[ieee 2014 ieee 64th electronic components and technology conference (ectc) - orlando, fl, usa...
TRANSCRIPT
A Stress-Based Effective Film Technique for Wafer Warpage Prediction of Arbitrarily Patterned Films
Gregory T. Ostrowicki, Siva P. Gurrum
Texas Instruments, Inc.
12500 T I Blvd, Dallas, TX 75243
Abstract
Initially flat silicon wafers are prone to warp due to the
high levels of intrinsic stress of deposited films, particularly
metallic films. Processing and handling of warped wafers in
the fab is a challenge. One of the ways to control the degree
of warpage is by limiting the amount of metallization allowed
on the wafer. However, this imposes a constraint on the
silicon designers, and can lead to decreased performance of
the IC. Therefore, there is a need to accurately predict the
amount of wafer warpage caused by a proposed layout in
order to give designers the most freedom to develop IC
solutions while ensuring that the processed wafers meet the
manufacturing equipment requirements.
The metal artwork (in addition to other materials, layer
thicknesses, processing parameters, etc.) is an important factor
in determining wafer curvature. Simple analytical methods,
such as Stoney’s Formula, cannot capture the non-uniform
warpage due to these patterned films. On the other hand,
numerical methods which require detailed modeling of the
film patterns across the whole wafer are computationally
expensive. Thus, a new finite element modeling technique
was developed in which the entire patterned film stack is
represented as a uniform effective orthotropic film bonded
onto a silicon substrate. The orthotropic properties are
determined from a small set of virtual experiments using a
unit-cell model that is characteristic of the actual pattern. The
resultant effective film, despite using a very course mesh, is
able to capture the non-uniform surface stress induced by a
patterned multi-layer film stack, and thus results in very
similar wafer warpage as in the conventional detailed model.
Several example film patterns will be presented here, where
the warpage difference between the detailed model and the
effective film model are less than a few percent across the
whole wafer.
1. Introduction
Wafers warp from the accumulation of stress within the
deposited films due to CTE mismatch, lattice mismatch,
impurities, recrystallization, creep, cure shrinkage, and other
phenomena [1]. It is a challenge to characterize all the
individual mechanisms that can contribute to the film stress
for each material in the stack. However, many of these
mechanisms can be lumped together into an “effective CTE”
(e.g. [2]). Therefore, this work focuses on thermal expansion
as the driving warpage mechanism.
As a silicon wafer goes through the IC manufacturing
flow, a series of patterned films are deposited across the wafer
to create a rectangular array of identical dies. Each die within
the wafer interior is essentially a unit-cell, with a
characteristic stress distribution through the film stack. Much
of the film stress is concentrated within the metal artwork, due
to the relatively high modulus and significant CTE mismatch
of metals with respect to the silicon substrate. Thus, the
patterning of the metal circuitry typically results in a non-
uniform biaxial stress depending on both the volume of metal
within the film as well as the design of the circuitry itself.
Any technique used to predict the global warpage of the wafer
after processing must therefore capture this non-uniform
stress in the film stack at the die level. For finite element
methods, this can require modeling the patterns exactly, using
element sizes that are on the order of the minimum feature
size in the film. Since the metal line width and thickness can
be in the micron to sub-micron scale, the necessary element
count can approach the millions for a single die depending on
the artwork, number of layers, and the die size. It can quickly
become computationally unreasonable to include this level of
detail across a whole wafer. For example, a 300 mm wafer
composed of 1x1 mm dies contains about 70,000 dies. To
simplify matters, one can model only a small portion of the
wafer and extrapolate the local curvature to estimate the
global warpage. However, this approach alone cannot capture
the effects of gravity, mechanical fixturing, and other global
body forces and constraints that can impact the warpage.
