Characterizing Molded Package Stress Utilizing Tensile Displacement Measurements

  • Slides: 23
Download presentation
Characterizing Molded Package Stress Utilizing Tensile Displacement Measurements Presented by Andrew Schoenberg Fairchild Semiconductor

Characterizing Molded Package Stress Utilizing Tensile Displacement Measurements Presented by Andrew Schoenberg Fairchild Semiconductor May 3, 2006 Symposium on Polymers www. fairchildsemi. com 1

Overview • • Conventional Epoxy Novolac Molding Compounds with silica filler content of approximately

Overview • • Conventional Epoxy Novolac Molding Compounds with silica filler content of approximately 75%, normally require a post mold curing process (PMC) of two to five hours at 175 C to reach full cure and result in a Tg of approximately 145 C. Significant package and environmental regulation requirements have considerably changed the formulation requirements of Mold Compounds. New molding compounds have silica filler loadings up to 92% and a broad spectrum of resin chemistries to facilitate the elevated filler loadings, new package requirements, and new flame retardants. Cost pressures on package assembly has in some cases necessitated the use of No PMC molding compounds to eliminate this time consuming process step and reduce the overall assembly cost 2 2 September 2005

Overview - Continued • • • Additionally some new packaging technologies are utilizing molding

Overview - Continued • • • Additionally some new packaging technologies are utilizing molding compounds in extremely thin film applications (film thickness of <0. 2 mm) which result in a package thickness too thin to be characterized with conventional Thermomechanical Analysis (TMA) for the determination of the first and second coefficients of thermal expansion. A new technique was required to characterize thermal mechanical behavior of these molded systems and verify the resulting stresses both from residual cure, processing conditions, and for applications modeling This presentation describes the use of Film Tension Displacement Measurements for characterization of mold compound stresses and the subsequent FEA modeling results indicating the implications of using No PMC mold compounds. 3 3 September 2005

Objective • • The primary objective of this presentation is to explore the use

Objective • • The primary objective of this presentation is to explore the use of film tension displacement characterization to quantify the stresses (shrinkage and expansion) in molding compounds that result from residual cure, and determine the implications of these stresses on the assembled package robustness (using FEA modeling). Additionally, this presentation will touch on the use of this technique for extremely thin molding applications (< 0. 2 mm thickness) for characterization of Alpha 1 and Alpha 2 CTE measurements, where standard conventional characterization techniques can not be utilized. 4 4 September 2005

Design – Background – PMC Vs. No PMC • • A specific package under

Design – Background – PMC Vs. No PMC • • A specific package under development looked to use a reported No PMC mold compound to reduce the assembly cost. Upon initial evaluation of the mold compound it was found through TMA and DSC that there was significant residual cure remaining in the mold compound after the transfer molding process. DSC after 1 hour 175 C PMC – Tg onset 129. 6 C DSC of No PMC Run 1 – Tg onset 99. 78 C 5 5 September 2005

Design – Background Continued. • • • TMA of mold compound only devices (dummy

Design – Background Continued. • • • TMA of mold compound only devices (dummy devices) molded using exact processing conditions and device configurations resulted in a first thermal ramp with an average slope intercept derived Tg of 91 C and for samples post mold cured for 1 hour at 175 C, an average calculated slope intercept derived Tg of 122 C. Correlative studies on full devices resulted in a calculated Tg average intercept derived Tg of 93 C and for samples post mold cured for 1 hour at 175 C an average calculated slope intercept derived Tg of 118 C. As noted in the adjacent image, the shape of the TMA curve for the no PMC samples exhibits a “shoulder” above the Tg region that was not present on the post mold cured samples. This type of “shoulder” is NOT uncommon for first thermal ramp molded samples. 6 6 September 2005

Design – Background – Failure Mechanism • In addition to the cure characteristics of

