© 2001 Amkor Technology, Inc. Ahmer Syed/ TMS 2001
Overview of Reliability Models andData Needs
Ahmer SyedAmkor Technology
Workshop on Modeling and Data Needs for Lead-Free Solders
Sponsored by NEMI, NIST, NSF, and TMS
© 2001 Amkor Technology, Inc. Ahmer Syed/ TMS 2001
Outline
Failure Mechanisms Related to Solder Joint
Life Prediction Model Requirements
Lessons Learned from Sn/Pb– Life Prediction Models– Material Behavior– Stress Analysis Approach– Test Data
Data Needs for Pb Free Solder
© 2001 Amkor Technology, Inc. Ahmer Syed/ TMS 2001
Failure Modes & MechanismsRelated to Solder Joint
Failure Modes– Failure in Bulk Solder– Failure at Intermetallic Layer– Trace Failures– PCB Failures
Failure Mechanisms– Temperature Related: T, dT/dt, ∆T– Displacement Related: ∆D– Acceleration: G, Grms
1
3
2
4
Component
Board
Laminate
Solder MaskTrace
1 2
Component
1
2
3
4
© 2001 Amkor Technology, Inc. Ahmer Syed/ TMS 2001
Causes of Failures
Thermal/Power Cycling– CTE Mismatch, ∆T, dT/dt, Tmax, Tmin, Time @ Tmax and Tmin
Failure in Bulk Solder - Creep-FatigueFailure at Intermetallic - Overstress
PCB Bend, Cyclic Bend, Vibration– Relative Displacement Between Package & Board
Failure in Bulk Solder - Fatigue, Creep RuptureFailure at Intermetallic Layers - OverstressTrace & PCB Failures - Solder Alloy/Intermetallic Strength
Shock & Drop– High Gs, Large Displacements
Failure at Intermetallic Layers - OverstressTrace & PCB Failure - Solder Alloy/Intermetallic Strength
Ball Shear– Intermetallic or Bulk Solder
How Well Can We Predict?
© 2001 Amkor Technology, Inc. Ahmer Syed/ TMS 2001
Life Prediction Model Requirements/Steps
LoadProfile
MaterialBehavior
Analysis- Analytical- FEA
Stress,Strain,Energy Density
Fatigue Test
ComponentDescription
Failure Definition&
Failure Data
Life Prediction Model
Model Validation
Predictions for NewDesigns and Conditions
© 2001 Amkor Technology, Inc. Ahmer Syed/ TMS 2001
Sn/Pb Solder Fatigue Life Prediction Models
Early Attempts (mid to late 80s)– Traditional Coffin-Manson Eqn
– Isothermal Mechanical Fatigue– Plastic Strain Range Controlled
TestsTemperature ModificationFrequency Modification
– Very Little Data on Real SolderJoints
Wen & Ross, ASME EEP-9Investigator C k Remarks
0.52 0.68 Torsion1.18 0.46 Lap Shear0.16 0.30 Tensile Joint0.6 0.39 1/15 CPM Shear Joint
0.565 0.30 5 CPM Shear JointShine 0.19 0.53 1 Hz, Tensile
0.1538 0.415 0.5Hz0.24 0.41
Enke 0.26 0.52Gua/Cutionco 0.34 0.49Solomon 1.32 0.52Guo/Conrad 3.00 0.70 TensileKluizenaer 0.39 0.51 Tensile
1.3 0.637 strain rate 0.1/sec4.72 0.653 4 x 10-4/sec
10.12 0.643 1 x 10-5/sec
Coomb
Wild
Kitano
Aldrich
kfp NC −=∆ε
© 2001 Amkor Technology, Inc. Ahmer Syed/ TMS 2001
Sn/Pb Solder Fatigue Life Prediction Models
