Electromigration in Flip Chip Solder Joints

Transcription

1 Electromigration in Flip Chip Solder Joints K.N. Tu Dept. of Materials Science & Engineering, UCLA 1. Introduction 2. Why does electromigration in solder joint become a reliability problem? 3. Electromigration in flip chip solder joints Effect of current crowding Effect of eutectic composition Effect of composite solder joints 4. Summary Support : NSF,SRC Co-workers : Prof. Chang-yi Liu, National Central University, Taiwan ; Prof. Taek Yeong Lee, Hanbat National University, Korea; Dr. Everett Yeh, Intel, Santa Clara, CA; Dr. Woojin Choi, Intel, Chandler, AZ; Ms. Hua Gan, Mr. Albert Wu, Mr. Jong-ook Suh, Miss Emily Ou, Miss Annie Huang, UCLA

2 125 µm Si J = 6 ~8 x 10 3 amp/cm 2 T = 150 o C, t = 100 hrs Peter Elenius, Flip Chip Technologies, (1998)

3 Why does electromigration in solder joints become a reliability problem? Calculation of Diffusivities Surface D l = 0.5 exp (- 34 Tm/RT) D gb = 0.3 exp (-17.8 Tm/RT) D s = exp (-13 Tm/RT) T m : Melting point of Metal Lattice Grain boundary Not due to diffusion! LogD vs. T m /T for FCC metals (N.A. Gjostein, in Diffusion, edited by H. I Aaronson, American Society for Metals, (Metals Park, OH, 1973))

4 Different Diffusion Mechanisms Melting point ( o K) Temperature ratio 373 o K/Tm Diffusivities at 100 o C [373 o K] (cm 2 /sec) Surface D s =10-12 Al Grain Boundary D gb =6x10-11 Pb Lattice D l =6x10-13 Eutectic Pb Lattice D l =2x10-9 to 2x10-10 Electromigration in Al is dominated by grain boundary diffusion Electromigration in is dominated by surface diffusion Electromigration in solder is dominated by lattice diffusion Fast lattice diffusion is not the reason of EM failure in solder joints

5 Critical Product in Short Stripe D dσω D J = C + C Z * ee kt dx kt x Cathode If J=0, there is no net electromigration flux. σω = x * Z e Critical product ρ j E= Electric Field (E = ρj ) σ Ω (j x) critical = * Z eρ Al TiN Jσ Jem Anode If j x < (j x ) c No electromigration damage

6 Electromigration in Al short strips on TiN By Dr. Alexander Straub, MPI Stuttgart, Germany, July 1999 Line width :10µm, Line Thickness : 100nm, rrent density 10 6 A/cm 2, Temperature : 225 o C

7 Small Critical Product in Solders ( σ Ω j x ) c = * Z e ρ ( Y ε Ω j x ) c = * Z e ρ σ =Y ε, ε=0.2% at elastic limit Y(Gpa) Z * ρ(µω-cm) Solder ~30 ~30 ~22 Al ~69 2~4 ~2 ~ Y ε Ω ( j x ) x Z e ρ 3 c_solder = 5 x10 ( j ) * c_/al Constant Δx, Δx, the the current density needed to to fail fail solder is is 2~3 2~3 orders smaller than than that that needed to to fail fail Al Al or or If If Al Al or or fails fails at at to10 to A/cm 2 2,, solder will will fail fail at at or or A/cm 2 2

8 rrent crowding in a flip chip solder joint The design rule is 0.2A per bump, so for a bump of 100µm 100µm The average current density is A/cm 2 or A/cm 2 The unique geometry of a solder joint leads to current crowding

9 Tim Potential rve Resistance does not change much Si (c) (b) 6 Potential (V) (b) (c) (d) Void propagation 2 Al or Interconnects Solder Bumps Time (hrs) Crosssection Resistivity 0.5x0.2 μm μm cm 100x100 μm μm cm Resistance 1 to 10 ohm 10-3 ohm rrent 10-3 amp 1 amp rrent Density 10 6 A/cm to 10 4 A/cm 2

