Microelectronics Reliability 150 (2023) 115079

Contents lists available at ScienceDirect

Microelectronics Reliability
journal homepage: www.elsevier.com/locate/microrel

Using X-ray imaging for the study of crack development in solder

reliability testing
P. Roumanille a, *, J. Lesseur a, J. Uzanu b, H. Le Trong a, E. Ben Romdhane c, A. Guédon-Gracia d,
H. Frémont d
a
Institut de Recherche Technologique Saint-Exupéry, Toulouse, France
b
Elemca, Toulouse, France
c
Mines Paris-Psl, MAT-Centre des Matériaux, CNRS UMR 7633, Evry, France
d
University of Bordeaux, IMS Laboratory, UMR 5218, Talence, France

A R T I C L E I N F O A B S T R A C T

Keywords: The possibility of monitoring the progressive damage and cracking in solder joints is investigated through a non-
X-ray destructive method. As X-ray 3D Computed Tomography is known to require small sample dimensions in order to
Nondestructive testing ensure a good resolution, Computed Laminography was used to get a high-resolution top view of various types of
Failure analysis
solder joints located on a large printed circuit board. In a first step, the assemblies were cross-sectioned after X-
Solder
ray imaging in order to verify the findings and confirm the crack morphology. Then, by avoiding a destructive
sample preparation step, it was possible to identify in a single sample the effects of solder fatigue during a
standard reliability test. Firstly, multiple crack initiation sites could be highlighted in a single X-ray analysis.
Moreover, the study gave an interesting insight of the time evolution of voids in solder joints. X-ray laminog­
raphy produces 3D-like images that contain more information than a conventional cross section. By improving
the recognition of horizontal cracks in this unusual point of view of solder joints, the refinement of this imaging
and monitoring method can result in a real progress in failure analysis time and efficiency.

1. Introduction difficult to find cracks in a single try as they are not necessarily in the
plane of the cross-section.
In operation, electronic packages may be exposed to a combination Due to the destructive nature of cross-section analysis, the study of
of mechanical and thermal loadings. The development and improve­ the solder crack propagation rate in thermal cycling has to be performed
ment of products requires reliability testing and suitable failure analysis with different samples that have been submitted to various levels of
techniques. Thermal cycling is a common and standard test for the thermomechanical damage [1]. The monitoring of the same specimen
qualification of board assemblies (second level interconnects) and the during a lifetime test implies a non-destructive method. For example, the
evaluation of solder fatigue lifetime. Fatigue cracks form in solder joints crack propagation in a solder layer was recently monitored during a
under the cyclic loading due to the thermal expansion mismatch be­ power cycling test with Acoustic Emission Monitoring and Scanning
tween the different materials of the board and the package. Reliability Acoustic Tomography [2]. Besides, large area solder joints were ana­
tests involve different tools as electrical monitoring for failure detection lysed through X-ray radiography during thermal cycling to reveal the
and cross-sectioning for failure analysis. These techniques may allow development of voids and cracks [3,4].
accurate identification and location of cracks in solder joints, while they X-ray imaging technologies are powerful non-destructive methods
require a dedicated design of test boards and dummy packages or a enabling the inspection of the internal structures of materials without
destructive analysis of the sample. The most common way to find cracks destroying the specimen. Widely used in various fields such as medicine,
is cross-section analysis as they often start to propagate under the geosciences, biology, material sciences or food industry, X-ray imaging
package and cannot be seen with visual inspection. However, it can be methods are particularly useful for defect and failure analysis in

* Corresponding author.
E-mail address: [email protected] (P. Roumanille).

https://doi.org/10.1016/j.microrel.2023.115079
Received 4 June 2023; Received in revised form 28 June 2023; Accepted 30 June 2023
Available online 1 October 2023
0026-2714/© 2023 Published by Elsevier Ltd.
P. Roumanille et al. Microelectronics Reliability 150 (2023) 115079

Fig. 1. Optical images of the R1206 and QFN68 packages (a, b) and laminographic images of their solder joints (c, d).