Models Incorporating Time & Temperature Dependent Behavior ofSolder (Mostly Analytical Treatment)– Damage Integral Method (Subrahmanyan et al, CHMT 1989)
Stress Based
– Energy Partitioning Approach (Dasgupta et al, ASME, EEP, 1993)Elastic + Plastic + Creep
– Fracture Mechanics Based (Pao, CHMT 1992)
– Matrix Creep Model (Shine & Fox, ASTM STP 942)Isothermal Test DataCalculated Creep Strain
– CSMR Model ( Clech et al, 43rd ECTC)Analytical ModelInelastic Strain Energy
© 2001 Amkor Technology, Inc. Ahmer Syed/ TMS 2001
Sn/Pb Solder Fatigue Life Prediction Models
Energy Density Based (Darveaux et al, Ball Grid Array Technology, Ed. J. Lau)
– Crack Initiation & Growth– Inelastic Constitutive Eqn– Finite Element Analysis
10 -210 -310 -410 -510 -510 -410 -310 -210 -110 010 110 210 310 410 510 6
99C129C133C27C100C68CMASTER
τ
γ
/G
(T/G
) ex
p(Q
/kT
) (
K/s
ec/p
si)
o
© 2001 Amkor Technology, Inc. Ahmer Syed/ TMS 2001
Sn/Pb Solder Fatigue Life Prediction Models
Partitioned Creep StrainBased (Syed, 1996 SEM)– Wong et al Constitutive
Eqn (CHMT, 1989)
two mechanisms
– Finite Element Analysis
10
100
1000
10000
100000
10 100 1000 10000 100000
Cycles to Mean Failure (Test)
Cyc
les
to M
ean
Fai
lure
(P
red
icti
on
s)
PBGAs
LCCCs
QFPs
Flip Chips *
PBGAs - Motorola
TSOPs *
Master
2X Above
2X Below
25% Above
25% Below
* Unpublished Test DataTest Data & Predictions Scaled by the Same Factor
Open Symbols : Model DevelopmentClosed Symbols : Model Validation
( ) 1063.002.0 −Ε+Ε= MCGBSf xxN
1.E-14
1.E-12
1.E-10
1.E-08
1.E-06
1.E-04
1.E-02
1.E+00
1.E-06 1.E-05 1.E-04 1.E-03 1.E-02
Normalized Stress (σ/E)
Ste
ady
Sta
te C
reep
Str
ain
Rat
e (1
/s)
-50 C
0 C
25 C
75 C
125 C
© 2001 Amkor Technology, Inc. Ahmer Syed/ TMS 2001
Material Property Characterization
Stress-Strain ( Shi et al, JEP, 1999)
Stress-Strain
Ductility
Modulus
Strength
© 2001 Amkor Technology, Inc. Ahmer Syed/ TMS 2001
Material Property Characterization
Creep Behavior– Aldrich & Avery, Kashyap & Murty, Grivas, Mohamed & Langdon, Lam et
al, Arrowood and Mukherjee, and others– Hall, Solomon, Wilcox, Wong, Shine & Fox, Darveaux, Busso, Hong, and
others
10 410 310 210 110 -10
10 -910 -8
10 -710 -6
10 -5
10 -410 -3
10 -210 -1
10 0
Shear Stress (psi)
Stea
dy S
tate
Str
ain
Rat
e (1
/sec
)
27C
67C
100C132C
Bulk Solder (Wong et al Model)Real Joints (Darveaux)
© 2001 Amkor Technology, Inc. Ahmer Syed/ TMS 2001
Material Property Characterization
Mechanical Test Fixture for Creep Test of Real Joints– (Darveaux et al, Ball Grid Array Technology, Ed. J. Lau)
Double Lap Shear Fixture
Steel Rods CeramicSubstrates
SolderJointArray
Adhesive
Tensile Fixture
LVDT
StainlessSteelGrips
© 2001 Amkor Technology, Inc. Ahmer Syed/ TMS 2001
Stress Analysis
MC : 14 ppm/oC
