strategy and feasibility of defect-free mask fabrication...
TRANSCRIPT
2009 EUV Symposium, Prague G-Number11
Strategy and Feasibility of
Defect-free Mask Fabrication to Enable EUVL
Ted Liang, Guojing Zhang, Sunny Son, Robert Chen, Sang Lee, Michael Leeson, Pei-yang Yan, Andy Ma, Long He,
Gilroy Vandentop
Firoz Ghadiali, Bennett Olson, Xiangyang Liu, Cheng-hsin Ma. Eric Lanzendorf, Barry Lieberman
Intel Corporation
Components ResearchTechnology and Manufacturing Engineering
Intel Mask Operations
2009 EUV Symposium, Prague G-Number22
OutlineOutline
Defect TypesDefect Types
Quantitative defect analysis with Quantitative defect analysis with existing tool setexisting tool set
Achieving zero defectAchieving zero defect
SummarySummary
2009 EUV Symposium, Prague G-Number33
EUV Mask Defect TypesEUV Mask Defect TypesDefects are more complex than those on 193nm mask
Cross-sectional view
Absorber (TaN, 80nm)
Ru cap (2.5nm thick)
Mo-Si ML (40-50 pairs, 280nm)
Substrate (LTEM, ¼”)
Conductive film (CrN, 70nm)
λ=13.5nm Schematic showing different types of defects
EUV mask defects are complexDifficult to differentiate and disseminate by wafer printNecessary to measure directly on the mask
2009 EUV Symposium, Prague G-Number44
Mask Defect ChallengesMask Defect ChallengesAchieving mask yieldAchieving mask yield
–– Detect every single defect above certain impact (Detect every single defect above certain impact (e.ge.g 10%10%ΔΔCD/CD)CD/CD)–– Go after every defectGo after every defect
Quantitative defect analysis is necessary to develop best Quantitative defect analysis is necessary to develop best known method (process, tool, procedure) to enable mask known method (process, tool, procedure) to enable mask yieldyield
–– Defect count, size, sources and dispositionDefect count, size, sources and disposition–– Only way to enable effective/efficient defect reductionOnly way to enable effective/efficient defect reduction
Challenges in fast yield learning for EUV maskChallenges in fast yield learning for EUV mask–– Blank defects: high count, majority with unknown propertyBlank defects: high count, majority with unknown property–– No tool at mask shop to measure phase or amplitude, materials coNo tool at mask shop to measure phase or amplitude, materials compositionmposition–– Insufficient Insufficient S/HS/H infrastructure to keep mask clean for printinginfrastructure to keep mask clean for printing
We must progress with what we have today in lieu of the We must progress with what we have today in lieu of the tool limitationstool limitations
–– Can we achieve zero defect to the limit of current tool capabiliCan we achieve zero defect to the limit of current tool capabilities?ties?
2009 EUV Symposium, Prague G-Number55
Blank ML Defect TrendBlank ML Defect Trend
~100 defects @50nm+~100 defects @50nm+Best blank ~10 defects @70nm+Best blank ~10 defects @70nm+
Size distribution
Need ~100X defect reduction to reach Need ~100X defect reduction to reach current tool limitcurrent tool limit
0
50
100
150
200
250
300
350
400
45 50 55 60 65 70 75SiO2 equivalent size (nm)
# of
Def
ects
at >
= gi
ven
size
LTEM/ML received in Q1-Q2with M7360 in spection
Total defect count
1.E+00
1.E+01
1.E+02
1.E+03
1.E+04
Jan-06
Jan-07
Jan-08
Jan-09
Jan-10
Jan-11
Jan-12
Def
ect c
ount
s
80nm defect size50nm defect sizeRM-for blank need by 2012
80nm50nm 45nm
35nm25nm2G tool1G inspection
tool
3G tool
M1350 M7360
2009 EUV Symposium, Prague G-Number66
Strategy of Defect StudyStrategy of Defect Study
Utilize full field mask as test vehicle: Utilize full field mask as test vehicle: –– ML blank: full plate inspection to map out defectsML blank: full plate inspection to map out defects
–– Pattern mask: clear field to sensitize defects and maximize detePattern mask: clear field to sensitize defects and maximize detectionction
Tools:Tools:–– Optimize current tool capability and develop Optimize current tool capability and develop BKMBKM
–– Identify critical needs for Identify critical needs for HVMHVM: Tools, process, procedure: Tools, process, procedure
Measurements Measurements –– inspection and characterization inspection and characterization metrology: metrology:
–– Focus on direct measurements on mask (mask qualification is doneFocus on direct measurements on mask (mask qualification is done at the at the mask shop, not at the wafer mask shop, not at the wafer fabfab))
–– Verification with wafer printVerification with wafer print
2009 EUV Symposium, Prague G-Number77
Mask Defect Disposition Flow at Mask Defect Disposition Flow at Mask ShopMask Shop
POR for 193nm maskPOR for 193nm mask–– All defects must pass AIMS in order to be All defects must pass AIMS in order to be dispositioneddispositioned
PatternInspection
