You are on page 1of 9

2680

IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 58, NO. 8, AUGUST 2011

Polyimide-Enhanced Stretchable Interconnects:


Design, Fabrication, and Characterization
Yung-Yu Hsu, Mario Gonzalez, Frederick Bossuyt, Jan Vanfleteren, and Ingrid De Wolf, Senior Member, IEEE

AbstractThis paper discusses the optimization of a stretchable


electrical interconnection between integrated circuits in terms of
stretchability and fatigue lifetime. The interconnection is based
on Cu stripes embedded in a polyimide-enhanced (PI-enhanced)
layer. Design-of-experiment (DOE) methods and finite-element
modeling were used to obtain an optimal design and to define design guidelines, concerning both stripe and layer dimensions and
material selection. Stretchable interconnects with a PI-enhanced
layer were fabricated based on the optimized design parameters
and tested. In situ experimental observations did validate the optimal design. Statistical analysis indicated that the PI width plays
the most important role among the different design parameters.
By increasing the PI width, the plastic strain in the Cu stripes is
reduced, and thus, the stretchability and fatigue lifetime of the
system is increased. The experimental results demonstrate that
the PI-enhanced stretchable interconnect enables elongations up
to 250% without Cu rupture. This maximum elongation is two
times larger than the one in samples without PI enhancement
[1]. Moreover, the fatigue life at 30% elongation is 470 times
higher [2].
Index TermsDesign for experiments, design optimization, failure analysis, finite element methods, flexible electronics, integrated
circuit interconnections, integrated circuit packaging, materials
reliability, strain, stress.

I. I NTRODUCTION

ONFORMAL electronics have attracted attention in recent years because of their deformability in bending,
twisting, and stretching. This unique deformability allows the
electronic systems to be used in biomedical applications such
as artificial skins [3], phototherapy, and implantable devices
[4]. Several technologies have been proposed in recent years
to achieve a high deformability. One of the most effective technologies is hybrid systems. These hybrid systems are usually
composed of rigid or bendable elements such as a thick or thin
Si IC chip, connected through stretchable electrical conductors.
The sensory or active components such as sensors or transistors
are placed on the rigid or bendable elements, and as such, when

Manuscript received February 18, 2011; revised April 13, 2011; accepted
April 20, 2011. Date of publication June 9, 2011; date of current version
July 22, 2011. This work was supported by the European Commission via the
PLACE-it research project under Contract 0248048. The review of this paper
was arranged by Editor A. Schenk.
Y.-Y. Hsu and I. De Wolf are with the IMEC, 3001 Leuven, Belgium, and also
with the Department of Materials Engineering, Katholieke Universiteit Leuven
(KUL), 3001 Leuven, Belgium (e-mail: Yung-Yu.Hsu@imec.be).
M. Gonzalez is with the IMEC, 3001 Leuven, Belgium.
F. Bossuyt and J. Vanfleteren are with the Center for Microsystems Technology IMEC, 9052 Gent, Belgium.
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TED.2011.2147789