Design – Background – Failure Mechanism • In addition to the cure characteristics of the No PMC process, a failure mechanism associated to die cracking was observed in the assembled package. This die cracking appeared to originate from two different locations; chip outs from the sawn edges, and from the center of the die. No evidence of impact defects were detected at the center die crack origins 7 7 September 2005

Design – Film Tension Dynamic Mechanical Analysis • • An alternative analytical method was

Design – Film Tension Dynamic Mechanical Analysis • • An alternative analytical method was explored to attempt to characterize the displacement stress that occurs due to residual cure shrinkage not accurately described by TMA This method utilized the Film Tension Clamp assembly of a Dynamic Mechanical Analyzer (DMA) in an oscillatory experimental Mode. The % strain was derived by ambient LVR characterization 8 8 September 2005

Design – DMA – Data: 0. 05% Strain • Due to the highly filled

Design – DMA – Data: 0. 05% Strain • Due to the highly filled nature of this molding compound (~80%) the ambient LVR characterization did not exhibit a yield. Therefore the first series of experiments were run with a % strain of 0. 05, and a Force Track of 150%. • The resulting data showed a negative Alpha 1 of ~ 10 ppm. • The No PMC exhibited a significantly larger expansion / contraction peak after Tg, however the final Alpha 2 for both samples was about + 37 ppm 9 9 September 2005

Design – DMA – Data: 0. 015% Strain • A second series of experiments

Design – DMA – Data: 0. 015% Strain • A second series of experiments were run with the % Strain set at 0. 015, again utilizing a Force Track of 150%. • The resulting curves again showed a significant expansion / contraction peak for the No PMC material, however now almost no measurable contraction was detected for the PMC (1 hr @ 175 C) sample. • Again there was a negative Alpha 1 of ~10. 5 ppm and a positive Alpha 2 of ~ 34 ppm. 10 10 September 2005

Design – Data – Correction Factor Determination • • 11 11 DMA Tensile data

Design – Data – Correction Factor Determination • • 11 11 DMA Tensile data indicated that the Alpha 1 of the mold compounds had a negative coefficient. Since this is not reasonable, a correction factor had to be developed to compensate for the AISI Type 330 Stainless Steel Clamps used in this technique. A known compliance sample of AISI 1010 Steel was analyzed using the same oscillatory experimental parameters as the mold compound samples, resulting in a negative CTE detected of approx. -9. 7 ppm (avg. of 2 runs in image). Since AISI 1010 Steel has a known CTE of +13 ppm from 20 C to 250 C, an offset correction factor of 23 PPM was used for mold compound CTE calculations throughout the temperature range of the experiments. September 2005

Results – 0. 05% Strain Corrected File # • No PMC With PMC data

Results – 0. 05% Strain Corrected File # • No PMC With PMC data 020206. 45 data 020106. 044 0. 05 150% 0. 37 No PMC With PMC . 01 N CTE Data Without Correction % Strain Force Track Poisson Ration used Preload Force . 01 N Offset CTE 20 C to 250 C based on above known standards 23 ppm Temperature Range of Contraction below Tg 40 C to 95 C 40 C to 119 C Corrected (adding 23 ppm) CTE to Temperature Range Below Tg 12 ppm 13 ppm -10. 9 ppm -9. 7 ppm 151 ppm 150 ppm -76 ppm -47 ppm 37 ppm 36 ppm Temperature Range of First Expansion after Tg 95 C to 131 C 119 C to 159 C Corrected (adding 23 ppm) CTE to first Expansion after Tg 174 ppm 173 ppm 131 C to 170 C 159 C to 180 C Corrected (adding 23 ppm) CTE to Contraction -53 ppm -24 ppm Temperature Range of Expansion after Contraction 170 C to 219 C 180 C to 220 C Corrected (adding 23 ppm) CTE to expansion after Contraction 60 ppm 59 ppm Temperature Range of Contraction after First Expansion after Tg 12 • • 12 Utilizing the correction factor of 23 ppm, all displacement ranges were recalculated. Alpha 1 is now ~12. 5 ppm and Alpha 2 is now ~59. 5 ppm Vendor reported Alpha 1 is ~13. 5 ppm and Alpha 2 is ~59 ppm A significantly higher shrinkage stress is calculated for the No PMC molded samples The Expansion immediately after Tg is consistent for both samples and is approx. 173 ppm September 2005

Results - 0. 015% Strain - Corrected • No PMC With PMC data 020106.

