edge placement error analysis for n7 logic patterning...

2015 International Symposium on Extreme Ultraviolet Lithography, Maastricht, NL

Edge placement error analysis for N7 logic patterning options

Oct-05-2015 ASML: Eelco van Setten, Eleni Psara, Friso Wittebrood, Dorothe Oorschot, Joep van Dijk, Guido Schiffelers, Jo Finders, Mircea Dusa IMEC: Vicky Philipsen, Eric Hendrickx



Slide 2

Public

On the edge

or over the edge…


Slide 3

Public

On the edge

or over the edge…

Photo: WFAA.Com Photo:nbcdfw.com


Slide 4

Public

Edge Placement Error combines all CD and OVL errors

Design intent

After patterning

EP tolerance Y

EP tolerance X


Slide 5

Public

Edge Placement Error combines all CD and OVL errors

EPEinter= (𝟑𝛔)𝐎𝐕𝐋,𝐢𝐧𝐭𝐞𝐫𝟐 +

𝟑𝛔𝐂𝐃𝐔𝐥𝐢𝐧𝐞𝐬

𝟐

𝟐

+𝟑𝛔𝐂𝐃𝐔𝐜𝐮𝐭𝐬

𝟐

𝟐

EPEintra= (𝟑𝛔)𝐎𝐕𝐋,𝐢𝐧𝐭𝐫𝐚𝟐 +


𝟐

𝟐


𝟐

𝟐

EPElocal= (𝟔𝛔)𝐋𝐏𝐄𝟐 +

(𝟔𝛔)𝟐𝐋𝐖𝐑𝐥𝐢𝐧𝐞𝐬

+(𝟔𝛔)𝟐𝐋𝐂𝐃𝐔𝐜𝐮𝐭𝐬

𝟐

EPEmax = EPEinter + EPEintra + EPElocal

die

die

die

die

die

…….

Uncorrelated edges

EPEglobal

EP tolerance Y

EP tolerance X


Slide 6

Public

Main overlay and CD error sources

Overlay errors

CD errors

Machine overlay

Image placement

Mask registration

Pattern transfer

Litho cluster CDU

OPC residuals

Mask CDU

Pattern transfer

Global (die/wfr scale) Local (pitch scale)

E-beam/resist stochastics

Resist/etch stochastics

E-beam/resist stochastics

Resist/etch stochastics

LCDU/resist stochastics

LPE/resist stochastics

die

die

die

die

die

Public

Contents

Edge Placement Error

Patterning options for N7

Overlay errors

CD errors

Summary & conclusions


Slide 8

Public

Design options for critical N7 Logic Metal layers

Uni-directional (1D) M1 Bi-directional (2D) M1

• Design freedom

• Design restrictions

• Additional layers required

• Litho friendly

BUT…

• Challenging for litho

BUT…

Complementary litho: 1D direct print / 2D direct print

ArFi SAQP EUV cut/block

+ =

EUV EUV


Slide 9

Public

Evaluation done on NXE:33x0B EUV scanner NXE:33x0B

NA 0.33

Illumination Conventional 0.9s,

6 off-axis pupil settings

Tip-to-tip (T2T)

Cut/block 1D direct print / 2D direct print

Staggered CHs (Irregular) cuts

Y. Borodovsky, Intel, 2012 Int’l

Workshop on EUVL, Maui, Hi

Goal: Compare relative performance using EPE model

Public

Contents



Overlay errors

CD errors



Slide 11

Public

MMO:

• X = 2.0 nm

• Y = 2.2 nm

DCO:

• X = 1.0 nm

• Y = 0.6 nm

NXE:3350: 2.0nm Matched Machine OVL and 1.0nm

Dedicated Chuck OVL demonstrated

On-product overlay NXE:3300B to NXT:1960Bi:

measured = 5-nm, correction towards 4‐nm possible

NXE:3350B: OPO < 3.5nm expected, to be demonstrated

NXE:3350B

Source P ~ 60-

80W

MMO

Jan Mulkens et al, SPIE 2015


-0.5

0

0.5

1

1.5

-100 -50 0 50 100

Cut-X Annular

Cut-Y Annular

Cut-X Conv

Cut-Y Conv

Pla

cem

en

t Er

ror

[nm

]

Focus[nm]

Slide 12

Public

Image placement errors through focus can be minimized

by mask and pupil optimization

For more details see ‘Experimental verification of phase induced M3D effects in EUV imaging’ – Friso Wittebrood

