 D5 B4 A4 D6 B5 A5 B6 A6 F8 D7 A7 E9 C8 B8 F10 E10 D10 C10 B9 A9 H10 G10 D9 C9 A8 I10 F9 D8 B7 J10 H9 G9 E8 C7 J9 H8 F7 K10 I9 G8 E7 C6 J8 H7 F6 K9 I8 G7 E6 C5 J7 H6 F5 K8 I7 G6 E5 C4 J6 H5 F4 K7 I6 G5 E4 C3 J5 H4 D3 K6 I5 G4 E3 C2 J4 F3 D2 K5 I4 G3 E2 C1 J3 H3 F2 D1 K4 B10 A10 Exposure re of 110 bems: 1.4 mm x 1.5 mm With the blnker rry module nd opticl dt pth, ech bem writes individul lyout. Stge position controlled by CAP sensors limited movement rnge Not yet scnning exposure not yet stitching 8

9 Bem Grid Error monitor Ech circle represents one bem intensity bove the bem stop perture. Bem grid error (BGE) is the positioning error of the bem center to the perture hole. It ws mesured nd corrected before ech exposure job. Only bem F2 not switching, BGE kept stble, with spec < 15µm, for longer thn 2 months Grid Error Monitor 30 Men+3σ, μm before lignment fter lignment 15 New ILO Implement ML New X-CAP sensor b b b b b 20 b b b b 01 b b b b b b b 29 0 New BSW Time (dte) 9

10 Bem Current monitor Bem Current Monitor 50 Vent chmber to fix stge Sensor clibrtion 180 Stge unstble b c b b b b b 0 b sigm[pA] Ib men [pa] 210 Dte Men-YAG Men-Frdy Cup 3σ-YAG The cthode hs been used for more thn 6 months. Individul bem current mesurement through YAG nd Frdy cup showed consistent trend. Men bem current ws yet lower thn spec 200pA. Bem-to-Bem current devition ws 3σ < 8pA. 10

11 Spot Size nd Focl Z monitor BMS problem Focl Z ws unstble for 1 dy fter venting chmber b b b b b b b b b Stge CAP mlfunction Focl Z Men, nm Spot Size Monitor nd Focl Z monitor Time (dte) Spot size nd Focl of ll 110 bems were mesured simultneously by using Knife-Edge (KE) sensor. The spot size mesurement is yet not enough ccurte due to influences of the KE profile nd stge dynmics. However, s the resolution 30-nm HP ws consistently observed, it s believed the ctul spot size should be ~35nm s designed. The POS hs just been upgrded to 25-nm spot size. Exposure dt is being collected. 11

12 Resolution by 35-nm spot size HP30nm HP45 HP36 HP32 HP30 Multiple CAR resist smples hve shown consistent resolution t 30nm HP. LWR (>10nm) is yet too high! 12

13 DoF > 45nm CD Lrge DoF for CD control. Precise leveling is nevertheless required for overly. 13

14 CDU test t 45-nm Holes Spec 45-nm dense holes 45-nm iso holes # of bems > Men (nm) MTT < 7% σ (nm) <10% Mx (nm) Min (nm) At the time of test, 4 bems were missing due to broken dt chnnels. 3 of them were recovered lter by correcting setting of the blnker rry nd mistke of bem mpping. 14

15 Outlines ITRS rodmp Performnce of MAPPER Pre-lph TSMC Chllenges nd solutions of MEB ML2 for HVM Conclusions 15

16 Cluster concept for 100WPH tool HVM clustered production tool: >13,000 bems per chmber (10WPH) 10WPH x 5 x 2 = 100WPH Footprint ~ArF scnner < 2/3 EUV scnner In-line to trck MAPPER single column tool t 10WPH Courtesy by MAPPER, Proc. of SPIE 2009,Vol. 7271, 72710O 16 1m Interfce to trck

