Microelectronics Reliability 63 (2016) 134–141

Contents lists available at ScienceDirect

Microelectronics Reliability

j ourna l homepage: www.e lsev ie r .com/ locate /mr

Modified single cantilever adhesion test for EMC/PSR interface in thinsemiconductor packages

Kenny Mahan a, Byung Kim a, Bulong Wu a, Bongtae Han a,⁎, Ilho Kim b, Hojeong Moon b, Young Nam Hwang c

a Mechanical Engineering Department, University of Maryland, College Park, MD 20742, USAb Package Development Team, Semiconductor R&D Center, Samsung Electronics, 1, Samsungjeonja-ro, Hwaseong-si, Gyeonggi-do, Koreac Measurement and Analysis group, Corporate R&D Institute, Samsung Electro-Mechanics Co., 150, Maeyoung-ro, Yeongtong-gu, Suwon-si, Gyeonggi-do, Korea

⁎ Corresponding author.E-mail address: bthan@umd.edu (B. Han).

http://dx.doi.org/10.1016/j.microrel.2016.05.0150026-2714/© 2016 Elsevier Ltd. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Article history:Received 2 April 2016Received in revised form 26 May 2016Accepted 26 May 2016Available online 29 June 2016

We propose and implement an adhesion test configuration called “modified single cantilever adhesion test” (M-SCAT) that can be employed to determine the adhesion strength of epoxy molding compound (EMC) and photosolder resist (PSR) interface in thin semiconductor packages. The proposed M-SCAT method is optimal for quickand quantitative in-situ testing of the interface with strong adhesion as sample preparation and testing are sim-ple while maintaining a lowmodemixity at the crack tip. Detailed sample preparation and experimental testingto determine the critical load required for delamination are presented. A numerical procedure is followed to as-sess the stress and strain fields around the crack tip at the point of delamination, thus allowing for the J-integralmethod to be employed to determine the critical energy release rate. The proposed approach is carried out fortwo different EMC/PSR interfaces. The results show excellent repeatability, which allows for the test method tobe used effectively to select the most ideal material set for given applications.

© 2016 Elsevier Ltd. All rights reserved.

Keywords:Adhesion strengthSingle cantilever adhesion testEnergy release rateJ-integralEMCEpoxy molding compoundPhoto solder resistPSR

1. Introduction

With the advancement of thin-profile designs, in-situ adhesionstrength testing of products directly off of manufacturing lines is a crit-ical challenge. If potential delamination failures due to inadequate adhe-sion strength can be detected early in the design cycle, unreliabledesigns can be ruled out leading to quicker product turn around. Thisis especially important in the fast-pace environment of the electronicpackaging industry where quick and effective adhesion strength testingmethods are required.

One such interface thatmust address these challenges is the high ad-hesion strength interface found between epoxy molding compound(EMC) and photo solder resist (PSR). The interface is shown in Fig. 1,where the PSR is sandwiched between the EMC and the printed circuitboard (PCB) substrate. The EMC/PSR interface can delaminate duringmanufacturing processes and/or operating conditions. Thus, it is criticalto assess the adhesion strength of any newly proposed material combi-nations to assure the adhesion strength of selected material sets isstrong enough for the intended product application.

The adhesion strength is often characterized using the energy re-lease rate of the interface [1]. The energy release rate, G, is a measureof the energy available for an increment of crack extension. When a

specimen with an interfacial crack is loaded, energy is stored insideuntil a critical loading state is reached. At this point, the crack propa-gates, and energy is released from the systemwhile creating a new sur-face area along the interface. For this case, G can be evaluated directlyfrom a typical load vs. displacement graph either analytically or throughthe aid of numerical methods. Two such methods are the J-integralmethod [1–6] and the virtual crack closure technique (VCCT) method[2,7–9].

