Dummy Feature Placement for Chemical-mechanical Polishing Uniformity in a Shallow Trench Isolation Process

1
Dummy Feature Placement for Chemical-mechanical
Polishing Uniformity in a Shallow Trench
Isolation Process

• Ruiqi Tian1,2, Xiaoping Tang1, D. F. Wong1

1. Dept. of CS, University of Texas at Austin,
Austin, TX 787122. Motorola Inc., 3501 Ed
Bluestein Blvd., Austin, TX 78721 ruiqi, tang,
wong_at_cs.utexas.edu
2
Outline

• The STI Process
• Derivations for CMP in STI
• Models used (a review)
• Assumptions and results
• Dummy Feature Placement
• Problem formulation
• Iterative Approach
• Computational Experience
• Conclusion

3
The STI Process
CMP in STI is a dual-material polish
Nitride Deposition
Etch
Oxide Deposition
CMP
Nitride Strip
4
Derivations for CMP in STI
Models Effective Density from Pad Bending
5
Derivations for CMP in STI
Models Local Pad Compression

• Polish rates of high and low areas are related by
  step height due to pressure re-distribution
• Initial contact height decreases with increasing
  density, no consideration for spacing

6
Derivations for CMP in STI
Models Dual-Material Polish

• Polish rates are similar to local pad compression
  
• Different blanket polish rate for different
  materials
• Intersection depends on contact height and density

7
Derivations for CMP in STI
Assumptions
Results

• Two stages identified for CMP in STI
• Overburden oxide removal
• Dual-material polish of nitride and oxide

8
Dummy Feature Placement for STI
Formulations as NLP Problems

• Min-Fill Formulation

• Min-Var Formulation

Min
Min
S.T.
S.T.
9
Dummy Feature Placement for STI
Iterative Approaches
10
Computational Experience
11
Computational Experience
Density and Post-CMP Topography Simulations for
L3
Density
Topography
Original
Tiled
12
Conclusion

• Formulation for CMP in STI
• Models for pad bending, pad compression, and
  dual-material polish are considered
• Dummy feature placement as an NLP problem
• Solution for dummy feature placement
• Iterative approaches proposed
• Experimental results are good
• Future studies needed
• Contact height dependence on feature spacing