Results - 0. 015% Strain - Corrected • No PMC With PMC data 020106. 40 data 013106. 039 % Strain 0. 015 Force Track 150% 0. 37 No PMC With PMC . 01 N CTE Data Without Correction File # Poisson Ration used Preload Force . 01 N Offset CTE 20 C to 250 C based on above known standards 23 ppm Temperature Range of Contraction below Tg 36 - 88 C 45 - 121 C Corrected (adding 23 ppm) CTE to Temperature Range Below Tg 12. 9 ppm 11. 1 ppm -10. 08 ppm -11. 86 ppm 9. 54 ppm 35. 21 ppm -26. 2 ppm 12. 2 ppm 34. 2 ppm 33. 6 ppm Temperature Range of First Expansion after Tg 88 - 127 C 121 - 162 C Corrected (adding 23 ppm) CTE to first Expansion after Tg 32. 54 ppm 58. 21 Temperature Range of Contraction after First Expansion after Tg 127 - 162 C 162 - 177 C Corrected (adding 23 ppm) CTE to Contraction -3. 2 ppm 35. 2 ppm Temperature Range of Expansion after Contraction 162 - 235 C 177 - 240 C Corrected (adding 23 ppm) CTE to expansion after Contraction 57. 2 ppm 56. 6 ppm 13 13 • • Again as with the 0. 05% strain, Alpha 1 corrected is ~12 ppm and Alpha 2 corrected is ~57 ppm (consistent with Vendor data) Now only the No PMC molded samples exhibit a negative shrinkage stress after Tg However with the 0. 015% strain the expansion immediately after Tg is different between the PMC and No PMC samples September 2005

reflow model/mesh with PCB Load: From Tg to reflow temperature 260 C in 30

reflow model/mesh with PCB Load: From Tg to reflow temperature 260 C in 30 secs PMC: 126 ->260 C No. PMC: 96. 2 -> 260 C @ 131 C, PMC t_pmc=1. 12 s No. PMc t_nopmc= 6 s @ 159 C, PMC t_pmc=7 s No. PMC t_nopmc= 11. 5 s Analysis: Elastic Simulation 14 14 September 2005

Die First Principal stress 0. 05% Strain @ 131 C & 260 C Max:

Die First Principal stress 0. 05% Strain @ 131 C & 260 C Max: 103. 5 MPa @ 1. 12 s Max: 576. 4 MPa @ 6 s 131 C No PMC (5. 6 times greater) With PMC 260 C Max: 398. 6 MPa 16. 7% greater Max: 341. 5 MPa @ 260 C 15 15 September 2005

Max EMC Von-Mises Stress at 0. 05% Strain @ 131 C & 260 C

Max EMC Von-Mises Stress at 0. 05% Strain @ 131 C & 260 C Max: 219 MPa @ 6 s Max: 44. 9 MPa @ 1. 12 s 131 C No PMC (4. 8 times greater) With PMC 260 C Max: 41 MPa @ 260 C Max: 46. 5 MPa @ 260 C 16 16 September 2005

Die First Principal stress at 0. 015% Strain @ 127 C & 162 C

Die First Principal stress at 0. 015% Strain @ 127 C & 162 C Max: 7 MPa @ 0. 22 s Max: 110 MPa @ 5 s 127 C With PMC No PMC 162 C Max: 163 MPa @12. 05 s @162 C Max: 170 MPa @ 8 s @162 C 17 17 September 2005

