Accelerating the next technology revolution Cost of Ownership Considerations for Maskless Lithography Lloyd C. Litt, SEMATECH Andrea F. Wüest, SEMATECH Copyright 2008 SEMATECH, Inc. SEMATECH, and the SEMATECH logo are registered servicemarks of SEMATECH, Inc. International SEMATECH Manufacturing Initiative, ISMI, Advanced Materials Research Center and AMRC are servicemarks of SEMATECH, Inc. All other servicemarks and trademarks are the property of their respective owners.
Overview Introduction / Motivation COO Modeled Comparisons ML2 Cost Considerations Tool/System requirements for Low COO Conclusions 2
Motivation Leading edge litho cost will increase dramatically for the 32 nm half-pitch (hp) node Miniaturization of devices is continuing at the same pace Identify parameters for cost-effective lithography at 32 nm and hp 3
Overview 1,000 Wafers/Mask 45 nm 32 nm Excludes Excludes data data prep prep and and additional additional inspection inspection costs. costs. 500% 450% 400% 350% 300% Normalized values to 45 nm SE 250% 200% 150% 100% Reticle Clean Etch Metrology Deposition Litho 50% 0% Technology (wph) 45 nm ArFi SE (125) 32 nm LELE (180) 32 nm Freeze (180) 32 nm Spacer (180) 32 nm HI ArFi SE (120) 32 nm EUVL (50) LELE (200) Freeze (200) Spacer (200) HI LELE (135) EUVL (100) ML2 (50) ML2 (100) 4
Overview 5,000 Wafers/Mask 45 nm 32 nm 350% 300% 250% 200% 150% 100% 50% Reticle Clean Etch Metrology Deposition Litho 0% Technology (wph) 45 nm ArFi SE (125) 32 nm LELE (180) 32 nm Freeze (180) 32 nm Spacer (180) 32 nm HI ArFi SE (120) 32 nm EUVL (50) LELE (200) Freeze (200) Spacer (200) HI LELE (135) EUVL (100) ML2 (50) ML2 (100) 5
Results - hp (1,000 w/m) 45 nm 500% 450% 400% 350% 300% 250% 200% Reticle Reticle cost cost large large for for low low WPM WPM (70 90 (70 90 %) %) 150% Reticle Clean Etch Metrology Deposition Litho 100% 50% 0% 45 nm ArFi SE (125 wph) LELE (200 wph) Freeze (200 wph) Spacer (200 wph) HI LELE (135 wph) EUVL (100 wph) ML2 cluster (50 wph) ML2 cluster (100 wph) Technology (wph) 6
Results - hp (5,000 w/m) 45 nm 350% 300% 250% Defect inspection cost is a variable for ML2 200% 150% 100% Reticle Clean Etch Metrology Deposition Litho 50% Technology (wph) 0% 45 nm ArFi SE (125 wph) LELE (200 wph) Freeze (200 wph) Spacer (200 wph) HI LELE (135 wph) EUVL (100 wph) ML2 cluster (50 wph) ML2 cluster (100 wph) 7
ML2 Cost of Ownership Because some of the cost reduction is due to mask cost savings, the attractiveness of ML2 technology is highly dependant on the type and volume of production under consideration. A factory with a diverse mix of low volume products will potentially see a significant reduction in mask costs if ML2 is used. However, very large volume producers of a small number of parts (mainstream microprocessors, SRAM, DRAM, etc.) may not realize a large enough savings to consider ML2. 8 SEMATECH CONFIDENTIAL L. Litt
ML2 Cost of Ownership There are potential applications of ML2 within a large volume manufacturing operation that have some real value. Prototyping and design verification Reduce initial device product learning Reduce mask respin orders to correct design errors 9 SEMATECH CONFIDENTIAL L. Litt
ML2 Cost of Ownership Most ML2 cost-of-ownership assessments center on the potential savings of removing mask/reticle costs from the cost equation. Mask costs can be considerable and these savings are real But other aspects of the technology may quickly reduce these savings 10 SEMATECH CONFIDENTIAL L. Litt
Process COO Implications of ML2 *Relative costs approximated by size scale of boxes Prep Write Process RETICLE Inspect Expose Process Inspect WAFER 11
Mask Fabrication Cost Much of the cost of mask fabrication is from mask inspection. A significant portion of this cost category is likely to be transferred to the wafer fab in an ML2 implementation. This reduces the savings from removing the mask from the ML2 COO assessment. 12