electronic products [5,6]. Digital Radiography (DR) and X-ray 2. Experimental procedure
Computed Tomography (XCT) are one of the well-known techniques.
The former provides a 2D radiographic projection of the internal fea­ 2.1. Test board and packages
tures thanks to the X-ray beam directly transmitted to a digital detector
through the object. The latter takes several 2D radiographic projections The test board was initially designed for the study of solder fatigue
while the object is rotated through 360◦ . A mathematical reconstruction lifetime [12]. It is a 160 × 144 mm, 1.6 mm thick FR-4 printed circuit
aims to generate a 3D map of the local densities inside the specimen. A board with Electroless Ni/Au (ENIG) surface finish. R1206 chip re­
third technique also called Computed Laminography (XCL) is especially sistors, BGA288 (19 × 19 mm) and QFN68 (10 × 10 mm) dummy
suited for flat objects such as circuit boards. It can be seen as an inter­ packages were assembled with a SnAgCu (SAC) solder paste. The daisy-
mediary method compare to DR or XCT by providing 3D-like images chained packages and the board design allowed the electrical moni­
containing higher defect sensitivity information on specimen’s depth as toring of the solder joints and the failure detection during a lifetime test.
compared to DR. In addition, XCL provides high-resolution images of the A fatigue crack in a solder joint induces an electrical open that can be
region of interest (ROI) on large flat specimens, which can be difficult to detected with a data logger.
achieve on most XCT laboratory systems where the achievable resolu­
tion is directly correlated to the size of the object, and the intensity of the 2.2. Thermal cycling
X-ray attenuation along the longitudinal direction.
X-ray imaging is already used for the inspection and the failure Thermal cycling was performed between − 55 ◦ C and 125 ◦ C with a
analysis of board assemblies. Several papers highlighted the interest of ramp of 10 ◦ C/min and a dwell time of 15 min, according to the TC4 test
XCL for manufacturing quality purposes. The defects that were detected condition of IPC-9701 standard [13].
in BGA solder joints with XCL are voids, opens, bridges, pad nonwetting A first test board was chosen for preliminary analyses after an elec­
and ball-pad misalignment [7–9]. Solder failure analysis with X-rays is trical failure had been measured in the solder joints of a QFN package at
less developed, as it is difficult to see narrow cracks without reducing the 1700 cycles [12]. A BGA package was also studied after 2500 thermal
sample size. Su et al. recently emphasized that fine cracks cannot be cycles, as the first electrical failures were measured in a previous work
detected with DR while XCL is too much time-consuming [10]. Lall et al. [14] in the solder joints under the die edge between 2100 and 2500
developed the 3D reconstruction of solder joints with XCT, including cycles. These specimens were used to consider the possibility of identi­
cracks, voids and Head-in-pillow defects and published images of full fying cracks in solder joints with different X-ray imaging methods and
cracks only [11]. parameters. The same thermal cycling test was also interrupted several
In the present paper, we propose to explore the possibilities offered times (after 400, 700 and 1100 cycles) for multiple XCL analyses of a
by XCL to characterise and monitor the damage and the propagation of second test board in order to monitor the damage and the cracking of the
cracks in the solder joints of large electronic boards subjected to thermal solder joints (QFN, chip resistor) before it can be detected with an
cycling. The first XCL results were compared to XCT and cross-section electrical measurement.
images. In the process of developing an efficient non-destructive imag­
ing method coupled with a long-term reliability test, XCL was chosen to 2.3. X-ray analysis
be faster than XCT and more accurate than DR in the measurement of the
dimensions of objects. The computed laminography [15] set up is similar to XCT, but
instead of a full rotation of the object, the acquisition is based on the
relative motions between the object, the X-ray tube and the detector.
Most of the commercial XCT systems propose, in addition to

2
P. Roumanille et al. Microelectronics Reliability 150 (2023) 115079

Table 1 were used during the full study.

