Page 53 - Wire Bonding in Microelectronics

32 Cha pte r T w o

have some (unknown) high-strain-rate-modified response. Compila-
tions of strain-rate hardening in Al [2-42] would not support the
proposed mechanism. Furthermore, Langenecker has verified his US
softening mechanism up to 1 MHz, implying that the strain-rate hard-
ening is not significant below that frequency. However, that paper
omits many measurement details and itself leaves questions. Shirai’s
proposed mechanism, which may be qualitatively correct, should be
further studied. In particular, its theory should be rederived around
the known properties of soft, polycrystalline, face-centered cubic
metals. Most of our understanding of the US softening and welding
mechanism is based on fundamental studies carried out in the early
to mid-1960s. These experiments should be repeated with current
measurement methods and high-speed computers for mathematical
computation (verification) of models. Full understanding is the
necessary basis for continued advancement of US bonding, as we
push towards the limits of speed, fine pitch (~20 µm), and especially,
high yield. Our present understanding is mostly empirical, and that
knowledge has been pushed to its limits.

2.5 In-Process (Real-Time) Bond Monitoring
There has long been a desire for an electrical or a mechanical real-
time (in-process) quality control system to increase the bond yield.
Efforts, which to date have been mostly empirical, include measuring
an electrical parameter from the US power supply, the transducer’s
impedance, the tool-drop (related to bond deformation) during bond-
ing, tool lift-up force after bonding, the amount of US energy trans-
mitted through the package, second and third harmonic output from
sensors on the transducer. In some cases, the bonding time can be
extended until the defined parameter is achieved or the power input
altered. Modern high-speed computers (incorporated in advanced
autobonders) make several of these possible and practical. Some of
these methods have been published, others patented, and others have
appeared only in internal company or military contract reports (see
[2-43 to 2-51]). Many are slight variations of earlier publications or
patents. The details of current systems are usually proprietary, and a
thorough review of these and other bond-monitoring systems is
beyond the scope and intent of this book. Bonder manufacturers will
discuss their particular system with customers, and some have more
detailed information on their Internet Web sites.
Bond-monitoring systems are essentially empirical, since there is
no accepted quantitative theory of ultrasonic/thermosonic bonding.
The electrical-signal systems require defining or establishing some
empirically determined bond quality window. (In general, a process
control based solely on windowing will fail some good bonds and/or
pass some weak ones.) Other systems measure and control the bond