Previous numerical approaches have been used to model
global wafer warpage based on a film patterns of parallel lines
of a single material [3]. However, actual IC wafers can have
very complex film stacks. In this work, a finite element
approach is outlined to simplify the detailed nature of the
composite film stack into a uniform material with effective
properties, as shown in Figure 1. This representative film can
be modeled with a very coarse mesh, but still develop an
equivalent biaxial stress (and thus, equivalent warpage)
resulting from its CTE mismatch with the silicon substrate. A
detailed unit-cell model with representative film pattern
(typically one die) is first modeled in order to calculate the
effective film properties. Then, a second model of the wafer
with effective film can be used to determine the global
warpage by using a very coarse mesh.
Effective Film
Substrate
hf
Substrate
Composite Film Stack
hf
Figure 1. Simplification of composite film stackup into an
effective film with uniform properties.
978-1-4799-2407-3/14/$31.00 ©2014 IEEE 821 2014 Electronic Components & Technology Conference
2. Theory
Stress and Warpage
Stoney famously related the curvature of a plate to the film
stress through
( ) (2.1)
where R is the radius of curvature, E is the Young’s modulus,
h is the thickness, ν is the Poisson’s ratio, and the substripts f
and s correspond to the film and substrate, respectively [4].
This relationship holds for the case where the substrate and
film materials are uniform, isotropic, the film is much thinner
than the substrate, and the radius of curvature is much greater
than the substrate thickness (i.e. hf ≪ hs ≪ R). The film stack
on an IC wafer, however, is a mixture of different layers,
materials, and patterns. These heterogeneities can create a
non-uniform film stress, and thus non-uniform warpage along
the different radial directions of the wafer. This work aims to
simplify the composite film stack into a homogeneous
effective film for warpage prediction. Since any warpage
model must capture the directional nature of the film stress, an
anisotropic material model is used for the effective film.
Orthotropic Thin Film
A linear orthotropic material model is used to represent
the effective properties of an arbitrary composite film stack.
Since the film is thin and at a free surface, it can be assumed
to be in a state of plane stress, for which the conventions of
the stress components are shown in Figure 2.
σxx
σyyτxy σ11
σ22
θ=θp
Y
X Figure 2. Conventions for element in plane stress.
Hooke’s law for a thin orthotropic material can be
expressed in matrix form as
[
]
[
]
[
] (2.2)
where E is the Young’s modulus, G is the shear modulus, and
ν is Poisson’s ratio [5]. The following constraint is also
required to make the stiffness matrix symmetric:
(2.3)
In addition, there is an orientation θp at which the biaxial
stresses are aligned with the principal stresses and there is no
shear stress (i.e. τ = 0). The principal stresses σ11 and σ22 can
be determined by
( )
√(
)
(2.4)
and the principal orientation angle θp is determined by
(
) (2.5)
Therefore, for an orthotropic material in plane stress, and
oriented along the principal stress directions (i.e. θ = θp),
Hooke’s law can be further simplified to just the following
relationships
{
(2.6)
where the subscripts 1 and 2 refer to the principal directions,
and the corresponding stiffness symmetry constraint is
(2.7)
This result indicates that only the biaxial Young’s moduli
and Poisson’s ratios are required to fully define the stiffness
of the effective film, provided the film stress is oriented in the
principal stress directions. In addition, the film can be
assumed to have biaxial thermal expansion coefficients α1 and
α2. Thus for thermo-mechanical stress analysis, the effective
orthotropic film can be fully defined by only six material
parameters: E1, E2, v12, v21, α1, and α2.
Thermally Induced Stress-Based Effective Film
In the modeling approaches, a stress is induced in the film
by subjecting the wafer to a uniform change in temperature.
