How good will EUV masks need
to be to meet LER requirements?
Patrick Naulleau, Simi George, and
Brittany McClinton
Lawrence Berkeley National Laboratory
1
Outline
Problem
Evidence
Implications
Complications
Summary
2
The Problem
3
Mask sources of LER
Absorber
LER
4
Multilayer with
replicated surface
roughness (RSR)
Imaging demagnifies and filters mask LER
0.32 NA
σ = 0.5
Mask LER PSD
Image LER PSD
40-nm
period
Appl. Opt. 42, 3390-3397 (2003)
5
Imaging transforms replicated surface
(phase) roughness to intensity speckle
230 pm RSR
In focus
defocus
50-nm
σ = 0.5
0.3
Contrast
Contrast==0.9%
6%
9%
See Goldberg et al, Tuesday 12:20PM for
experimental demonstration
6
Experimental
evidence
7
Exposure-to-exposure correlation observed
LER ~ 4.3 nm
Average correlation = 61%
Correlated LER = 3.4 nm
Correlated LER = (Full LER)*sqrt(correlation)
8
Good agreement between
measured correlated LER and
modeled mask-induced LER
Configuration
Measured
correlated
LER (nm)
Modeled
mask-induced
LER (nm)
Mono, F=100 nm
3.4 ± 0.2
3.0
Ann, F=100 nm
2.7 ± 0.3
2.5
Ann, F=0 nm
2.0 ± 0.3
1.4
* Correlated LER = (Full LER)*sqrt(correlation)
Uncertainty based on limited extent of correlation
9
measurement relative to bandwidth
Implications
10
Modeling assumptions
0.32 NA
22-nm HP
Ideal optic
assumed in
all cases
Disk σ = 0.5
0.32 NA
0.42 NA
16-nm HP
From Canon’s
SPIE AL09
presentation
Cross-pole
σout = 0.76
σin = 0.57
11
2/3 Annular
σout = 0.6
Error budget allocation
assumptions
Half pitch (nm)
22
16
Total image plane LWR (nm)1
1.8
1.3
Mask LWR contribution (nm)2
0.7
0.5
Allowable DOF reduction (%)3
30
30
1
8% of CD (from ITRS)
2 10% contribution to total in quadrature
3 Reduction from the NILS = 1 DOF
12
Mask absorber LER coupling
depends on mask LER PSD
Lc == 14.17
24.32 nm
Lc
Roughness
Roughnessexp.
exp.==1.530
0.852
-2
Lens LER cut-off
PSD (nm2/µm-1)
10
-4
10
Using PSD
measured from
a high resolution
EUV mask
0
1
2
10
10
10
Spatial Frequency (lines per unit length) (µm-1)
13
Mask LER magnitude based
on 2008 ITRS
Half pitch (nm)
3σ LWR (nm)
22
2.0
16
1.4
22 nm
16 nm
14
Modeled image plane LWR
resulting from ITRS spec mask LER
15
What if we use expected mask LWR
values?
Half pitch (nm)
3σ LWR (nm)
22
8.0
16
6.0
16
What if we use expected mask LWR
values?
22-nm
target
16-nm
target
17
Multilayer replicated roughness is
generally low frequency
10
PSD (nm4 )
10
10
10
3
Characteristic
feature width
~125 nm
2
1
0
-1
10 -4
10
-3
-2
-1
10
10
10
Spatial Frequency (lines/nm)
18
10
0
Roughness sensitivity @ 22-nm HP
46-pm RMS surface
46-pm RMS surface
roughness requirement if
roughness requirement
mask LER also considered
Litho Parameters
• 0.32 NA
• Disk σ = 0.5
• 22-nm half pitch
• Ideal optic
RMS surface
roughness (pm)
LWR Limits
• Total: 1.8 nm
(8% of CD)
• Mask: 0.7 nm
(10% impact on total)
NILS
DOF Requirement
• 130 nm
(70% of NILS=1 DOF)
Mask LWR limit
19
Mask roughness limits summary
Configuration
RSR limit
(pm)
RSR limit
with mask
LER (pm)
22-nm, 0.32 NA
46
46
16-nm, 0.32 NA
77
77
16-nm, 0.42 NA
77
57
20
Rough Capping
Layer
21
With capping layer roughness, phase shift
is no longer geometric, but refractive
Impact of capping layer
roughness depends on
capping material refractive
index
RSR is geometric effect
∆θ = 4πh/λ
h
22
Sensitivity to capping layer roughness
highly dependent on material and much
lower than RSR
Capping
Material
Si
Double Pass
Roughness
Phase Shift
Equivalent to
per nm of
50 pm RSR*
material
730 nm
0.002°
Ru
6°
0.44 nm
C
2°
1.25 nm
23
Roughness correlation width plays
important role
22-nm, 0.32 NA, disk
LER (nm)
LER (nm) @ -50 nm defocus
2.0
0.7
0.6
Corr. W idth
24 nm
52 nm
76 nm
100 nm
128 nm
200 nm
320 nm
1.5
1.0
0.5
0.5
0.0
-100
0.4
-50
0
50
Focus (nm)
0.3
50pm
RSR
0.2
0.1
0
0
100
200
300
Correlation width (nm)
24
400
100
Summary
• Replicated mask substrate roughness leads to
image plane LER
• Current LER requirements indicate replicated
roughness limits near 50 pm
• Predicted 50-pm RSR limit relies on achieving
stringent absorber LER specs
Acknowledgements
• Tom Pistor, Panoramic: modeling support
• Warren Montgomery, SEMATECH: All printing data
obtained using the SEMATECH MET @ Berkeley
• Paul Denham, Gideon Jones, Brian Hoef, and
Lorie-Mae Baclea-an, LBNL
25