 EMC Von-Mises Stress at 0. 015% Strain @ 127 C & 162 C

EMC Von-Mises Stress at 0. 015% Strain @ 127 C & 162 C Max: 43. 5 MPa @ 5 s 3 MPa @ 0. 22 s 127 C No PMC (14. 5 times greater) With PMC (@1 s) 162 C 52. 6 MPa @ 8 s @ 162 C Max: 21. 2 MPa @ 12. 05 s @ 162 C 18 18 September 2005

Die First Principal stress and EMC Von. Mises Stresses at 0. 015% Strain @

Die First Principal stress and EMC Von. Mises Stresses at 0. 015% Strain @ 260 C Max: 377. 3 MPa @ 260 C 16% Greater Max: 325. 6 MPa @ 260 C With PMC No PMC Max: 39. 7 MPa Max: 45 MPa 19 19 September 2005

Discussion – Ansys Modeling Results • • • For the 0. 05% strain, the

Discussion – Ansys Modeling Results • • • For the 0. 05% strain, the Max stresses appear when EMC second curing begins, while for the 0. 015% strain the Max die stress (and total max stress) appears at reflow 260 C For the 0. 05% strain, the die max first principal stress without PMC is about 41. 7% greater than with PMC when the second EMC curing begins. For both strain experimental conditions at the reflow temperature of 260 C, the die max first principal stress without PMC is about 16% greater than with PMC. For the 0. 05% strain, the max von-Mises stresses of EMC without PMC is about 48 % greater than with PMC when the second EMC curing begins. However at the reflow 260 C, the max von-Mises stress of EMC without PMC is a little lower than with PMC. For the 0. 015% strain, the max von-Mises stresses of EMC without PMC is smaller than with PMC @ reflow (due to the shrinkage assoc. to cure). • It appears that the shrinkage stresses resulting from residual cure move the maximum 1 st principle stress from the corners of the die (PMC samples) to the center of the die (no PMC samples) 20 20 September 2005

Discussion - Tensile Displacement Measurement Method • The use of the correction factor of

Discussion - Tensile Displacement Measurement Method • The use of the correction factor of 23 ppm results in good correlation with the vendor published data of Alpha 1 and Alpha 2. This correlation was also realized with several other very different molding compounds (filler loadings and chemistry). • The Ansys modeling results highlight the significance of properly selecting the appropriate oscillatory strain for the entire experiment (throughout the entire temperature range). The use of multiple LVR evaluations bracketing the temperature range of interest will allow for improved selection of a % strain that consistently lies within the LVR. • The results of the Ansys modeling pertaining to the change in 1 st principle stress maximum moving from the corners of the die to the center appears to correlate to the failure mechanism observed during production trials. While many factors can exacerbate die crack propagation, this is a new learning regarding the effects of residual cure shrinkage 21 21 September 2005

Discussion – Application for Thin Film, Highly Filled (~90% by wt. ), Mold Compound

Discussion – Application for Thin Film, Highly Filled (~90% by wt. ), Mold Compound – after PMC TMA 1 st thermal ramp 0. 5 mm Thick film Second Thermal Ramp 1. 0 mm Thick film Std. TMA Second thermal ramp Tensile Eval. of 0. 14 mm film 22 22 September 2005

Conclusions • It appears that this Tensile Displacement measurement technique is viable for measuring

Conclusions • It appears that this Tensile Displacement measurement technique is viable for measuring displacement changes associated to residual cure. However additional work is required to identify the appropriate % strain for each new material to most accurately quantify the actual shrinkage displacement. • It also appears that this technique can be used to measure alpha 1 & 2 CTE’s for molded thin film applications that cannot be measured with conventional TMA. • Ansys modeling has identified a change in 1 st principle stress concentration from the corners (normal anticipated location due to standard mis-match of CTE between different materials) to the die center. This relates to the inclusion of multiple CTE stresses to FEA model generation (beyond the standard Alpha 1 and Alpha 2). • When evaluating a non-PMC molding compound process, additional characterization should be required to verify that residual cure shrinkage stresses to not induce secondary failure mechanisms. 23 23 September 2005