Silicon : 2.6 ppm/oCBT/Cu Composite ~ 16 ppm/oC
PCB : 15 - 18 ppm/oC
25 to -40oC
Analytical Models– CTE Mismatch– Pure Shear
Finite Element
25 to 125oC
© 2001 Amkor Technology, Inc. Ahmer Syed/ TMS 2001
Finite Element Modeling
3-Dimensional Models
Inelastic Constitutive Models
Accurate Loading Conditions
Multiple Responses– Stress, Strain, Energy Density
© 2001 Amkor Technology, Inc. Ahmer Syed/ TMS 2001
Failure Data
Failure Definition - Electrical Open
Thermal Cycle Fatigue Test Data– Different Cycling Conditions– Test Board Variables– Component Design Variables
1000.00 10000.001.00
5.00
10.00
50.00
90.00
99.00
Cycles to Failure
Cum
ulat
ive
Fai
lure
s (%
)
WeibullTC1_20 mils Brd
W2 RRX - SRM MED
F=11 / S=23
β1=12.80, η1=3161.34, ρ=0.96
TC1_62 mils Brd
W2 RRX - SRM MED
F=38 / S=1
β2=7.30, η2=1968.64, ρ=0.97
1000.00 10000.001.00
5.00
10.00
50.00
90.00
99.00
Cycles to Failure
Cum
ulat
ive
% F
aile
d
WeibullSn-Pb_TC1W2 RRX - SRM MED
F=12 / S=3
β1=10.40, η1=3164.00, ρ=0.98
Sn-Pb_TC3W2 RRX - SRM MED
F=14 / S=1
β2=12.95, η2=6194.94, ρ=0.97
© 2001 Amkor Technology, Inc. Ahmer Syed/ TMS 2001
Solder Joint ReliabilityTemperature Cycle Test Data
wsCSP– 54 Lead Center Pad,
9x11 mm– Wafer Thickness
45% and 60%Reduction in Life withWafer Thickness of0.5 and 0.625 mm
– Ball SizeMounted Height <1mm for 0.33mm Balls30% Improvement inFatigue Life with0.45mm Solder Balls 1.0
5.0
10.0
50.0
99.0
500.0 5000.0
54 Lead wsCSP, 20 mils Boards, TC2 Condition
Cycles to Failure
Cum
mul
ativ
e %
Fai
led
0.33ball/0.35waferP=2, A=RRX-S F=9 | S=29
β1=14.5, η1=3279.6, ρ=0.8
0.33ball/0.50waferP=2, A=RRX-S F=30 | S=0
β2=7.8, η2=2505.9, ρ=1.0
0.33ball/0.625waferP=2, A=RRX-S F=45 | S=0
β3=7.7, η3=1508.9, ρ=1.0
0.45ball/0.50waferP=2, A=RRX-S F=43 | S=1
β4=11.0, η4=2910.2, ρ=1.0
© 2001 Amkor Technology, Inc. Ahmer Syed/ TMS 2001
Life Prediction Model Correlationfor wsCSP
Predictions within 25% Except for 2 Cases
Same Trend Predicted as Observed from Tests
0
1000
2000
3000
4000
5000
6000
7000
1 2 3 4 5 6 7 8 9 10
Cyc
les
to M
ean
Fai
lure
TestPredictions
Ball Size (mm) 0.33 0.33 0.45 0.33 0.33 0.33 0.45 0.33 0.33Wafer Thk. (mm) 0.35 0.50 0.50 0.625 0.35 0.50 0.50 0.625 0.50PCB Thickness (mm) 0.50 0.50 0.50 0.50 1.60 1.60 1.60 1.60 0.50Temp Cycle TC2 TC2 TC2 TC2 TC2 TC2 TC2 TC2 TC1
© 2001 Amkor Technology, Inc. Ahmer Syed/ TMS 2001
1721
768
2738
983
913
1420
664
1568
734
1750
750
550
1578
630
0
500
1000
1500
2000
2500
3000
FlexB
GA 132,
6.4 m
m Die
FlexB
GA 132,
9.5 m
m Die
FlexB
GA 144,
3.65 m
m Die
FlexB
GA 144,
7.4x8
.3 mm D
ie
FlexB
GA 144,
9.6 m
m Die
FlexB
GA 160,
7.2 m
m Die
FlexB
GA 160,
8.9 m
m Die
FlexB
GA 180,
4.45 m
m Die
FlexB
GA 180,
9.25 m
m Die
Fai
lure
Fre
e L
ife
(Cyc
les) Estimated, 75% of Predicted CMF
Measured First Failure