Experimental details of the various X-ray imaging techniques.
X-ray method HR tomography HR laminography Laminography 2.4. Cross-section
(1) (1)

Voltage (kV) 100 100 130 Some packages that had been submitted to X-ray analysis were then
Voxel size (μm) 0.8 × 0.8 × 0.8 0.8 × 0.8 × 3.0 2.0 × 2.0 × 8.0 characterized through cross-sectioning. The boards were cut and the
Scanning time 4.5 2.0 1.5 packages were mounted in resin. The sample preparation procedure
(h) involved mechanical grinding with SiC paper and polishing with dia­
X-ray imaging RX Solutions EasyTom 150–160 RX Solutions
system EasyTom 230
mond suspensions. The cross-section images were made with a Keyence
VHX-7000 digital microscope.
(1) Performed on a package that had been cut out of the board for HR tomog­
raphy (0.8 μm / voxel).
3. Results

conventional tomography, laminography acquisition procedures. The 3.1. Preliminary study on a QFN package
method used in this paper consists in applying a linear translation of the
object (with multiple intermediate randomized positions during the The first part of the study aimed to visualize an open crack with
displacement) while the X-ray tube and the detector remain stationary. various imaging techniques in a QFN solder joint after 1700 thermal
During the process, mosaic-like radiographic areas are recorded. Thanks cycles.
to the conic geometry of the X-ray beam and the associated angular The laminographic slices give a top view of the solder joints, which is
aperture, each translation point is equivalent at this local position to a an unusual way of looking for cracks that propagate horizontally in the
small rotation of the object. XCL can be seen as a limited angle tomog­ joints. In comparison with an as-reflowed state (Fig. 1.d), the damage in
raphy. The reconstruction produces depth slices images with a resolu­ the 68 solder joints of the package was obvious then the analysis was
tion which is less accurate than the lateral resolution of the image. focused on the most damaged one (Fig. 2). The CTE mismatch between
Artefacts of blurred features are produced due to the limited angle the board and package materials induces stresses and strains in solder
projections (Fig. 1). In a single acquisition of a whole package, images of joints when the assemblies are submitted to temperature variations. The
tens of solder joints may be recorded. global shape of the solder joint is very far from a pristine one due to the
The preliminary study for the characterisation of the cracks inside plastic deformation that accumulates with thermal cycling. The top
QFN solder joints using XCT and XCL where performed on a RX Solu­ surface that is in contact with the package copper lead seems rough and
tions EasyTom 150–160 at ELEMCA (Table 1). The system is equipped grainy. Some voids can be identified but they do not have the common
with a transmission nanofocus tube used at a voltage of 100 kV and a rounded shape that is induced by the reflow soldering process. Partial
2025 flat panel detector. For XCT analysis, the board was carefully cut cracks are visible. One starts from the external surface of the joint. The
around the ROI, using a diamond wire saw, in order to enable high cracks could be found in the images, whatever the tested resolution of
resolution. For XCL, the specimen was placed very close to the tube with the XCL analysis. However, the 0.8 μm resolution (Fig. 2.b) gives more
the longitudinal part parallel to the X-ray tube output window. In order details on the solder joint condition than the 2 μm resolution (Fig. 2.a).
to detect thin crack features, the final reconstructed voxel size was set to We can see that the crack connects a void to the outside of the solder.
0.8 μm × 0.8 μm × 0.8 μm for XCT and 0.8 μm × 0.8 μm × 3.0 μm for This was confirmed by the XCT results.
XCL (depth). By cutting out the board around the QFN package with a diamond
For the monitoring of the thermal cycling test, the XCL measure­ wire saw, it was possible to obtain a high resolution 3D reconstruction of
ments where performed at IRT Saint Exupéry on a RX Solutions EasyTom the solder joint. An horizontal slice was recorded to make a comparison
230 equipped with a reflection microfocus X-ray tube and a 2520 flat with the laminographic images (Fig. 2.c). The solder damage and
panel detector. The voltage was set to 130 kV and the voxel size to 2.0 cracking were identified as well, but a total crack cannot be highlighted
μm × 2.0 μm × 8.0 μm (depth). This resolution is considered as sufficient with a single slice as it is not perfectly horizontal. A cross-sectional slice
for crack detection, and is more common for standard XCT laboratory of the solder joint in an appropriate plane revealed the open crack that
systems. For optimal comparisons the same acquisition parameters were goes from one end of the joint to the other (Fig. 2.d). Finally, a fully
used at each step of the thermal cycling study. No X-ray spectrum filters destructive cross-section analysis was performed in the same plane as

Fig. 2. Identification of a crack in the same QFN solder joint in different steps: (a) laminographic slice (resolution: 2 μm), (b) laminographic slice (resolution: 0.8 μm)
without cutting the board, (c) tomographic slice, top view (resolution: 0.8 μm) after cutting the package out of the board, (d) tomographic slice, lateral view
(resolution: 0.8 μm), (e) cross-sectional view in optical microscope after polishing the package.