the electronic systems are subjected to a large deformation, the


active devices feel no or limited strain, whereas the stretchable
conductors have to hold most of the deformation. For this
reason, to avoid losing electrical performance and structural
integrity during stretching, a robust design of the stretchable
conductors is essential.
Coplanar patterned metal conductors [5], [6] and out-ofplane wavy or wrinkling conductors [6], [7] are reported
as effective ways to realize stretchable interconnects. These
unique technologies require sophisticated component transferring (pick-and-place) processes [8] and a premechanical
stretch of the elastomeric substrate during metal deposition
[6], which are both incompatible with the current conventional coplanar technology. Moreover, reliability studies of such
stretchable devices are very limited. It is reported that the
electrical resistance is unstable: it is a function of elongation
and of stretching cycles due to metal microcracking [9] and is
thus limited to certain applications.
An alternative technology, which is both compatible with
conventional planar microfabrication and shows a stable electrical performance when stretching, has been proposed [10][12].
This technology uses an in-plane patterned metal conductor
which can act like a spring when stretching. An example of such
a spring is a Cu conductor patterned in the shape of horseshoe, as shown in Fig. 1 [2]. This way, the induced permanent
damage in the metal is very low even when a large deformation
is applied to the sample. Although these in-plane patterned
metal conductors have been proven to show a stable electrical performance, the cyclic-stretching performance still
requires further improvement. According to our latest study,
the horseshoe-type in-plane patterned metal conductor can be
stretched for only a few tens of cycles at 30% elongation [2].
In this paper, a stretchable metal conductor (horseshoe
shape) embedded in polyimide (PI-enhanced) with high cyclicstretching performance and maximized one-time-stretching
elongation is proposed. By combining the design-ofexperiment (DOE) technique with finite-element modeling
(FEM), design guidelines are derived. These guidelines are
mainly obtained from information on the plastic strain in the
Cu within a specific design space. Indeed, this plastic strain
will determine when the metal will start cracking or break.
Using the optimized design parameters obtained from DOE and
FEM, PI-enhanced stretchable interconnects are fabricated for
experimental validation. The validation was done by both onetime-stretching and cyclic-stretching in situ electromechanical
measurements. To study the deformation mechanisms, the
stretching behavior of the PI-enhanced metal conductor was
recorded in situ and correlated to numerical analysis.

0018-9383/$26.00 2011 IEEE

HSU et al.: PI-ENHANCED STRETCHABLE INTERCONNECTS: DESIGN, FABRICATION, AND CHARACTERIZATION

Fig. 1. (a) Schematic illustration of the factorial design of a PI-enhanced


stretchable interconnect (not in scale). (b) Dimension of one repeated unit of
the Cu-horseshoe structure.

2681

Fig. 3. Fabrication steps. (a) Cu/PI laminate adhered on carrier. (b) Photo
lithography. (c) Reactive ion etching. (d) Spin-on PI layer. (e) Laser cutting.
(f) PDMS casting. (g) Final encapsulation with PDMS.

TABLE I
H IGH AND L OW M AGNITUDE L EVELS FOR E ACH OF
THE D ESIGN FACTORS AS S HOWN IN F IG . 1

Fig. 4. Complete PI-enhanced stretchable interconnect.

Fig. 2. Finite-element meshes of a sample with PI-enhanced stretchable


interconnection. Forty percent elongation is applied on one end surface, and
another end surface is assumed to be a symmetrical plane.

II. S TRUCTURAL D ESIGN OF A PI-E NHANCED


S TRETCHABLE I NTERCONNECT AND
N UMERICAL M ODELING
Fig. 1(a) illustrates the PI-enhanced horseshoe-patterned Cu
conductor encapsulated in an elastomeric substrate. The struc-

tural parameters of the metal conductor, i.e., angle (), width


(WCu ), and radius (rCu ), are 120 , 0.1 mm, and 0.75 mm,
respectively, as shown in Fig. 1(b). To optimize the dimensions
and the materials properties, four design parameters, i.e., PI
width, PI thickness, elastic modulus of PI, and substrate thickness, are selected for the DOE matrix. Table I shows these
four design factors and their corresponding low and high levels.
These low and high levels are defined according to the capabilities of the processing and the material availability. Moreover,
these low and high levels formulate a so-called design space,
in which the numerical results from the FEM analysis can be
interpolated for response surface calculation. It should be noted
that a PI width of 0.1 mm (low level) is the extreme case
indicating that the PI material has the same width as the Cu
and, as such, covers only the top and bottom sides of the metal
conductor. There is, in this case, no sidewall coverage of PI
material on the metal conductor. This extreme configuration
appears to be the worst case in the later DOE and FEM analysis,