The second critical property to consider is the mode mixity at thecrack tip. The mode mixity can be defined as the ratio of in planeshear (mode II) to opening (mode I) loading at the crack tip [1,7,8,10,11]:

tan−1Ψ ¼ffiffiffiffiffiffiGII

GI

sð1Þ

where GI and GII represent the mode I and II energy release rate contri-bution at the interface, respectively. The energy release rate is known toincrease with respect to themodemixity due to the increased presenceof in-plane shear loading; when comparing different material sets, it isimportant to compare them at similar mode mixity values [1,7,8,10,11]. The larger the mode mixity, the more susceptible the interfacialcrack will be to kinking out of the interface [12–15]. Additionally,

Fig. 1. Close-up SEM image of thin multilayer stacks consisting of EMC, PSR, and PCB.

Fig. 2. Schematic of (a) test configuration for M-SCAT specimen with a precrack and (b)load vs. displacement graph for a M-SCAT test.

135K. Mahan et al. / Microelectronics Reliability 63 (2016) 134–141

crack kinking is of greater risk when dealing with high adhesionstrength interfaces and adherends with low fracture toughness.

Energy release rate testing methods, such as the double cantileverbeam (DCB) method [11,16–22] and the four point bending (4 PB)method [10,11,23–39], have been practiced widely, but when facedwith increasingly strong interfaces and thinner layers of interest, com-plications in testing arise. For instance, the DCB testing offers nearmode I energy release rates by applying opening mode loading to theend of a beam specimen. However, DCB and unsymmetric double canti-lever beam (UDCB) methods require complex sample preparation tomaintain nearmode I loading [20,21]when dealingwith thinmultilayerstructures. Due to the elasticmismatch at the interface, even in a config-uration with a global mode I loading condition, such as DCB and UDCB,there is still a non-zero mode mixity at the crack tip [21,40].

The 4PB test method offers the advantage of crack length indepen-dent energy release rates due to the constant moment region insidethe loading pins. Due to the global loading condition, however, themode mixity at the crack tip is much larger in this configuration [10].When faced with large crack tip mode mixity and a high adhesionstrength interface, it can be difficult to obtain delamination along the in-terface before the crack kinks into an adjacentmaterial. This issue is ap-parent for the EMC/PSR interface due to the increasingly large high-fillercontent of EMCs that are being used in newer products. High-filler con-tent leads to a reduced EMC fracture toughness, whichmakes a precrackat the PSR/EMC interface more susceptible to kinking out of theinterface.

The single cantilever beam (SCB) method was inspired by the DCBmethod and focused on assessing the higher adhesion strengths of thin-ner interface materials between flexible and rigid layers [41–44]. In thecases described, however, the flexible and rigid layers were the samematerial with the rigid layer reinforced in some manner to increase itsflexural strength. This allowed the closed form solutions of the energyrelease rate.

Although effective for a certain set of applications, the methods de-scribed above are not ideally applicable to the particular challenges ofthe EMC/PSR interface (i.e., very high adhesion strength and low frac-ture toughness of EMC). This paper proposes and implements an adhe-sion test configuration called the “modified single cantilever adhesiontest” (M-SCAT) that can handle the two challenges effectively. Follow-ing detailed sample preparation and experimental testing to determinethe critical load required for delamination, a numerical procedure to de-termine the critical energy release rate and the mode mixity of the

interface is described. The results obtained from the proposed approachare presented to discuss the effectiveness of the proposed M-SCATmethod.

2. Modified single cantilever adhesion test (M-SCAT) method

2.1. Physical description

The M-SCAT configuration comprises of a thin multilayer specimenwith a precrack at the interface-of-interest that is secured to asupporting block. The material above the interface-of-interest extendsfurther so that a cantilever arm is offset past the fixed area to allowfor direct loading. A schematic of a typical M-SCAT specimen and testconfiguration can be seen in Fig. 2a.

The specimen is secured to amechanical test stand and pin loading isapplied to the offset cantilever arm. A typical load vs. displacementcurve that can be expected from the M-SCAT is shown in Fig. 2b. Asthe displacement increases monotonically, the load increases until acritical load, PCrit, is reached and delamination begins at the end of thepredefined area. From a load vs. displacement graph the critical loadwill be the maximum load on the graph. Once delamination begins,stored energymust be released from the body to create the new surface

Fig. 4.Mode mixity vs. moment arm length for E1/E2 = 5 and h1 = 1mm in M-SCAT testconfiguration.

