Post on 25-Mar-2018
© IMEC 2010
COMPARISON BETWEEN EXISTING
INSPECTION TECHNIQUES FOR EUV
MASK DEFECTS
DIETER VAN DEN HEUVEL, RIK JONCKHEERE,
ERIC HENDRICKX, SHAUNEE CHENG, KURT RONSE,
TSUKASA ABE (A), JOHN MAGANA (B), TRISTAN BRET (C)
A: DNP, B: INTEL, C: CARL ZEISS SMS
© IMEC 2010
INTRODUCTION
Goal of this investigation: Is it possible to find printing, natural
reticle defects (32nm node) with wafer inspection, that were
missed by existing blank inspection or patterned mask inspection.
Printing: only defects verified printing on wafers exposed on ADT
32nm node: to assure appropriate process window throughout ADT full-field
Natural:
– Focus is on defects that are in the ML or absorber; excluding particles
– Opposite to programmed defects
Existing: the most state-of-the-art available at tool vendors and/or in use in the field
for state-of-the-art applications, both for patterned mask inspection (PMI), blank
inspection (BI) and wafer inspection (WI).
2DIETER VAN DEN HEUVEL, EUVL SYMPOSIUM, 18OCT 2010, KOBE (JAPAN)
© IMEC 2010
RETICLE LAYOUT: DEFECT RETICLES
DIETER VAN DEN HEUVEL, EUVL SYMPOSIUM, 18OCT 2010, KOBE (JAPAN) 3
DEFECT32
DEFECT40FF(A+B)
40 30 35
32 32 32
32 32 32
40 40 40
40 40 40
40 40 40
40 40 40
40 40 40
• Sub-modules to allow die-to-die inspection (PMI)
• Vertical lines and spaces => maximum printability
• Support bars to prevent pattern collapse
• Cell-to-cell WI is possible => WI can detect repeater
defects (=reticle defects)
• Various pitch-dimensions
• Matrix of programmed defects (known sizes and types) to
verify sensitivity of each technique
Lines and spaces
Support bar
© IMEC 2010
5 different WI:
• Standard KLA2800 inspection
at IMEC
• 4 more state-of-the-art tools
DIETER VAN DEN HEUVEL, EUVL SYMPOSIUM, 18OCT 2010, KOBE (JAPAN) 4
INTRODUCTION: DEFECT32
METROLOGY BENCHMARKING
MAP (reference wafer)
48 printing defects in total
BI after ML-deposition
Lasertec M1350 (standard BI)
BI after absorber deposition
Lasertec M1350
4 different PMI:
• Standard inspection mask
shop
• 3 more state-of-the-art tools
All locations
reviewed on
1 wafer
© IMEC 2010
DEFECT REVIEW FOR 48 DEFECTS
DIETER VAN DEN HEUVEL, EUVL SYMPOSIUM, 18OCT 2010, KOBE (JAPAN) 5
Vertical scan
Average profile
3nm bump
Defect on wafer
Defect on mask
Horizontal scan
© IMEC 2010
81 81
15 10 0
45
12
31
65 67
8880
0
20
40
60
80
100
BI M
L
BI ab
sorb
er
PM
I A
PM
I B
PM
I C
PM
I D
WI A
WI B
WI C
WI D
WI E
be
st case P
MI + B
I
Pe
rce
nta
ge o
f to
tal d
efe
cts
fou
nd
0 6
57
93 89100
4456
6779 83
100
0
20
40
60
80
100
BI M
L
BI ab
sorb
er
PM
I A
PM
I B
PM
I C
PM
I D
WI A
WI B
WI C
WI D
WI E
be
st case P
MI + B
I
Perc
enta
ge o
f to
tal d
efec
ts f
ou
nd
SENSITIVITY FOR ML AND
ABSORBER DEFECTS SEPARATELY
DIETER VAN DEN HEUVEL, EUVL SYMPOSIUM, 18OCT 2010, KOBE (JAPAN) 6
ML DEFECTS absorber DEFECTS
Conclusion: absorber defects are more likely to be
detected (and also more likely to be repairable )
=> ML defects is biggest concern
BI PMI WIBI PMI WI
gap
© IMEC 2010
SUMMARY DEF32 METROLOGY
BENCHMARKING
Standard M1350 Blank Inspection failed to find
certain known printing ML defects
Capture rate Patterned mask inspection:
ForAbsorber defects 100% is possible
is low for ML-defects
With AFM-review on mask we found proof of a natural
3nm bump in ML causing killer defect.
