Enabling Full Profile CMP Metrology and Modeling

5/24/2001

1

Enabling Full Profile CMP Metrology and Modeling

SFR WorkshopMay 24, 2001

Runzi Chang, Costas SpanosBerkeley, CA

2001 GOAL: Develop periodic grating metrology to support integrated CMP model by 9/30/2001.

5/24/2001

2

Motivation• First-principle based modeling can help optimize the operation

of CMP and drive the technology further.

• The ability to measure profile evolution at various polish stages is a key ingredient in establishing and confirming models.

• Formal CMP modeling can the be used to address issues relasted to non-uniformity, pattern loading effects, etc.

5/24/2001

3

Key idea: measure the evolution of a 1-D periodic pattern at various polish stages

SubstrateOxide

• Use scatterometry to monitor the profile evolution• The results can be used for better CMP modeling

5/24/2001

4

Mask Designed to explore Profile as a function of pattern density

• The size of the metrology cell is 250m by 250m

• Periodic pattern has 2m pitch with 50% pattern density

5/24/2001

5

Sensitivity of Scatterometry (GTK simulation)

0 500 1000 1500 20000

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

0.5Profile Evolution during CMP

Oxide (nm)

pro

file

(m

icro

n)

• We simulated 1 m feature size, 2 m pitch and 500nm initial step height, as it polishes.

• The simulation shows that the response difference was fairly strong and detectable.

Tan PSI Response to Profile Evolution

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

240

280

320

360

400

440

480

520

560

600

640

680

720

760

Wavelength(nm)

tan

PS

I

tan PSI 500nm

tan PSI 400nm

tan PSI 300nm

tan PSI 200nm

tan PSI 100nm

tan PSI Flat Surface

Cos DEL Response to Profile Evolution

-1.5

-1

-0.5

0

0.5

1

1.5

240

280

320

360

400

440

480

520

560

600

640

680

720

760

Wavelength(nm)

cos

DE

L

cos DEL 500nm

cos DEL 400nm

cos DEL 300nm

cos DEL 200nm

cos DEL 100nm

cos DEL Flat Surface

5/24/2001

6

Characterization Experiments

• Three one-minute polishing steps were done using the DOE parameters

Initial profilesSopra/AFM

CMP NanospecThickness

measurement

SopraSpectroscopicellipsometer

AFM(AMD/SDC)

Wafer cleaning

Wafer #

Down Force

(psi)

Table Speed (rpm)

Slurry Flow

(ml/min)

1 4 40 50

2 8 40 50

3 4 40 150

4 8 40 150

5 8 80 50

6 4 80 50

7 8 80 150

8 4 80 150

9 6 60 100

10 6 60 100

11 6 60 100

5/24/2001

7

Library-based Full-profile CMP Metrology

Reference: X. Niu, N. Jakatdar, J. Bao, C. Spanos, S. Yedur, “Specular spectroscopic scatterometry in DUV lithography”, Proceedings of the SPIE, vol.3677, pt.1-2, March 1999.

• Five variables were used in describing the oxide profile to generate the response library: bottom oxide height (A), bottom width (B), slope 1 (C), slope 2 (D) and top oxide height (E).

Substrate

AB

CD

E

oxide

5/24/2001

8

-3 -2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2

x 104

0

2000

4000

6000

-2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 3

x 104

-5000

0

5000

-3 -2 -1 0 1 2 3

x 104

0.9

1

1.1x 10

4

AFM

Results

• Extracted profiles match SEM pictures with 10nm precision

• Scatterometry is non-destructive, faster and produces more descriptive than competing methods.

SEM Scatterometry

5/24/2001

9

Conclusions and 2002 / 2003 Goals

• We have demonstrated that scatterometry, in conjunction with specialized profile libraries, can be used for complete profile evaluation during polish.

• This method has been demonstrated on clean, dry samples. It will be interesting to examine the feasibility of using wet samples, for in-situ/in-line deployment of full-profile CMP metrology.

Our next goals are:

Integrate initial chemical models into basic CMP model; Validate predicted pattern development (with Dornfeld and Talbot), by 9/30/2002.

Develop comprehensive chemical and mechanical model (with Dornfeld and Talbot); Perform experimental and metrological validation, by 9/30/2003.