136 K. Mahan et al. / Microelectronics Reliability 63 (2016) 134–141

area. In the experimental results this is seen as a load drop correspond-ingwith the beginning of delamination. Since the test is run in displace-ment control, the displacement will continue to increase, resulting insteady delamination along the interface of interest until the test isstopped.

2.2. Numerical analysis

A numerical analysis is employed to extract the energy release rateandmodemixity from the experimental results. A 2Dmodel of the spec-imen is first established with plane strain elements. A meshing strategycan be used to minimize the total number of elements while stillretaining a dense mesh at the crack tip. Maintaining a concentratedmesh near the crack tip is essential for thin structures as it allows formore accurate stress and strain fields to be assessed. A sensitivity anal-ysis should be run to assure that the element size near the crack tip issmall enough to be mesh independent. The model can then be run toevaluate the critical state by applying the critical load from the experi-ment. From the results a post-processing routine can be run to evaluatethe critical energy release rate, GCrit, and the mode mixity, Ψ, of theinterface.

2.3. Mode mixity in M-SCAT method

When establishing the specimen dimensions for the test, it is impor-tant to consider the mode mixity at the crack tip during testing. For theM-SCAT, themost critical parameters that influence themodemixity atthe crack tip are the modulus, E, and thickness, h1, of the top materialand the moment arm of the applied loading. When designing a speci-men, the material properties of the top layer are typically fixed due tothe interface-of-interest being tested. The specimen designer can stilloptimize the height of the top layer to minimize the mode mixity atthe crack tip. By increasing the height of the top layer the flexural rigid-ity of the loaded beam is greatly increased, thus resulting in a greatermode I opening loading conditionwhen pin loading is applied to the off-set cantilever beam. Using FEAmodeling this can be clearly shown for agiven material pair.

Consider a three-layer specimen with a predefined area precrackplaced between the top and middle layers creating a moment arm,larm, of 5 mm. The bottom layer has a thickness of h3 = 300 μm and aneffective modulus of E3 = 27.5 GPa. The middle layer is a thin filmlayer of thickness, h2 = 25 μm and the modulus of the top layer is

Fig. 3.Modemixity vs. upper layer thickness for severalmaterial combinations inM-SCATtest configuration.

E1=24 GPa. These parameters were set as a baseline for later consider-ation of a specific case on interest, while the middle layer modulus, E2,and the upper layer height, h1, were allowed to vary to highlight differ-ent potential specimen configurations. Using the plane strain numericalmodel, the mode mixity at the crack tip vs. the increasing top layerthickness for several material combinations with different elastic mod-uli mismatches was calculated and plotted in Fig. 3. It is clear from thefigure that in each case increasing the height of the top layer willallow for a smaller crack tip mode mixity to be evaluated.

The second factor to consider is themoment arm. To evaluate the ef-fect of moment arm length on the mode mixity, the mode mixity wasplotted against multiple moment arm lengths in Fig. 4. For the plot thecase closest to the real sample from Fig. 3 was selected with themoduliE1/E2 = 5, and an upper layer thickness of h1 = 1 mm. As the momentarm length increases, the mode mixity at the crack tip also increases.The results suggest that the shortest moment arm length should beused during testing to minimize mode mixity.

3. Test preparation

Two separatematerial sets consisting of an EMC/PSR interface on topof a thin PCB were selected to implement the M-SCAT procedure. Eachmaterial set is consistent except for two potential PSRmaterials of inter-est (will be referred to as PSR1 and PSR2).

3.1. Specimen design

While the PSR and PCB thicknesses were fixed due tomanufacturingspecifications, the thickness of the EMCwas a variable to consider whendesigning the specimen. The previous numerical analysis (Fig. 3) clearlyindicated that the largest possible EMC thickness should be selected tominimize the mode mixity. Due to height restrictions of the mold, themaximum allowable EMC height was 950 μm. Following this logic, the

Table 1Material properties for the EMC/PSR specimen.

Material Thickness Modulus Poisson's ratio

EMC 950 μm 24 GPa 0.28PSR 25 μm 4.2 GPa 0.38PCB 300 μm 27.5 GPa 0.22

Fig. 5. Plot of 3D energy release rate normalized by plane strain energy release rate vs.specimen width for a EMC/PSR M-SCAT sample.