Solder Joint Reliability PredictionPrediction Vs. Measured
© 2001 Amkor Technology, Inc. Ahmer Syed/ TMS 2001
Realistic Realistic - Excessive Excessive
Condition Specified Reliability RequirementsChamber
Zones1 Year Life 5 Years Life
0 to 100 C Single 180 900 1500 Cycles -25 to 100 C Single 125 625 700 Cycles -40 to 100 C Single 120 600 800 Cycles -40 to 125 C Single 90 450 500 Cycles -40 to +85 C Dual 130 650 300 - 500 Cycles -40 to 100 C Dual 90 450 800 Cycles -40 to 125 C Dual 70 350 500 Cycles -55 to 125 C Dual 60 300 300 Cycles
Realistic Reliability Requirements
Solder Joint Reliability PredictionField Conditions
Application : Cell PhoneAssumed Worst Case Field Conditions
Sales Person; May - October : Arizona, November - April : AlaskaArizona Cycling : +20 to +55 C, 6 Cycles/Day, 1000 Cycles in 6 MonthsAlaska Cycling : -20 to +20 C, 6 Cycles/Day, 1000 Cycles in 6 Months
Required Life/Year : 2000 + 20% = 2400 Cycles
© 2001 Amkor Technology, Inc. Ahmer Syed/ TMS 2001
1000
1400
2000
800
1300
1000
1450
0 500 1000 1500 2000 2500
TC1
TC2
TC3
TC4
TC5
TC6
TC7
Tem
per
atu
re C
ycle
Co
nd
itio
n
Relative Cycles to Failure
-40 to 125, 15 minutes Ramps and Dwells, 1 Hr Cycle
-40 to 100, Dual Zone, 1 Hr Cycle
-40 to 100, 15 minutes Ramps and Dwells, 1 Hr Cycle
-40 to 125, Dual Zone, 30 minutes Cycle
0 to 100 C, 10 minutes Ramps 5 Minutes Dwells, 30 minutes Cycle
-25 to 100, 15 minutes Ramps and Dwells, 1 Hr Cycle
-40 to 85, Dual Zone, 30 minutes Cycle
Single Zone : Slow RampsDual Zone : Fast Ramps (2-3 Sec Transfer), Steady State at Board Level within 2-3 minutes
Solder Joint Reliability PredictionTest Condition Comparison
© 2001 Amkor Technology, Inc. Ahmer Syed/ TMS 2001
Cyclic 3-Point BendingPCB Strain vs. Life– 12mm-132 lead fleXBGAs– 0.85mm thick Board (h)– Measured Strain Level
400
800
1200
1600
1,000 10,000 100,000
Cycles to Failure
PC
B S
trai
n L
evel
(µε
)
© 2001 Amkor Technology, Inc. Ahmer Syed/ TMS 2001
3-Point Bend Cycle Simulation
Bend Cycle Fatigue
0
20
40
60
80
0 0.1 0.2 0.3
Strain
Str
ess
(MP
a)
Darveaux’ Data
0
0.005
0.01
0.015
0.02
0.025
0.03
0 0.2 0.4 0.6 0.8 1 1.2Time (second)
Str
ain
EPPLEQVTotalEPEQ
0.0001
0.001
0.01
0.1
1
1000 10000 100000
Cycles to Mean Failure (Test)
Str
ain
(S
imu
lati
on
s)
EPPLEQV (Range)Total
EPEQ (Accumulated)
fNf = 42.66(EPEQ)-1.09
Thin Brd3mm 2mm
ThkBrd.2mm
© 2001 Amkor Technology, Inc. Ahmer Syed/ TMS 2001
What is Needed for Pb Free Solder
LoadProfile
MaterialBehavior
Analysis- Analytical- FEA
Stress,Strain,Energy Density
Fatigue Test
ComponentDescription
Failure Definition&
Failure Data
Life Prediction Model
Model Validation
Predictions for NewDesigns and Conditions
© 2001 Amkor Technology, Inc. Ahmer Syed/ TMS 2001
What is Needed for Pb Free Solder
