Content uploaded by Kevin Distelhurst
Author content
All content in this area was uploaded by Kevin Distelhurst on Mar 15, 2025
Content may be subject to copyright.
Comparison of Magnetic Field Imaging and Thermal Laser Stimulation for
Detecting Electrically Shorting Defects within a Die BEOL
Kevin A. Distelhurst
GLOBALFOUNDRIES, Essex Junction, VT, USA
kevin.distelhurst@globalfoundries.com
Abstract
Magnetic Field Imaging (MFI) and Thermal Laser Stimulation
(TLS) failure analysis (FA) techniques (e.g. OBIRCH, XIVA,
ect.) both have advantages and disadvantages. The obstacles
encountered from these techniques may hinder further fault
isolation (FI), lengthen turn-around-time and/or detract from
actionable results. MFI using a Giant Magneto Resistance
(GMR) sensor is compared to TLS techniques to understand
the capability of the MFI technique at finding shorting defects.
A short within a capacitor bank is successfully isolated using
both techniques.
Introduction
Having a variety of tools for die level FI and analysis provides
administrative options since no single technique can provide
universal coverage for all types of applications. For TLS
techniques, two criteria are required to detect the defects: the
sensitivity of defects to laser stimulation and adequate path for
the laser energy to reach the defects. MFI with a GMR is
capable of overcoming the TLS obstacles that may hinder
detection. However, certain criteria are also needed to detect a
defect with the MFI/GMR technique: sufficient current
(>50uA) passing through the defect, no obstruction by
ferromagnetic material, and the proximity of GMR within
roughly 10µm z-direction of the current. Using a
Superconducting Quantum Inference Device (SQUID), which
is also an MFI technique, can overcome most of the obstacles
presented by the GMR at the cost of reduced resolution. The
SQUID has demonstrated the ability to localize shorting
defects within TSV [1], on multichip packages [2] and on die
[3]. A comparison of both MFI and TLS is needed to
determine the feasibility of seamlessly interchanging which
technique is used.
MFI with a GMR sensor is capable of isolating electrically
shorting defects on a die [4], [5]. An example of a typical
GMR sensor setup is found and described in [4]. This setup is
used to map the current density, which is derived from the
magnetic field map, through a verified shorting defect on the
BEOL of an experimental die. An overlay and comparison of
the current density map (CDM) with the IR/NIR image in
combination with a CAD image of the die reveals the location
of the defect. TLS is also used to identify the location of the
defect by mapping the locations on a die sensitive to laser
energy. Likewise, overlay and comparison with the IR/NIR
and CAD image of the die localize the defect.
In this paper, a comparison between TLS and MFI with a
GMR is presented. In both cases, each technique is sufficient
in isolating the defect to a desirable area of interest (AOI).
MFI is a useful technique as an alternative to TLS and when
TLS is not capable of isolating the defect.
Figure 1: (a) CAD image of the capacitor bank. The blue
represents the upper capacitor plate conductor in the BEOL
stack and the green represents the lower capacitor plate
conductor. The plates are separated by a dielectric and
connected to probe pads off image. Note that the entire bank is
not represented in this image. (b) XIVA signal (arrow)
overlaid with the IR image. Fill patterns exist above the
capacitor bank and have no impact to this particular analysis.
Setup and Results using TLS
The sample used for analysis consists of various test
structures, one of which is a capacitor bank. A CAD image of
the capacitor bank is in Fig.1a. The bank consists of numerous
capacitors in parallel with one another. Each plate of each
capacitor is tied together and directly connected to a pad for
probe contact. The bank is localized to the first few levels of
the BEOL; that is, the furthest metallization levels from the
FEOL.
(a)
(b)
Figure
2: (a)Overlay of the CDM and CAD image; the scan lasted less than 30 minutes and has an XY spatial resolution of 4 µm. (b)An
overlay of the CAD image and peak localization of the same scan as 2a. It provides a more defined path of the current. (c)Overlay of
the CDM and CAD image; the scan lasted about 12 hours and has an XY spatial resolution of 0.3µm. (d)An overlay of the CAD image
and peak localization of the same scan as 2c. It provides a more defined path of the current.
Electrical 2-point DC measurement by contacting the two pads
with micromanipulator probes indicated an electrical short
when an electrical open is expected.
The sample was provided as a diced, die-level piece. The top
level passivation was removed prior to analysis using a dry
etch recipe. Several other samples with similar failures existed
and were used to identify the defect mechanism which was a
broken-down capacitor dielectric. The sample was loaded into
a laser scanning microscope setup for TLS. A pair of
micromanipulator probes contacted the pads connecting
between upper and lower plates of the capacitor bank. A slight
bias of 15 mV was applied which achieved about 600µA of
current.
The biasing readout also confirmed the electrical short still
existed. A 1340 nm laser was used to both image and
stimulate the defect. A reflected light image was first captured
to provide a structural (optical) image of the capacitor bank.