137K. Mahan et al. / Microelectronics Reliability 63 (2016) 134–141

mode mixity of the specimen configuration was ~28.5°. Material prop-erties of the individual layers can be found in Table 1.

Another key consideration for creating samples is to make sure thespecimen is under the plane strain condition. Plane strain testing is re-quired for assessing the critical energy release rate as an intrinsic mate-rial property since a specimen in plain strain has themost concentratedtri-axial stress state [1]. After the cross-section of the specimen has been

Fig. 6. Sample preparation steps: a) schematic of board layout with dicingmarkers and plannedand c) finished silicone oil/aluminum sputtered predefined area on PSR surface.

designed to obtain thedesiredmodemixity at the crack tip, a 3Dnumer-ical analysis was performed to assess the minimum specimen thicknessrequired to maintain a plane strain state in the specimen.

The energy release rate was assessed at the center of themodel. Thisprocesswas repeated over several specimenwidths. Fig. 5 shows the re-sults normalized by the energy release rate obtained from the 2-D planemodel (G3D/GPlane-strain) and plotted as a function of the specimenwidth.As expected, the specimen approaches a plane strain condition as thesamplewidth increases. For this specific sample configuration, the sam-ple is within 1% of a plane strain configuration with a width of 16 mm.Since any further increase in width only fractionally increases theplane strain condition and considering the difficulty in testing verywide specimens, a 16 mm sample width was selected for all samples.

3.2. Specimen preparation

Sample sets were prepared in individual boards designed to fit witha typical molding process in a manufacturing environment. Fig. 6a de-picts a schematic of the board layout used for creating a sample batch.Each PCB board was created with a top PSR layer and with dicingmarks etched on the bottom of the board to facilitate laser cutting of in-dividual specimens at the end of specimen preparation.

The most important function of this specimen fabrication is the cre-ation of a low-adhesion predefined area at the interface-of-interest. Thepredefined area provides a thin, natural crack front that is essential toaccurate testing using the M-SCAT method. To create the predefinedarea (PDA), the following approach was implemented.

A temporary mask was placed on the PSR surface to establish thelength of the predefined area for two specimens on the row (Fig. 6a).A thin layer of silicone oil was applied in between the masked area

predefined area, b) temporarymasked area on top of PSR/PCB layers for predefined area,

138 K. Mahan et al. / Microelectronics Reliability 63 (2016) 134–141

and allowed to dry. An aluminum layer was then deposited bysputtering aluminum over this area. The masked area was then re-moved, revealing the finished predefined area as seen in Fig. 6c.

The silicone oil greatly reduced the adhesion in this area, which re-sulted in very small loading requirements to delaminate this area. Thealuminum sputtering layer acted as a barrier layer between the siliconerubber and the top adherend. This barrier layer was critical since thespecimen went through the high pressure and temperature EMCmold-ing process.Without the aluminum sputtered layer, the predefined areacould be modified undesirably (non-ideal plane crack).

The board was then put through the molding process to create theupper EMC layer. A cross-sectional view of the PCB/PSR/EMC specimenis shown schematically in Fig. 7 along with a close-up SEM image of thesilicone oil/aluminum layers of the predefined area. It is clear throughthe SEM image that a thin, continuous silicone oil layer remains intactafter the molding process on top of the PSR layer due to the sputteredaluminum layer. Specimens were then laser cut using the dicingmarks shown in Fig. 6a to establish individual specimens of the desiredlength and width.

Fig. 8a highlights two specimens from a given row on the sampleboard. Cut A in Fig. 8a was made via the laser to arrive at the individualspecimen. Prior to testing, the PCB side of the specimenwas bonded to arigid aluminum support block using a thin layer of a structural epoxy. Inthis configuration, the end of the specimen is aligned with thesupporting block with part of the predefined area offset over the edgeof the fixed support (Fig. 8b). During this process, the sides of the spec-imenwere temporarily coveredwith tape to prevent epoxy frompoten-tially bonding to the interface-of-interest. Any excess epoxy at theinterface-of-interest could potentially lead to artificially large adhesionstrength results. After the epoxy cured over a twelve-hour period, the

Fig. 7. Cross-sectional schematic view of predefined area creation for EMC/PSR M-SCATspecimens along with an inset of a zoomed in SEM image of silicone oil and aluminumsputtered layers of predefined area.