M3D simulation 21/22nm HP cuts

Annular Conventional

Cut-X

Cut-Y

-0.5

0

0.5

1

1.5

-100 -50 0 50 100

Trench Dip90Y

Trench FlexPupil

Gap Dip90Y

Gap FlexPupil

Pla

cem

en

t Er

ror

[nm

] Focus[nm]

Dip90Y FlexPupil

M3D simulation T2T @ 16nm HP

Gap Trench

Pattern placement aware SMO, ‘EUV Reticle Enhancement Techniques for k1 0.4 and below’ SPIE 2015 – Stephen Hsu et al


Slide 13

Public

Local Placement Error on mask gives systematic

overlay error at pitch scale

Pitch scale

Mask local placement error:

Deviation from designed position at pitch scale

Method:

• Measure center of gravity for 9x9 cuts within array (CD SEM) • Average over ~64 fields ~ 8x noise reduction

• Calculate distance in X/Y between cuts

• Calculate delta w.r.t. design distance (pitch)

• Calculate 3σ over rel. X/Y placement error

Δ=0.5 Δ=-1.1

Δ=-0.5 Δ=0.8

Δ=0.7 Δ=-0.1 Δ=0.1


Slide 14

Public

Systematic Local Placement Error (mask) ~

1.0nm for both CHs/ cuts and T2T

0.0

0.5

1.0

1.5

2.0

22p44 staggeredCHs

21/22nm HP cuts

X

Y

LPE

[nm

3σ

; 1

x]

Mask LPE for CHs and cuts

0

0.5

1

1.5

2

12 20

22p44 T2T

LPE

[nm

3σ

; 1

x]

Design gap [nm; 1x]

Mask LPE of T2T gap

Derived from wafer measurements

Pitch scale

Public

Contents



Overlay errors

CD errors



Slide 16

Public

NXE:3350: 0.6nm FWCDU demonstrated for horizontal

16nm HP L/S and 20nm iso trenches

16nm dense L/S 20nm iso trenches

CDU CD FW CDU IF CDU

H line 16.1nm 0.6nm 0.5nm

CDU CD FW CDU IF CDU

H trench 21.0nm 0.6nm 0.4nm

Reticle and shadowing error corrected


Slide 17

Public

Good CD control for both cut and T2T features Cut-X CDU is 1.0nm 3σ

CD (X/Y) FW CDU

(X/Y)

IF CDU

(X/Y)

21.9nm

71.3nm

1.0nm

2.0nm

0.8nm

1.6nm

CDU 21/22nm HP cuts

NXE:3350B

X-cut Y-cut


Slide 18

Public

Good CD control for both cut and T2T features Cut-X CDU is 1.0nm 3σ

CD FW CDU IF CDU

30.9nm 2.9nm 2.2nm

CDU T2T @ 16nm HP

NXE:3300B NXE:3350B

0

0.5

1

1.5

2

2.5

3

0

20

40

60

80

100

120

22p44CHS

Feat -1 Feat -2 Feat -3 Feat -4

Mean X Mean YFW CDU X FW CDU YIF CDU X IF CDU Y

CD

[n

m]

CD

U [

nm

3σ

]

CDU 21/22nm HP cuts


0

2

4

6

8

10

22p44CHS

Feat -1 Feat -2 Feat -3 Feat -4

Slide 19

Public

Cuts show better LCDU performance than line ends

0

2

4

6

8

10

12 14 16 18 20 25 30

Meas

Sim

LGU

[n

m 3σ

]

Design gap [nm; 1x]

LCDU 21/22nm HP cuts LGU for T2T – 16nm HP

LCD

U [

nm

3σ

]

Pitch scale

Estimated SEM contribution ~ 3nm 3σ

LCDU Diameter

Estimated SEM contribution ~ 1nm 3σ


Slide 20

Public

Local CDU can be minimized by maximizing contrast

LCDU CH data LGU T2T data

0

5

10

15

20

8 18 28 38

p32 - DY90

p40 - Q45

p44 - Q45

LGU

[n

m 3σ

] 1/ILS [nm]

NXE:3300 CAR data

High contrast Low contrast

𝐿𝐺𝑈 𝑛𝑚 ~ 0.35

𝐼𝐿𝑆+ 1.5

0.0

1.0

2.0

3.0

4.0

5.0

0.10 0.20 0.30 0.40 0.50 0.60

20

22

24

30

35


𝐿𝐶𝐷𝑈 𝑛𝑚 ~6.6

𝑁𝐼𝐿𝑆+ 1.1

NXE:3300 CAR data HP (nm)

