Molded Electronic Package Warpage Predictive

Modelling Methodologies

Ong, Kang Eu Ron W. Kulterman

Wei Keat Loh Flex Ltd, Austin Tx.
Intel Technology SdnBhd USA
Malaysia [email protected]
[email protected]
[email protected]

Chih Chung Hsu Haley Fu

Jenn An Wang iNEMI
CoreTech System (Moldex3D) Shanghai
Taiwan [email protected]
[email protected]
[email protected]
 iNEMI: Warpage Characteristics of
Organic Packages, Phase 4

Phase 4 (2016-2017)
•Continue
establishing
Phase 3 (2015-2016) technology package
•Continue dynamic warpage of
establishing new package
Phase 2 (2013-2014) technology package technology (all
•Establish current dynamic warpage kinds)
technology package (all kind) •Leverage previous
Phase 1 (2011-13) dynamic warpage •Dynamic warpage effort to study the
•Literature survey (POP, PBGA, measurement impact of Low
•Establish metrology FCBGA) metrology Temperature Solder
iNEMI Technology
correlation between •Measurement assessment. on dynamic warpage
Roadmap 2013 requirement
sites and increase protocol (effect of •(Performing)
•Identify package awareness “As Is”, “Bake” and •Collaborate with
warpage as key •Reaching out to Moisture Exposure simulation software
challenge industry for Time (MET)) and to derive modeling
component donation sample size needed. approach in better
•(Forming & •(Norming) predict the dynamic
Storming) warpage
 Project Scopes in Phase 4 SOW

Characterization of latest packaging technology

• Characterize emerging electronic packaging technology dynamic warpage behavior to develop a

better understanding of the current development of package construction and material
development as listed in Table I.
• Silicon Interposer and Embedded Silicon Bridge with different package stiffeners and
constructions for Heterogeneous Packaging Solution
• Next generation of POP packages that leverages wafer level process for package construction
which include Panel and Wafer level molding. These also include the use of new fiber reinforced
mold material for warpage control.
• System In Package/Multi Chip Package (BGA) with different configuration and layout.
• Embedded Package (embedded silicon, actives and passives)

Assess the impact of Low Temperature Solder (LTS) on package dynamic

warpage requirement

• To assess the impact of lower temperature solder on package warpage for those packages
collected in Phase 2 and 3.
• To establish the risk level based on package technology with respect to warpage only.

Assessment the use FEA in optimizing package warpage

• Establish modeling optimization approach and tools requirement to enable higher accuracy
prediction technique
• Compare modeling with experiment data and identify potential gaps for further development in FEA
technique

3
 Content

• Introduction
• Implementation details
• Results and discussions
• Conclusions

Disclaimer: The work here covers some feasibility study which collaborated
with iNEMI and we are not endorsing any software in particular.

4
 Introduction: Multi-physics in Assembly
Process
Incoming Materials Electronic Package SMT Assembly
Assembly

•Package Design – •Process Steps •Reflow

manufacturing variation •Temperature exposure Temperature/Transient
•Substrate warpage •Time effect (Visco and Thermal exposure and
variation creep) distribution
•Die Warpage •Mechanical loading •Mechanical interaction
•Raw materials (different •Board level SJR
material characteristics) •Yield

Chip Attach Encapsulation Lid Attach (if Ball Attach Test and Final
& Molding applicable) Inspection

•Mass Reflow •Underfill •Mechanical •Mechanical •Mechanical

•TCB •Injection or loading loading loading
•Non- Compression •Adhesive •Reflow •Coplanarity/
Conductive Molding •Thermal temperature Warpage
Paste or Film •Molded Interface
Underfill Material

Predicting electronic package warpage and managing dynamic warpage still remain as a heavy
lifting effort and difficult for industrial.
A practical warpage prediction model is critical to drive more advance package design and risk
management.
 Introduction: Current Modeling Approach is it
Adequate?
Apply linear and
Create FEA visco-elasticity Compute the
Geometry, model from the Post processing to
material
Meshing and apply initial stress free extract predicted
constitutive
constraint condition package warpage
properties and
boundaries for risk assessment
include different How to account
condition with account for
processing steps of for the individual process, material
and include how materials are residual stresses of and incoming
incoming added to the the components variation.
variations construction of the during assembly
package

• There are a lot of work done in different scope and assumption and with different level of
success.
• Focus here include the study of the impact of
• visco-elastic constitutive property
• with time-temperature superposition (TTS) were included with a two steps shift
factor model of Arrhenius and WLF functions.
• include the chemical shrinkage (Pressure Volume Temperature Cure PVTC)
implementation. PVTC function is not widely included in general FEA tools. Some
researches found out that the pressure-volume-temperature-cure (PVTC) can
influence the amount of warpage of the package.