3
P. Roumanille et al. Microelectronics Reliability 150 (2023) 115079

Fig. 3. XCL analysis of BGA solder joints (top view) performed after 2500 thermal cycles (first step) and followed by a cross-section analysis of the cracked solder
joint G4 (second step).

Fig. 4. Evolution of QFN solder joints with the number of thermal cycles, laminographic slices (resolution: 2 μm).

the tomographic slice to confirm the solder crack features in a more

usual way (Fig. 2.e).
This case study revealed the potential of XCL for solder failure
analysis, at the same time as it showed the challenge of identifying 3D
cracks from a top view. The various depth slices need to be analysed for
an efficient crack recognition. XCL is faster than XCT and does not
require any cutting of the electronic board. Unlike cross-sectioning, it
brings information in more than a single vertical plane.

3.2. Preliminary study on a BGA package

As the resistance of the daisy-chain in the BGA package was not

monitored, there was no information on electrical failures in the solder
joints. At least, partial cracks were expected after 2500 thermal cycles.
Fig. 5. Cross-section of a QFN solder joint after 500 thermal cycles showing
Cracks propagate in BGA solder joints near the interfaces with copper
interfacial crack initiation sites.
pads on the package side and the board side. Once again, the XCL depth
slices were processed to deduce the potential crack location. The top
surface in contact with the package copper pad was grainy in some of the sample preparation. Fig. 4 shows the progressive damage in three QFN
considered joints (Fig. 3). The solder joint identified as G4 was cross- solder joints before they reached the same state as in Fig. 2. The solder
sectioned and observed with an optical microscope and a partial shape gets deformed as the damage accumulates and the surface changes
cracking was confirmed in the upper part of the joint (Fig. 3). from smooth to grainy. Small cavities form at the interface between the
The possibility to locate cracks in BGA solder joints with XCL in an solder and the package copper lead. They have already been identified
important result, as it is well known that the exact location of a cracked as crack initiation sites in cross-section analyses (Fig. 5) that were
joint under a BGA package is hard to find with conventional failure published in a previous paper about the solder failure mechanism in
analysis techniques. thermal cycling [12].
The progressive deformation of initial voids is also obvious, as they
tend to lose their rounded shape. Two neighbouring voids merge
3.3. Monitoring of solder damage in thermal cycling together due to the large deformation induced by the thermal cycles. The
biggest void gradually opens towards the outside of the solder joint.
The second part of the study was based on the outcomes of the Voids are considered as process-induced defects in electronic assem­
preliminary work. XCL was used as a monitoring tool in a solder fatigue blies. Studies revealed that they can influence the fatigue life of solder
test. The same solder joints were analysed at different intervals by joints, depending on their location and distribution [16–18]. These
periodically removing the board from the thermal chamber, without any

4
P. Roumanille et al. Microelectronics Reliability 150 (2023) 115079

Fig. 6. Comparison between Digital Radiography (DR) and Computed Lam­

inography (XCL) with QFN solder joints after 1100 thermal cycles.

Fig. 8. Cross-section of partially cracked R1206 solder joints after ther­

mal cycling.

details in the object’s depth, thus in the solder joint features, without
any rotation of the specimen (Fig. 6). All the small voids and cracks are
not easy to see in the DR image. Radiography remains a technique that
can be used as a quick preselection of a ROI to be examined in XCL.
The investigation was extended to 1206 resistor solder joints. Pack­
ages with a low CTE ceramic body are known to produce earlier solder
failures than plastic packages. The damaging process involves the same
features as in QFN solders but seems to be faster (Fig. 7). It can be
noticed that numerous cavities are created in the part of the solder
which is located under the resistor termination, where the fatigue cracks
often initiate for this kind of assembly. By comparing XCL results with
cross-sectioned resistor solder joints after thermal cycling (Fig. 8), the
difficulty to see cracks forming at the angle of the resistor is obvious. But
the laminographic top view gives an information through all the joint
width, while a cross-section leads to an image of the sample in a single
plane.