A 1-D approximation of such a bi-material strip undergoing
thermal loading is shown in Figure 3. Here the total strain of
each material must be equivalent as described by
(2.8)
where ΔT is the change in temperature from an initial stress-
free condition, and the subscripts f and s refer to film and
substrate, respectively. When the film is very thin (and
relatively soft) with respect to the substrate, the mechanical
strain in the substrate can be neglected (i.e. εs = 0) and (2.8)
can be expressed as
(2.9)
αf, Ef
αs, Es
Figure 3. 1-D illustration of bi-material strip under
thermal expansion
Although crystalline silicon is an anisotropic material, it
has been shown that a typical (100) oriented wafer has a
symmetric biaxial modulus [6], and its thermal expansion can
also be treated as isotropic [7]. Thus assuming an orthotropic
film on an isotropic substrate, (2.9) can be expanded along the
film’s principal directions to result in
822
{
(2.10)
where the subscripts 1 and 2 correspond to the film along the
respective principal directions. Substituting (2.6) into (2.10),
after some algebra results in
{
(2.11)
where
(
) (
)
(
) (
)
(2.12)
Equation (2.11) shows that for a given temperature
difference, the film’s principal stress is simply a linear
function of the substrate CTE, where the constants m and b
are purely functions of the film material properties. This
implies that effective orthotropic properties of a composite
film stack could potentially be determined if the average
thermally induced stress was measured for at least two
samples, on substrates with different CTEs. While this can be
impractical to exercise in a real experiment, it is
straightforward to implement in a numerical experiment. The
proposed process to determine the effective film properties is
thus outlined below:
1. Create detailed unit-cell model of the wafer with
representative film pattern (e.g. a single die).
2. Subject unit-cell to a characteristic temperature
change, and calculate the resulting volume-averaged
film stress along the principal directions.
3. Modify the CTE of the substrate in the unit-cell
model and repeat Step 2.
4. Solve for the constants m and b using (2.11) and the
results from Steps 2-3.
5. Find an appropriate property set for the film which
satisfies (2.7) and (2.12). The final step in the process outlined above presents a
potential caveat with this approach. As mentioned before, the
film is assumed to have six independent material parameters
(E1, E2, v12, v21, α1, α2). However, (2.7) and (2.12) provide for
only five constraint equations from which to solve for the six
properties. This in an unconstrained problem for which there
can be an infinite number of property sets which satisfy the
conditions. In the absence of a physical constraint, it is
possible to define an additional condition in an effort to obtain
a unique solution, such as assuming an isotropic CTE (i.e. α1
= α2). However, the authors feel that such a constraint may be
arbitrary, and in their experience may result in a poorly
defined material property set. Instead, an evolutionary
algorithm was used in EXCEL® to find a suitable material
property set, where the bounds of the effective modulus and
CTE were set to be within the maximum values of the
component materials in the film stack (i.e. 0 ≤ E1, E2 ≤ Emax
and 0 ≤ α1, α2 ≤ αmax), and the Poisson’s ratios were set to be
realistic values (i.e. 0 ≤ ν12, ν21 ≤ 0.5).
3. Numerical Models
A hypothetical wafer design was devised for use as a
baseline in order to demonstrate and compare the warpage
modeling approaches. A unit-cell of this wafer is shown in
Figure 4, and is representative of the patterning on a single
die. This die features alternating parallel strips of metal and
polymer film over a silicon substrate. Representative material
properties for the film and substrate are shown in Table 1.
Here hf = 10 um, hs = 725 um, b = d = 1 mm, wm = 71.4 um,
and vm = 0.5, where hf is the film thickness, hs is the substrate
thickness, b is the die width, d is the die depth, wm is the metal
line width, and vm is the volume fraction of metal in the film,
respectively.
For warpage modeling, only a small section of the wafer is
modeled due to element restrictions, but which is sufficient to
capture the wafer curvature in the absence of external forces.
As shown in Figure 5, the warpage model is a disk with radius
of 5 mm and is composed of a rectangular array of the
baseline die. Symmetry is utilized so that only ¼ of the model
is represented, where appropriate symmetry conditions are
imposed on the cut faces. Body forces such as gravity are
neglected such that the resulting curvature is purely a function
of the CTE mismatch between the film and substrate.