1/NILS [-]

LCD

U [

nm

3σ

]

Data courtesy: Sander Wuister, ASML


0

5

10

15

20

5 15 25 35

p32 - DY90

p40 - Q45

p44 - Q45

CH - All data

LCD

U /

LG

U [

nm

3σ

]

1/ILS [nm]

Slide 21

Public

Local CDU can be minimized by maximizing contrast

LCDU CH and LGU T2T data

NXE:3300 CAR data



2.0

2.2

2.4

2.6

2.8

3.0

3.2

3.4

12 13 14 15 16 17 18 19 20 21 22

0.33NA 38% PFR

0.33NA 20% PFR

NXE:33x0

NXE:3400

L/S HP [nm]

NIL

S [-

]

Horizontal L/S

Slide 22

Public

Contrast optimization by OPC, pupil and absorber

optimization

BF shift & Bossung tilt

Pa

rtly

fa

din

g

Laurens de Winter et al. EMLC 2015

Contrast optimization:

• Pupil optimization: • Pupil fill ratio NXE:33x0 = 38%

• Pupil fill ratio NXE:3400 = 20%

• OPC / SMO, eg. SRAFs

• Mask stack optimization Resolution enhancement

Stephen Hsu et al. SPIE 2015

No AF

With AF

Public

Contents



Overlay errors

CD errors



Slide 24

Public

Main EPE contributors are Local CDU and on product OVL

• Process / etch optimization expected to reduce Local CDU up to

30-50% for CHs/cuts and 10-30% for line ends

Main contributors to EPE budget (after litho)

Cut (X/Y) T2T

OPO (TBC) [nm 3σ] ~3.5 ~3.5

LPE [nm 6σ] ~2.0 / 2.4 ~2.4

CDU/2 [nm 3σ] ~0.5 / 1.0 ~1.45

LCDU/2 [nm 6σ] ~4 ~8

EPEglobal= (𝟑𝛔)𝐎𝐕𝐋𝟐 +


𝟐

𝟐


𝟐

𝟐

EPElocal= (𝟔𝛔)𝐋𝐏𝐄𝟐 +

(𝟔𝛔)𝟐𝐋𝐖𝐑𝐥𝐢𝐧𝐞𝐬

+(𝟔𝛔)𝟐𝐋𝐂𝐃𝐔𝐜𝐮𝐭𝐬

𝟐

die

die

die

die

die

Cut (X/Y) T2T

OPO (TBC) [nm 3σ] ~3.5 ~3.5

LPE [nm 6σ] ~2.0 / 2.4 ~2.4

CDU/2 [nm 3σ] ~0.5 / 1.0 ~1.45

LCDU/2 [nm 6σ] ~4 ~8

LCDU/2 [nm 6σ] AEI ~2-3 ~5.5-7


Slide 25

Public

Conclusions

EPE model:

• Includes overlay and CD errors at global and local scale

• Allows balancing of overlay and (L)CDU requirements

EPE model used to compare relative performance difference for cutmask and tip-to-tip:

• Local CDU and on-product OVL dominate the EPE budget

• Local gap uniformity more difficult to control than local CDU of contacts and cuts driven by contrast (ILS)

OPC/ SMO, pupil and mask stack optimization are powerful knobs to improve:

• Contrast LCDU reduction

• Placement errors through focus OPO reduction


Slide 26

Public

The authors would like to thank:

• NXE Applications team

• Thijs Hollink

• Vidya Vaenkatesan

• Cheuk-Wah Man

• John McNamara

• Frank Horsten

• Rik Hoefnagels

• Marieke Meeuwissen

• Kees Ricken

• Others

• Sander Wuister

• Leon Levasier

• Marcel Beckers

• Martien de Rooij

• Martin Verhoeven

• Pieter Woltgens

• Jan Mulkens

• Gian Lorusso

• Lieve Van Look

• Dan Corliss

• NXE simulation team

• Gijsbert Rispens

• Paul Colsters

• Laurens de Winter

• Kateryna Lyakhova

• Cemil Kaya

• Gerardo Bottiglieri


Thank you for your

attention !

edge placement error analysis for n7 logic patterning...

Documents