1. LED Assembly, Reliability & Testing Symposium
Failure in LED Interconnects: Effect of Substrate Material and Solder Alloy
LED Assembly, Reliability & Testing Symposium
November 28-30, 2017
Maxim Serebreni1, Patrick McCluskey2, Gyan Dutt3, Ranjit Pandher3 1DfR Solutions, 9000 Virginia Manor Rd #290, Beltsville MD CALCE/ Department of Mechanical Engineering, University of Maryland, College Park MD Alpha Assembly, 300 Atrium Drive, Somerset NJ

2. Outline Introduction Fatigue failure in solder joints
Solder Fatigue in LED on MCPCB
Influence of Solder Alloy
Thermal cycling & Die shear strength
Fatigue Life Predictions
Methodology
Results
Conclusion and Future Work

3. Introduction SMD Package-on-board (POB) & Chip-on-board (COB) LEDs.
Coefficient of thermal expansion mismatch between metal board and ceramic submount under fluctuating thermal loads during operation.
Dielectric material, Solder alloy and substrate material influence thermal fatigue life of solder interconnects.
POB
COB
CTE = 3-8 ppm/C
CTE = ppm/C

4. Fatigue Failure in LED Solder Joints
Solder interconnect fatigue failure is greatly influenced by the substrate material and solder alloy.
Solder fatigue failure is influenced by the joint height as much as the effective CTE mismatch!
LED on FR4
LED on Cu Heat Spreader
Zhao, X. J., J. F. J. M. Caers, Sander Noijen, Ying Zhong, M. De Jong, Harry Gijsbers, Gorden Elger, and Harald Willwohl. "Potential interconnect technologies for high power LEDs assemblies." In Electronic System-Integration Technology Conference (ESTC), th, pp IEEE, 2012.
LED on MCPCB

5. Solder Fatigue in LED on MCPCB
LED on MCPCB show dependence on metallic substrate (aluminum vs. copper), solder alloy and dielectric material
Reference:
Effect of Metal Clad Substrate
Effect of Thermal Profile
1600 vs cycles, 2.8X difference
1000 vs cycles, 2.8X difference
Effect of solder alloy
Effect of Dielectric Material
1600 vs cycles, 2.1X difference

6. Influence of Solder Alloy
High Power ceramic submount LEDs assembled on Al MCPCBs with Pb-free Creep Resistant Alloy (CRA) and SAC305
Thermal cycling from -40°C to 125°C with 30 minute dwell
2X improvement in fatigue life under accelerate life testing!
CRA
β~1600 cycles
Percent of LED population with the lights ON
SAC305
β~850 cycles

7. Die Shear Strength
The improved thermal fatigue properties of Creep Resistant Alloy can be seen from die shear strength after 1000 thermal cycles
CRA alloy demonstrated highest shear strength in comparison to SAX305 and SACX0807
CRA
SAC305
SACX0807

8. Analytical vs. FEA Prediction Methodology
Each fatigue life prediction method has its strength and weaknesses. Analytical model calibrated for specific package type and substrate.
FEA model provides high fidelity and can account for complex material interaction and geometry. Requires more material characterization and user inputs.
Finite element Model
Analytical Model
3D Stress-Strain Behavior
1D Stress-Strain Behavior
Strain Energy density
Strain Energy density
Damage Law
Damage Law
Cycles to Failure
Cycles to Failure

9. Fatigue Life Prediction
Analytical approach: First order approximation of stress and strain in solder joints.
Combine empirical data with linear elastic prediction.
Assumption include simplified geometry and temperature dependent material properties.
Follows same damage law as FEA.

10. Fatigue Life Prediction
Finite element approach: enables accurate representation of LED and solder joint geometry.
Provides insight to the localized effects and global influence from system level.
Highly sensitive to simulation methodology and material input.
Die
Epoxy Lens
Metallization layer
Copper pads
Metal
Dielectric layer
Solder joints anode and cathode
Thermal pad Solder

11. Thermal Cycle & LED Geometry
Temperature cycles with 10 °C/min ramp rate and 30 minute dwell at each temperature extreme.
Temperature change applied to entire LED and IMS package as experienced during accelerated thermal cycling conditions.
Philips LUXEON Rebel white LED.

