EUV lithography: status, future · Maarten van Kampen, Andrei Yakunin, Vladimir Ivanov and many...
Transcript of EUV lithography: status, future · Maarten van Kampen, Andrei Yakunin, Vladimir Ivanov and many...
Vadim Banine
with the help of Rudy Peters, David Brandt, Igor Fomenkov,
Maarten van Kampen, Andrei Yakunin, Vladimir Ivanov and
many other people of ASML and Cymer
EUV lithography: status, future
requirements and challenges
EUVL Dublin
November 2013
Introduction
Why EUVL
Status of the source
Summary and acknowledgements
Public
Slide 2
Contents
Introduction
Why EUVL
Status of the source
Summary and acknowledgements
Public
Slide 3
Contents
EUV is a cost effective solution Public
Slide 4
EUV enables 50% Scaling for the 10 nm logic node Layout restrictions and litho performance limit shrink to ~25% using immersion
Reference
N20/16
double
patterning
triple
patterning
EUV
No
rma
lize
d d
ie s
ize
[%
]
Source: ARM
EUV meets all litho requirements
Triple patterning does not show a process window
Public
Slide 5
ASML’s NXE:3100 and NXE:3300B Public
Slide 6
NXE:3100 NXE:3300B
NA 0.25 0.33
Illumination Conventional 0.8 s Conventional 0.9 s
Off-axis illumination
Resolution 27 nm 22 nm
Dedicated Chuck Overlay /
Matched Machine Overlay
4.0 nm / 7.0 nm 3.0 nm / 5.0 nm
Productivity 6 - 60 Wafers / hour 50 - 125 Wafers / hour
Resist Dose 10 mJ / cm2 15 mJ / cm2
The NXE:3100 has exposed >46,000 wafers Public
Slide 7
0
5000
10000
15000
20000
25000
30000
35000
40000
45000
50000
20
10
27
20
10
32
20
10
37
20
10
42
20
10
47
20
10
52
20
11
05
20
11
10
20
11
15
20
11
20
20
11
25
20
11
30
20
11
35
20
11
40
20
11
45
20
11
50
20
12
03
20
12
08
20
12
13
20
12
18
20
12
23
20
12
28
20
12
33
20
12
38
20
12
43
20
12
48
20
13
01
20
13
06
20
13
11
20
13
16
20
13
21
20
13
26
Acc
um
ula
ted
waf
ers
exp
ose
d o
n N
XE:
31
00
• Cycles of learning
• Stable performance
• Steady wafer output
Introduction
NXE:3300B
Public
Slide 8
Contents
Public
Slide 9
Eleven NXE:3300B systems in various states of integration
System 3
System 1 System 10
System 4
System 8
System 5
System 7
System 6
System 2
System 9 System
Training
EUV cleanroom
System 11
Qualified
Qualified
Building extension started
Resolution shown on NXE:3300B for dense line spaces, regular
and staggered contact holes; all single exposures Public
Slide 10
13nm HP
14nm HP
Dipole30,
Chemically Amplified
Resist (CAR)
Quasar 30 (CAR)
17nm HP
18nm HP
18nm HP
19nm HP
Large Annular (CAR)
Dipole45,
Inpria Resist
13nm HP
14nm HP
EUV ArF immersion
Node: N10 (23nm HP)
Target insertion point for EUV
Node: N20 (32nm
HP)
Single Exposure
Conventional illumination
Double Patterning
(design split)
Best focus difference ~10nm Best focus difference
up to 40-60nm
Overlapping DoF
current 100..120nm (expected to improve after