6
 Bi-material model: Modeling implementation
approach
• An arbitrary bi-material model evaluation was conducted to compare the material
model of linear elastic, visco-elastic, time-temperature superposition shift factor and
chemical shrinkage.
• Commercially available modeling software considered here were based on the
Moldex3D R17 and a general FEA tool called FEA-A.
• Stress analysis was computed for post mold cure induced shrinkage and followed by
time dependent thermal exposure loading to capture the evolution of warpage or
deformation.
• PVTC model is engaged to approximate the dynamic shrinkage based on transient
molding condition to predict the package behavior.

Bi-material model geometry Bi-material mesh part Bi-material material properties

Materials Mold Silicon
Modulus (MPa) 18190 1172800
14.5 (T < 443 K)
CTE (ppm) 2.3
53.6 (T≧443 K)

Mesh
1000um global size
2 elements across thickness
Total 1600 elements

7
 Stress-strain constitutive properties

• Stress-strain constitutive properties coupled with non-linear

function of time, temperature and other field dependent variables
like cure conversion rate.

Stress equilibrium equation Stiffness tensor

E (t )
 =  C = Cijkl =
E (t )
( ik  jl +  il  jk ) + 
2(1 + ) (1 + )(1 − 2 ) ij kl

Stress strain relationship Thermal induced strain

 = C( -  Thermal -  Cure )  Thermal =  CTET

1
 = (U + U T )
2

8
 Visco-elastic constitutive behavior

• The visco-elastic properties of epoxy molding compound are in the

form of Prony series with time-temperature shift factor.
• A Prony series (also known as Generalized Maxwell model) can be
expressed as follows:
t
Shear modulus G(t) = G  + i =1 Gi exp(−
n
) Shift factor a = aT (T )
i (T0 )  a
Bulk modulus K(t) = K +  n K exp(− t
 i =1 )
i (T0 )  a
i

1.2E+11 1.2E+11
t = 0.1 s
T = 30 ℃
1.0E+11 1.0E+11 t=1s
T = 60 ℃

Shear Modulsu (dyne/cm2)

Shear Modulsu (dyne/cm2)

t = 10 s
T = 150 ℃ 8.1E+10
8.1E+10

6.1E+10 6.1E+10

4.1E+10
4.1E+10

2.1E+10
2.1E+10

6.0E+08
6.0E+08 25 50 75 100 125 150 175 200
1.0E-10 1.0E-02 1.0E+06 1.0E+14 1.0E+22
Temperature (℃)
Time (s)
 Visco-elastic TTS

• There are two models to describe the time-temperature

superposition (TTS):
– Arrhenius type equation
– WLF equation
• Most FEA tools can only considered one kind of TTS model.
Subroutine needed if TTS is impacted by Tref .
• Tref was 423 K (150C) in this case.
− C1 (T − T0 )
WLF equation log aT = T  Treference
1.0E+10
C2 + T − T0
Moldex3D

WLF
1.0E+06
H T 1 1
Arrhenius type equation aT = ( ( - )) T  Tref
Shift Factor aT

User Subroutine
1.0E+02
R T T0

1.0E-02
C1=17.44, C2=51.6
1.0E-06

1.0E-10 ∆HT: chemistry activation energy

25 50 75 100 125 150 175 200
Temperature (℃) R : gas constant

10
 Curing Kinetics Mold Material Properties

• The combined model to investigate the curing kinetics of the given

mold because of its ability to accurately predict the experimental
data. The combined model can be expressed as follows:

Conversion rate

d
dt
( )
= k1 + k 2 m (1 −  )
n

 E 
k1 = A1 exp − 1 
 RT 

 E 
k 2 = A2 exp − 2 
 RT 
A1, A2, E1, E2, m, n are model parameters

11
 Modeling for chemical shrinkage without
curing percentage effect

Pseudo-CTE shrinkage
approach

Equivalent curing
shrinkage CTE Tan, Lin, et al. "Study of viscoelastic effect of EMC on FBGA block warpage by FEA simulation." 2013
14th International Conference on Electronic Packaging Technology. IEEE, 2013.

Effective curing shrinkage

CTE as function of Tg and
Filler %
12
 Modeling for chemical shrinkage with
curing percentage effect

Mixing law based on

CTEraw vs CTErubbery stage
Nawab, Yasir, et al. "Determination and modelling of the cure shrinkage of epoxy vinylester resin and
associated composites by considering thermal gradients." Composites Science and Technology 73
(2012): 81-87.

PVTC - Pressure Volume Hong, Li-Ching, and Sheng-Jye Hwang. "Study of warpage due to PVTC relation of EMC in IC
packaging." IEEE Transactions on Components and Packaging Technologies 27.2 (2004): 291-295.

