Professional Documents
Culture Documents
net/publication/249574453
CITATIONS READS
19 7,045
2 authors, including:
Erol Sancaktar
University of Akron
318 PUBLICATIONS 2,686 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Erol Sancaktar on 10 October 2017.
Abstract—This paper describes a novel fatigue life prediction methodology aimed at providing the
design engineer an easy fatigue life predictive tool using experimental data for thermo-mechanical
load cyclic fatigue under constant maximum load (Pmax ) and load ratio (R = Pmin /Pmax =
σmin /σmax ). This encompasses an integrated approach to joint testing, analysis and modeling.
Utilizing the proposed methodologies, we aim to predict the changes in fatigue life of the adhesive,
based on the whole spectrum of test variables including temperature, humidity and load ratio. For
this purpose, joints were prepared using stainless steel adherend specimens and a commercial silver-
filled electronically conductive adhesive, and tested under monotonic and cyclic fatigue conditions,
at 28◦ C, 20% relative humidity, 50◦ C, 90◦ C and elevated humidity levels. Load–number of cycles
(P –N ) curves were generated using two specimen geometries at two different load ratios (R), at a
cyclic frequency of 150 Hz. Using the experimental data, a life predictive methodology was developed
and validated. Furthermore, the usefulness of the above-mentioned fatigue life predictive capability
was extended to varying stress states.
Keywords: Electronically conductive adhesives; adhesive joints; fatigue damage; cyclic loading; P –N
curves; fatigue life predictive model.
1. INTRODUCTION
One of the major questions that needs to be addressed for electronically conductive
adhesive joints is the estimation and prediction of the real service life. Exposing the
adhesive to the major stresses and determining how these stresses affect its service
life can answer this question. In order to conduct service life evaluations, samples
potential sites of fatigue failure by calculating the average induced shear stress
occurring at the interfaces of the adhesive joint.
Several authors have employed fracture mechanics approach to predict
fatigue failure in adhesively bonded microelectronic devices [9]. Abdel Wahab
et al. [10] proposed a model for predicting the life of adhesively bonded joints.
In this study, they used the concept of continuum damage mechanics (CDM) and
fracture mechanics (FM) as tools in predicting the life of the joints, using carbon-
fiber-reinforced composite (CFRP) substrates bonded with an epoxy film adhesive.
Furthermore, a number of research articles have been published in recent years
dealing with fatigue life prediction of engineering structures [11 –25].
For solder bumps, much work has been carried out to establish a relation-
ship between cycling conditions, and the number of cycles to failure. Several
lifetime models, many of them based on the Coffin–Manson fatigue relationship,
have been proposed [26, 27]. Later work extended these models to circuits with
underfill [28].
In this paper, we propose a fatigue life prediction methodology for electronically
conductive adhesive joints cycled under constant maximum load (Pmax ) and load
ratio (R = Pmin /Pmax = σmin /σmax ). For this purpose, first we used a commercially
available electronically conductive adhesive (ECA), and generated load–number
of cycles (P –N ) curves under a wide spectrum of environmental conditions and
stress states. Using the experimental data obtained from the above experiments,
we proposed a fatigue life predictive model for constant-cycle loading conditions.
Furthermore, a comprehensive analysis and model of ECA joints subjected to
variable loading can be found in Ref. [29].
2. EXPERIMENTAL
In this study, stainless steel 304 and Ablebond MA-2 (Ablestik Laboratories,
Rancho Dominguez, CA, USA) ECA were used as the model adherend and
adhesive, respectively. A complete description of the bonding procedure, optical
fracture surface analysis and monotonic and fatigue testing procedures can be
found in Ref. [30]. Furthermore, complete fatigue and failure analyses of ECA
joints fabricated using Ablebond MA-2, and subjected to varying environmental
conditions, namely, ambient, elevated temperature and elevated humidity, can be
found in Refs [30 –32].