2682

IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 58, NO. 8, AUGUST 2011

TABLE II
ANOVA TABLE

and therefore, the figures as shown throughout this paper focus


on the metal interconnect fully embedded in PI material with
sidewall coverage. The PI thickness in Table I represents the
overall thickness of PI material including three parts, i.e., top,
middle, and bottom parts, as shown in the inset of Fig. 1(a).
The top and bottom parts always have the same thickness, and
this thickness is a variable. The middle part of PI material is a
constant and has the same thickness as the Cu layer (0.009 mm).
For instance, a PI thickness of 0.027 mm (low level) in Table I
indicates that the PI material of both top and bottom parts is
0.009 mm thick, whereas a PI thickness of 0.059 mm (high
level) in Table I indicates that the PI material of both top and
bottom parts is 0.025 mm thick.
The commercial finite-element code MSC.MARC was used
to simulate the deformation process and to calculate the different stress and strain components. Fig. 2 shows an example
of the mesh used for the samples in the 3-D FEM model.
The change of color intensity in the thickness direction is
because of the expansion of the meshes. Taking advantage of
the symmetry of the structure, only one stretchable interconnect
unit is modeled. A uniaxial elongation of 40% is applied to
the substrate at one end, and the other end is assumed to be
a symmetrical plane. The metal used in this research is copper
and is modeled as a plastically deformable solid, obeying the bilinear isotropic hardening rule, with an elastic Youngs modulus
E0 = 117 GPa, the yield point at y = 0.1723 GPa, and the tangent modulus Et = 1.0342 GPa. The elastomeric substrate is
poly-dimethyl-siloxane (PDMS) modeled as an incompressible
hyperelastic Neo-Hookean solid [13] with C10 = 0.157 MPa.

Fig. 5. Plot of normal probability versus the effect of the design factor on
plastic strain.

III. S AMPLE P REPARATION


The processing steps for the fabrication of the samples are
shown in Fig. 3. First [Fig. 3(a)], a commercially available
Cu/PI laminate foil (Upisel-N SE 1410) with 9-m-thick Cu
and 2- m-thick PI underneath the Cu layer was adhered on
top of a ceramic carrier, with the PI layer facing the carrier
to ensure surface uniformity when the copper is chemical
etched. A standard lithography process was carried out with
AZ4562 photoresist and a CuCl2 etching solution [Fig. 3(b)],
followed by reactive ion etching to etch away the uncovered
PI [Fig. 3(c)]. To fully embed the copper, another PI layer was
spin-coated on top of the patterned structure [Fig. 3(d)]. The
PI is then cured in a vacuum oven under nitrogen atmosphere.
Two-step curing processes were carried out as follows: heating

Fig. 6.

Plot of normal probability versus residual.

from room temperature to 200 C, with a ramp rate of 4 C


per minute; 30 minutes at 200 C for precuring; heating from
200 C to 375 C, with a ramp rate of 2.5 C per minute; and
cooling down to room temperature.
With this two-step curing process, a thick layer of PI with few
micrometers can be achieved. Next, the horseshoe-patterned
metal conductor, which is fully embedded in PI, is adhered
on a temporary FR-4 carrier and cut by an excimer laser
[Fig. 3(e)]. Next, it is placed on a stainless steel mold with

HSU et al.: PI-ENHANCED STRETCHABLE INTERCONNECTS: DESIGN, FABRICATION, AND CHARACTERIZATION

Fig. 7.

2683

Interaction plot. (a) Interaction between PI width and PI thickness. (b) Interaction between PI width and PI elastic modulus.

an opening of the same size as the carrier. The following


casting of PDMS [Fig. 3(f)], removal of the carrier, and final
encapsulation [Fig. 3(g)] steps are the same as the processing
steps discussed in [12]. Fig. 4 shows a completed PI-enhanced
stretchable interconnect. It should be noted that according to
our analysis in the later sections, the optimized structure is with
sidewall coverage of PI to Cu. Therefore, Figs. 3 and 4 show the
fabrication steps and a PI-enhanced interconnect sample with
sidewall coverage.
IV. R ESULT AND D ISCUSSION
A. Statistical Analysis: Factorial Analysis and ANOVA
To obtain information on the sensitivity of the different
factors defined in Fig. 1 and Table I, a two-level and fourfactor full factorial (24 ) analysis is performed by means of
FEM simulations. The maximum equivalent plastic strain in the
Cu as calculated from FEM is used for the statistical analysis.
This strain ranges from 1.70% up to 5.29%, while the applied
elongation is 40% for all studied combinations. In general, a
DOE is done by experiments, giving experimental results and
error bars. When doing this through simulations, there are no
error bars (each repeated simulation with the same input will
give the same output). To assign an error to our simulation
results, we can do error estimation. This is done by using highorder terms (interactions) that are negligible or factors whose
contribution is very low as experimental error.
The percentage contribution measures the contribution of
each term (design factors and their interactions) to the total sum
of squares of all terms. Table II shows the analysis of variance
(ANOVA) [14], in which the sum of squares of each design
factor is calculated. In Table II, one can calculate the percentage
contribution of each factor by dividing the corresponding sum
of squares to the total sum of squares. It is noted that the PI
width contributes most (85%) and is identified as the most
important design factor, followed by the PI thickness (6%),
the interaction of the PI width and thickness (4%), and the
PI elastic modulus (3%). The PDMS thickness and all other
high-order cross terms have negligibly small effects. For every
design factor listed in Table II, the p-value is smaller than
0.0001, which indicates that not only the model but also the
listed factors are strongly significant. Moreover, because R2

