transient liquid phase bonding for power electronic modulesstisrv13.epfl.ch/masters/img/238.pdf ·...
TRANSCRIPT
CH-RD. S2 Packaging & Reliability Power and productivity
for a better world TM abb
Motivation & Objectives Optimization of TLP Emitter Bonding
Theory & Model
Bonding technologies play an indispensable role in the assembly of power electronic (PE)
modules. With increased requirements for higher operation junction temperatures, longer life-
time and environmental considerations, transient liquid phase (TLP) bonding is becoming a
possible promising alternative packaging technology.
References: [1]. L. Feller et al., "Lifetime Analysis of Solder Joints in High Power IGBT Modules for Incresing the Reliablity for Operation at 150°C," 2009. [2]. O. Grant et al., "Overview of transient liquid phase and partial transient liquid phase
bonding," Journal of Materials Science, no. 5305-5323, p. 46, 2011. [3]. L. Deillon et al., "Growth of Intermetallic Compounds in the Au-In System: Experimental Study and 1D modelling," Acta Materialia, 2012.
Transient Liquid Phase Bonding
for Power Electronic Modules Quanbo Jiang
1, Franziska Brem
2, Chunlei Liu
2, Michel Rappaz
1
1Institute of Materials Science and Engineering, Swiss Federal Institute of Technology Lausanne (EPFL), 1015 Lausanne, Switzerland.
2ABB Swiss Corporate Research Center (CHCRC), 5405 Baden-Dättwil, Switzerland.
Contact: Quanbo Jiang via [email protected]
Conclusions & Outlook
Acknowledgements
AlSiC Baseplate
Substrate (ceramic insulator)
Metallization
Metallization
Solder
Chip
Silicone
gel Plastic
cover
Gate Al bond wire
Cooler
Transient liquid phase (TLP) bonding
is a process for joining components
via the dissolution of a solid-state
material into a liquid-state interlayer
and the formation of intermetallic
compounds (IMCs) by interdiffusion
of materials[2].
Assumptions of the model[3]:
Chemical potential varies linearly through
the IMC phases.
Sn is initially saturated with Cu.
No nucleation stage occurs.
Local equilibrium reached at every interface.
Gradient of chemical potential:
𝐽 = −𝑀𝐶𝜕µ
𝜕𝑥
Fick’s first diffusion law:
𝐽 = −𝐷𝜕𝐶
𝜕𝑥
Fick’s second law:
𝜕𝐶
𝜕𝑡= 𝐷
𝜕2𝐶
𝜕𝑥2
Arrhenius equation:
𝐷𝐵 = 𝐷𝐵0𝑒(−
𝐸𝑎,𝐵𝑅𝑇 )
Velocity of interface:
𝑣𝛼/𝛽 =𝑑𝑥
𝑑𝑡=
1
(𝐶𝐵𝑏−𝐶𝐵
𝑎)𝐷 𝛼
𝜕𝐶𝐵𝑎
𝜕𝑥− 𝐷 𝛽
𝜕𝐶𝐵𝑏
𝜕𝑥
𝐶𝑢6𝑆𝑛5 (𝜂) 𝐶𝑢3𝑆𝑛 (𝜀)
𝐶𝑢 𝑆𝑛
𝑆𝑛 𝜺 𝜼 𝐶𝑢 𝐶𝑆𝑛 𝜇
𝜇2
𝜇3
𝜇1
𝐶𝑆𝑛𝑐𝑢
𝐶𝑆𝑛𝑆𝑛
𝑣1∗ 𝑣2
∗
𝑣3∗
Parameters:
𝐽 : diffusive flux
𝑀 : atomic mobility
𝐶 : Concentration
µ : Chemical potential
𝐷 : Diffusion coefficient
𝐸𝑎,𝐵 : Activation energy
𝑣𝛼/𝛽: Velocity of interface
Supporting Experiment for Model
Variac Thermocouple
Ar flow
He flow
Chip
Chip
Top Metal Plate Leadframe
Motivation of leadframe emitter bonding by TLP:
Replace the wire bonds with full area bonds.
Improve for higher power density and higher
junction temperature.
Reduce the inductance.
Objectives:
Develop 1D finite model for the
growth of intermetallic
compounds (IMCs).
Determine diffusion coefficients
in Cu/Sn system.
Optimize process parameters for
TLP leadframe emitter bonding.
Reduce void percentage of the
bonding areas.
Analyse and improve the
properties of resulting joints.
Schematic structure of high power module package[1].