4. Conclusion

This study deals with the use of X-ray imaging for solder failure
analysis. In particular, the potential of Computed Laminography is
investigated for the development of a method that would allow the
monitoring of the damage and crack development in solder joints.
Fig. 7. Evolution of chip resistor solder joints with the number of thermal Although XCL offers an uncommon top view of the solder, it represents a
cycles, laminographic slices (resolution: 2 μm). powerful tool to highlight the multiple crack initiation sites that cannot
be found with conventional cross-section analysis. It implies the devel­
results confirm at least that they take part into the solder failure opment of efficient crack and defect recognition, as cracks are usually
mechanism. Fatigue cracks can propagate with or without voids but the identified in the plane of a lateral cross-section. While XCL is mostly
evolution of voids is an indicator of the solder deformation under cyclic used for board quality inspection, it is a faster alternative to 3D X-ray
loading. As the daisy-chained package was electrically monitored with a methods for the preliminary selection of a region to be cross-sectioned
data logger, no failure was measured in the solder joints along the 1100 for failure analysis. This methodology is still limited by its processing
thermal cycles. time (acquisition, reconstruction, investigation). It also requires an
The comparison of the XCL image of QFN solder joints with an appropriate resolution that is not fulfilled by every X-ray equipment.
equivalent 2D radiographic projection confirms that XCL brings more Radiography remains faster and easier to use in the industry, as well as

