1. Sensitivity of Spectroscopic Scatterometry: Sub-100nm Technology
SFR Workshop
November 8, 2000
Ralph Foong, Costas Spanos
Berkeley, CA
2001 GOAL: To fully characterize the capabilities of scatterometry in fulfilling the metrology needs of the 100nm technology node.
11/8/2000

2. Motivation
Capabilities of scatterometry and required equipment specifications need to be formalized for 100nm metrology.
Commercial ellipsometers have been identified as being able to perform spectroscopic scatterometry. Hence, the focus of this study is on these equipment.
Precision of current generation commercial ellipsometers in measuring profiles consistent with 100nm technology node has to be confirmed.
Scalability of scatterometry towards 70nm and 50nm metrology has to be explored.
Minimum commercial ellipsometer specifications necessary to successfully implement 70nm and 50nm metrology need to be determined.
11/8/2000

3. Which part of the spectrum contains the most information?
Overall Framework of Sensitivity Analysis
Commercial Equipment Analysis
Profile Parameters
Simulations for variation in parameter X
[X(-),X(Nominal), X(+)]
Cos D
Lambda
Tan Y
Determine Noise Contributions
Tan Y, Cos D Noise Spectrum
Are Variations Detectable?
NoYes
EM Response Variations
Which part of the spectrum contains the most information?
11/8/2000

4. Methodology
Electromagnetic simulations are conducted for small changes in profile parameters to measure variations in EM response.
Noise analysis of commercial ellipsometers is carried out to determine detectability of EM response variations.
Rounding
d(IDetector)
d(ISource)
Slope
Angle
Height PR CD
d(qAnalyzer)
Footing
d(qPolarizer)
ARC
d(Beam Divergence)
Poly-Si
Sample Si
11/8/2000

5. Signal-to-Noise Ratio for SOPRA Ellipsometer
Intensity fluctuation is the main contributor of measurement noise in ellipsometers.
Monte-Carlo simulations incorporating intensity fluctuations are used to determine the final distributions of Tan Y and Cos D.
The ‘Minimum Detectable Variation’ lines represent the sum of the 3s errors of each of the 2 profiles measured to obtain the variation.
The graphs demonstrate a trend toward significant information contained in a narrow band in the lower wavelength spectrum.
Signal averaged over 30 measurements
Noise represents 1s standard deviation for each wavelength
Empirical formula for signal-to-noise ratio:
Noise = 0.412(Intensity)0.632
(R2 Value = 0.937)
11/8/2000

6. 100nm Technology Simulations
100nm Dense Lines (ASIC)
65nm Isolated Lines (MPU)
Detectable (Above yellow line)
Undetectable (Below yellow line)
Detectable (Above yellow line)
Undetectable (Below yellow line)
spectrum of information
content
11/8/2000

7. Detectable (Above yellow line) Undetectable (Below yellow line)
70nm Technology Simulations
70nm Dense Lines (ASIC)
45nm Isolated Lines (MPU)
Detectable (Above yellow line)
Undetectable (Below yellow line)
11/8/2000

8. Detectable (Above yellow line) Undetectable (Below yellow line)
50nm Technology Simulations
50nm Dense Lines (ASIC)
30nm Isolated Lines (MPU)
Detectable (Above yellow line)
Undetectable (Below yellow line)
11/8/2000

9. 2002 & 2003 Goals
Study the feasibility of building 100nm capable profile extraction using small footprint, in-line spectroscopic ellipsometry, by 9/31/2002
Implement lithography controller that merges full profile in-line information with available metrology, by 9/31/2003
Profile Diagnostics
DUV Photolithography
PR Deposition, Focus, Exposure, Bake Time, Development Time, etc
Process Flow
In-Line Scatterometry
Wafers
Feedback Control Loop
11/8/2000