Temperature Cure
approaches Two domain modified Tait model

13
 Chemical Shrinkage through PVTC with
curing percentage effect
• The two domain modified Tait model is used to formulate the
specified volume of resin as below (Moldex3D implementation):

Conversion rate

1 1 1
= (1 −  ) + 
V Vuncured Vcured
  p 
Vcured / uncured = V0 1 − C ln 1 + 
  B 
b1S + b2 S T , if T  Tt
V0 = 
b1L + b2 LT , if T  Tt
b3S exp(− b4 S T ) , if T  Tt
B=
b3L exp(− b4 LT ), if T  Tt
Tt = b5 + b6 P
T = T − b5 , C = 0.0894
 Boundary conditions & steps

• Transient Coupled Temp-Displacement Analysis

– 2 steps; 100s each
– Convective heat transfer with 20 W/m2/K coefficient
– 5 faces (top horizontal + 4 vertical faces)
– Initial temp. 448 K / 175 ℃
– Step 1 ambient temp. 298 K / 25℃
– Step 2 ambient temp. 533 K / 260 ℃

Ambient temperature Temperature on chip top center node

 Bi-material: Linear elastic and visco-elastic
effects – no chemical shrinkage considered
Z displacement on top center node
250
Visco-elastic effect
Moldex3D linear elastic
Linear elastic
200
FEA-A linear elastic

150 Moldex3D VE-no TTS

Z displacement (μm)

FEA-A VE-no TTS

100
Visco elastic
50 (VE)

-50

-100

-150
0 50 100 150 200
Time (s)

• At lower temperature, the model deformed into concave shape because of the higher
contraction of the mold.
• For visco-elastic material model without the TTS, the z-displacement reduced significantly
compared to linear elastic.
• The difference is due to the modulus relaxation and stress relaxation which counters the
thermal strain.
16
 Bi-material: Visco-elastic with/without TTS – no
chemical shrinkage considered
Z displacement on top center node
80
Visco-elastic with TTS
Moldex3D VE-no TTS
Visco elastic
60 (VE) without
FEA-A VE-no TTS TTS
40 Moldex3D VE-TTS
Visco elastic
Z displacement (μm)

FEA-A VE-TTS
20
(VE) with TTS
0

-20

-40

-60

-80

-100
0 50 100 150 200
Time (s)

• The concave warpage increases to -80um while convex warpage

reduces to 19.32um.
• As the temperature reduce, the modulus increases and hence the
warpage magnitude increases towards the linear elastic case.
17
 Bi-material: Effects of PVTC based chemical
shrinkage
Z displacement on top center node Temperature and conversion
40 570 110

20 Temperature (K)
520
0
Conversion % 100

-20 470
Z displacement (μm)

Temperature (K)

Conversion (%)
-40 VE-TTS
420 90
-60

VE-TTS with 2% chemical

-80 370
shrinkage
80
-100
VE-TTS with 1% chemical 320
shrinkage
-120

-140 270 70
0 50 100 150 200 0 50 100 150 200
Time (s) Time (s)

• Including chemical shrinkage (1% or 2%) increase the bi-material z displacement during
the cool down.
• During the first 100 second, no significant difference in the z-displacement due to the
conversion rate only increased by 1%.
• The heating cycle speeds up the chemical reactions to reach the 100% conversion rate.

18
 Conclusions

1. The use of bi-material system in this paper demonstrates the impact of

transient thermal structural analysis coupled with the effect of the visco-
elastic, time-temperature superposition (TTS) and the pressure-volume-
temperature-cure (PVTC) material model implementation.
2. The structural and thermal solutions are consistent or closely match
between a general FEA and Moldex3D for both linear and non-linear
elastic cases.
3. The visco-elastic implementation on the bi-material model captures the
reduction of z-displacement compared to linear model.
4. The TTS principle which interpolated the material properties based on
time and temperature field must be taken into consideration for the
actual assembly process history.
5. PVTC should not be neglected as the predicted value can be highly
influenced by conversion percentage of the mold.
6. The logical step for this industrial project is to validate the effect seen on
the visco-elastic behavior and the chemical shrinkage impact and
evaluate all these different approaches

19
Two additional chemical shrinkage
values of 1% and 2%

0.555

Cured state, 0.1MPa

0.550
Uncured state (1.0%), 0.1MPa

Uncured state (2.0%), 0.1MPa

0.545
Specific volume (cm3 /g)

0.540

△V (2.0%), 260oC
0.535

0.530

0.525
△V (1.0%),
260oC
0.520

0.515
25 50 75 100 125 150 175 200 225 250 275 300

Temperature (∘C)