1. Improving Dimensional Stability of Microelectronic Substrates by Tuning of Electric Artworks Parsaoran Hutapea Composites Laboratory Department of Mechanical Engineering Temple University, Philadelphia, PA 19122 hutapea@temple.edu ASME IMECE Orlando, FL November 10, 2005

2. Overview  Background  Reducing Warpage by Tuning  Reducing Warpage of Microelectronic Substrates  Summary

3. Background

4. Drivers for warpage  glass/epoxy (FR4) and copper have different CTEs => (iso)thermal warpage  thermal gradients => thermal warpage even if CTEs are the same  transient temperature effects => transient thermal warpage  anisotropy non-uniform over PCB => in-plane stress induced warpage Problems  large components & small pitch => small tolerance for warpage  high stresses => low reliability  fabrication (improper soldering), component drop-offs challenge area

5. Analytical Model  X-Y routing for complex PCB’s  homogenized properties calculated X-Y Routing Homogenized Homogenization Plate Theory Model PCB 3D Analysis 2D Lamination Theory

6. Lamination Theory Strain Energy Goal is to use numerical simulation to solve for the coefficients of the A, B, and D matrices and then use the coefficients to back calculate the effective reduced stiffness matrices

7. FR4 FE unit cell models to calculate homogenized properties  different models developed to include effects of complex stress distribution  periodic boundary conditions applied  electric trace width varied  obtained lamina stiffness Grenestedt and Hutapea, J. Appl. Phys., 94(1), 686, 2003

8. Stiffness (Q   CTE’s (   )  compared the homogenized properties with the Voigt and Reuss Estimations  analytical models developed as a function of copper volume fractions:, where e.g., Traces

9. Reducing Warpage by Tuning

10. Note on Copper Balance  copper amounts in layers placed on opposite sides of the mid-plane, and at the same distance from it, are made equal  achieved by adding ground planes, perforating ground or power planes, etc.  copper balancing is NOT sufficient to minimize warpage  trace directions and anisotropy influence warpage as much as copper amount OriginalTuned 0.159E-041.580E-04 0.159E-041.580E-04 00 (1/mm) 0.210.0046 (1/mm) -0.21-0.0046 00 warpage reduced by ~ 98% ! Grenestedt and Hutapea, J. Appl. Phys., 94(1), 686, 2003 Original Tuned 100% copper balanced

11. Board B Board C Board A original tuned Experiments  experimental using a scaled-up size (due to manufacturing issues)  size effects observed warpage reduced! manufactured 10 times larger than typical

12. Applied on PCB  divide PCB into 14x19 areas  calculate effective PCB properties in each area  make FE model with 14x19 areas (1064 shell elements) and calculate warpage  “tune” the artwork by allowing e.g. 10% change in copper amount in each of the 6x14x19 layers (1596 design variables) 0.025’ 0.023’ original tuned Using shape optimization to search for the optimal 1596 copper amounts Six-layer PCB A, B, D N T, M T

13. Y-Z View Tuned Original X-Z View Original Tuned  10% changes in trace width  warpage decreased by 80%! Original Tuned original tuned X-Z View Y-Z View

14. Reducing Warpage of Microelectronic Substrates

15. Objectives  calculate homogenized properties of electric traces, PTH’s,  vias, and adhesion holes  mechanics-based explanation of warpage  develop numerical tools PTH  via adhesion holes Traces Substrate 1 Substrate 2

16. Method  Lamination Theory  need homogenized properties of substrate features such as traces, PTH’s,  vias, adhesion holes, etc. 4F3F2F1FC 1BC 2B 3B 4B PTH Homogenized Substrate

17. PTH Stiffness (Q   CTE’s (   ) Homogenized Properties substrate features: traces, PTHs,  vias, adhesion holes

18. Substrate 2 Exp. (avg)Pred. (avg) CTE X (ppm/°C) A: 18.7218.12 B: 18.3317.90 C: 20.5719.75 D: 18.1617.91 AB CD Substrate 2 Exp. (avg)Pred. (avg) CTE Y (ppm/°C) A: 16.9918.03 B: 17.1417.81 C: 19.4619.67 D: 17.5717.74 Experimental & Numerical Comparison  measure CTE’s using digital image correlation (Intel)  predict CTE’s using lamination theory

19. Stiffness using Tensile Test more PTH and  vias

20. Stress-Strain Curves specimen 3 (die shadow region)

21. Spec. 1Spec. 2Spec. 3Spec. 4Spec. 5 Pred. (avg) 26.2024.8721.0524.4025.79 Exp.** (avg) 26.61 (29.51, 23.71) 25.45 (25.54, 25.35) 18.42 (17.83, 19.00) 25.17 (25.16, 25.18) 24.06 (24.22, 23.90) ** Stiffnesses are calculated from stress-strain data up to 0.2% (~ linear part) Stiffness Comparison stiffness was the lowest in the die shadow region (specimen 3)

22. Summary  developed analytical models to estimate effective properties with (arbitrary) traces  developed tuning strategy, indicates that warpage can be reduced by approx. 80%  analyses and experiments showed different warpage in the substrates  homogenized properties of substrate features  experimental verification  future work: tuning and software and numerical tools

23. Joachim L. Grenestedt (Lehigh University) Mitul Modi (Intel Corporation) Kristopher Frutschy (Intel Corporation) Acknowledgements

24. Questions?