further optimization (e.g. OPC))
Overlapping DoF
typical ≈ 60nm
Public
-80nm
-60nm
-40nm
-20nm
0nm
20nm
40nm
60nm
80nm
focus
Slide 11
NXE:3300B enables single exposure random logic metal
layer with large DoF minimum HP 23 nm (N10 logic cell)
-12mm 0mm +12mm
Excellent print performance over the full exposure slit
EUV enables aggressive shrink on 2D logic shrink possible beyond N10 node requirement Public
Slide 12
1D
EUV (SE)*
ArFi (SE) 75nm
50nm
22nm
31nm MEEF Tip2Line
* using high dose resist @ ~50mJ
EUV (SE) ArFi (DPT) 16nm
Node
NXE3300B: Line ends imaging - tip2tip and tip2line -
supports 10nm logic node with Quasar illumination
15 mJ/cm2 16.5 mJ/cm2 18 mJ/cm2
20nm mask gap 22nm L/S grating
22nm mask gap
28nm
36nm
28nm
36nm Conv. Quasar
Logic 10nm
Logic 10nm
Public
Slide 13
Ded
icate
d C
hu
ck
Ov
erl
ay [
nm
]
1 2 30
2
4
6
8Lot (1.3,1.3)
1.31.0
1.21.4
1.41.3
X
Y
Day
5 nm
99.7%x: 1.3 nmy: 1.3 nm
Filtered S2F Chuck 1 (S2F)
NXE:3300B imaging and overlay beyond expectations matched overlay to immersion ~3.5nm
Matc
hed
Mach
ine O
verl
ay
NX
E-
imm
ers
ion
[n
m]
1 2 30
2
4
6
8
Lot (3.4,3.0)
3.52.7 3.0
2.3
3.23.3
X
Y
Day
5 nm
99.7%x: 3.4 nmy: 3.0 nm
Filtered S2F (S2F)
XT:1950i reference wafers
EEXY sub-recipes
18par (avg. field) +
CPE (6 par per field)
Full wafer CDU = 1.5nm
22nm HP
BE = 15.9 mJ/cm2
EL = 13%
DoF = 160 nm
Scanner qualification
Scanner capability
18 nm HP 13 nm HP 23 nm HP
9 nm HP
Single exposure EUV Spacer
Public
Slide 14
>900,000 wafer cycled on NXE:3300B for integration and
reliability testing Public
Slide 15
Introduction
Why EUVL
Status of the source
Summary and acknowledgements
Public
Slide 16
Contents
Source Pedestal
Scanner
Pedestal
Fab Floor Fab Floor
Sub-fab Floor
Scanner
metrology for
source to
scanner
alignment
CO2-droplet Metrology
Main Pulse / Pre Pulse
split
Focusing
CO2 system
Tin
catch
Vessel
Droplet
Generator
Collector
EUV&droplet
Metrology
Beam
Tra
nsp
ort
Power Amplifiers PP&MP Seed unit
Inte
rmed
iate
Fo
cu
s U
nit
Plasma+Energy
Control
Machine Control
x z
x=droplet stream direction, z=CO2 light direction, y=orthogonal
EUV source system cross-section Public
Slide 17
Power
Availability
Dose control
Key components:
• Drive Laser
• Collector
• Droplet generator
• Vessel
• Controls (E,x,y,z,t)
• Final Focus Assembly
Source Cymer ASML company
Cymer/ASML LPP Source Development History Public
Slide 18
Cymer/ASML
cooperation
established
3300
Source
Shipment
Coatings for
collector
lifetime
Collector
protection
to 60W
CO2/Sn for
high CE
3300 Vessel
shipment
30mm droplets
with wide spacing
1st source
Development
First 5sr
collector
Proto source
integrated
with scanner
Active gas
debris
mitigation
MOPA + PP
Development
SPF
collector
-1.0 -0.5 0.0 0.5 1.0
0
1
2
3
4
-1.0 -0.5 0.0 0.5 1.0 0
1
2
3
4
CE
, %
Lens position, a.u.