Repair qual byAIMS
MaskRepair
Disposition by Simulation and AIMS
PatternInspection
As a workAs a work--around without EUV AIMS, use physical tools around without EUV AIMS, use physical tools ––SEM, AFM, blank and pattern defect overlaySEM, AFM, blank and pattern defect overlay
EUV mask, currently no AIMSEUV mask, currently no AIMS
PatternInspection*
Mask Repair;Qual by SEM/AFM
CharacterizationBy SEM, AFM
PatternInspection
ML blank defect tracing/separation
2009 EUV Symposium, Prague G-Number88
Pattern Inspection and Defect SizingPattern Inspection and Defect Sizing● Defect sizing with SEM vs. pattern inspection tool
– Inspection tool does not size defects
● SEM and AFM are sufficient for pattern defect disposition with aid from printability modeling
2009 EUV Symposium, Prague G-Number99
● Defect from pattern and blank – 19 total defects from pattern inspection– 16 matched to ML blank defects– 3 are real pattern defects
Automatic Defect Overlay for Defect ClassificationAutomatic Defect Overlay for Defect Classification
Defect size and height
● Pattern inspection detects most large blank defects (embedded particles)
Pattern inspection19 defects
ML blank inspection ~100 defects
16 defects traced to ML blank 3 real absorber defects
50
100
150
200
250
300
350
400
3 5 7 9 11 13 15 17 19 21 23
Inspection pixels
SEM
siz
ing
(nm
)
10
45
80
115
150
185
220
255
AFM
Ht (
nm)
SizeHeight
50
100
150
200
250
300
350
400
3 5 7 9 11 13 15 17 19 21 23
Inspection pixels
SEM
siz
ing
(nm
)
10
45
80
115
150
185
220
255
AFM
Ht (
nm)
SizeHeight
2009 EUV Symposium, Prague G-Number1010
Assessing Potential Impact to PrintAssessing Potential Impact to Print● Total of 19 total defects at pattern inspection
● Impact depends on location – Must eliminate every large defect
3 absorber defects are all repairable
9 of 16 ML blank defects are printable
7 of 16 blank defects (covered by absorber)
2009 EUV Symposium, Prague G-Number1111
Total Pattern Defect ReductionTotal Pattern Defect Reduction
● >4X reduction in total defects >40nm● Best achieved: 8 total defect >40nm
Jan’09
1st pass defect reduction trend
Lot/time sequence
Def
ect c
ount
1
10
100
1000
Dark field test maskClear field test mask
Total count >40nmTotal count >70nm
1st pass defect reduction trend
Lot/time sequence
Def
ect c
ount
1
10
100
1000
Clear field maskDark field mask
Jan’09
1st pass defect reduction trend
Lot/time sequence
Def
ect c
ount
1
10
100
1000
Dark field test maskClear field test mask
Total count >40nmTotal count >70nmTotal count >40nmTotal count >70nm
1st pass defect reduction trend
Lot/time sequence
Def
ect c
ount
1
10
100
1000
Clear field maskDark field mask
2009 EUV Symposium, Prague G-Number1212
Achieving Zero Defect Achieving Zero Defect –– Pattern RepairPattern RepairExample 1: Defect caused 10% Example 1: Defect caused 10% ΔΔCDCD/CD/CDFully repaired as by Fully repaired as by ADTADT printprintSEM + AFM sufficient for SEM + AFM sufficient for patternpattern repair repair validationvalidation
2009 EUV Symposium, Prague G-Number1313
Achieving Zero Defect Achieving Zero Defect –– Pattern RepairPattern RepairExample 2: Defect caused 15% Example 2: Defect caused 15% ΔΔCDCD/CD/CDFully repaired as by Fully repaired as by ADTADT printprint
2009 EUV Symposium, Prague G-Number1414
Repairing ML BlankRepairing ML Blank--caused Defectcaused DefectComplex defectComplex defect
Defect caused by large CDefect caused by large C--containing containing particulate on ML blank resulting in thin particulate on ML blank resulting in thin absorberabsorber
SEM + AFM are used to guide repairSEM + AFM are used to guide repair–– Defect natureDefect nature–– Repair by depositionRepair by deposition
2009 EUV Symposium, Prague G-Number1515
120nm tall particle embedded in ML blank; Caused bridging120nm tall particle embedded in ML blank; Caused bridgingSEM + AFM not sufficient to guide repairSEM + AFM not sufficient to guide repairNeed composition to dial in repair etch chemistryNeed composition to dial in repair etch chemistry
Repairing Embedded ML Blank DefectRepairing Embedded ML Blank Defect
Resist print improvedResist print improved–– Pass/fail depends on device layerPass/fail depends on device layer
Compensation is needed for Compensation is needed for incomplete defect removalincomplete defect removalML blank must be free of ML blank must be free of large embedded particleslarge embedded particlesAIMS is a mustAIMS is a must--have to have to monitor and qualify repairmonitor and qualify repair
2009 EUV Symposium, Prague G-Number1616
Repair Qualification DiscussionRepair Qualification DiscussionNonNon--pure absorber defect, caused pure absorber defect, caused bridgingbridgingIs repair good enough?Is repair good enough?
