Time-Resolved Thermoreflectance Imaging for Thermal Testing and Analysis
Kazuaki Yazawa, Dustin Kendig
Microsanj LLC., Santa Clara, CA USA
kaz@microsanj.com, +1-408-256-1255
Ali Shakouri, Kazuaki Yazawa
Purdue University, West Lafayette, IN USA
Abstract
High speed, time-resolved, thermoreflectance imaging is a
novel way to locate defects or regions of potential failures in
microelectronic devices. This paper reports on our
thermoreflectance technique for dynamic imaging of circuit
temperature distributions. This transient imaging method is
based on a precise electrical lock-in technique with image
processing similar to an old fashioned animation movie. An
ordinal shutter speed camera is used in conjunction with an
illumination LED that is pulsed for sampling the temperature
distribution. This paper presents the method and gives a
description of the system hardware. A theoretical comparison
to lock-in thermography, which is based on infrared emission
imaging, will be given. Limitations of thermoreflectance and
the driving factors for spatial and time resolution will be
discussed. Finally, we highlight and provide examples of near
infrared (NIR) wavelength imaging, to enable both through-
silicon thermal imaging and emission imaging in the same
system. The combination of these two techniques is expected
to enable hotspot temperatures and any anomalous emission
sites to be correlated, hopefully leading to a better
understanding of the nature of the defect.
Introduction
Detecting a time-dependent defect and identifying it as a
potential failure or gaining an understanding of the failure
mechanisms related to the defect is a challenge in failure
analysis for today’s complex high-speed electronic devices.
The scaling of device features results in a significant reduction
in their time response and an increased sensitivity to transient
events [1]. For example, a transient temperature change can
result in a functional failure in a circuit with a tight design
margin because of the timing perturbation caused by a small
capacitance drift [2]. As device features shrink, the probability
of detecting defects is also greatly reduced, thus leading to the
development of statistics-based, large-area sampling
methods [3]. Weaker signals and a lower probability of failure
detection have made the understanding of the failures more
difficult, but yet still very important.
Some failures and defects can only be effectively isolated with
the aide of thermal information that is obtained while the
semiconductor devices are operating, or, at a minimum, while
a bias is applied to the defective circuit. A “short circuit”
defect caused, for example, by a whisker, or by a very minor
local misalignment of circuit features can produce highly
concentrated Joule heating, even with a very small amount of
dissipated power. While some detection methods have been
developed for these cases, they are effective only under limited
conditions [5]. One major problem is the rapid diffusion of
heat, which blurs the hotspot and makes localization difficult
using DC/static techniques.
Using a high speed transient imaging technique such as time-
resolved thermoreflectance instead of static imaging enables
weak heat signals to be captured and localized before the heat
completely diffuses. This technique is capable of analyzing the
small changes in temperature of the detailed features of a
microelectronic device, while the device is driven under
specific time-varying conditions such as a work load vector or
a time varying bias.
Besides providing better localization capability, time-resolved
thermoreflectance is also useful for analyzing the time-
dependence of hotspot locations. By acquiring a time sequence
of images in accordance with a time-varying workload,
unexpected transient hotspot locations can be identified. These
hotspots could be an indication of a logic failure, a timing
failure, or a circuit design anomaly.
The thermoreflection concept itself has been known for some
time, but its utilization for thermal analysis and defect
detection has been limited. The current availability of turn-key
systems, reduced complexity and cost, and innovations with
CCD imaging and synchronization techniques has greatly
enhanced the utility of this approach [4]. One recent
innovation is the incorporation of photon emission imaging in
a NIR thermoreflectance system. This combination enables
comparisons to be made between hotpot and photon emission
sites and makes it possible to analyze their correlations.
Examples of backside imaging and the application of this
combination will be shown. This paper will also provide the
ISTFA 2013: Conference Proceedings from the
39th International Symposium for Testing and Failure Analysis
November 3–7, 2013, San Jose, California, USA
Copyright © 2013 ASM International®
All rights reserved
www.asminternational.org
194