equals 0.99 (R2 = SSmodel /SStotal = 29.48/29.55 = 0.99),


where SSmodel is sum of squares of the model and SStotal
is sum of squares of the overall effect combination, the model
explains about 99% of the variability.
The contribution of the various effects of design factor on
plastic strain can be better examined using a normal probability
plot, as shown in Fig. 5. The effects that are negligible follow
normal distribution and tend to fall along a straight line on
this plot, whereas the significant effects are far from that line.
As indicated before, the apparently negligible effects can be
combined as an estimate of the error. In Fig. 5, the significant
effects that emerge from this plot are the same as those obtained
from Table II. In addition, the interaction of PI width and PI
elastic modulus has a small influence. To ensure that there are
no missing significant factors during the analysis, the normal
probability plot of the residuals is shown in Fig. 6. The points on
this plot lay reasonably close to a straight line, lending support
to the conclusion that the PI design factors width, thickness,
elastic modulus, interaction of width and thickness, and interaction of width and elastic modulus are the only significant factors
and that the underlying assumptions of the analysis are correct.
The interaction of PI width and thickness and the interaction
of PI width and elastic modulus are plotted in Fig. 7(a) and
(b), respectively. Considering each single design factor in both
Fig. 7(a) and (b), it is found that these three factors (PI width,
PI thickness, and elastic modulus of PI) have a negative effect.
In other words, by increasing the width, thickness, and elastic
modulus of PI, the plastic strain in the Cu can be reduced by
different levels of magnitude. Moreover, it should be noted that
the PI width has a stronger effect when the PI thickness is
0.027 mm and the elastic modulus of PI is 3.3 GPa.
B. Regression Model and Response Surface Analysis
The response of the plastic strain in the Cu to variations in the
design factors can be predicted by a regression model in terms
of the actual factors
Plastic strain in metal
= +3.03 1.25 P I width 0.33 PI thickness
0.22 elastic modulus of PI + 0.28
PI width PI thickness + 0.19
PI width elastic modulus of PI.
(1)

2684

IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 58, NO. 8, AUGUST 2011

Fig. 8. Contour plot and response surface. (a) Contour plot of PI width and PI thickness. (b) Contour plot of PI width and PI elastic modulus. (c) Response
surface of PI width and PI thickness. (d) Response surface of PI width and PI elastic modulus.

Fig. 9. Cubic plot of PI width, PI thickness, and PI elastic modulus and its
corresponding plastic strain.

This equation shows that the plastic strain in the metal reduces
with an increase in the PI width, the PI thickness, and the
PI elastic modulus. The width has the most significant effect
because of the rather large coefficient (1.25) compared to
the other coefficients in the equation. Fig. 8(a) and (b) shows

Fig. 10. Top view of the PI-enhanced stretchable interconnect. The dimensions are according to the optimized parameters from DOE analysis.

contour plots of the PI width and thickness and the PI width and
elastic modulus, respectively. Corresponding to these contour
plots, the response surfaces for the plastic strain in the Cu
obtained from the regression model are shown in Fig. 8(c) and
(d). Because of interactions, the contour lines of the plastic

HSU et al.: PI-ENHANCED STRETCHABLE INTERCONNECTS: DESIGN, FABRICATION, AND CHARACTERIZATION

2685

Fig. 11. Optical images taken in situ while stretching to (a) 0%, (b) 40%, (c) 80%, (d) 120%, (e) 160%, and (f) 200%.