SEM alone not sufficientSEM alone not sufficient
AFM difficult to measure AFM difficult to measure at the edgeat the edge
AIMS is needed to AIMS is needed to provide immediate provide immediate feedback to ensure feedback to ensure successful repair successful repair
2009 EUV Symposium, Prague G-Number1717
ZeroZero--defect Handling and Storage defect Handling and Storage Zero particle adders can be achieved with RSP, but not always zeroStandard EUV reticle carrier sPodwith inner pod to further prevent particlesNeed to build and test infrastructure around ADT to facilitate learning
– Including cleaning tool
ReticleFacing down
Transfer between buildings
2009 EUV Symposium, Prague G-Number1818
0.0E+00
5.0E+03
1.0E+04
1.5E+04
2.0E+04
2.5E+04
3.0E+04
0 10 20 30 40 50 60
Height (nm)
Are
a (n
m^2
)
Small 3D-like defectsMetal oxides (Ti, Cr, Ta)
Thin flat 2D-like defects (C-containing)
50nm eq-Dia.
80nm eq-Dia.
Challenges in Mask CleaningChallenges in Mask Cleaning
-505
101520
0 0.5 1(μm)
Hei
ght (
nm)
Defect #24, pixel 7
-505
101520
0 0.5 1(μm)
Hei
ght (
nm)
Defect #222, pixel 12
Frequent cleaning poses additional challenge to mask lifetime– Avoid Ru-ML damage
Biggest challenge: avoiding adders
Common adders are 2D-like carbonaceous and 3D-like oxidesKey focus: Particle filtration and removal efficiency
2009 BACUS paper
2009 EUV Symposium, Prague G-Number1919
SummarySummaryEUV mask defects are complex in types and root causesEUV mask defects are complex in types and root causes
–– Direct measurements is necessary for defect reduction and mask qDirect measurements is necessary for defect reduction and mask qualificationualification
We have developed best known methods for We have developed best known methods for quantitativequantitative defect defect understanding and reduction by optimal use of current tool setunderstanding and reduction by optimal use of current tool set
We have demonstrated full loop device pattern repair as verifiedWe have demonstrated full loop device pattern repair as verifiedwith with ADTADT printprint
–– SEM + AFM capable to guide repair for majority of defects, but nSEM + AFM capable to guide repair for majority of defects, but not sufficientot sufficient
–– AIMS is in critical need, the first priority tool for defect AIMS is in critical need, the first priority tool for defect dispositiondisposition
–– Repair + compensation is necessary for ML blank defectsRepair + compensation is necessary for ML blank defects
ML blank defects must be reduced to <<10 and all need to be ML blank defects must be reduced to <<10 and all need to be smaller than primary mask patterns to smaller than primary mask patterns to be be compensatablecompensatable
–– Current inspection tool is deemed capable to support this effortCurrent inspection tool is deemed capable to support this effort for blank suppliersfor blank suppliers
–– Highest risk for ML blanks will likely be large embedded particuHighest risk for ML blanks will likely be large embedded particulate defects vs. small late defects vs. small phase defectsphase defects
2009 EUV Symposium, Prague G-Number2020
AcknowledgementsAcknowledgements
Intel Components Research – Armando Cobarrubia, John Magana, Seh-Jin Park
Intel Mask Operations–– Nathan Wilcox, Andrew Nathan Wilcox, Andrew JamielsonJamielson, Chang, Chang--Ju Choi, Ju Choi,
YiYi--Ping LiuPing Liu
Carl Zeiss for supporting repair test– Thorsten Hofmann, Klaus Edinger, Kinga Kornilov