Discrete Voids Formation in Flip-Chip Solder Joints due to Electromigration Using In-Situ 3D Laminography and Finite-Element Modeling

Transcription

1 Discrete Voids Formation in Flip-Chip Solder Joints due to Electromigration Using In-Situ 3D aminography and Finite-Element Modeling Yuan-Wei Chang 1,2*, Yin Cheng 1,3,. Helfen 1,3,4, C. Chen 2, K. N. Tu 5, and T. Baumbach 1,3 1 ANKA/IPS, Karlsruhe Institute of Technology (KIT), ermany 2 MSE, National Chiao Tung University, Taiwan (R. O. C.) 3 AS, Karlsruhe Institute of Technology (KIT), ermany 4 The European Synchrotron (ESRF), France 5 MSE, University of California, os Angeles (UCA), US KIT/ANKA-IPS Yuan-wie Chang 1

2 Outlines Introduction Electromigration (EM) reliability in flip-chip solder joints In-situ three-dimensional (3D) laminography FEM basing on 3D images Experimental aminography observation & processing of images Results & Discussion Discrete voids formation & coalescence Correlation between current density & voids distribution Conclusion KIT/ANKA-IPS Yuan-wie Chang 2

3 Economical Scale of Flip-Chip Packaging 23.5 Billion 32 Billion KIT/ANKA-IPS Yuan-wie Chang 3

4 Traditional In-Situ Observation Substrate Si Polish Solder joints Solder joints F. Ouyang, AP 91, (2007) 1. The surface oxidation 2. The change of heat dissipation 3. The relief of stress 4. The surface diffusion Z. Chen, JEM, 44(12), KIT/ANKA-IPS Yuan-wie Chang 4

5 3D aminography Imaging rotation axis specimen detector rotation stage X-ray computed tomography (CT) 1960s/1970s 3D imaging nondestructive testing (NDT) specimen rotation stage rotation axis detector Synchrotron-radiation computed laminography (SRC) flat, laterally extended objects with high spatial resolution. KIT/ANKA-IPS Yuan-wie Chang 5

6 FEM Basing on 3D Images SiC in Al (eo-material) N. Chawla, Scripta Materialia 51.2 (2004): N. Chawla, Acta Materialia 54.6 (2006): Ag 3 Sn in Sn (Electronics-material) R. S. Sidhu, Scripta Materialia 54.9 (2006): Super-austenitic stainless steel A. C. ewis, Scripta Materialia 55.1 (2006): Bainitic steels N. Osipov, Philosophical magazine (2008): Bones (Bio-material) P.. Young, PTRS-A: MPES (2008): Heart (Bio-material) Y. Zhang, Medical image analysis 16.6 (2012): KIT/ANKA-IPS Yuan-wie Chang 6

7 Specialties aminography observation High-resolution Real-time 3D Non-destructive Quantitative analysis FEM basing on 3D images Real electronic packages Very common solder (SAC: SnAgCu) KIT/ANKA-IPS Yuan-wie Chang 7

8 Sample Dimensions Plane view Al trace Bump 1 Bump μm e - UBM Ti 1000 Å / Cu 3000 Å / Cu 7.5 μm 270 μm in diameter 210 μm Solder Cu trace ROI Cross-sectional view Al trace Ti/Cu/Cu 7.5 μm e μm e - IMC Solder: SAC1205 Si chip 270 μm UBM Al trace Bump 1 Bump 2 Solder Cu trace 210 μm Substrate Region of interest (ROI) Solder Temp. 130 C Current Density SAC1205: (Sn, Ag 1.2%, Cu 0.5%, and Ni 0.05%) A/cm 2 KIT/ANKA-IPS Yuan-wie Chang 8

9 aminography (during EM) detector x rotation axis optis z y θ slits rotation stage goniometer system filters wiggler ID15A in ESRF in renoble, France. θ 65 continuously rotate by projections of pixels Exposure time = 600 ms. Effective pixel size = 0.84 μm Field of view = 0.86 mm 0.86 mm. Setups:. Helfen, Appl. Phys. ett. 86, (2005). Helfen, Appl. Phys. ett. 94, (2009). Helfen, Rev. Sci. Instrum. 82, (2011) KIT/ANKA-IPS Yuan-wie Chang 9

10 Procedures aminography before EM EM testing ( A/cm 150 C) aminography Stop EM Fail? SEM Experiment 3D Images reconstruction Images processing Quantitative analysis Void labeling & rowth tracing FEM construction Solving & Post-process developed by ESRF beamline scient Aligned images by multi-level squar Segmented images by local thresho Analyzing material changes & quant Tracing void growth by Hierarchical Submodeling in ANSYS Bottom-up construction by direct pi Solving & post-processing by ANSY Correlating material changes & curr KIT/ANKA-IPS Yuan-wie Chang 10