For the initial case, the entire film/substrate system is
assumed to be stress-free at elevated temperature and then
subjected to a drop in temperature to ambient conditions,
where ΔT = -100°C. In a more realistic scenario, however,
each individual film material is likely to have a unique stress-
free temperature depending on the material’s processing
history. This situation is addressed later as described in
Section 4, and is only briefly neglected here in order to
demonstrate the methodology of the approach. It should also
be noted that the film pattern was specifically chosen so that
the resulting principal stress directions in the film are aligned
with the coordinate system (i.e. σ11, σ22 = σxx, σyy). Die
designs where this is not the case are also discussed in Section
4.
Finite element models were developed using ANSYS®
14.5. First, a detailed model was constructed to establish a
benchmark of the wafer warpage. Then, a new approach was
pursued which first determined the effective properties of the
film based on stress within a unit-cell, and then applied the
effective properties to a simplified global warpage model.
823
silicon
metalpolymer
b
d
hs
hf
wm
Figure 4. Baseline die design with patterned film of
alternating metal and polymer strips on a silicon
substrate.
XY
Z
Figure 5. Baseline warpage model domain.
Table 1. Material properties of film and substrate.
Metal film Polymer film Si substrate
E 100 GPa 5 GPa 131 GPa
v 0.33 0.33 0.28
CTE 20 ppm/°C 40 ppm/°C 2.6 ppm/°C
Benchmark Approach: Detailed Model
A detailed finite element warpage model was constructed
using solid quadratic elements (SOLID186) with element
sizes of ~5x5x1 um in the film and ~5x5x145 um in the
substrate (see Figure 8). A sufficiently fine mesh was used to
capture the stress distribution through the film. The resulting
warpage is shown in Figure 6 for ΔT = -100°C. The radius of
curvature R was calculated along the X- and Y- axis by fitting
a circular arc to the out-of-plane displacement profiles shown
in Figure 7, which was found to be Rx = 11.492 m and Ry =
20.401 m. As a side note, if this curvature was extrapolated to
a 200 mm wafer, that would translate to a maximum warpage
of 435 um and 245 um along the X- and Y- axis, respectively.
X
Y
Uz (mm)
Figure 6. Benchmark warpage contours of detailed model.
-2.0E-04
0.0E+00
2.0E-04
4.0E-04
6.0E-04
8.0E-04
1.0E-03
1.2E-03
0 1 2 3 4 5 6
Z-D
isp
lacem
en
t u
z(m
m)
Distance from center (mm)
X-Axis
Y-Axis
Figure 7. Warpage profile of benchmark model.
The disparity of the warpage along the two axes can be
explained by examining the stress within the composite film
as shown in Figure 8. Here the stress is concentrated in the
metal, and is much higher in the direction of the strips than
perpendicular to them. This is due to the fact that the metal is
less constrained near the sidewall adjacent to the low stress
polymer. In order to determine the effective film stress across
the wafer, the volume-averaged stress σ(avg)
in the film was
calculated to be σxx(avg)
= 123 MPa and σyy(avg)
= 90 MPa. It is
interesting to note that although the film stress in the X-
direction is only about 1.4X higher than in the Y-direction, the
curvature 1/R is almost 1.8X higher. Only a relatively small
non-uniformity in the biaxial stress is necessary to create a
significant difference in the warpage along the two
perpendicular axes.
824
Si
Polymer
Z
Y
PolymerMetal
σxx
σyy
Stress
(MPa)
Figure 8. Cross section of the detailed model near the
film/substrate interface. The corresponding contour plots
of σxx and σyy indicate that the overall film stress is more
tensile in the direction of the strips than perpendicular to
them.
New Approach: Stress-Based Effective Orthotropic Film
Whereas the benchmark approach involves using a
detailed model for the entire warpage domain, the new
approach employs a two-step process. First, a detailed model
is constructed only for one representative unit-cell (or die),
from which the smeared effective film properties are
calculated as described in Section 2. Then, a separate
simplified model, composed of just the substrate and effective
film, is used to determine the warpage over the entire domain.
Since the effective film has uniform properties across the
wafer, a relatively coarse mesh can be used for the warpage
model, which enables a much more efficient solution than the
benchmark approach.