Subsequently, XIVA imaging was performed and then
overlaid with the NIR structural image. The overlaid image
shows that one of the capacitors in the bank is sensitive to the
laser energy. The results are provided in Fig. 1b. The defect is
likely in the nearest adjacent capacitor, highlighted by a box in
Fig. 1a.
Setup and Results using MFI
The sample was then transferred to the MFI system. A pair of
micromanipulator probes contacted the pads of the capacitor
bank. An AC bias of about 0.8 Vp-p was applied with about
450µA peak-to-peak current. Since the probes have about a
1300 Ω resistor in series with this setup, a higher bias voltage
was used. In all magnetic scans the sensor was approximately
1µm above the die surface observed in Fig. 1b. An initial low
resolution scan of 20 µm of the entire capacitor bank was
performed. A CDM was derived from the magnetic field map
and used to isolate and focus upon the AOI. A quick scan of
about 30 minutes at the AOI with a resolution of 4 µm showed
a strong current flow through one of the capacitors in the
bank. The higher resolution CDM was then overlaid with the
CAD image as shown in Fig. 2a. A peak localization of the
CDM was also obtained and overlaid with the CAD image in
Fig. 2b. A longer, 12-hour scan was performed with a
resolution of 0.3 µm. Likewise, the CDM was overlaid with
the CAD image resulting in Fig. 2c with a peak localization
shown in Fig. 2d. Although the longer scan provided a higher
resolution CDM, no significant differences between defect
localization was observed in the varying resolutions; a longer
scan was not required to provide accurate results. Both scans
localized the defect.
The CDM was compared to the CAD drawing to predict the
location of the defect by observing at what point the current
density crosses an unexpected boundary. In this case, a defect
was likely bridging the upper capacitor plate to the lower at
the point marked by an arrow in Fig. 2. The bridge was likely
through the dielectric as indicated in similar samples. An
overlay of the TLS NIR/XIVA image and the MFI peak
localization was produced and provided in Fig. 3 to further
illustrate this point.
The current density in both Fig. 2 and Fig. 3 doesn’t
completely align with the expected metal traces. The current
density appears to exist in a non-conducting region. This is
(a)
(c)
(b)
(d)
due to the magnetic fields from the upper and lower plate
interacting and shifting the apparent location of the current.
The close proximity of the conducting traces and their 3D
separation should be considered during analysis.
Figure 3: Overlay of the MFI peak localization illustrating the
current flow through the defective capacitor and the TLS
NIR/XIVA image illustrating the location of laser sensitivity.
Conclusion
Both TLS and MFI techniques were capable of isolating the
defect to a single capacitor in the bank. MFI is capable of
isolating defects on the die using a GMR sensor. In certain
cases, MFI would be able to find the defect when TLS could
not. For example, when obstacles are blocking the defect from
laser energy, the CDM can be aligned with a CAD image to
achieve fault isolation. MFI also has the added benefit of
mapping the current density thus providing deeper insight into
how a defect may be altering the state of the circuit. In this
experiment, the current was shown to flow directly down one
path through the distributed upper plates of the bank, through
the defect, then directly back to the probe attached to the
lower plates.
Further work may identify unique instances when TLS fails to
isolate the defect. For instance, MFI can reveal the circuits
that should be the primary area of focus in cases where a
regional defect insensitive to TLS is causing abnormally large
current demands. MFI provides the analysis another technique
to isolate the diverse variety of defects for a successful,
actionable analysis.
References
[1] J. Gaudestad, A. Orozco, I. D. Wold, T. Wang, T. Webers, R. Kelley, T. Morrison and S. Madala, "Failure Analysis Work
Flow for Electrical Shorts in Triple Stacked 3D TSV Daisy Chains," ISTFA 2014: Proceedings of the 40th International
Symposium for Testing and Failure Analysis, p. 5, 2014.
[2] J. Gaudestad, A. Orozco, M. Kimball, K. Gopinadhan and T. Venkatesan, "Short Localization in a Multi Chip BGA
Package," IEEE 21st International Symposium on the Physical and Failure Analysis of Integrated Circuits, p. 4, 2014.
[3] D. P. Vallett, "Scanning SQUID Microscopy for Die Level Fault Isolation," Proceedings from the 28th International
Symposium for Testing and Failure Analysis, 2002.
[4] J. Gaudestad, N. Gagliolo, V. V. Talanov, R. H. Yeh and C. J. Ma, "High Resolution Mangetic Current Imaging for Die
Level Short Localization," 20th IEEE International Symposium on the Physical and Failure Analysis of Integrated
Circuits (IPFA), p. 4, 2013.
[5] D. Vallett, J. Gaudestad and C. Richardson, "High-Resolution Backside GMR Magnetic Current imaging on a Contour-
Milled Globally Ultrathin Die," ASM International, p. 5, 2014.