11 Procedures 3D Images reconstruction developed by ESRF beamline scientists Images processing Quantitative analysis Void labeling & rowth tracing FEM construction Aligned images by multi-level squaring difference method Segmented images by local threshold value method Analyzing material changes & quantity Tracing void growth by Hierarchical Tree structure Submodeling in ANSYS Bottom-up construction by direct pixel-to-element method Solving & Post-process Solving & post-processing by ANSYS (APD) Correlating material changes & current density KIT/ANKA-IPS Yuan-wie Chang 11

12 FEM Basing on 3D aminography 3D Images Alignment Image Segmentation abeling Materials File Conversion lobal Model Construction Statics Analysis of Void rowth Correlation between Current Density and Microstructure Evolution Submodel Construction Postprocessing Solution of lobal Models Solution of Submodel Direct Image-to- Element Method Objective Oriented Finite- Element (OOF) KIT/ANKA-IPS Yuan-wie Chang 12

13 Discrete Voids Formation initial 26 hr 103 hr 303 hr 540 hr void Void growing void 3D rendering of voids and solder top Discrete voids form and grow simultaneously during the entire EM testing. KIT/ANKA-IPS Yuan-wie Chang 13

14 Projection of Discrete Voids Testing Time (hr) UBM opening coalescence coalescence coalescence coalescence e - A new EM failure mechanism, discrete voids formation and coalescence, is found. 1. Many voids simultaneously nucleated and grew on the entire IMC/solder interface. 2. There was a higher probability for voids to form in the high current density region (the current crowding site). 3. The voids obviously coalesced each other after 47 hours of EM testing. 4. The coalescence caused the formation of large voids in the late stage of testing. e - coalescence KIT/ANKA-IPS Yuan-wie Chang 14

15 Coalescence Enhanced Void rowth The growth rate exponential n = 0.58 Sn diffusion dominated. DZ* = cm 2 /s Volume growth rate of all voids (μm 3 /hr) Volume growth rate per voids (μm 3 /hr) N total (green) = N new (black) N coalescence (red) The rate of new void formation was controlled by nucleation rate of voids magnitude of applied current The coalescence rate between voids depends on growth rate of voids density of voids (the number of voids) non-uniformity of void distribution KIT/ANKA-IPS Yuan-wie Chang 15

16 Current Density Distribution by FEM aminography 3D image Segmented 3D image Finite Element Model Current Density in Solder UBM 2.15 IMC 1.80 void 1.45 Solder A/cm 2 The finite-element model came from the 3D images by the direct pixel-to-element method. KIT/ANKA-IPS Yuan-wie Chang 16

17 Discrete Voids Formation does not Affect the lobal Current Density Distribution Stage# aminography 3D image Segmented 3D image Finite Element Model Current Density Distribution in Solder initial UBM Solder IMC Al trace 13 hr 26 hr 47 hr 78 hr 90 hr 103 hr 108 hr 303 hr 503 hr 540 hr void A/cm 2 KIT/ANKA-IPS Yuan-wie Chang 17

18 Discrete Voids Formation does not Affect the lobal Current Density Distribution Time initial 47 hr 103 hr 303 hr 503 hr 540 hr Current Density Distribution in Solder 10 4 A/cm KIT/ANKA-IPS Yuan-wie Chang 18

19 Differences between Pancake Void and Discrete Voids pancake void initiation in the literature voids initiation and evolution observed by 3D imaging ROI of Bump 1 ow Current Density High KIT/ANKA-IPS Yuan-wie Chang 19

20 Current Density ocal Current Crowding Induced Nucleation, rowth, or Coalescence High N N N N 0.0 hr a C N 13.0 hr C N C C 26.0 hr ow N C C C N 47.0 hr N: nucleation; : growth; C: coalescence; : local current crowding 77.5 hr KIT/ANKA-IPS Yuan-wie Chang 20

21 Probability of new void formation (%) Probability of new void formation (%) Correlation between Current Density and Discrete Voids Formation Current Crowding Region Correlation Coefficient = Plan View solder Cross-sectional View solder e - e - UBM trace KIT/ANKA-IPS Yuan-wie Chang 21

22 Correlation between Current Density and Discrete Voids Formation P void formation (%) = 100 E void in current stage E solder in previous stage Only when the current density was higher than the threshold value, the probability of void formation was affected by the current density. KIT/ANKA-IPS Yuan-wie Chang 22

23 Conclusion 3D aminography + FEM real electronic packages Non-destructive observation High quality 3D images Correlation between void growth and current density An EM failure mechanism, discrete voids formation & coalescence, is found. Discrete voids simultaneously nucleated and grew. The coalescence between voids started after 47 hours of EM testing. Voids coalesced each others into large ones in the late stage of testing. The discrete voids induced χ change of global current density distribution. local current crowding surrounding them. The correlation between the discrete voids formation and the current density distribution is high, and a threshold current density is observed. KIT/ANKA-IPS Yuan-wie Chang 23

24 Thanks for your attention. Contact Information Yuan-wei Chang: Yin Cheng: KIT/ANKA-IPS Yuan-wie Chang 24