dave guest lecture - people @ eecs at uc berkeley · λ >= 193 nm λ = 13.5 nm multilayer...
TRANSCRIPT
1
1© Patrick Naulleau, 2008
EUV Lithography
Patrick Naulleau
Center for X-ray Optics, Lawrence Berkeley National Laboratory
2© Patrick Naulleau, 2008
Outline
• Introduction
• Review of key optics concepts for lithography
• EUV Capabilities/Challenges
• EUV patterning and resists
• Mask defects and printing
2
3© Patrick Naulleau, 2008
EUV: extension of optical lithography
Wafer
Mask
λ
Mask
λ
Wafer
PhotoresistPhotoresist
λ >= 193 nmλ = 13.5 nm
Multilayer
coatings used
to render
surfaces
reflective at
EUV
4© Patrick Naulleau, 2008
The mask serves as the circuit master in lithography systems
wafer
mask
Projection optics
3
5© Patrick Naulleau, 2008
Review of key optics concepts
for lithography
• Diffraction
• Resolution
• Depth of focus
• Partial coherence
• Etendue
6© Patrick Naulleau, 2008
Diffraction
W WW
λ = 13.5 nm
W = 300 nm
z =
0.2 µm
z =
1 µm
z =
5 µm
z =
25 µm
Angular spread, θ = asin(λ/W)
4
7© Patrick Naulleau, 2008
Image formation: inverse diffraction
• Function of lens is to invert diffraction process
• θ = asin(λ/W) => Wmin = λ/sin(θmax)
θmax
Resolution ∝ Wmin = λ/sin(θmax) = λ/NA
Res = k1 λ/NA
NA = sin(θmax)
8© Patrick Naulleau, 2008
Depth of focus
• Let DOF be proportional to distance where
single-sided blur = resolution
θmax
dmax NA = λ/NA
DOF ∝ dmax = λ/NA2
DOF = k2 λ/NA2
Blur = d NA
d
5
9© Patrick Naulleau, 2008
Concept of partial coherence factor (σσσσ)
• Most fundamentally, σ is the ratio of mask-side
resolution limit to coherence width (Wc)
Wc
• Large Wc = small σ
• Determines area over which object components
add coherently
10© Patrick Naulleau, 2008
Critical illumination: σσσσ = ratio of NAs
NApo
Projection
optic pupilCondenser
optic pupil
Projection
optic object
plane
NAci
Wc ∝ 1/NAci Reso ∝ 1/NApo
σ = Reso/Wc = NAci/NApo
6
11© Patrick Naulleau, 2008
Pupil fill: alternative view of σσσσ
σ = NAci/NApo = rs/rL
NApo
Projection
optic pupilCondenser
optic pupil
Projection
optic object
plane
NAci
rsrL
Example view of pupil fill for σ = 0.35
Projection lens pupil
12© Patrick Naulleau, 2008
Etendue: space-bandwidth product
NA=0.25
NA=0.0625Field size = 6x96 mm2
• The optic etandue (maximum accepted space-
bandwidth product) is
• Field of view area X acceptance solid angle
• 6*96*0.01 mm2 sr ~ 7 mm2 sr
Projection optic
7
13© Patrick Naulleau, 2008
NA=0.7Source size = 1 mm2
• A 1 mm source with 45-degree collection would
have an etendue of ~2 mm^2 sr (min σ = 0.53)
• A 0.1 mm source with 70-degree collection would
have an etendue of ~0.04 mm^2 sr (min σ = 0.08)
Projection optics etendue limits usable
source etendue and thus power
Collector optic
14© Patrick Naulleau, 2008
6 mirrors8 mirrors
NA = 0.1
4 mirrors
Increasing NA demands more mirrors
• More mirrors means
lower throughput
• But, larger NA means
larger etendue
8
15© Patrick Naulleau, 2008
EUV Patterning Capabilities: modeling
16© Patrick Naulleau, 2008
Extendibility of EUVBinary amplitude mask, σ = 0.7, no OPC, no bias correction
9
17© Patrick Naulleau, 2008
Optical transfer function vs NA
Binary amplitude mask, σ = 0.5, no aberrations
18© Patrick Naulleau, 2008
Optical transfer function through sigma
Binary amplitude mask, NA = 0.32, no aberrations
10
19© Patrick Naulleau, 2008
Optical transfer function through illumination type
Binary amplitude mask, NA = 0.32, no aberrations
20© Patrick Naulleau, 2008
8-nm possible at 0.5 NA with
conventional mask
8-nm elbow pattern
Binary amplitude mask, dipole illumination, no OPC, no bias correction
Pole radius = 0.2Pole offset = 0.8
11
21© Patrick Naulleau, 2008
EUV-specific challenges
Long-range “proximity” correction required
Short wavelength -> high scatter -> large position dependent flare
Position dependent mask bias correction
required
Tilted mask plane: shadowing by 3D mask structure
Strict reticle flatness requirementsTilted mask plane: system not telecentric at reticle
Use electrostatic clamping, mag. Lev.