Material Charaterization– Stress-Strain Behavior
strain rates dependent, andtemperature dependent
– Ductility & Strength– Temperature Dependent
Modulus– Temperature Dependent
Inelastic BehaviorCreep & Stress RelaxationStress to Rupture
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.5 1 1.5 2 2.5S
n/P
b
Sn
/Ag
Sn
/Cu
Sn
/Ag
/Cu S
n/B
i
-40 to 125oC Cycle Range
Pb Free Alloys In ConsiderationHave Homologous Temperature of ~ 0.5 at -40oC
MaterialBehavior
© 2001 Amkor Technology, Inc. Ahmer Syed/ TMS 2001
What is Needed for Pb Free Solder
Of all Pb free Alloys, Sn/Ag hasbeen Characterized the most– Not as much as Sn/Pb
Very Little data on other alloys– Recent data on Strength and
Ductility on Sn/Ag, Sn/Cu,Sn/Ag/Bi,Sn/Ag/Cu by Xiao et al (J.of Electronic Materials, 2000)
– Time &Temperature dependentmaterial behavior is of mostImportance.
10 410 310 210 -10
10 -9
10 -8
10 -7
10 -6
10 -5
10 -4
10 -3
10 -2
10 -1
Shear Stress (psi)
Stea
dy S
tate
Str
ain
Rat
e (1
/sec
)
132C
80C
27C
Steady State Creep Data on Sn/Ag by Darveaux et al
(Ball Grid Array Technology, Ed. J. Lau)
MaterialBehavior
© 2001 Amkor Technology, Inc. Ahmer Syed/ TMS 2001
What is Needed for Pb Free Solder
Bulk versus Joint Behavior– Data from bulk solder samples might not be directly applicable to solder joints
Constraining effect of the solder / substrate interfaces,
Precipitation strengthening from dispersed intermetallics, and
Difference in grain structure, grain size, or grain / specimen size ratio.
– Data from Real Solder Joint Samples is Preferred
10 510 410 310 210 -9
10 -8
10 -7
10 -6
10 -5
10 -4
10 -3
10 -2
10 -1
10 0
Darveaux - 60Sn40PbDarveaux - 62Sn36Pb2AgDarveaux - 60Sn40Pb S->TDarveaux - 62Sn36Pb2Ag S->TSubrahmanyan - 60Sn40Pb S->TSkipor - 63Sn37PbKashyap - 62Sn38Pb 28.4umKashyap - 62Sn38Pb 9.7umSchmidt - 62Sn38Pb
Tensile Stress (psi)
Stea
dy S
tate
Str
ain
Rat
e (1
/sec
)
Joint
Bulk
ture
Steel Rods CeramicSubstrates
SolderJointArray
Adhesive
Tensile Fixture
StainlessSteelGrips
MaterialBehavior
© 2001 Amkor Technology, Inc. Ahmer Syed/ TMS 2001
What is Needed for Pb Free Solder
Analysis Tools and Methodologies are in Place
Constitutive Equations Need to be developed– Consideration must be given on how to implement a particular
constitutive Equation in FEA Software packages.Provide guidelines or User subroutines
Analysis- Analytical- FEA
© 2001 Amkor Technology, Inc. Ahmer Syed/ TMS 2001
What is Needed for Pb Free Solder
Sn/Pb vs. Sn/Ag/Cu (fleXBGA Package)
No Difference in two Sn/Ag/Cu Compositions– Sn/Ag/Cu Better than Sn/Pb
25% for -55 to 125oC Cycle80% for 0 to 100oC Cycle
1000.00 10000.001.00
5.00
10.00
50.00
90.00
99.00
Cycles to Failure
Cum
ulat
ive
% F
aile
d
WeibullSn-3.4Ag-0.7Cu
F=25 / S=5