CO2 PP
40W stable
MOPA+PP
Automated 50W
MOPA+PP
3100 source
shipment
2006 2007 2008 2009 2010 2011 2012 2013
EUV Power Scaling
Top Technology Challenges Public
Slide 19
High power CO2 Laser
High repetition rate:
fast droplet generator
Collector lifetime:
debris protection
EUV output:
Conversion efficiency
Dose control:
targeting and
laser stability
EUV Power Scaling
Top Technology Challenges Public
Slide 20
High power CO2 Laser: to 47 kW
High Power Seed Laser is Key to the Drive Laser
350W at 80kHz Demonstrated Public
Slide 21
• Seed Laser power delivery to the amplifiers is critical to achieving saturation
and maximum power extraction from the amplifiers
• 350W target design power at 80 kHz repetition rate
– Already achieved in system-level bench testing
40 50 60 70 800
100
200
300
400
H
igh-
Pow
er S
eed
Mod
ule
Out
put (
W)
Pulse Repetition Rate (kHz)Source Cymer ASML company
New Amplifier for Increase Power is Operational
Reached maximum of 35kW with good laser mode profile Public
Slide 22
• Higher power amplifier development completed at supplier
• Repeatable, stable operation at ~35kW (increased from 20kW)
– Single amplifier continuous (cw) output power
– Pulsed mode operation is ~30% of cw
• Good beam quality measured good focusability
Good beam quality Source Cymer ASML company
EUV Power Scaling
Top Technology Challenges Public
Slide 23
EUV output:
Conversion efficiency
To 3.3%
MOPA pre-pulse explained Public
Slide 24
Pre-pulse
Expansion of pre-
pulsed target Shooting with the
main pulse
Shadow-gram
from a side
Shadow-gram
from the front
Diagnostics is at place to understand and tune the MOPA pre-pulse process
Video
Modeling of the particle break down
More on this in the presentation of V. Ivanov at this conference
Model Experimentl
Model is being created to predict the droplet expansion behaviour
Public
Slide 25
CE: RZLINE vs measured for various target and laser
configurations
Slide 26 |
• No fitting parameters are used in the model
• RZLINE predicts well CE trend for various target geometries
• Maximum predicted CE=5.5% (for λ=10.6 um)
Data points:
Lab N1
1. MOPA+PP; Type A
2. MOPA+PP; Type B
3. TEA; Plane target (E=0.3 J)
Lab N2:
4. MOPA+PP Type C
5. MOPA+ PP Type D
6. MOPA+ PP Type E
Lab N3:
7. Other wavelength
1
2
3
4
5
6
1 2 3 4 5 6
Mea
sure
d CE
, %
Calculated CE, %
Cymer
GPI
PowerLase
Maximum predicted for optimal pulse
shape CE=5.5%
1.
3.
5. 2.
4.
6.
7.
Lab N2
Lab N1
Lab N3
Public
Slide 26
RZLINE vs experimental CE data on plane target
(variation of laser spot size)
Slide 27 |
Cymer experiment (SPIE data)
RZLINE parameters
• Sn flat target
• E=300 mJ
• F=190 mm
• CO2 laser
Cymer experiment vs RZLINE
RZLINE predicts well optimal power density and general CE trend on plane target
Optimum
Public
Slide 27
RZLINE vs Cymer CE data on plane target
(Z-variation)
Slide 28 |
Cymer experiment (SPIE data) RZLINE with laser caustics
RZLINE reproduces “double hump” behavior of CE
This module is developed to account for back reflection of laser
0
1
2
3
4
5
6
7
-5 -3 -1 1 3 5
CE, %
in 2
pi o
f con
dens
er
Focusing lens position, mm
CE as a function of focusing lens position; plane target, Cymer 175ns pulse, 70 um (1/e^2)
CO2 beam
MOPA
Public
Slide 28
MOPA Prepulse Technology for High Power Sources
Improved Prepulse shows 3.7% CE, driven by target size and
stability (droplet and expanded target)
720W in 2p 176W raw at IF 140W at IF
(calc dose controlled)
LT1 EUV power at low DC
40 50 60 70 80 900
2
4
6
8
10
12
EU
V E
nerg
y S
tabi
lity
(%) (
s/m
ean)
EUV Sensor Angle (degree)
10 mm PP
1 mm PPType 1 Type 2
Source Cymer ASML company
Public
Slide 29
EUV Power Scaling
Top Technology Challenges Public
Slide 30
Dose control:
targeting and
laser stability
Public
Slide 31
0 10 20 30 40 50 60-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
Time [min]
Do
se
Err
or
[%]
0 10 20 30 40 50 60-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
Time [min]
Do
se
Err
or
[%]
0 10 20 30 40 50 60-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
Time [min]
Do
se
Err
or