ML2 COO Considerations Transfer of costs from mask fab to wafer fab Mask-based lithography relies on known-good masks. Each mask is inspected and verified for defects, CD, IP, etc. ML2 will not have known-good starting material. Additional inspection is needed in a wafer fab unless high confidence in writing quality. 13
Defect Inspection Sensitivity Analysis 5,000 Wafers per Mask Normalized cost per wafer (to 45 nm hp ArFi SE) 350% 325% 300% 275% 250% 225% 200% 175% 150% 125% 100% 75% 50% 25% 0% LELE (200 wph) EUV (100 wph) 45 nm ArFi SE (125 wph) 0% 20% 40% 60% 80% 100% Inspected wafer area Cross over for ML2 cluster (50 wph) with EUV NO cross over for ML2 cluster (100 wph) with EUV) 5,000 wpm 2 wafer/lot, 50% area Every 5 th th lot 2 wafer/lot, Every 5 th th lot 2 wafer/lot, Every lot 10 wafer/lot, Every lot 20 wafer/lot, Every lot 25 wafer/lot, Every lot 14
Defect Inspection Sensitivity Analysis 275 250 225 200 Effective defect inspection throughput Throughput (wph) 175 150 125 100 75 50 25 1 inspection tool supports 50 wph ML2 cluster Multiple Inspection tools required 10 inspection tools required (50 wph) ML2 cluster 100 wph ML2 cluster 50 wph 0 0% 5% 10% 15% 20% 25% 30% 35% 40% Inspected wafer area 2 wafer/lot, 50% area Every 5 th th lot 2 wafer/lot, Every 2 nd nd lot 2 wafer/lot, Every lot 5 wafer/lot, Every lot 10wafer/lot, Every lot 0.2 inspection tools required to meet 50 wph 15
ML2 COO Considerations Error rate Mask writer error rate is low but ML2 includes??,000x beams writing??,000x wafers(area) with addition of cross talk interactions. 1 mask writer = 150x150mm * 3/day = 67.5K mm 2 /day 1 wafer writer = 3.14*150 2 * 10wph * 24 = 17000K mm 2 /day 250X more area written per day. 16
Technical Requirements for Low COO Obvious Requirements Low capital cost, footprint, operating costs High throughput Other Issues Consumables Component reliability and costs Built-in quality checks to minimize inspection In situ beam and data transfer verification ability to verify data from memory actually written at tool. (proportional to inspection requirements) 17
Conclusions 100 wph cluster is approximately equivalent to conventional litho costs not including mask. Mask cost transfer Inspection costs cannot be neglected and may lead to additional inspection capacity required and thus less potential savings. Data prep can reduce effective throughput. Inspection capability needed is not currently available. Tool design must address key issues to realize low COO. 18
Acknowledgements Greg Hughes, Frank Goodwin, Dennis Fandel, Jackie Ferrelll, Jacque Georger, Chawon Koh, Bob Rulliffson, Phil Seidel, Larry Smith, Robert Wright (SEMATECH) Andrew J. Hazelton (Nikon), Will Conley (Freescale), Rob Crowell (TEL),Hiroyuki Mizuno (Toshiba), Nick Stacey (Molecular Imprints), Obert Wood (AMD) Céline Lapeyre (CEA-LETI Minatec), Gary Zhang (Rohm and Haas) 19
Backup Slides 20
Defect Inspection Sensitivity Analysis 1,000 Wafers per Mask 500% Normalized cost per wafer (to 45 nm hp ArFi SE) 450% 400% 350% 300% 250% 200% 150% 100% 50% 0% 45 nm ArFi SE (125 wph) 0% 20% 40% 60% 80% 100% Inspected wafer area LELE (200 wph) EUV (100 wph) 1,000 wpm NO cross over for ML2 cluster (50 wph) NO cross over for ML2 cluster (100 wph) 2 wafer/lot, 50% area Every 5 th th lot 2 wafer/lot, Every 5 th th lot 2 wafer/lot, Every lot 10 wafer/lot, Every lot 20 wafer/lot, Every lot 25 wafer/lot, Every lot 21
Candidate Technologies hp : LELE, Freeze, Spacer High-index : LELE EUVL ML2 SE: Single Exposure, DPL: Double Patterning LELE: Litho-Etch-Litho-Etch 22
Process Flows (schematic) Deposit hardmask(s) Coat, expose, develop ArFi SE Etch hardmask, Strip resist Coat, expose, develop DPL LELE (Line) Freeze Freeze resist Coat, expose, develop Etch hardmask Deposit spacer, Etch back spacer Remove hardmask lines Coat, expose, develop (cut mask) Spacer EUV ML2 e-beam write For all flows at end: Etch hardmask, Strip resist, Etch pattern, Strip hardmask 23