stages
Vacuum clamping cannot be used. Air-bearing motion
mechanisms are complex
Reflective mask leads to potential “buried” phase
defects (< 3-nm tall)
EUV sources are inefficient producers of radiation
Source chamber cannot be physically separated from
imaging optics chamber
Hydrocarbons & water vapour are cracked by EUV,
contaminating mirror surfaces – C deposition &
oxidation of coatings
All solid materials strongly absorb EUV radiation
EUV radiation is not transmitted through the atmosphere
Challenge
Extremely high sensitivity mask blank
inspection required
Efficient thermal management of waste
heat from high input powers required
Contain any debris produced by source –
particles & ions
Minimize hydrocarbons and water vapour
content in the tool. Needs cleanliness of
ultra high vacuum (UHV)
Refractive optics not possible. Use only
reflective mirrors & reticles
Tool must operate in vacuum environment
Consequence
22© Patrick Naulleau, 2008
Top 3 Critical Issues
1. Reliable high-power source and collector module
• Largely industrial effort
2. Availability of defect-free masks and mask infrastructure
3. Resist resolution, sensitivity & Line Edge Roughness (LER) met simultaneously
Ref: 2008 International EUVL Steering Committee
EUVL critical issue list
12
23© Patrick Naulleau, 2008
• Numerical aperture = 0.3
• Spatial resolution = 12 nm*
• Supports approximately 200 user shifts per year with nearly 100% uptime
• Unique programmable coherence illuminator enables world’s finest projection EUV resolution
• Supports:
• Resist development
• Mask development
• Mask defect studies
ALS Undulator BL12.0.1.3
Coherence control module
Mask stage
2-mirror projection optics
Wafer stage and height sensor
Coherence monitor
Berkeley EUV Nano-patterning tool designed to support advanced EUV research
24© Patrick Naulleau, 2008
Worldwide user base including industry, academia, and research institutes
13
25© Patrick Naulleau, 2008
Simultaneously meeting resolution, sensitivity, and LER crucial issue for EUV resists
10 mJ/cm2
10 mJ/cm210 mJ/cm2
1.2 nm0.8 nm0.6 nm
32-nm half pitch (21-nm iso) - 2013*22-nm half pitch (15-nm iso) -201616-nm half pitch (11-nm iso) -2019
* 2007 ITRS RoadmapLER: Line Edge Roughness
26© Patrick Naulleau, 2008
Although most EUV resists are based on 193 or 248 nm systems, EUV interaction with resist these materials is fundamentally different
• EUV energy (92 eV) many times higher than Photo Acid Generator (PAG) activation energy
• EUV interacts with all atoms, cannot be made to preferentially interact with PAG
• Photons do not directly activate PAG but rather generate secondary electrons upon interaction with first encountered atom
–Secondary electrons eventually activate PAG
14
27© Patrick Naulleau, 2008
Although most EUV resists are based on 193 or 248 nm systems, EUV interaction with resist these materials is fundamentally different
H
O
C
Photo Acid Generator (PAG)
Visible photons, only enough energy to interact with PAG
hνVis
hνEUVe-
28© Patrick Naulleau, 2008
Chemically amplified resists now approaching 20 nm
22 nm HP 20 nm HP
Resist C
12.7 mJ/cm2
50-nm resist thickness
24 nm HP
Resist D
15.2 mJ/cm2
15
29© Patrick Naulleau, 2008
High pattern fidelity at small feature sizes
22 nm HP 20 nm HP24 nm HP
30 nm 1:1 contacts
Resist D, film thickness = 50 nm
30© Patrick Naulleau, 2008
Rinse agents for LER reduction without resolution or sensitivity trade-offs
Data courtesy of Tom Wallow, AMD
Baseline Process Rinse Agent Process
CD = 41.7 +/- 0.8 nm
LER = 4.3 +/- 0.4 nmCD = 40.6 +/- 0.6 nm
LER = 3.2 +/- 0.4 nm
• Rinse agent applied instead of water after development
• ~1-nm reduction in LER observed
• No effect on resolution or sensitivity
XP5494-C resist,
Y-Monopole
Berkeley METBerkeley MET
16
31© Patrick Naulleau, 2008
Repeated printing of 35 nm contacts shows
variation NOT dominated by photon noise
RHEM ResistBerkeley METAnnularE0 = 10mJ/cm2
focus
+50 nm
Esize +15%
• 35-nm 1:2 contacts
• RMS size variation = 3.2 nm
• Reproducible size variation through dose and focus
• Contact variation must be dominated by mask
32© Patrick Naulleau, 2008
Optical mask error enhancement factor (MEEF) does not explain observed contact variation
Mask Ideal Resist
MEEF = 12.8 nm error on 35 nm contacts
(wafer coordinates)Aerial-image
modeling includes full EUV wavefront
Actual Mask
• 35 nm 1:1 contacts on 5x EUV mask
• RMS 1x diameter variation = 1.1 nm
• Resist var. = 1.1 nm
17
33© Patrick Naulleau, 2008
Modeling Resist Using Simple
Point-Spread-Function (PSF) Method
“Deprotection blur” function PSF
* C. Ahn, H. Kim, K. Baik, “A novel approximate model for resist process,” Proc. SPIE 3334, (1998).