Fig. 8. Schematics of typical (a) two M-SCAT specimen with predefined area prior todicing, and (b) test configuration for a M-SCAT test.

tapewas removed. Finally the offset area under the interface-of-interestwas removed to allow for direct pin loading to the EMC(Cut B in Fig. 8a).

4. Experiment and results

4.1. Testing

For testing, the specimen was secured on a mechanical test standwith a specially designed test fixture (Fig. 9a). The test fixture shouldallow for direct loading of the EMCwith a greasedpinwhilemaintainingalignment throughout testing. A fixture that allowed for self-alignmentin the out-of-plane direction was ideal to prevent twisting during test-ing which would result in erroneous test results. A camera was alsosetup to capture the initial pin position on the offset arm and to takevideo of the following delamination.

Once the specimen was set up, a displacement rate of 100 μm/s wasapplied to the pin monotonically until delamination occurred at theinterface at a critical load, Pcrit. As seen in Fig. 9b, the camera capturedthis initial crack propagation and the following steady state delamina-tion along the interface as the load decreased with increasing pindisplacement.

Following the test, the moment arm, larm, from the experiment wasdetermined for each specimen to use in the numerical model. Therewas a slight variation in the moment arm from specimen to specimenresulting from alignment of the loading pin and from the dicing of indi-vidual specimens. The initial position of the moment arm was deter-mined from a still image from the camera prior to testing. The offsetdistance from the center of the pin to the edge of the specimen wasthen recorded. Next by evaluating the delaminated side of the specimen

Fig. 9. (a) M-SCAT test setup with EMC/PSR specimen, test stand, loading fixture, andcamera used to monitor delamination; and b) in-situ camera image of delaminationoccurring at EMC/PSR interface during testing.

139K. Mahan et al. / Microelectronics Reliability 63 (2016) 134–141

the extent of the predefined areawas determined. Using the critical loadand moment arm, the numerical model is used to evaluate GCrit and Ψ.

4.2. Evaluation of adhesion strength

Specimens from both EMC/PSR1 and EMC/PSR2 sample sets weretested until a critical load was reached, resulting in a significant loaddrop and delamination along the interface. Each specimenwas reloadedseveral times, resulting in several distinct load drops corresponding toincreased delaminated area. Fig. 10 shows a load vs. normalized

Fig. 10. Comparison of load vs. normalized displacement curves for both EMC/PSRmaterial.

displacement curve for each material set where the relationship isbased off of simple cantilever beam displacement:

P ¼ EEMCwBh3

EMC

4l3arm

!ð2Þ

where EEMC represents the modulus of the EMC layer. This figure nor-malizes the data qualitatively, and quickly identify any potential testingoutliers due to bad sample preparation or testing alignment.

In Fig. 10, the first three critical loads and subsequent load drops aredenoted for each specimen shown. As seen in Fig. 11, by examining theEMC surface of a tested specimen, clear delamination can be seen at theonset of crack propagation corresponding to each peak load during test-ing. The delamination patterns varied between material sets. The EMC/PSR2 set showed much larger load drops following delamination whilethe PSR1 specimens showed very small incremental delamination cor-responding to each load drop. After testing the EMC layer was yankedoff to create the distinct change in pattern after the test. Since the onlydifference between sample sets is the PSR material used, this behaviorcan be attributed to differences in the PSR materials.

Testing for the EMC/PSR1 material set was repeated over six speci-mens. The average critical load and moment arm from the testing was50.5 N and 7.5 mm, respectively. Each sample was then evaluatedusing the 2D plane strain numerical model to determine the energy re-lease rate and mode mixity. The average GCrit was 147 J/m2 and the av-erage Ψ = 29.2°. The variation in the sample was only ±3% showingvery good repeatability for an adhesion strength test.