Assumptions All technologies are equally reliable and support equal yield. All technologies meet manufacturing requirements. Double Patterning Stepper overlay Mask registration Mask yield (with 30 hour write time) Yield even though with more processing steps EUVL Defect-free masks Source power, tool transmission, and resist sensitivity enable throughput Tool reliability supports uptime Mask and optics meet lifetime requirements Maskless??? 24
Calculation Procedure $ / yr General ( C fixed + Crecur ) COO = + Cmaterials + C T U Y 24 365 other $ / wafer wafer / h h / yr $ / wafer $ / wafer C fixed Depreciation, Floor space ($/yr) C materials Resist, etc. ($/wafer) C recur Utilities, Consumables, Labor ($/yr) C other Other ($/wafer) T Throughput (wafer / h) U Utilization (%) Y Yield (%) 25
Calculation Procedure Linear relationship Lithography ( C fixed + Crecur ) COO = + Cresist + T U Y 24 365 C N reticle wpr Inversely proportional: large effect C fixed Depreciation, Floor space ($/yr) C resist Resist, etc. ($/wafer) C recur T Utilities, Consumables, Labor ($/yr) Throughput (wafer / h) C reticle N wpr Reticle ($) Wafer / reticle U Y Utilization (%) Yield (%) 26
Calculation Procedure Calculate COO for each process step Litho Deposition Etch Metrology Clean Process flow # of different process steps Total COO = Σ (Process Cost #Process Steps) Only one critical layer calculated Normalized to 45 nm ArFi SE 27
Model Parameters 45 nm hp 32 nm hp ArFi SE HI ArFi EUVL LELE Freeze Spacer Tool Cost $40M $49M $49M $49M $50M $54M Throughput / wph 125 180 180 180 120 50 Tool Cost / TPT (M$/wph) 0.3 0.3 0.3 0.3 0.4 1.1 Reticle Cost $200k $584k $584k $466K $396k $178K 45 nm hp hp ArFi SE HI EUVL ML2 LELE Freeze Spacer LELE (cluster) Tool Cost $40M $52M $52M $52M $53M $89M $50M $50M Throughput / wph 125 200 200 200 135 100 50 100 Tool Cost / TPT (M$/wph) 0.3 0.3 0.3 0.3 0.4 0.9 1.0 0.5 Reticle Cost $200k $1176K $1176K $752K $1176K $252K N/A N/A Tool cost based on historical extrapolation, scales with throughput and resolution ML2 tool cost independent of throughput 28
Model Parameters Reticle cost (SEMATECH model, G. Hughes) Mask Cost = [ (Capital Cost term Write Time) + Material Cost ] / Yield 2.5 data growth per node for optical 2 data growth for EUV and Imprint Mask yield based on ITRS difficulty 45 nm: 70% EUVL: 77% (Yield targets looser) DPL: 63% (Mask registration tighter) Imprint: 54% (Defect specs much tighter, minimum feature 1 ). Utilization fixed at 83%, yield at 98% for all technologies Non-litho process costs based on SEMATECH data ~ 250 total parameters (~ 25 per litho technology) 29
Error Estimation (20,000 Wafers/Mask) 300% 300% 250% 250% 200% 200% 10 % error on parameters ~15 % error on COO (U and Y kept constant) 150% 150% 100% 100% 50% 50% 0% 45 nm ArFi SE (125) 32 nm LELE (180) 32 nm Freeze (180) 32 nm Spacer (180) 32 nm HI ArFi SE (120) 32 nm EUVL (50) LELE (200) Freeze (200) Spacer (200) HI LELE (135) EUVL (100) 0% 30
Results - 32 nm hp (20,000 w/m) 45 nm 32 nm 200% 200% DPL most expensive because of reticle, 160% deposition, and etch costs 140% DPL most expensive 180% 120% 120% 100% 100% 80% 80% 60% 60% Reticle Clean Etch Reticle Clean Etch Metrology Deposition Litho Metrology Deposition Litho 40% 40% 20% 20% 0% Technology (wph) 0% 45 nm ArFi SE (125 wph) ArFi SE (125) 32 nm LELE (180 wph) (180) 32 nm Freeze (180 wph) Freeze (180) 32 nm Spacer (180 wph) Spacer (180) 32 nm HI ArFi SE (120 wph) HI ArFi SE (120) 32 nm EUVL (50 wph) EUVL (50) High EUV litho cost because of capital cost 31
Results - hp (20,000 w/m) 45 nm 250% 200% High reticle costs for DPL Cost advantage of EUVL 150% 150% 100% 100% Reticle Clean Etch Metrology Metrology Deposition Deposition Litho Litho 50% 50% 0% 0% Technology (wph) 45 nm ArFi SE (125 wph) ArFi SE (125) LELE (200 wph) LELE (200) Freeze (200 wph) Freeze (200) Spacer (200 wph) Spacer (200) HI HI ArFi LELE (135 wph) DPL LELE (135) EUVL (100 wph) EUVL (100) 32