For the purpose of incorporating varied stress states in our model, two single-
lap joint configurations with varying stress-states were designed, referred to as
overhanging and non-overhanging joint geometries. Figures 1 and 2 show the
dimensions of the two joint geometries. In the case of the overhanging joint
geometry, the single-lap joint was gripped only partially, as shown in Fig. 1, leaving
a 6.18 mm wide portion of the specimen faces ungripped.
90 R. R. Gomatam and E. Sancaktar
First, in order to assess the state of stress in the joints, Non-Linear Elasto-Plastic
finite element analyses (FEA) were performed on both the overhanging and non-
overhanging type joints, using ALGOR© Finite Element Software.
Figures 3 and 4 show the FEA mesh patterns for the non-overhanging and
overhanging joints, respectively. The mesh pattern across the thickness of the joint
is shown in Fig. 5. These patterns involve finer meshing over the overlap region,
followed by a gradual change into a coarser mesh density, away from the overlap
region as shown in Figs 3 and 4. Similarly, Fig. 5 shows much finer meshing near
the adhesive layer with a gradual change to a coarser mesh density away from the
adhesive layer. It should be noted that, in the finite element models generated, the
mesh patterns were refined and optimized to obtain consistent stress values.
The adherend and adhesive material parameters used in the non-linear elasto-
plastic finite element analyses are given below:
Adherend:
Elastic modulus = 179 GPa
Poisson’s ratio = 0.3
Yield stress = 1.4 GPa
Strain hardening modulus = 95.8 GPa.
Adhesive:
Elastic modulus = 33 MPa
(a)
(b)
Figure 5. Finite element model meshing sequences. (a) Across the thickness of joint and (b) along
the adhesive layer.
values, respectively. The mathematical basis for the maximum principal stress to
replace the maximum cyclic load in equation (1) can be illustrated by considering
the following relations:
C3 C3
= , (2)
C1 C1
94 R. R. Gomatam and E. Sancaktar
and
C2 C2
= . (3)
C1 C1
Table 2 shows the values of the intercepts and their ratios computed based on
maximum cyclic load and maximum principal stress values, validating the use
Comprehensive fatigue life predictive model for ECA joints 95
Table 1.
Slope and intercept values calculated from P –N curves based on equation (1) for specimens tested
under different environmental conditions
Figure 10. Pictorial representation of the method for slope and intercept determination based on
maximum load and maximum principal stress.
Table 2.
Computed values of intercepts and their ratios based on maximum cyclic load (C ) and maximum
principal stress (C)
Environ. C1 C2 C3 C1 C2 C3 C3 /C1 C2 /C1 C3 /C1 C2 /C1
condition (kN) (kN) (kN) (MPa) (MPa) (MPa)
28◦ C, 3.095 3.175 3.42 108.73 114.09 120.58 1.105 1.03 1.109 1.049
20% RH
50◦ C 2.74 2.78 2.83 61.36 62.05 63.2 1.04 1.018 1.029 1.011
90◦ C 2.23 2.25 2.27 23.79 23.837 24.26 1.02 1.009 1.018 1.009
new environmental condition. Figures 12–15 show the design charts for slopes
and intercepts based on maximum cyclic load or maximum principal stress for the
stainless steel adherends adhesively bonded using Ablebond MA-2 ECA.
To test the efficiency of the methodology proposed, tests were performed at
R = 0.1, Pmax = 2.1 kN, under 28◦ C, 50◦ C and 90◦ C test temperatures. Compari-
son between experimental and predicted values shown in Table 4 for the test cases
indicated above clearly validates the accuracy and efficiency of the life predictive
methodology proposed.
The modeling procedure described above was also extended for application to
varying state of stress. For this purpose, a non-linear elasto-plastic finite element
98 R. R. Gomatam and E. Sancaktar
Table 3.
Slope and intercept values based on maximum principal stress (equation (4)) calculated from S–N
curves for specimens tested under different environmental conditions
Table 4.