strain in both Fig. 8(a) and (b) are curved (and the response
surfaces in Fig. 8(c) and (d) are twisted planes). It is desirable
to have an as low as possible plastic strain in the metal when
stretching. Therefore, from both contour plots and response
surfaces, one can determine the options for reducing the plastic
strain in the metal, i.e., by increasing the PI width and the PI
thickness and by using a PI with a higher elastic modulus. This
trend can be confirmed and elaborated through a cubic plot, as
shown in Fig. 9. At the lower left corner of the cube, where the
PI width (0.1 mm), thickness (0.027 mm), and elastic modulus
(3.3 GPa) all have a low magnitude, one obtains the highest
plastic strain in the Cu. This corner is clearly undesirable for
designing a reliable stretchable interconnect. In contrast, in the
opposite corner of the cube, where all three design factors have
a high magnitude, one can obtain the lowest plastic strain in the
metal. Therefore, design parameters with a PI width of 0.3 mm,
a PI thickness of 0.059 mm, and a PI elastic modulus of
8.5 GPa are optimal for reducing the plastic strain in the Cu
upon stretching, i.e., they are the desired parameters. These
three factors with high level are preferable for having a good
reliability, and therefore, they are chosen for sample fabrication
and for further experimental studies. It should be noted that the
aforementioned trend is valid only within the specific design
space, as specified in Table I.
C. In Situ Observation of the Deformation Behavior
To verify the design and investigate the performance,
a stretchable interconnect with a PI-enhanced layer was
fabricated according to the optimized parameters obtained from
the DOE analysis. The PI layer is 300 m wide and 59 m thick
and has an elastic modulus of 8.5 GPa. Fig. 10 shows a top view
of a sample. The Cu resides in the center of the meander and is
fully embedded in the PI layer. The deforming processes, failure mechanisms, and electromechanical performance of these
samples subjected to uniaxial stretching were investigated using
a home-built tensile tester and in situ monitored by a highresolution Hirox KH-7700 optical microscope. All tests were
performed at room temperature with a constant displacement
rate of 2.5 104 s1 . During tensile testing, the electrical
resistance of the patterned metal circuits was recorded every

0.5 s by an Agilent 34420A multimeter. The four-point resistance measuring technique was used so that the resistance of
the connecting wires can be neglected.
Fig. 11(a)(f) shows a sequence of in situ observations of
samples, starting at 0% up to 200% elongation. The images
show that the deformation processes (geometrical opening and
substrate thinning) are similar as the ones observed in sample
without a PI layer [1], [12]. However, it is found that there
is a periodic color change at every crest of the meander, as
illustrated in the inset of Fig. 11(b). This color change is identified as a severe out-of-plane deformation of the interconnect
and will be discussed in detail further on. In addition to the
out-of-plane deformation, it is found that the PDMS substrate
has a cleavage failure when stretching for 200% elongation.
Surprisingly, at this extreme elongation, the Cu track is still
conducting. No interfacial delamination was observed between
the Cu and the PI layer. In other words, the Cu is very well
protected by the PI layer, although the substrate already failed.
D. In Situ Electromechanical Measurements: One-Time
Stretching and Cyclic Stretching
Fig. 12 shows the result of an electromechanical measurement of a PI-enhanced stretchable interconnect sample during
one-time stretching. The electrical resistance was continuously
recorded until metal rupture. Unlike our prior investigation
[1], [12], the measured electrical resistance of the PI-enhanced
stretchable interconnect is not constant but changes with elongation. Starting from 110% elongation, the electrical resistance
gradually increases. This phenomenon can be explained by
the fact that multiple microcracks initiate and propagate in
the metal. This microcrack propagation is retarded and held
from complete metal rupture by the PI layer. This gives the
possibility to extend or maximize the ultimate elongation of
the stretchable interconnect. In this investigation, the ultimate
elongation is 250%, which is two times the maximum magnitude of elongation shown in our prior research [1], where no PI
layer was used. It should be noted that the onset of interfacial
delamination between the PI layer and the PDMS substrate is
marked at 65% elongation, as shown in the inset of Fig. 12.
Even though this delamination exists, the metal is still fully

2686

IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 58, NO. 8, AUGUST 2011

Fig. 12. Normalized resistance as a function of elongation by in situ electrical


measurement.