[%]
Repeatable 50W MOPA PrePulse EUV Power and Dose Stability Dose Stability <±0.5%, Die Yield >99.7%
0 10 20 30 40 50 6049
49.2
49.4
49.6
49.8
50
50.2
50.4
50.6
50.8
51
Time [min]
Po
wer
[W]
0 10 20 30 40 50 6049
49.2
49.4
49.6
49.8
50
50.2
50.4
50.6
50.8
51
Time [min]
Po
wer
[W]
0 10 20 30 40 50 6049
49.2
49.4
49.6
49.8
50
50.2
50.4
50.6
50.8
51
Time [min]
Po
we
r [W
]
Power roadmap enabled by higher power
CO2 and optimization of pre pulse system
• Total 32 hours 40W & 50W runs with
good dose reproducibility:
• 99.7% of the dies < 0.5% dose
• representing ~ 1330 exposed
wafers @ 15 mJ/cm2
• 55W run (97.5% of dies in spec)
to test peak performance
Public
Slide 32
MOPA PrePulse sources demonstrated repeatable, stable
performance & dose controlled
EUV Power Scaling
Top Technology Challenges Public
Slide 33
Collector lifetime:
debris protection
3100 Collector Lifetime in the
Field
34
Champion lifetime in the field ~11 months
(~120 billion pulses)
Six collectors with >6 months lifetime
Update given at SPIE in February
SPIE 2011 SPIE 2013 May 2013
Type Average Lifetime
(sample size)
Best lifetime
Uncapped 7 Gp (8) 15 Gp (1)
Current Cap Layer 42 Gp (19, 5 still going) 120 Gp (1)
Cap layer development has
increased average collector
lifetime by >5X since initial
installations
Source Cymer ASML company
Modeling of flow and collector contamination
More on this in the presentation of V. Ivanov at this conference
Target:
Minimization of collector
contamination
Public
Slide 35
In-situ collector cleaning has been demonstrated and
is a key enabler for availability and CoO
• Cleaning on standard MLM capped NXE:3100 Collector
• Tin deposited during normal EUV operation was removed with reactive gas
Public
• Start of test • Collector after cleaning in-situ (in the
vessel) reflectivity fully recovered Area to be cleaned
Slide 36
Collector Cleaning using RF Plasma
37
Cymer funded project with University of Illinois at Urbana – Champaign
Demonstrated 200nm Sn cleaning from Si sample placed on collector surface
Demonstrated 25nm Sn cleaning from the entire 300mm dummy LT-1 collector
200nm Sn cleaning from Si samples
placed on collector surface
25nm Sn cleaning from 300mm dummy
collector
Source Cymer ASML company
Public
Slide 37
MOPA +PP collector protection demonstrated on
NXE:3100 source with full reflectivity recovery
After cleaning (15 Gp) 1.2 Mpulses Bottom
Top
Bottom
Top
15 Gpulses Bottom
Top
R/R0 = 100% R/R0 ~ 95% R/R0 ~ 100%
• Source operated at 40W and all loops closed
Public
Slide 38
EUV source power roadmap with dose control
NXE:3300B NXE:3300B NXE:3300B
EUV dose controlled
power (in-burst) 80W 125W 250W
Drive Laser 26kW 33kW 47kW
CE 3.0% 3.0% 3.3%
NXE:3300B Drive Laser NXE:3300B
Vessel
Power scaling is achieved with increased
CO2 laser power and conversion efficiency
Public
Slide 39
Introduction
Why EUVL
Status of the source
Summary and acknowledgements
Contents
Public
Slide 40
Summary
• NXE:3100 in use for process and device development at customers
• NXE:3300B performance fit for customer development 10nm Logic and sub-20nm DRAM
• Overlay performance of DCO<2nm and MMO<4nm demonstrated
• Resolution of 13nm LS and 18nm Contact Holes demonstrated. Further process optimization to be done
• Good imaging performance for 1D (Line Space), 2D (Contact Holes and Metal 1), and Tip-to-Tip / Tip-to-Line have been shown
• Dose reduction achieved by utilizing contrast enhancement with off-axis illumination
• 50W repeatable source power demonstrated with good dose control
Slide 41
Public 13nm HP
1 2 30
2
4
6
8
Lot (3.4,3.0)
3.52.7 3.0
2.3
3.23.3
X
Y
Ma
tch
ed
Ma
chin
e
Ove
rlay [
nm
]
0 10 20 30 40 50 6049
49.2
49.4
49.6
49.8
50
50.2
50.4
50.6
50.8
51
Time [min]
Po
wer
[W]
50W repeatable
power
Acknowledgements
Slide 42
The work presented today, is the result of hard work and dedication
of teams at ASML , Cymer and many technology partners
worldwide including our customers
Special thanks to our partners and customers for allowing us to use
some of their data in this presentation
Public