Six specimens were tested from the EMC/PSR2 material set with anaverage critical load andmoment arm of 48.2 N and 6.9mm, respective-ly. The average GCrit was 115 J/m2 and the average mode mixity was29.0°. The variation in the PSR2 set was noticeably larger at b12% varia-tion. This can be attributed to the difference in crack growth on thedelaminated surfaces. The delamination in the PSR2 set was larger andless stable than the PSR1 set. Additionally, for comparison purposes be-tween the two sets, the PSR1 set has a definitively stronger energy

Fig. 11. Delaminated surfaces from M-SCAT tested specimens with crack propagationcorresponding to load drops during testing for the a) EMC/PSR1 and b) EMC/PSR2material sets.

Fig. 12. Comparison of adhesion strength results for both EMC/PSR material sets.

140 K. Mahan et al. / Microelectronics Reliability 63 (2016) 134–141

release rate than the PSR2 set. A comparison of the energy release rate ofeach set can be seen in Fig. 12.

5. Discussion–validity of LEFM

In order to employ numerical approaches to assess the interfacial en-ergy release rate, namely the J-integral and VCCT methods, it is criticalto assure that the linear elastic fracture mechanics assumption is not vi-olated. For LEFM to bemaintained any plasticity near the crack tip mustbe limited to a very small area around the crack tip [1]. To verify that thisis the case duringM-SCAT testing a moiré interferometry approachwasemployed.

Fig. 13. Moiré fringe patterns on EMC/PSR specimen after delamination along interface

Moiré interferometry allows for the in-plane displacement fields to beassessed during testing by replicating a high frequency grating on thesample prior to testing [45]. Tuning this system prior to testing allowsfor the establishment of a null field of zero strain on the sample. After ap-plying load to the sample the interferometer can be used to assess the in-plane displacement field that occurs after calibrating a null field.

To assess if the plasticity that occurs is confined to a near crack tip re-gion, the M-SCAT specimen with the gratingwas tested until delamina-tion occurred at the interface-of-interest and then unloaded. Thespecimen was then returned to the moiré interferometer and the uxand uy displacement fieldswere assessed. Fig. 13 shows the residual de-formation after delamination seen in the sample for both the ux and uydisplacement fields. The fringe lines in the delaminated portion of theEMC cantilever beam mean that there is some rigid body rotation ofthe cantilever beam. This is an indication of plastic deformation at thecrack tip and increased surface roughness of the delaminated area thatis preventing the beam from returning to its original position.

An increased magnification was used to examine the moiré fieldnear the crack tip in Fig. 14 and it is apparent that there is a small plasticwake that does occur in the area where the interface has alreadydelaminated due to the change in the fringe pattern where at thedelaminated edge. The development of this plastic wake prevents thecantilever arm from returning to its initial position. This is why rigidbody rotation of the cantilever arm is seen in Fig.13.

However, for assessing the energy release rate, the concern is thatthe extent of the plastic deformation near the crack tip is small enoughso as to use linear elastic fracture mechanics concepts. The J-integralmethod is calculated using a contour away from the crack tip. As longas the selected contour is not within the near crack tip plasticity regionthen the J-integral can be used to represent the energy release rate ofthe interface. In Fig. 14, carrier rotations were applied to the image toprovide a better contrast near the interface. Deviations from the con-stant contour lines near the crack tip are an indication of plastic defor-mation. Using this micro moiré setup, deviations from the carrierfringe patterns were measured to assess the extent of near the cracktip. Calculating the potential extent of the near field plasticity usingthe camera and digital software, the plastic region was found to be con-fined to 10 μmaround the crack tip. Based on a contour sensitivity studyrun in the FEAmodel for this specimen, a path-independent J-integral is

, obtained after unloading: (a) ux displacement field and (b) uy displacement field.

Fig. 14. Near crack tip moiré images obtained after unloading: (a) ux displacement field and (b) uy displacement field.

141K. Mahan et al. / Microelectronics Reliability 63 (2016) 134–141

evaluated at a contour that is roughly 50 μm away from the crack tip.Thus, it is safe to assume that LEFM is still valid and the numericalmethods can be used to calculate G and Ψ for the M-SCAT method.