12. Material Properties for Simulation
Materials in the LED assumed to be linear elastic except for solder alloys
SAC305
Linear elastic materials
Material
Elastic Modulus (GPa)
Poisson’s ratio
CTE (ppm/ºc)
PCB 24
0.138 17
Ceramic sub
300
0.28
8.2
Silicon
185
0.18
2.6
Silicone
130
150
Metallization 79
0.37
14.2
Copper
120
0.3
16.9
Aluminum 70
0.35
SAC305
Properties defined by creep equation 20
CRA
CRA
Creep model
Solder A1 α n H1
SAC305
277984
6.41
54041
CRA
15E9
0.1538
4.0
123000

13. Dielectric Material
Temperature dependent properties of the dielectric material.
Hyperbolic tangent function fits storage modulus data.
High Temperature (HT), MP (Multi-purpose), CML (Circuit Material Laminate) dielectrics.
Coefficient of thermal expansion
Storage modulus < Tg
Before Tg: HT < CML < MP
After Tg: HT < MP < CML
MP < HT < CML HT MP
CML a
0.5
0.425 b
1.2
1.4
1.5
S (°C) 55 25 35
Tg (°C)
140 85
CTE1 40 30
CTE2 95
110
120

14. Accumulated Strain Energy Density
Accumulated creep strain energy density is lower for CRA compared to SAC305 in all simulated conditions. Lower energy >> Higher fatigue life
Accumulated creep strain energy density magnitude higher for LED on MCPCB compared to FR4 due to larger in plane CTE mismatch.
The higher temperature extreme of 125°C accumulates more creep energy compared to 80°C for SAC305 solder alloy per thermal cycle.
LED FR4
LED MCPCB/HT Dielectric
-40°C to 125°C
-40°C to 125°C
20°C to 80°C
20°C to 80°C

15. FEA Results: LED on MCPCB
Von-Mises stress and shear strain contours for LED on MCPCB with CML Dielectric. Stress distribution more uniform in anode and cathode pads.
Peak strain occurs at the corner of solder joints. Location of fatigue crack initiation.
SAC305: 80°C
CRA: 80°C
Stress
Stress
Shear Strain
Shear Strain

16. FEA Results: LED on FR4 SAC305 CRA
Accumulated creep strain energy contours for Luxeon Rebel LED on two substrates and solder alloys. -40°C to 125°C. Maximum solder creep strain occurs on ceramic side of the package.
End of second thermal cycle
Stress & strain at 125°C
cathode
SAC305
anode
CRA

17. Fatigue Life Predictions
LED on MCPCB from -40°C to 125°C for CRA and SAC305.
Fatigue life improvement due to CRA solder alloy overcomes contribution of dielectric material type.
FEA models does not account for dielectric influence as much as analytical model due to insufficient material properties at higher temperatures.
Analytical model shows largest improvement in fatigue life by combining soft dielectric on MCPCB with CRA over stiffer dielectric and SAC305.
FEA
Analytical

18. Fatigue Life Predictions
LED on MCPCB from 20°C to 80°C between CRA and SAC305.
Analytical model predicts fatigue life improvements of 5X with CRA compared to SAC305. FEA predicts fatigue life improvement of 2.5-3X over SAC305.
Analytical model accounts for empirical behavior correlated from test data. This is indicative to contribution of higher creep resistance at lower temperatures of CRA over SAC305.
Influence of dielectric in fatigue life improvement diminished over lower temperature fluctuations.
FEA
Analytical

19. Fatigue Life Prediction
LED on FR4 with SAC305 and CRA solder alloys for two thermal profiles.
Lower CTE mismatch between submount (8.2ppm/°C) and FR4 (17ppm/°C) compared with aluminum (24ppm/°C).
Under 20°C to 80°C profile CRA shows 4X fatigue life improvement over SAC305. Analytical model predictions more conservative than FEA. Real world fatigue data could fall between two prediction methods.
45,258
60,433
14,830
8728
767
4381
730
1567

20. Conclusion and Future Work
Fatigue life of LED solder interconnects can be greatly increased by selecting Creep Resistant Alloy over SAC305.
The benefit of solder alloy selection is further enhanced at lower operating temperatures in comparison to accelerated thermal cycling.
Fatigue life of LED on MCPCB greatly depend on the dielectric material properties. Lower modulus dielectric greatly increase fatigue life for MCPCB applications.
Fatigue life prediction models can account for the influence of solder alloy type, LED and substrate geometry.
Empirical and FEA models enable fast integration of LED design and process parameters (i.e. solder alloy type) on fatigue life predictions.
Thermal conductivity analysis will be included in future simulations to account for thermal gradient and heat sink influence on solder fatigue.

21. Thank you for your attention!
Questions?
Contact:
Maxim Serebreni Gyan Dutt
DfR Solutions Alpha Assembly Solutions
(301) (908)