10 Unique characteristics of electromigration in a flip chip solder joint Typically, it is a two-phase eutectic alloy It has UBM on chip side and bond-pad on substrate side: - chemical force interacts with electrical force Flip chip solder joints are under thermal stress: - mechanical force interacts with electrical force Its geometry leads to current crowding Solder joint is low resistance as compared to Al and interconnects:- it is insensitive to resistance change and hard to detect failure by resistance change Pb-free vs. Pb-containing solders Thick UBM vs. thin film UBM Composite solder joint interdiffusion is an issue

11 Test Samples Flip chip solder joints (from Flip Chip Technologies) UBM: Thin film Al/Ni(V)/ Solder: Eutectic Pb and eutectic Ag Test condition: 125 to 140 C and 2.25 to 3.8 x 10 4 A/cm 2 V-groove solder lines Electrodes: wires or Ni wires Solder: Eutectic Pb, eutectic Ag and eutectic Ag Test condition: Eutectic Pb: 150 C and 2.8 x 10 4 A/cm 2, or RT and 5.7 x 10 4 A/cm 2 Eutectic Ag: 120 C to 180 C and 3.2 to 4x 10 4 A/cm 2 Eutectic Ag: 150 C and 3.0 x 10 4 A/cm 2

12 Sample Structure of Flip Chip Solder Joint (a) Side View (b) Top View 2nd cross section 1st cross section Fig. Schematic diagram of the electromigration test sample

13 Failure Mode in Eutectic-Pb Solder Joints (a) 37 hrs (b) 38 hrs contact window (c) (d) 40 hrs 43 hrs Sequence of the void propagation in e-pb solder joint at 125 o C, and A/cm 2.

14 Failure Mode in Eutectic Ag Solder Joints 0 hr 14 hrs J = A/cm 2 T = 140 o C

15 Effect of Solder Composition EM in eutectic Pb > eutectic Ag (wt %) e-pb 63% e-ag 97% 3.8x10 4 A/cm 2 at120 o C. (a) A eutectic Pb solder joint after 40 hours, (b) A eutectic Ag solder joint after 200 hours.

16 EM-driven IMC Formation in Eutectic Pb Solder Bump Si Chip conducting trace 0min 15mins 90mins PCB I = 1.27A, T = 100, 30mins e - 45mins dissolution 65 IMC e - e- open From professor R.Kao in National Central Univ. in Taiwan e-

17 EM in Composite Pb-Containing Solder Joints (a) Before current stressing (b) After 3 hr J= A/cm Temp. of Si Backside : 155 o C From KAIST (c) After 12 hr (d) After 20 hr

18 Three kind of UBM for high-pb 1. IBM : Cr/Cr-(phased in)/ 2. Delco: Al/Ni(V)/ 3. KAIST: Al/TiW/ (thick, 5 µm) 97Pb3 Chip

19 Substrate side 37Pb63 Substrate Solder Mask or Ni

20 Electromigration Test Sample Interconnect Chip UBM 97Pb/ electron flow, 37Pb/ Solder resistor 97Pb/ underfill Substrate side Pad Substrate

21 Simulation Al line TiW 3 97Pb3 Solder Cross sectional Structure rrent density distribution (A/cm 2 )

22 The magnified cross-sectioned SEM images of cathode side in the downward electron bump (a) Before current stressing (b) After 3 hr From KAIST (c) After 12 hr (d) After 20 hr

23 Proc. 6 th Internl. Vacuum Congr Japan. J. Appl. Phys. Suppl. 2, Pt. 1, 1974 Room Temperature Interaction in Bimetallic Thin Film Couples King-Ning Tu and Robert Rosenberg IBM thomas J. Watson Research Center Yorktown Hights, New York Table. Intermetallic compounds formed at room temperature Ag Au Ni Pd Pt Pb - - AuPb 2 - PdPb 2 PtPb Ag 3 Au 4 Ni 3 4 Pd 4 Pt 4, Ag, Au, Ni, Pd, Pt Si, Ge,, Pb Interstitial Diffusion

24 Electromigration- Driven Phase Separation J= A/cm Eutectic Pb Sample before EM T=160 o C T=82hrs Pb From ASE, Taiwan

25 Effect of Eutectic Composition Pb- Ni- α β A B No chemical potential gradient as a function of composition below the eutectic temperature.