The detailed unit-cell model was created for the die shown
in Figure 4, again using SOLID186 elements and a fine mesh.
Conventional unit-cell boundary conditions were applied on
the cut faces (i.e. out-of-plane displacements were coupled
across each of the four respective cut faces). The model was
subjected to ΔT = -100°C, and the resulting volume-averaged
biaxial film stress was determined. Then, a second identical
unit-cell model was made except that the CTE of the Si was
artificially changed to a different value, and again the
resulting volume-averaged biaxial stress was determined.
Although these two unit-cell models are sufficient to
determine m and b, a total of four models were created with Si
substrate CTEs of 1, 2.61, 5, and 10 ppm/°C in order to
clearly demonstrate that the trend is indeed linear as predicted
by (2.11) and shown in Figure 9. A suitable set of effective
film properties was found using the strategy described in
Section 2 with values listed in Table 1.
y = 67536x - 1.4403
y = 46347x - 1.0471
-1.6
-1.4
-1.2
-1.0
-0.8
-0.6
-0.4
000.E+0 4.E-6 8.E-6 12.E-6
σf(a
vg) /Δ
T (
MP
a/
C)
αsubstrate (1/ C)
X-DirectionY-Direction
Figure 9. Biaxial film stress using detailed unit-cell model
as a function of substrate CTE. This data is from four
separate models in order to demonstrate that the trends
are linear. In general, two models are sufficient.
Table 2. Calculated effective film material property set.
Effective Film
Ex 56378 MPa
Ey 35973 MPa
νxy 0.2408
νyx 0.1536
αx 21.08E-6 °C-1
αy 22.96E-6 °C-1
Next, a warpage model was constructed using SOLID186
elements for the Si substrate and quadratic shell elements
(SHELL281) for the film. Unlike heterogeneous films which
can have a large variation in biaxial stress through the
thickness of the film (e.g. see Figure 8), the stress in a
homogeneous film is uniformly distributed through the film
thickness. Thus it is appropriate to use shell elements for the
effective film and this further reduces the model complexity.
Using the calculated effective film properties in Table 2, the
resulting warpage for this model is shown in Figure 10, and
can be seen to match very closely with the results of the
benchmark model in Figure 6. A comparison of the two
models is shown in Table 3, where the difference in the total
film stress and warpage is only 0-1.3%.
825
X
Y
Uz (mm)
Figure 10. Warpage contours of simplified model using
effective film properties.
Table 3. Comparison of baseline warpage models.
Detailed Model
Effective
Model % Difference
σxx(avg)
123.14 MPa 123.25 MPa 0.09%
σyy(avg)
90.27 MPa 91.05 MPa 0.86%
Rx 11.49 m 11.63 m 1.21%
Ry 20.40 m 20.13 m 1.32%
4. Additional Model Considerations
Arbitrary Patterns
For the example die design used in the previous analysis,
the principal stress in the film is naturally oriented with the
global coordinate system since the patterning is symmetric
with respect to the imposed unit-cell boundary conditions.
Therefore, it was sufficient to calculate the effective film
properties along the defined X- and Y- axes. However, for an
arbitrary die layout, the principal film stress may not be
aligned with the defined coordinate system. Since the
maximum warpage is typically in the direction of the
maximum principal stress, it is then necessary to rotate the
coordinate system so that the effective properties are
determined in the proper orientation.
To illustrate this procedure, consider the die layouts in
Figure 11. Design (a) is the same unit-cell as shown in Figure
4, with horizontally aligned strips where φ = 0°. Design (b)
has the same strip pattern except that the strips are at an angle
of φ = 45°. In fact, these two designs are really just
alternative unit-cells for the same repeating pattern of parallel
film strips. Therefore, both dies ought to have equivalent
principal stress magnitudes, and thus equivalent warpage.