Si chip
Cu lead frame
Galvanic Cu
Sn foil
Cu metallization
Cu metallization AlN substrate
Si chip
Cu lead frame
Program profile
Pgm2:
Pgm2:
real
temperature
profile
Pgm3:
Pgm3:
real
temperature
profile
Pgm4:
Pgm4:
real
temperature
profile
cycles VP_L
(%) B1_S2
VP_R
(%)
VP_L
(%) B2_S2
VP_R
(%)
0 1.6 1.5 2.7 3.9
200 broken 5.6 10.8 24.5
400 broken 7.0 14.9 34.0
800 broken 17.1 33.2 broken
1000 broken 21.5 31.0 broken
0
10
20
30
40
50
60
70
80
90
100
0 500 1000
Vo
id p
erce
nta
ge (
%)
Cycles (time)
B1_S2_L
B1_S2_R
B2_S2_L
B2_S2_R
Properties Cu6Sn5 Cu3Sn Copper
Melting point (°C) 415 676 1083
Density (𝑔 𝑐𝑚3 ) 8.28 8.90 8.94
Hardness (HV) 417±36 343±12 155±20
Shear strength (MPa) 32.9±10.9 210.0
0
20
40
60
0 5 10 15 20 25
Shea
r st
ren
gth
(M
Pa)
Void percentage (%)
Shear strength (Mpa) VS Void percentage (%)
The work presented in this thesis was conducted at ABB CHCRC, Baden and LSMX, Lausanne. I
first would like to thank my supervisors, Prof. M. Rappaz, Dr. F. Brem and Dr. C. Liu, for their kind
encouragement and patient guidance throughout my thesis work. Furthermore, I would like to
thank Dr. L. Deillon for her help with the model and experiment at LSMX. Moreover, I am very
grateful to all the employees at ABB for cultivating a friendly and motivating working environment.
The schematic diagram of experimental
furnace for dipping Cu wire inside Sn bath is
shown above. The growth of IMCs was
measured at liquid state (300°C) and at solid
state (200°C) over time.
@300°C.
@200°C.
400s.
3600s.
6400s.
1h.
9h.
100h.
Factors of Sn in phases 𝐷0(𝑚2/𝑠) 𝐸𝑎(kJ/mol)
𝐶𝑢 5.1 × 10−9 103.5
𝐶𝑢3𝑆𝑛 4.1 × 10−8 92.0
𝐶𝑢6𝑆𝑛5 8.7 × 10−9 78.8
D of Sn in phases D @300°C(𝑚2/𝑠) D @200°C(𝑚2/𝑠)
𝐶𝑢6𝑆𝑛5 6.0 × 10−16 1.8 × 10−17
𝐶𝑢3𝑆𝑛 1.8 × 10−16 3.0 × 10−18
𝐶𝑢 2.0 × 10−18 2.0 × 10−20
The growth kinetics of IMCs (above) follows
parabolic law. The diffusion coefficients
(below) were obtained from the simulation by
fitting to the experimental results. The
temperature dependency was determined.
Advantages and disadvantages of TLP
emitter bonding process for PE modules:
+ The Cu-Sn system is suitable for
packaging of power electronic modules
and low void percentage joints can be
achieved.
+ Low pressure is applied for clamping the
joining parts without damaging chips.
+ Higher mechanical strength and melting
point than Sn and soft solders are reached.
+ Lower cost for raw materials than SAC
solder and other new bonding materials.
- High requirement for the surface finish,
flatness and cleaning process of joining
partners.
- Long process times with vacuum furnace.
- Uniform bond line thickness is difficult to
achieve.
Process parameters were improved for TLP
process during 6-month work at ABB
CHCRC. The results demonstrate that TLP
emitter bonding by is a promising and
reproducible process for replacing the wire
bonding and for other small-area bonding.
Si
½ hard Cu
Sputtered
Cu
𝐶𝑢3𝑆𝑛
𝐶𝑢3𝑆𝑛
𝐶𝑢6𝑆𝑛5 Sn
𝐶𝑢3𝑆𝑛
𝐶𝑢6𝑆𝑛5 𝐶𝑢3𝑆𝑛 Galvanic
Cu
Initial setting parameters in the furnace
Pgm1 2h holding time @320°C, vacuum @230°C, inner water cooling
Pgm2 2h holding time @320°C, vacuum @260°C, N2 cooling
Pgm3 2h holding time @320°C, vacuum @320°C, 5min waiting time, N2
pgm4 2h holding time @320°C, vacuum @240°C and start H2 flow, N2
The TLP process parameters for leadframe
emitter bonding including load on the chips,
heating ramp, and vacuum steps were
optimized. Cu joining partners with different
surface roughness, grain size and thickness
range were compared.
SAM, SEM, EDX, hardness, thermal shock
cycling and shear tests were used for
characterizing the joint properties.
Optimal
Results:
Pgm3
(profile)
Thinner
Galvanic Cu
plating
0.06MPa
(load)
Si
½ hard Cu
Sputtered Cu
Galvanic Cu
Program 2 Program 4
280°C 300°C
Cu fully comsumed,
Sn left
Residual Sn
0.02MPa 0.04MPa 0.06MPa 0.08MPa
Batch Number B7_G1_Left B8_S1_Left B9_G2_Left B11_G4_Left
Optical
Microscope
SAM images
Void
percentage best
samples (%)
26.4 2.0 1.9 2.4
Average void
percentage (%) 35.6 11.8 8.8 7.8
pressed-
out Sn
pressed-
out Sn
020406080
100120140160180200220240260280300
0 15 30 45 60 75 90 105 120 135 150 165 180 195 210
Tem
pe
ratu
re (°
C)
Time (min)
020406080
100120140160180200220240260280300320
0 15 30 45 60 75 90 105 120 135 150 165 180 195 210
Tem
pe
ratu
re (°
C)
Time (min)
020406080
100120140160180200220240260280300320
0 15 30 45 60 75 90 105 120 135 150 165 180 195 210
Tem
pe
ratu
re (°
C)
Time (min)
Impact of void percentages for thermal
shock resistance and shear strength.