26 Sample Preparation Si V-groove Si (001) 100 µm <110> Electrode Electrode SiO 2 (0.12 µm) 69.4 µm Cr or Ti (.05 µm) (2µm) or Ni/Au (0.3µm/0.05µm) Solder line Si Wafer Electrode Solder: Eutectic Pb Pb Eutectic Ag Eutectic Ag Ag Electrode: or or Ni Ni wire wire Reflow Reflow temperature: ~ C Reflow Reflow time: time: ~1min ~1min

27 Typical Sample View Solder V-groove sample after reflow Polished sample ready for EM test Ni Si wafer Solder line Ni Ni Si wafer Solder line Ni Width of of solder line: line: ~ 100µm Length of of solder line: line: ~ µm Ni Ni Cathode Anode

28 Experiment Setup DC Power Supply Advantage of of V-groove sample: --High High current current density density --Easy Easy to to observe --No No expected current current crowding Cathode - Si Substrate Solder + Anode Electrode Furnace Experiment temperature : 120ºC, 150ºC, 180ºC rrent Density: ~ 10 4 A/cm 2

29 EM in E-Pb V-groove Lines Pb rich rich 2.8 x 10 4 A/cm o C 8 days 5.7 x 10 4 A/cm 2 Room Temp 12 days Volume of mass transport V J em em = = C Ω J D kt em Z At * ee Calculated Z*=33 at at high T. T. Reported value for for bulk bulk Pb Pbis is Calculated Z*=39 at at room T. T. Reported value for for bulk bulk is is

30 Redistribution of Pb and Redistribution of Pb for EM at 150 o C After EM Anode Cathode 151 C 210 Pb 113 Redistribution of for EM at RT concentration (wt %) Pb C Anode length (um) Cathode D.Gupta, K. Vieregge, and W. Gust, Acta Mater., 47, 5-12 (1999)

31 Electromigration J = amp/cm 2 T = 150 º C t = 96 hrs E-Pb J = amp/cm 2 T = 180 º C t = 96 hrs E-Ag J = amp/cm 2 T = 150 º C t = 72 hrs Ni E-Ag Ni 100µm

32 Polarity Effect of EM on IMC Thickness 3.2x10 4 A/cm 2, 180 º C IMC Solder 10hr Cathode Solder From Dr. Hua Gan, UCLA Solder IMC 10hr Anode

33 EM Effect on Thickness Change of IMC Anode (3.2x10 4 A/cm 2, 180 o C) Cathode (3.2x10 4 A/cm 2,180 o C) (a) Solder Anode (c) Solder Cathode hr 0 hr hr (b) Solder Anode (d) Solder Cathode Anode 10hr 10hr 10hr Cathode

34 EM Effect on Thickness Change of IMC (cont.) Anode (3.2x10 4 A/cm 2, 180 o C) Cathode (3.2x10 4 A/cm 2,180 o C) (a) Solder Anode (c) Solder Cathode hr 21 hr (b) Solder Anode (d) Solder Void Cathode Anode 6 5 Cathode 3 87hr 3 87hr

35 EM in Eutectic Ag Solder Lines with Ni Electrodes Solder Anode Solder Void Cathode Ni 3 4 Ni 3 4 Anode Ni hr 10µm Ni 234 hr 10µm Cathode 6 Anode side Cathode side Thickness, um J= A/cm 2 T=150 ºC Time, hrs

36 Summary I In the flip chip solder joints, failure is due to the void propagation. Void forms at the cathode near the corner of current entrance into the solder bump. There is no chemical potential gradient force against phase separation and fast IMC formation in eutectic alloys. Diffusion of Pb and dominates at temperature above 120 o C and below 120 o C respectively. EM enhances IMC formation at anode and inhibits it at cathode.

37 Summary II Flip Chip Solder Joints vs. V-groove Solder Lines: 1. No current crowding is expected in V-groove solder lines. 2. It is easy to observe the polarity effect at anodes and cathodes in V-groove solder lines. 3. It is easy to get the redistribution of species at different depth after EM in V-groove solder lines. Eutectic Pb Solder vs. Eutectic Pb-free Solder: 1. Failure modes of eutectic Pb and eutectic Ag flip chip solder joints are the same. 2. Eutectic Pb-free solders have longer lifetime. Electrodes vs. Ni Electrodes: 1. Both and Ni dissolve into solder due to electrical force and chemical force. 2. The reactions between Ni electrodes and Pb-free solders are slower compare with electrodes.