6. Conclusion

The M-SCAT method was proposed and implemented to determinethe adhesion strength at strong interfaces in thin multilayer stacks. Themethod allowed for quick and quantitative adhesion strength evaluationof two sets of EMC/PSR interfaces. Detailed sample preparation, experi-mental testing, and numerical processing steps were presented. Excel-lent repeatability was obtained due to the well-defined precrack andthe testing procedure in spite of the very high adhesion strength at theinterface and the low fracture toughness of the EMC. The small testingscatter allowed for the test method to clearly distinguished that onema-terial set had larger adhesion strength than the other. The proposed tech-nique is ideally suited to various thin, multilayer structures encounteredin semiconductor packaging. More applications, including evaluation ofthe adhesion strength degradation caused by accelerated reliability test-ing and manufacturing conditions, are anticipated.

References

Anderson, T.L., 2005]. Fracture mechanics: fundamentals and applications: CRC.Irwin, G.R., 1956]. Onset of fast crack propagation in high strength steel and aluminum al-

loys. N. R. Laboratory, Ed., ed. vol. 2, pp. 289–305.Rice, J., 1968]. A path independent integral and the approximate analysis of strain concen-

tration by notches and cracks. J. Appl. Mech. 35, 379–386.Rice, J.R., 1988]. Elastic fracture mechanics concepts for interfacial cracks. J. Appl. Mech.

55, 98–103.Sun, C.T., Wu, X.X., 1996]. On the J-integral in periodically layered composites. Int. J. Fract.

78, 89–100.van der Sluis, O., et al., 2008]. Advances in delamination modeling. In: Morris, E.J. (Ed.),

Nanopackaging: Nanotechnologies and Electronics Packaging. Springer US, Boston,MA, pp. 61–91.

Agrawal, A., Karlsson, A.M., 2006]. Obtaining mode mixity for a bimaterial interface crackusing the virtual crack closure technique. Int. J. Fract. 141, 75–98.

Schlottig, G., 2012]. Reliability at the chip interfaces: delaminating the silicon die frommolding compound. Delft University of Technology, The Netherlands, Delft, PhD.

Xiao, A., 2012]. Interface characterization and failure modeling for semiconductor applica-tion (PhD doctoral thesis) Precision and Microsystems Engineering. Delft Universityof Technology, Delft.

Charalambides, P.G., et al., 1989]. A test specimen for determining the fracture resistanceof bimaterial interfaces. J. Appl. Mech. 56, 77.

Evans, A.G., et al., 1990]. The fracture energy of bimaterial interfaces. Metall. Trans. A 21,2419–2429.

He,M.Y., et al., 1994]. Crack deflection at an interface between dissimilar elasticmaterials:role of residual stresses. Int. J. Solids Struct. 31, 3443–3456.

He, M.Y., Hutchinson, J.W., 1989]. Kinking of a crack out of an interface. J. Appl. Mech. 56,270–278.

He, M.Y., Hutchinson, J.W., 1989]. Crack deflection at an interface between dissimilar elas-tic materials. Int. J. Solids Struct. 25, 1053–1067.

Wunderle, B., et al., 2012]. Advanced mixed-mode bending test: a rapid, inexpensive andaccurate method for fracture-mechanical interface characterisation. Thermal andThermomechanical Phenomena in Electronic Systems (ITherm), 2012 13th IEEE In-tersociety Conference on, pp. 176–186.

Williams, J.G., April 1987]. Large displacement and end block effects in the ‘DCB’ interlam-inar test in modes I and II. J. Compos. Mater. 21, 330–347.

Cao, H.C., Evans, A.G., Jun 1989]. An experimental study of the fracture-resistance ofbimaterial interfaces. Mech. Mater. 7, 295–304.

Dai, X., 1998]. Materials Study for Interfacial Adhesion and Reliability of MicroelectronicsPackaging Structures. The University of Texas, Austin, Doctor of Philosophy.

Dai, X., et al., 1998]. Polymer Interfacial Adhesion in Microelectronic Assemblies.pp. 132–137.

Shin, D.K., et al., 2009]. Chemical and mechanical analysis of PCB surface treated by argonplasma to enhance interfacial adhesion. IEEE Trans. Electron. Packag. Manuf. 32,281–290.