Fig. 14. (a) Surface topography obtained from the optical interfometry and
its corresponding optical image at 40% elongation. (b) Deformation of PIenhanced stretchable interconnect at 40% elongation obtained from FEM.

Fig. 13. Power-law fitting of elongation versus number of stretching cycles


(E-N curve).

encapsulated and protected by the PI layer, which functions as


a barrier against failure. It is clear that the PI layer provides an
extra protection of the metal conductor.
Fig. 13 shows a power law fitting of the elongation versus
the number of stretching cycles to failure. The experiments
were performed with different levels of elongation, starting
from 30% up to 60%. The strain rate was controlled at 1%
per second and the electrical resistance was measured every
5 s. It was observed that the PI-enhanced stretchable interconnect withstands more than 40 000 stretching cycles at 30%
elongation. This number of stretching cycles to failure is a
prominent improvement (470) compared to our prior results
[2] on samples without a PI layer, in which the number of
stretching cycles to failure at the same elongation was only 85.
Even for elongations up to 60%, the PI-enhanced stretchable
interconnect can still survive more than 1000 cycles. This
prominent improvement of both one-time stretching and cyclic
stretching proves the robustness and effectiveness of the PIenhanced stretchable interconnect, designed through the DOE
analysis.

Fig. 15. Plastic strain distribution along the curve length of one unit of
stretchable interconnect.

E. Numerical Modeling
To confirm the aforementioned conclusion of out-of-plane
deformation (see Fig. 11), both optical profilometry (white
light interferometry) and finite-element simulations are done.
Fig. 14(a) left shows the surface topography of the stretched
PI-enhanced interconnect at 40% elongation as obtained from
the optical interferometer. Fig. 14(a) right is the corresponding
optical image. It is clear that the PI-enhanced stretchable interconnect emerges from the substrate due to substrate thinning.
Furthermore, it is found that every crest deforms out of plane
from the substrate, creating a noncontinuously emerged pattern
(or M-shape pattern). This out-of-plane deformation of every
crest can be explained by the numerical simulation. Fig. 14(b)

HSU et al.: PI-ENHANCED STRETCHABLE INTERCONNECTS: DESIGN, FABRICATION, AND CHARACTERIZATION

2687

Fig. 16. Stress (xx ) distribution in the longitudinal direction (stretching direction) and its local free-body diagram at the crest.

shows the iso-view and top view (inset) of the deformed sample
at 40% elongation. The crest deforms out of plane, and the arms
deform by local twisting. This simulated deformation explains
the emerged noncontinuous pattern.
The prominent improvement of one-time stretching and
cyclic stretching compared to samples without PI can be explained by the redistribution of the plastic strain in the metal due
to the PI presence. Fig. 15 shows the maximum plastic strain
distribution along the curve length of one unit of stretchable
interconnect at 40% elongation. Taking advantage of symmetry,
one-fourth of the plastic strain distribution along the one unit
of horseshoe meander is plotted. It is found that the plastic
strain is widely distributed along the curve length, instead of
concentrating on the local area of the crest as is the case for
samples without PI [2]. This indicates that the PI layer effectively redistributes the plastic strain in the metal and therefore
improves the reliability of the stretchable interconnect. In our
previous research [2], [12], we showed that in the samples
without PI, the failure is located at every crest of the horseshoe.
However, the PI-enhanced stretchable interconnect is expected
to fail (metal fracture) around the arms. This failure location
of the PI-enhanced meander can be explained by the stress
distribution in the metal. By plotting the longitudinal tensile
stress component (xx ) at 40% elongation on the left side of
Fig. 16, it is shown that the crest of the PI-enhanced meander
has a compressive stress, which is known not to cause fast
metal rupture (tensile stress is more dangerous). Furthermore,
this compressive stress is also responsible for the out-of-plane
deformation at the crest. Fig. 16 right shows the free-body
diagram. When the PI-enhanced stretchable interconnect is subjected to elongation, not only tensile stress but also a bending
moment is generated in the metal. The tensile stress (xx )
mainly comes from the longitudinal stretching, and the bending
moment (M) is caused by the geometrical opening. It has been
proven that the geometrical opening contributes to the majority
of the elongation [1], [12]. Consequently, the bending moment
dominates the stress distribution in the metal. Because of the
PI layer (dashed lines in Fig. 16 upper right), the bending
moment from the geometrical opening generates a gradient of