β1=16.07, η1=2886.19, ρ=0.98
Sn-4.0Ag-0.5Cu
F=10 / S=20
β2=19.28, η2=2809.33, ρ=0.91
Sn-Pb
F=30 / S=0
β3=15.16, η3=2409.32, ρ=0.99
2000.00 20000.001.00
5.00
10.00
50.00
90.00
99.00
Cycles to Failure
Cum
ulat
ive
% F
aile
d
WeibullSn-3.4Ag-0.7Cu
F=11 / S=4
β1=13.95, η1=9983.61, ρ=0.99
Sn-4.0Ag-0.5Cu
F=9 / S=6
β2=15.70, η2=10368.55, ρ=0.84
Sn-Pb
F=14 / S=1
β3=12.95, η3=6194.94, ρ=0.97
-55 to 125 C Cycle 0 to 100 C Cycle
FailureData
© 2001 Amkor Technology, Inc. Ahmer Syed/ TMS 2001
What is Needed for Pb Free SolderEffect of Package Type
PBGA– 2X Higher Life for Sn/4.0Ag/0.5Cu (A14) Compared to Sn/Pb
fleXBGAs– 25% Higher Life for A14
20 Lead LCCCs– NCMS TMF Test
-55<>125oC, 70 minute Cycle– 2X Reduction in Life for A14!
Performance is Highly Dependent on Package Type– Solder Deformation Behavior is a Strong Function of Stress, Strain
Rate, and Temperature
Sn/Ag/Cu More Creep Resistant at Low Stresses, Less CreepResistant at High Stresses!
Will a Ceramic Component Soldered with Sn/Ag/Cu Performworse than Sn/Pb in Actual Field Conditions?
0
2000
4000
6000
8000
PBGA fleXBGA LCCC
Package Type
Mea
n L
ife
(Cyc
les) Sn/Pb
Sn/4.0Ag/0.5Cu
FailureData
© 2001 Amkor Technology, Inc. Ahmer Syed/ TMS 2001
What is Needed for Pb Free SolderEffect of Test Conditions
Acceleration Factors Depend onAccelerated Test Condition & Alloy
– Different for each Alloy– -40<>125C 0<>100 C
Sn/Pb: 2X Higher LifeSn/Ag/Cu: 3.5X Higher Life
Field Conditions Much More Benignthan Accelerated Test Conditions
A Package-Alloy CombinationPerforming Worse in Accelerated TestCondition May Actual Perform Same orBetter in Field Conditions
Performance Comparison from OnlyOne Accelerated Test MaybeMisleading– At Least two test conditions should be
used
0
2000
4000
6000
8000
10000
12000
Sn/Pb Sn/Ag/Cu (A14)Alloy
Mea
n L
ife
(Cyc
les) TC1
TC3
Acceleration Factors from TC3 to TC1
0
1
2
3
4
5
6
A1 A11 A14 A21 A66 B63Alloy
Acc
eler
atio
n F
acto
r
FailureData
© 2001 Amkor Technology, Inc. Ahmer Syed/ TMS 2001
Life Prediction for Pb Free SolderMaterials need to be characterized for time and temperaturedependent behavior– Creep deformation will still play a dominant role for temperature cycle
failures– Time independent plasticity more relevant for vibration and other high cycle
fatigue simulation– Data from realistic joint samples is more useful
Temperature cycle data on real components is needed– Isothermal fatigue data is not useful for life prediction model development– Publish as much as you can, don’t normalize– Use multiple cycling conditions & components
Modeling Techniques Exist– Easy implementation of Constitutive Equation in FEA software is the key
Guidelines or user subroutines should be provided for complex stress-strainbehavior