DIETER VAN DEN HEUVEL, EUVL SYMPOSIUM, 18OCT 2010, KOBE (JAPAN) 7
© IMEC 2010
SECOND RETICLE: DEFECT40FF-A
INVESTIGATION FOR ML DEFECTS
DIETER VAN DEN HEUVEL, EUVL SYMPOSIUM, 18OCT 2010, KOBE (JAPAN) 8
BI after ML-deposition on
Lasertec M1350
BI after absorber deposition on
Lasertec M1350
Standard PMI with Mask vendor
KLA2800 WI in IMEC
Remark: No PMI or WI on more
state-of-the-art inspection tools.
BI after ML-deposition on
Lasertec M7360 (= state-of-the-art)
110 defects (standard KLA2800 WI only)
© IMEC 2010
CORRELATION BI (M1350), PMI AND WI
WITH DEFECT SOURCE ANALYSIS (DSA)
DIETER VAN DEN HEUVEL, EUVL SYMPOSIUM, 18OCT 2010, KOBE (JAPAN) 9
53 defects only found by wafer
inspection => good candidates to
be similar ML-defects as on
DEF32
Focus-test to get extra indication
which defects could be ML-defects :
+0.10um+0.15um 0.00um -0.10um
21/53 show very strong focus-
behavior
Ref: Chris H. Clifford et al,
EUVL symposium 2009
© IMEC 2010
RETICLE REVIEW OF CANDIDATE ML-
DEFECTS MISSED BY M1350 INSPECTION
DIETER VAN DEN HEUVEL, EUVL SYMPOSIUM, 18OCT 2010, KOBE (JAPAN) 10
SEM-review on mask
Impossible to visualize
any of the 21 defects
AFM-review on mask
Impossible to visualize
any of the 21 defects
COORDINATE
ACCURACY?
Reticle marks
Print wafer on IMEC
ADT and check
defect location with
marks
X-offset
defect-mark
Y-offset
New attempt with
AFM with better
alignment
(reduced search
range)
© IMEC 2010
2ND ROUND OF AFM: RESULTS
DIETER VAN DEN HEUVEL, EUVL SYMPOSIUM, 18OCT 2010, KOBE (JAPAN) 11
3nm pit
Wafer review Reticle review
© IMEC 2010
MORE EXAMPLES
DIETER VAN DEN HEUVEL, EUVL SYMPOSIUM, 18OCT 2010, KOBE (JAPAN) 12
Out of 21 candidate ML defects, 14 were checked with
AFM and all were ML pits between 3-7nm!!
© IMEC 2010
WHAT ABOUT MORE ADVANCED
BLANK INSPECTION (M7360)?
DIETER VAN DEN HEUVEL, EUVL SYMPOSIUM, 18OCT 2010, KOBE (JAPAN) 13
1. Did M7360 detect these 21 defects?
2. Reticle review revealed not 21, but in total 41 defects
that were related to ML (no focus effect).
Did M7360 find all these defects?
3. Review of additional detections by M7360 on wafer
=> how many print?
4. Review of additional detections by M7360 on wafers
=> how many don’t print?
5. Important remark: state-of-the-art wafer inspection
tools might reveal smaller, even more-challenging ML-
defects that might have been missed by M7360-
inspection (future work)
All 21 defects were detected (red dots)
All 41 defects were detected (red + blue dots)
An additional 50 printing defects were
detected (yellow dots)
The amount of detections of non-printing
defects (black dots) is unacceptableNote: locations were only reviewed in BF
© IMEC 2010
SUMMARY DEFECT40FF-A RESULTS
Confirmation that standard Lasertec M1350 Blank
Inspection misses certain printing ML-defects
More advanced Lasertec M7360:
All known printing ML-defects were detected
(no data of state-of-the-art WI available)
Too many nuisance detections
With AFM-review on mask we found proof of a natural
3nm pit in ML causing killer defect.