compressive stress (M ) in the metal. As a result, combining


the tensile stress from stretching with the compressive stress
from bending, a resultant pure compressive stress (rst ) with
gradient at the crest of the PI-enhanced meander is found.
V. C ONCLUSION
In summary, by adding a PI layer around the stretchable Cu
interconnect, higher stretchability and reliability is achieved.
The optimal design parameters were obtained through a finiteelement-simulation-based DOE. From the DOE analysis, it has
been found that the width of PI has the largest effect among
other significant parameters (thickness and elastic modulus of
PI). By increasing the width, thickness, and elastic modulus
of the PI, one can effectively reduce the plastic strain in the
Cu metal. In addition, the response surfaces and the cubic plot
provide a pathway for evaluation and optimization of the design
parameters.
PI-enhanced stretchable interconnects were fabricated according to the optimized design parameters. The experimental results demonstrate that the PI layer indeed effectively
improves the ultimate elongation (250% elongation) and reliability (40 000 cycles at 30% elongation). Moreover, there
is no interfacial delamination between the PI layer and the
metal during stretching. In other words, the metal interconnect
is completely protected by the PI layer, even after substrate
cleavage failure at extreme elongation.
R EFERENCES
[1] Y. Y. Hsu, M. Gonzalez, F. Bossuyt, F. Axisa, J. Vanfleteren, and
I. De Wolf, The effect of encapsulation on deformation behavior and failure mechanism of stretchable interconnects, Thin Solid Films, vol. 519,
no. 7, pp. 22252234, Jan. 2011.
[2] Y. Y. Hsu, M. Gonzalez, F. Bossuyt, F. Axisa, J. Vanfleteren, and
I. De Wolf, Reliability assessment of stretchable interconnects, in Proc.
IEEE IMPACT, 2010, pp. 14.
[3] M. Y. Cheng, C. M. Tsao, and Y. J. Yang, An anthropomorphic robotic
skin using highly twistable tactile sensing array, in Proc. IEEE ICIEA,
2009, pp. 650655.
[4] D. H. Kim, Y. S. Kim, J. Amsden, B. Panilaitis, D. L. Kaplan,
F. G. Omenetto, M. R. Zakin, and J. A. Rogers, Silicon electronics on

2688

[5]
[6]
[7]

[8]

[9]
[10]
[11]

[12]

[13]
[14]

IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 58, NO. 8, AUGUST 2011

silk as a path to bioresorbable, implantable devices, Appl. Phys. Lett.,


vol. 95, no. 13, p. 133701, Sep. 2009.
D. S. Gray, J. Tien, and C. S. Chen, High-conductive elastomeric electronics, Adv. Mater., vol. 16, no. 5, pp. 393397, Mar. 2004.
H. Jiang, Y. Sun, J. A. Rogers, and Y. Huang, Mechanics of precisely
controlled thin film buckling on elastomeric substrate, Appl. Phys. Lett.,
vol. 90, no. 13, p. 133119, Mar. 2007.
L. Lin and K. Jain, Design and fabrication of stretchable multilayer selfaligned interconnects for flexible electronics and large-area sensor arrays
using excimer laser photoablation, IEEE Electron Device Lett., vol. 30,
no. 1, pp. 1417, Jan. 2009.
S. Kim, J. Wu, A. Carlson, S. H. Jin, A. Kovalsky, P. Glass, Z. Liu,
N. Ahmed, L. Elgan, W. Chen, M. Ferreira, M. Sitti, Y. Huang, and
J. A. Rogers, Microstructured elastomeric surfaces with reversible adhesion and examples of their use in deterministic assembly by transfer
printing, Proc. Nat. Acad. Sci. U.S.A., vol. 107, no. 40, pp. 17 095
17 100, 2010.
S. Wagner, S. P. Lacour, J. Jones, P. I. Hsu, J. C. Sturm, T. Li, and Z. Suo,
Electronic skin: Architecture and components, Phys. E, vol. 25, no. 2/3,
pp. 326334, Nov. 2004.
T. Li, Z. Suo, S. P. Lacour, and S. Wagner, Compliant thin film patterns
of stiff materials as platforms for stretchable electronics, J. Mater. Res.,
vol. 20, no. 12, pp. 32743277, 2005.
Y. Y. Hsu, M. Gonzalez, F. Bossuyt, F. Axisa, J. Vanfleteren, and
I. De Wolf, In situ observation on deformation behavior and stretching
induced failure of fine pitch stretchable interconnect, J. Mater. Res.,
vol. 24, no. 12, pp. 35733582, 2009.
Y. Y. Hsu, M. Gonzalez, F. Bossuyt, F. Axisa, J. Vanfleteren, and
I. De Wolf, The effect of pitch on deformation behavior and the
stretching-induced failure of a polymer-encapsulated stretchable circuit,
J. Micromech. Microeng., vol. 20, no. 7, p. 075036, Jul. 2010.
MSC Marc User Manual, MSC Software, Santa Ana, CA, 2008.
D. C. Montgomery, Design and Analysis of Experiments. New York:
Wiley, 2001.

Yung-Yu Hsu received the M.Sc. degree from the


National Tsing-Hua University, Hsinchu, Taiwan, in
2001. He is currently working toward the Ph.D.
degree with the Department of Metals and Applied
Materials Engineering, Katholieke Universiteit Leuven, Leuven, Belgium.
He was with the Industrial Technology Research
Institute, Hsinchu, as a Research Engineer for five
years. During which, he also served as a Project
Leader for several government-funded national-level
projects. He has been with IMEC, Leuven, since
2008. He is the author of more than 20 peer reviewed papers in the areas of
microelectronic packaging, MEMS, and nanotechnology. He is the holder of
four U.S. patents and nine patents granted in Taiwan and China.

Mario Gonzalez received the M.Sc. degree in materials science from the University of Nuevo Leon,
Nuevo Leon, Mexico, in 1996 and the PhD degree in
mechanical and materials engineering from the Ecole
Centrale Paris, Paris, France, in 2001.
Since then, he has been with IMEC, Leuven,
Belgium, as a Research Scientist. He is the author or
a coauthor of more than 80 papers in areas including
numerical FE simulation, mechanical and thermomechanical reliability analysis of packaging, MEMS,
and stretchable interconnections.

Frederick Bossuyt was born in Kortrijk, Belgium,


on September 15, 1983. He received the degree
in electrical engineering and the Ph.D. degree in
electrical engineering from Ghent University, Ghent,
Belgium, in 2006 and 2011, respectively.
Since 2006, he has been with the Center for
Microsystems Technology, which is part of Ghent
University and IMEC, Leuven, Belgium, where he
is involved in research on stretchable electronics
technologies.

Jan Vanfleteren received the Ph.D. degree in electronic engineering from the Ghent University, Ghent,
Belgium, in 1987.
He is currently a Senior Engineer and Project
Manager with the Center for Microsystems Technology, IMEC, Leuven, Belgium, and is involved in
the development of novel interconnection, assembly
and substrate technologies, particularly in the field
of flexible and stretchable electronics. In 2004, he
was appointed as Part-Time Professor with Ghent
University. He is the author or a coauthor of more
than 100 papers in international journals and conferences. He is the holder of
six patents/patent applications.
Dr. Vanfleteren is a Member of International Microelectronics and Packaging
Society.

Ingrid De Wolf (SM07) She has been with IMEC,


Leuven, Belgium, since 1989. She performed research on reliability physics of semiconductor devices, with special attention for mechanical stress
aspects and failure analysis. Since 1999, she heads
the group REMO (Reliability and Modelling), where
research is focused on test and reliability of MEMS
and IC-interconnect and packaging, including 3-D.
She is a part-time Professor with the Department of Metals and Applied Materials Engineering,
Katholieke Universiteit Leuven, Belgium.

You might also like