Results from fatigue predictive model utilizing design charts
analysis was performed for ambient condition (28◦ C, 20% RH) and a non-linear
thermo-elasto-plastic finite element analysis was performed for elevated tempera-
ture conditions to represent the overhanging and non-overhanging joints geometries
as two distinct states of stress. Figure 16 shows a pictorial representation of ambient
condition P –N curves for the two different stress states.
Utilizing the results from the finite element analyses, the relation between the
stress components, slopes, and intercepts were established, as shown by the relations
Comprehensive fatigue life predictive model for ECA joints 99
Figure 11. Pictorial representation of the shifting methodology used in establishing design charts for
predicting fatigue life.
Figure 12. Design chart for slope based on maximum load (equation (1)).
Figure 13. Design chart for intercept based on maximum load (equation (1)).
and could be superposed onto the entire P –N curve. On the other hand, the
maximum shear stress, τxz , affects the slope of the P –N curve, which reflects the
time-dependent behavior of crack propagation. This can also be inferred from the
FEA analyses results, which indicate higher τxz stress values in the overhanging
region of the overhanging joint geometry. We believe that this behavior results
in a higher slope for the S–N curve for the overhanging joint geometry, which
Comprehensive fatigue life predictive model for ECA joints 101
Figure 14. Design chart for slope based on maximum principal stress (equation (4)).
Figure 15. Design chart for intercept based on maximum principal stress (equation (4)).
diminishes for the non-overhanging counterpart. Thus, by using the above relations
along with a knowledge of the states of stress for any two joint configurations, one
of which has already been characterized for P –N behavior, the slope and intercept
for the other joint with unknown P –N behavior can be computed. These values
can subsequently be substituted into equations (5) and (6), to predict the fatigue life
of the joint with unknown fatigue (P –N ) behavior. Tables 5 and 6 show the ratios
of intercepts compared with the ratios of maximum principal and normal stresses
102 R. R. Gomatam and E. Sancaktar
Figure 16. Pictorial representation of P –N curves for two different stress states.
Table 5.
Computed ratios for intercepts, principal stresses and the stress components, σxx ,
σyy and σzz
The ratios are between the overhanging (OH) and non-overhanging (NOH) joint
geometries.
Table 6.
Computed ratios for slopes and shear stress components, τxy and τyz
The ratios are between the overhanging (OH) and non-overhanging (NOH) joint
geometries.
Comprehensive fatigue life predictive model for ECA joints 103
(Table 5), as well as the ratios of the slopes compared with the ratios of all shear
stress components (Table 6), validating the usage of equations (5) and (6) in our
fatigue life predictive model for different states of stress.
A novel life prediction methodology has been proposed in this paper, using slope
and intercept values computed from the P –N curves on a linear scale and under
constant-cycle loading condition. For this purpose, an easy-to-use life predictive
methodology was proposed, which utilizes a set of design charts. The design charts
are comprised of two parts, one for shifting the slopes and the other for shifting the
intercepts between different environmental conditions. Using these design charts,
the slope and intercept values for an unknown condition can be computed, which
would predict the total fatigue life of the joint. Furthermore, the proposed model
was extended to samples with varying stress states. For this purpose, a non-linear
elasto-plastic finite element analysis (FEA) for ambient condition, and a non-linear
thermo-elasto-plastic FEA for elevated temperature conditions were performed
using overhanging and non-overhanging joint geometries to represent two distinct
states of stress. Utilizing the FEA results, relations between the stress components,
slopes, and intercepts were established. The intercepts were found to be inversely
proportional to the maximum stresses, and the slopes directly proportional to the
shear-stress component τxz . Thus, by using these relations along with a knowledge
of the stress states for only two joint configurations, one of which has already been
characterized for the P –N behavior, the slope and intercept for the other joint with
unknown P –N behavior could be computed, from which the fatigue life of the joint
could be predicted.