DIETER VAN DEN HEUVEL, EUVL SYMPOSIUM, 18OCT 2010, KOBE (JAPAN) 14
© IMEC 2010
THIRD RETICLE: DEFECT40FF-B
INVESTIGATION FOR ML DEFECTS
DIETER VAN DEN HEUVEL, EUVL SYMPOSIUM, 18OCT 2010, KOBE (JAPAN) 15
Substrate inspection on Lasertec
M1350
BI after ML deposition on Lasertec
M7360
Standard PMI with Mask vendor
KLA2800 WI in IMEC
+
Optimized WI procedure on
more state-of-the-art WI-tool
BI after absorber deposition on
Lasertec M1350
KLA2800: 27 defects
© IMEC 2010
DEFECT40FF-B: DSA RESULTS
DIETER VAN DEN HEUVEL, EUVL SYMPOSIUM, 18OCT 2010, KOBE (JAPAN) 16
Substrate defects
Mostly particles, 1-2
defects with minor focus
behavior
What if we try to
improve WI?
© IMEC 2010
FOCUS BEHAVIOR ML-DEFECTS
DIETER VAN DEN HEUVEL, EUVL SYMPOSIUM, 18OCT 2010, KOBE (JAPAN) 17
-0.025um-0.075um-0.125um +0.025um +0.075um
Prints better
in + defocus
Prints better
in - defocus
variation
through focus
stable
through focus
© IMEC 2010
IMPROVEMENT OF WAFER
INSPECTION TECHNIQUE
DIETER VAN DEN HEUVEL, EUVL SYMPOSIUM, 18OCT 2010, KOBE (JAPAN) 18
• KLA 2800 WI in best focus (BF)
• KLA2800 WI on focus-skew wafer
•Advanced WI (AMAT UVI 4) in BF
•Advanced WI (AMAT UVI 4) on focus-skew
wafers
© IMEC 2010
DEFECT40FF-B:
DSA WITH OPTIMIZED WI
DIETER VAN DEN HEUVEL, EUVL SYMPOSIUM, 18OCT 2010, KOBE (JAPAN) 19
Nr. of defects
only found by WI
also increases
significantly
Number of defects
correlating with BI
increases
significantly
© IMEC 2010
NUISANCE RATE M7360
DIETER VAN DEN HEUVEL, EUVL SYMPOSIUM, 18OCT 2010, KOBE (JAPAN) 20
Sensitivity of 1 pixel is
needed to find all
printing defects
Is it possible to remove nuisance defects without loosing
real printing defects?
© IMEC 2010
REVIEW OF DETECTIONS MADE BY 7360
DIETER VAN DEN HEUVEL, EUVL SYMPOSIUM, 18OCT 2010, KOBE (JAPAN) 21
Defect of 92 pixels
Random review of 50
defects with 1 pixel
through entire focus
range => only 1/50 found
Major challenge for optical blank inspection will be
to differentiate between defects that are likely to
print and defects that are not likely to print
© IMEC 2010
DEFECTS ONLY FOUND BY WAFER
INSPECTION
DIETER VAN DEN HEUVEL, EUVL SYMPOSIUM, 18OCT 2010, KOBE (JAPAN) 22
40 defects only found by wafer inspection
From – to + focus offset
At least 10 defects show through-
focus behavior
Currently AFM analysis is ongoing
© IMEC 2010
OVERALL CONCLUSIONS
Patterned Mask Inspection:
all known printing absorber defects can be found, but most
advanced PMI-tools are needed
BI standard Lasertec M1350:
on all 3 reticles printing ML defects were missed
Blank inspection M7360 :
1. Nuisance rate (=non-printing defects) is too high
2. Strong evidence for M7360 failures, yet working on a proof
via visualization by AFM
Proof of both natural bumps and pits with only 3nm
height distortion on ML-surface , causing killer defects on
wafer => Maybe optical inspection techniques are limited for these types
of defects, because they cannot penetrate inside ML
DIETER VAN DEN HEUVEL, EUVL SYMPOSIUM, 18OCT 2010, KOBE (JAPAN) 23
© IMEC 2010
ACKNOWLEDGEMENTS
Tool vendors:
Applied Materials
KLA-Tencor
Nuflare
Hermes Microvision
INTEL: Sang Lee, Michael Leeson
Carl Zeiss: Thorsten Hofmann, Markus Waiblinger
All partner companies in imec’s Advanced Litho Program
IMEC colleagues: Jan Hermans, Bart Baudemprez, Rudi De Ruyter, ...
ASML ADT-team in IMEC
All of you, for your attention
DIETER VAN DEN HEUVEL, EUVL SYMPOSIUM, 18OCT 2010, KOBE (JAPAN) 24
© IMEC 2010
Thank you!