** Gregg Gallatin, “Resist Blur and Line Edge Roughness,” Proc. SPIE 5754, (2005)
• PSF resist modeling* is fast and convenient
• Model easily generated
• Provides intuitive link to resist resolution limit
• Few parameters makes model less susceptible to extrapolation errors
• Resist process well approximated by deprotection function**
34© Patrick Naulleau, 2008
Resist blur dominates MEEF
Mask 20-nm Blur Resist
MEEF = 3.62.8 nm error on 35 nm contacts
(wafer coordinates)Aerial-image
modeling includes full EUV wavefront
Actual Mask
• 35 nm 1:1 contacts on 5x EUV mask
• RMS 1x diameter variation = 1.1 nm
• Resist var. = 4.0 nm
18
35© Patrick Naulleau, 2008
Resist blur dominates MEEF
0 50 100 150 200 250 300 350 400 4500
0.2
0.4
0.6
0.8
1
Image Position (nm)
No
rma
lize
d D
ep
rote
ctio
nAerial image aloneResist-blurred deprotection image
36© Patrick Naulleau, 2008
The mask serves as the circuit master in lithography systems
wafer
mask
Projection optics
19
37© Patrick Naulleau, 2008
Pattern defects on become replicated problems on
the wafer, although attenuated by optic and resist
Data courtesy of Ted Liang, Intel
Details published at Photomask Japan, 2006
Mask: 60-nm defect Resist: 9% CD Change
38© Patrick Naulleau, 2008
Buried (phase) defects become intensity after band-limited imaging and further enhanced by defocus
Modeled aerial image of
programmed 100-nm isolated
defects through focus (100-nm
steps)
100-nm defects not expected
to be printable at Esize)
Modeled aerial image of
programmed 100-nm isolated
defects through focus (100-nm
steps)
100-nm defects not expected
to be printable at Esize)
Modeled defect
at surface:
parameters
based on latest
LLNL process
Modeled defect
at surface:
parameters
based on latest
LLNL process
Focus-100 nm +100 nm-200 nm-300 nm-400 nm
-600 nm-700 nm -500 nm-800 nm-900 nm-1000 nm
+600 nm+500 nm +700 nm+400 nm+300nm+200 nm
+800 nm
20
39© Patrick Naulleau, 2008
Programmed “buried” defects developed
to study critical mask issues
buried defectsTop view of
programmed defect cell
absorber
pattern
substrate
resulting phase defect
reflective multilayer
absorber pattern
60-nm defects
70-nm defects
buried defect
40© Patrick Naulleau, 2008
off-axis ZP (µµµµscope objective)
CC
D
mask
Illumination window
The world’s highest-performance EUV microscope dedicated to photomask research
• International EUV program’s primary tool for at wavelength mask defect inspection and cross correlation with complimentary techniques
• Spatial resolution approaching 90 nm at the mask (23 at wafer)
• Working with partners regarding future upgrades. Special attention to supporting the development of commercial tools.
NA = real-time selectable
0.25–0.35 (4x)
Res. ≥ 93 nm (23 @ 4x)
Mag = 800–1000x
See poster
Ken Goldberg / [email protected]
ALS BL 11.3.2
21
41© Patrick Naulleau, 2008
The CXRO mask inspection microscope provides an unparalleled view into defect printability
AerialImages
5 µm
0.00 µm
0.78
1.64
2.43
3.26
4.14
4.96
z1 µm
amplitude-defect repair sites
G. Yoon, et al.
SAMSUNG
Aerialimages
through focus
1 µm1 µm
Defects:
phase
absorber
through-focus
W. C
ho
200
7S
EM
AT
EC
H
Materials Sciences Division 41MSD Retreat August 12, 2008
Ken Goldberg / [email protected]
42© Patrick Naulleau, 2008
NA=0.7Source size = 1 mm2
• A 1 mm source with 45-degree collection would
have an etendue of ~2 mm^2 sr (min σ = 0.53)
• A 0.1 mm source with 70-degree collection would
have an etendue of ~0.04 mm^2 sr (min σ = 0.08)
Implications of etendue limits for
mask inspection microscopy
Collector optic