REFERENCES
1. E. Sancaktar, in: Engineered Materials Handbook Volume 3: Adhesives and Sealants,
H. F. Brinson, Technical Chairman, pp. 349–372. ASM International, Materials Park, OH (1990).
2. A. Rotem and H. G. Nelson, in: Fatigue of Fibrous Composite Materials, ASTM STP. 723,
K. N. Lauraitis (Ed.), p. 152. ASTM, Philadelphia, PA (1981).
3. Y. Miyano and M. K. McMurray, J. Composite Mater. 28, 1250 (1994).
4. C. T. Sun and W. S. Chan, in: Composite Materials, Testing and Design, STP 674, S. W. Tsai
(Ed.), p. 418. ASTM, Philadelphia, PA (1979).
5. R. A. Schapery, Int. J. Fracture 11, 141 (1975).
6. E. Dan-Jumba, S. G. Zhou and C. T. Sun, in: Advances in Thermoplastic Matrix Composite
Materials, ASTM STP 1044, G. M. Newaz (Ed.), p. 113. ASTM Philadelphia, PA (1989).
7. E. Suhir, in: Proceedings of the 37th IEEE Electronic Components Conference, p. 143 (1987).
8. A. Gladkov and A. Bar-Cohen, in: Proc. 3rd Int. Conference on Adhesive Joining and Coating
Technology in Electronic Manufacturing, Binghampton, NY, J. H. Constable (Ed.), p. 116
(1998).
9. T. Hattori, S. Sakata and G. Murakami, J. Electron. Packag. 111, 243 (1989).
104 R. R. Gomatam and E. Sancaktar
10. M. M. Abdel Wahab, I. A. Ashcroft, A. D. Crocombe and S. J. Shaw, J. Adhesion Sci. Technol.
15, 763 (2001).
11. P. J. Singh, D. R. G. Achar, B. Guha and H. Nordberg, Int. J. Fatigue 25, 1 (2003).
12. N. Himmel, Int. J. Fatigue 24, 349 (2002).
13. J. Schon and A. Blom, Int. J. Fatigue 24, 361 (2002).
14. M. M. Abdel Wahab, I. A. Ashcroft, A. D. Crocombe and P. A. Smith, Int. J. Fatigue 24, 705
(2002).
15. M. D. Halliday, C. Cooper, P. Poole and P. Bowen, Int. J. Fatigue 25, 709 (2003).
16. M. M. K. Lee and D. Bowless, Fatigue Fract. Eng. Mater. Struct. 25, 1025 (2002).
17. H. Adib and G. Pluvinage, Int. J. Fatigue 25, 67 (2003).
18. C. S. Lee and W. Hwang, Polym. Composit. 21, 798 (2000).
19. A. Singh, J. Mech. Design, Trans. ASME 125, 540 (2003).
20. S.-B. Lee and J.-K. Kim, Int. J. Fatigue 19, 85 (1997).
21. N. Hong and L. Shabo, Int. J. Fatigue 19, 517 (1997).
22. F. Morel, Fatigue Fract. Eng. Mater. Struct. 21, 241 (1998).
23. P. J. Laz and B. M. Hillberry, Int. J. Fatigue 20, 263 (1998).
24. J. S. Huh and W. Hwang, Composit. Struct. 44, 163 (1999).
25. Y. I. Kwon and B. S. Lim, Metals Mater. Intl. 7, 311 (2001).
26. H. Solomon, IEEE Trans. CHMT 9, 423 (1986).
27. W. Engelmaier, IEEE Trans. CHMT 12, 284 (1989).
28. V. Gektin, A. Bar-Cohen and S. Witzman, IEEE Trans. CPMT Part A 21, 577 (1998).
29. R. R. Gomatam, Modeling fatigue behavior of electronically conductive adhesives, PhD
Dissertation, The University of Akron, Akron, OH (2002).
30. R. R. Gomatam and E. Sancaktar, J. Adhesion Sci. Technol. 18, 731 (2004).
31. R. R. Gomatam and E. Sancaktar, J. Adhesion Sci. Technol. 18, 849 (2004).
32. R. R. Gomatam and E. Sancaktar, J. Adhesion Sci. Technol. 18, 1833 (2004).