The resulting volume-averaged film stress for each of these
unit-cell designs is shown in Figure 12, where each of the
three planar stress components were calculated in the rotated
X’-Y’ coordinate system, for 0°≤θ≤90°. Indeed, the principal
stresses for design (a) were found at θp(a)
= 0°, and are in close
agreement with those of design (b) where θp(b)
= 45°.
X′
YY′
θX
φ = 0° φ = 45°
φ
(a) (b)
Figure 11. Two alternative unit-cells that result in the
same global wafer design.
-40
-20
0
20
40
60
80
100
120
140
0 15 30 45 60 75 90
Fil
m S
tress
(M
Pa
)
θ (deg)
σxx' (φ=0°) σxx' (φ=45°)σyy' (φ=0°) σyy' (φ=45°)τxy' (φ=0°) τxy' (φ=45°)
Figure 12. Volume-averaged plane stress of the film in the
rotated X’Y’ coordinate system, where θ=0-90°. Two
cases are shown, for unit-cells (a) and (b).
The von Mises stress for the two designs is shown in
Figure 13. While the stress distribution is very similar for
both cases, there is some distortion near the cut boundary of
the die with angled strips. This is due to the fact that the
imposed unit-cell boundary conditions do not allow for global
shear deformation, and this enforces symmetry constraints
along the bounding faces even when the geometry is not
symmetric. Fortunately, this stress artifact is isolated to the
periphery of the die, and does not significantly affect the
volume-averaged stress calculations.
σvm
(MPa)
Figure 13. Von Mises film stress for the two unit-cell
designs. There is some artificial distortion in the stress
near the cut boundary of the angled strips due to the
asymmetric nature of the design with respect to the
imposed unit-cell constraints.
826
Multiple Stress-Free Temperatures
In all the previous models, the wafer was considered to be
stress-free at an initial temperature and then subjected to a
temperature ramp. In an actual wafer processing flow,
however, film materials are serially deposited, layer-by-layer,
at a wide range of temperatures and environmental conditions.
It is unlikely the wafer is stress-free at any given temperature.
In order to properly capture the component stresses of such a
heterogeneous film, the thermal assembly process can be
mimicked so that the appropriate film material is activated at
its respective stress-free temperature. This can be
accomplished using the element birth capability within the
finite element model.
Using the unit-cell from Figure 4 as an example, the
stress-free temperature Tref is assumed to be 120°C for the
metal film and 220°C for the polymer film, respectively. The
resulting stress distribution for the die at an ambient
temperature Tamb of 20°C can be determined using the strategy
shown in Figure 14, where the results of each load step from
the analysis are shown in Figure 15. In order to find the
effective film properties for this case, the model was repeated
using a modified Si substrate CTE as was done in Section 3.
Again the trends are clearly linear as shown in Figure 16,
from which appropriate effective film properties can be
calculated.
TrefA
Tamb
0 1
Birth (film A)
Birth (film B)
TrefB
2 3
Load Step
Tem
pera
ture
Figure 14. Example illustration of a multi-step film
deposition sequence, where each component film material
is activated through element birth at its respective stress-
free temperature.
Load Step 1, T=120°C
Load Step 2, T=220°C
Load Step 3, T=20°C
Figure 15. Progression of stress as the temperature is
ramped according to the film deposition sequence, and the
appropriate elements in the film are activated upon
reaching the stress-free temperatures of metal and
polymer, respectively.
y = 36722x - 0.8235
y = 28107x - 0.682
-0.9
-0.8
-0.7
-0.6
-0.5
-0.4
000.E+0 4.E-6 8.E-6 12.E-6
σf(a
vg) /Δ
T (
MP
a/
C)
αsubstrate (1/ C)
σx/ΔTσy/ΔT
X-Direction
Y-Direction
ΔT = -200°C
Figure 16. Biaxial film stress for unit-cell design with
different stress-free temperatures for metal and polymer
film.
5. Realistic Wafer
The methods outlined above were used to model the
warpage of a patterned wafer with a realistic die layout shown
in Figure 17. Here the metal artwork is partially covered by a
polymer film except for the circular regions that define an
exposed metal pad. The wafer thickness is 15 mil, and the
material properties used were the same as specified in Table
1, with stress-free temperatures of 120°C and 220°C for the
metal and polymer film, respectively.