5
P. Roumanille et al. Microelectronics Reliability 150 (2023) 115079

cross-sectioning is the standard destructive tool for solder failure anal­ [6] C. Andersson, J. Ingman, E. Varescon, et M. Kiviniemi, « Detection of cracks in
multilayer ceramic capacitors by X-ray imaging », Microelectron. Reliab., vol. 64,
ysis. All of these techniques are complementary and can be combined
p. 352–356, sept. 2016.
depending on the technical purposes. [7] T. D. Moore, D. Vanderstraeten, et P. M. Forssell, « Three-dimensional x-ray
laminography as a tool for detection and characterization of BGA package defects
CRediT authorship contribution statement », IEEE Trans. Compon. Packag. Technol., vol. 25, no 2, p. 224–229, 2002.
[8] S. M. Rooks, B. Benhabib, et K. C. Smith, « Development of an inspection process
for ball-grid-array technology using scanned-beam X-ray laminography », IEEE
P. Roumanille: Conceptualization, Methodology, Investigation, Trans. Compon. Packag. Manuf. Technol. Part A, vol. 18, no 4, p. 851–861, 1995.
Writing – original draft, Visualization. J. Lesseur: Conceptualization, [9] V. Sankaran, A. R. Kalukin, et R. P. Kraft, « Improvements to X-ray laminography
for automated inspection of solder joints », IEEE Trans. Compon. Packag. Manuf.
Methodology, Investigation, Writing – original draft, Visualization. J. Technol. Part C, vol. 21, no 2, p. 148–154, avr. 1998, doi: https://doi.org/10.1109/
Uzanu: Investigation. H. Le Trong: Conceptualization, Methodology. E. 3476.681394.
Ben Romdhane: Investigation. A. Guédon-Gracia: Writing – review & [10] L. Su, X. Yu, K. Li, et M. Pecht, « Defect inspection of flip chip solder joints based on
non-destructive methods: a review », Microelectron. Reliab., vol. 110, p. 113657,
editing. H. Frémont: Writing – review & editing. juill. 2020, doi: https://doi.org/10.1016/j.microrel.2020.113657.
[11] P. Lall, S. Deshpande, J. Wei, et J. Suhling, « Non-destructive crack and defect
Declaration of competing interest detection in SAC solder interconnects using cross-sectioning and X-ray micro-CT »,
in 2014 IEEE 64th Electronic Components and Technology Conference (ECTC),
Orlando, FL, USA: IEEE, mai 2014, p. 1449–1456. doi: https://doi.org/10.110
The authors declare that they have no known competing financial 9/ECTC.2014.6897484.
interests or personal relationships that could have appeared to influence [12] E. Ben Romdhane, P. Roumanille, A. Guedon-Gracia, S. Pin, P. Nguyen, et H.
Fremont, « QFN (Quad Flat No–lead) SAC Solder Joints under Thermal Cycling:
the work reported in this paper.
Identification of Two Failure Mechanisms », in 2022 IEEE 72nd Electronic
Components and Technology Conference (ECTC), San Diego, CA, USA: IEEE, 2022,
Data availability p. 716–722.
[13] IPC - Association Connecting Electronics Industries, « IPC9701 - Performance Test
Methods and Qualification Requirements for Surface Mount Solder Attachments ».
Data will be made available on request. 2002.
[14] E.B. Romdhane, P. Roumanille, A. Guedon-Gracia, S. Pin, P. Nguyen, H. Fremont,
References Evaluation of SAC solder joint thermomechanical fatigue in different types of
components, in: In 2022 23rd International Conference on Thermal, Mechanical and
Multi-Physics Simulation and Experiments in Microelectronics and Microsystems
[1] J. H. Lau, C. Chang, et S.-W. R. Lee, « Solder joint crack propagation analysis of (EuroSimE), St Julian, IEEE, Malta, avr. 2022, pp. 1–6, https://doi.org/10.1109/
wafer-level chip scale package on printed circuit board assemblies », IEEE Trans. EuroSimE54907.2022.9758868.
Compon. Packag. Technol., vol. 24, no 2, p. 285–292, 2001. [15] J. Zhou, M. Maisl, H. Reiter, et W. Arnold, « Computed laminography for materials
[2] Z. Zhang et al., « Development of Solder Deterioration Diagnosis System of a Power testing », Appl. Phys. Lett., vol. 68, no 24, p. 3500–3502, 1996.
Module via the Acoustic Emission Monitoring (AEM) Technique », in 2021 33rd [16] M. Yunus, K. Srihari, J. M. Pitarresi, et A. Primavera, « Effect of voids on the
International Symposium on Power Semiconductor Devices and ICs (ISPSD), reliability of BGA/CSP solder joints », Microelectron. Reliab., vol. 43, no 12, p.
Nagoya, Japan: IEEE, mai 2021, p. 179–182. doi: 10.23919/ISPSD50666.2021. 2077–2086, 2003.
9452263. [17] V. N. Le, L. Benabou, V. Etgens, et Q. B. Tao, « Finite element analysis of the effect
[3] G. Muralidharan, K. Kurumaddali, A. K. Kercher, L. Walker, et S. G. Leslie, « An X- of process-induced voids on the fatigue lifetime of a lead-free solder joint under
ray radiography study of the effect of thermal cycling on damage evolution in thermal cycling », Microelectron. Reliab., vol. 65, p. 243–254, oct. 2016, doi:
large-area Sn-3.5Ag solder joints », J. Electron. Mater., vol. 42, no 2, p. 240–248, https://doi.org/10.1016/j.microrel.2016.07.098.
févr. 2013, doi: https://doi.org/10.1007/s11664-012-2341-x. [18] P. Wild, D. Lorenz, T. Grözinger, et A. Zimmermann, « Effect of voids on thermo-
[4] G. Muralidharan et E. Cakmak, « X-Ray Radiography Study of The Evolution of mechanical reliability of chip resistor solder joints: experiment, modelling and
Defects in Solder Joints », 2023. simulation », Microelectron. Reliab., vol. 85, p. 163–175, juin 2018, doi: https://
[5] H. Villaraga-Gómez et J. D. Bell, « Modern 2D & 3D X-ray technologies for testing doi.org/10.1016/j.microrel.2018.04.014.
and failure analysis », ISTFA: ASM International., 2019, p. 14–19.