17 Scn Speed Comprison MB Scnner: 90 fields, 178 cc & dec MAPPER: 90 fields, 20 cc & dec Scn speed ~ wph Reticle scn speed ~2000 mm/sec 17 Scn speed <30 10 wph, Less dynmics problem Bems trce stge position error by dt correction. Slower nd lighter stge, smll footprint

18 Benchmrk of MEB ML2 Strtegies Source: Elmr-Pltzgummer, SPIE-2010-keynote_ Source Br & uniformity! Coulomb Effect Smll field high scn speed High Br source or? 18 Multi-columns required

19 Ptterned Bem Concept 7x7 sub-bems in one lenslet, So the required subbem current reduced from 13nA to 0.26nA! chllenge: Uniformity in full field! Source: M.J. Wielnd, MAPPER, SPIE

20 Why 5keV is better choice thn 50keV Use of MEMS micro-column E-Optics MEMS gives the higher precision of E-optics elements Esy nd low cost for spres Lower heting power (1/10x to 50keV) onto wfer High sensitivity of the resist (~5-10x to 50keV) Ambit <1 µm due to scttering effect EPC similr to OPC Shllow penetrtion depth less impct to underneth devices 20 Low k dielectric curing effect

21 250 Assumptions: Power thru resist Power in resist EB simultion by CASINO v2.42, incident electrons onto 50nm PMMA on Si. For 5 kev, Ein resist:ethough resist ~10%:90% For 50keV, Ein resist:ethrough resist ~0.1%:99.9% EUV@30mJ/cm2, 180wph 50keV 20wph ArF, ssume resist bsorbnce α ~ 1 µm-1 with perfect BARC. 5keV 20wph Power onto wfer (mw/cm 200 ArF@30mJ/cm2, 200wph 2) Heting Power onto Wfer nd into Resist EUV, ssume α ~ 4 µm-1. Trnsmission rte (pttern density): EB ~ 50%, EUV ~ 50%, ArF ~ 75%. For 5keV EBem, the heting power of the substrte wfer is the lowest. 21

22 Concerns of dt hndling in HVM Cluster 1 Hours? OASIS/GDSII Tpeout Dt Pth 1-1 6min/tool EPCed Lyout Librry (~200GB/lyer) LO & EPC Cycle time of LO nd EPC! File uplod speed Blnker rrys WPH Dt Pth 1-3 Blnker rrys WPH Cluster 2 Best cse to be < 6min for individul dt pth Bndwidth vs compression rtio(file size to uplod), decompression speed (computing in rel time) Dt Pth 2-1 >10TB/sec 13,000 chs Blnker rrys WPH Blnker rrys WPH Blnker rrys WPH Dt Pth 2-3 Blnker rrys WPH Dt Pth 2-10 Cluster 3 Dt Pth 2-2 Cost! 22 Blnker rrys WPH Dt Pth 1-2 Dt Pth 1-10 >10TB/sec 13,000 chs Blnker rrys WPH

23 Conclusions MAPPER Pre-Alph tool with 110-bem rster scn hs been stbly opertionl in the mnufcturing Gig-Fb environment. Ptterned bem concept hs been proposed to reduce the required source Br. However, it needs to be demonstrted Using CAR resist, resolution t ~ 30-nm HP hs been chieved t 35nm spot size nd 5keV. However, LWR is yet too high. Contour-bsed EPC for 5keV EBL hs been demonstrted. The 5keV ML2 solution is preferred for lower heting power, strightforwrd EPC, nd less impct to device dmge. Dt hndling methodology for HVM, especilly for foundry, needs ttention. HVM cluster tool for insertion in 2013 is required, however more funds nd industril supports re essentil to crry it out. 23

24 Acknowledgements Thnks to Tem members of ML2 Progrm in TSMC: Dr. Roy Lo, Dr. Rymond Chen, Fruk Krecinic, T.C. Chien Supports from MAPPER Lithogrphy 24 Merlijn Hjenius, Gerrd Brouwer, Murits Weed, Guido de Boer, Bert-Jn Kmpherbeek, nd other people from Delft.