827
Y
X
Figure 17. Example die layout, with metal artwork
covered by polymer film.
Both detailed and effective-film warpage models were
constructed for this wafer pattern in the same manner as
above, and subjected to the temperature history as shown in
Figure 14. The detailed model was composed of only a 4x6
die array to limit the computation expense, whereas the
effective film could be modeled for the entire wafer domain
due to its very low element count requirement. The results for
both models are compared in Table 4, where they are in
excellent agreement with a difference of just a few percent.
The principal stress orientation for this pattern resulted in the
maximum curvature to be in the Y-direction due to the pattern
of thin metal strips which run in parallel along the same
direction.
Table 4. Comparison of warpage models with realistic
pattern.
Detailed Model
Effective
Model % Difference
σxx(avg)
112.03 MPa 110.04 MPa 1.77%
σyy(avg)
118.85 MPa 116.98 MPa 1.58%
Rx 2.762 m 2.699 m 2.29%
Ry 2.399 m 2.421 m 0.91%
6. Conclusions
Wafer warpage is due not only to the amount of highly-
stressed metal content in the patterned film, but also the
design of the artwork. These patterns can have tendencies to
cause greater warpage in certain directions more so than
others, and cause handling issues during the manufacturing
flow. Conventional modeling of the warpage thus requires
modeling the fine details of the patterns, which is
computationally prohibitive for large areas of the wafer. A
better approach is to use an effective film with properties that
represent the cumulative effect of all the individual layers and
materials that make up the film stack on a patterned wafer.
A method for modeling the warpage of arbitrarily
patterned wafers was thus developed which uses an effective
orthotropic film to represent the composite film stack. A
detailed model of a representative unit-cell of the wafer was
first used to extract the material properties for an effective
film, based on thermal mismatch with the silicon substrate and
the resulting volume-averaged stress within the film stack.
Then, a global warpage model was constructed using the
effective film without having to capture the details of the
pattern. The stress-based effective film approach was also
demonstrated to be capable of capturing the unique stress-free
temperatures of each of the patterned film components.
The difference between the resulting biaxial warpage of
the simplified model with effective properties with that of the
detailed model was less than a few percent. Furthermore, the
proposed approach can be used to model the global warpage
across the entire wafer domain, which would be virtually
impossible to do with a detailed model. This approach
therefore allows for including the effects of external forces
such as gravity and mechanical fixturing which can impact the
wafer curvature.
References
1. P. A. Flinn et al., “Measurement and interpretation of
stress in aluminum-based metallization as a functino of
thermal history,” IEEE Trans. Electron Devices, vol. 34,
no. 3, pp. 689-699, Mar. 1987.
2. C.-H. Hsueh et al., “Residual stresses in thermal barrier
coatings: effects of interface asperity curvature/height
and oxide thickness,” Mater. Sci. Eng. A, vol. 283, pp.
46-55, 2000.
3. Y.-L. Shen et al., “Stresses, curvatures, and shape
changes arising from patterned lines on silicon wafers,” J.
Appl. Phys., vol. 80, no. 3, pp. 1388-1398, 1996.
4. G. G. Stoney, “The tension of metallic films deposited by
electrolysis,” Proc. R. Soc. Lond. A, vol. 82, no. 553, pp.
172-175, 1909.
5. N. E. Dowling, Mechanical Behavior of Materials 2nd
Ed., Prentice Hall, Upper Saddle River, 1999, pp. 190-
192.
6. M. A. Hopcroft, et al., “What is the Young’s modulus of
silicon?,” JMEMS, vol. 19, no. 2, pp. 229-238, Apr. 2010.
7. C. A. Swenson, “Recommended values for the thermal
expansivity of silicon from 0 to 1000 K,” J. Phys. Chem.
Ref. Data, vol. 12, no. 2, 1983.
828