Shin, D.K., et al., 2010]. Effect of PCB surfacemodifications on the EMC-to-PCB adhesion inelectronic packages. IEEE Trans. Compon. Packag. Technol. 33, 498–508.

Dai, X., et al., 2000]. Adhesion measurement for electronic packaging applicationsusing double cantilever beam method. IEEE Trans. Compon. Packag. Technol. 23,101–116.

Charalambides, P.G., et al., Feb 1990]. Development of a test method for measuring themixed-mode fracture-resistance of bimaterial interfaces. Mech. Mater. 8, 269–283.

Matos, P., et al., 1989]. A method for calculating stress intensities in bimaterial fracture.Int. J. Fract. 40, 235–254.

Suo, Z., 1990]. Failure of brittle adhesive joints. Appl. Mech. Rev. 43, S276–S279.Suo, Z., Hutchinson, J.W., 1990]. Interface crack between two elastic layers. Int. J. Fract. 43,

1–18.Ma, Q., et al., 1995]. Quantitative measurement of interface fracture energy in multi-layer

thin film structures. MRS Annual Meeting, San Francisco, CA, pp. 91–96.Becker, T.L., et al., 1997]. Limitations on the use of the mixed-mode delaminating beam

test specimen: Effects of the size of the region of K-dominance. Mech. Mater. vol.25, 291–308 (1997/6//).

Lane, M., et al., 1997]. Progressive debonding of multilayer interconnect structures. Pro-ceedings of MRS Annual Meeting. CA, San Francisco, pp. 21–26.

Ma, Q., 1997]. A four-point bending technique for studying subcritical crack growth inthin films and at interfaces. J. Mater. Res. 12, 840–845.

Dauskardt, R.H., et al., 1998]. Adhesion and debonding of multi-layer thin film structures.Eng. Fract. Mech. 61, 141–162.

Hofinger, I., et al., 1998]. Modified four-point bending specimen for determining the in-terface fracture energy for thin, brittle layers. Int. J. Fract. 92, 213–220.

Yan, X., Agarwal, R.K., 1998]. Two test specimens for determining the interfacial fracturetoughness in flip-chip assemblies. J. Electron. Packag. 120, 150.

Wang, B., Siegmund, T., 2008]. A modified 4-point bend delamination test. Microelectron.Eng. 85, 477–485.

Roham, S., et al., 2011]. Crack penetration and deflection at a bimaterial interface in afour-point bend test. J. Mater. Res. 19, 3019–3027.

Shirangi, M., et al., 2013]. Determination of Copper/EMC interface fracture toughness dur-ing manufacturing, moisture preconditioning and solder reflow process of semicon-ductor packages. ICF12, Ottawa 2009.

Tran, H., et al., 2014]. Temperature, moisture and mode-mixity effects on copperleadframe/EMC interfacial fracture toughness. Int. J. Fract. 185, 115–127.

van Driel, W.D., et al., 2005]. Prediction of delamination related IC & packaging reliabilityproblems. Microelectron. Reliab. 45, 1633–1638.

van Gils, M.A.J., et al., 2004]. Characterization and modelling of moisture driven interfacefailures. Microelectron. Reliab. 44, 1317–1322.

Dundurs, J., 1969]. Elastic interaction of dislocations with inhomogeneities. Math. Theor.Disloc. 70–115.

Chauffaille, S., et al., 2011]. Elasto-plastic analysis of the single cantilever beam adhesiontest. Eng. Fract. Mech. 78, 2493–2504.

Chauffaille, S., et al., 2011]. Pre-cracking behaviour in the single cantilever beam adhesiontest. Int. J. Fract. 169, 133–144.

Budzik, M.K., et al., 2014]. Experimental investigation of mesoscale crack front triple line.Appl. Phys. A Mater. Sci. Process. 114, 495–501.

Shin, D.K., et al., 2015]. Development of single cantilever beam method to measure theadhesion of thin film adhesive on silicon chip. Eng. Fract. Mech. 133, 179–190.

Post, D., et al., 1997]. High sensitivity moiré: experimental analysis formechanics andma-terials. Springer.