Page 52 - Wire Bonding in Microelectronics

Ultrasonic Bonding Systems and Technologies 31

was very narrow at such low temperatures, requiring much closer
control over machine parameters, heat stage temperature, and mate-
rials. Such low temperature studies have not been implemented in
production. Charles [2-38] has published the only carefully designed
experiments comparing 60 and 100 kHz Au ball bonding. His studies
concluded that 60 kHz had a larger bonding window than 100 kHz
for Au ball bonds, but was not as effective in bonding pads that were
difficult to bond, such as ones over soft substrates (e.g., PTFE). One
hundred kHz bonded in a shorter time for both Au ball and Al wedge
bonds. Outside of these studies it has been found that the Au crescent
(second bond) of a ball bond has a very narrow bonding window at
high frequencies, sometimes requiring a higher temperature for high
yield. At least one company produced a dual frequency transducer
(second harmonic used for the ball and the fundamental frequency
for the crescent bond).
Another paper reported wedge bonding with 100 µm diameter.
Al wire to Cu plates using 60, 190, and 330 kHz bonding frequencies
[2-39]. The authors reported that 330 kHz produced stronger bonds in
a shorter time and with lower vibration amplitude. Other workers
reported that 90 to 120 kHz US energy resulted in better ball bonds to
pads on polyimide that were placed over active areas of IC chips
[2-40]. There have also been statements (unpublished) that HF improves
bonding to pads over soft polymers such as PTFE. If verified, it may
imply that the polymer absorbs less energy at high frequencies, leav-
ing more for the bond interface. However, little is known about US
energy absorption in polymers. Thus, a particular frequency could be
either more or less effective than another when used for bonding over
a particular polymer. Hundreds of DOE comparison studies will be
required before all materials used as substrates are characterized and
a general understanding is achieved. Since the use of such HF is rela-
tively new in 1996, one can expect that limitations as well as advan-
tages will appear in the future. This is a very dynamic area.
One explanation of the differences in wedge bonding at higher
US frequencies has been proposed by Shirai [2-35] and incorporated
in [2-21]. In this explanation, the HF tool-to-wire vibration produces
a higher strain-rate and, therefore, a much higher stress in the Al wire.
The wire becomes strain-rate hardened, deforms less, and more
energy transmits to the weld interface. This results in a strong Al
wedge bond with lower deformation. From Ramsey [2-34], the higher
frequency increases the rate of metallic interdiffusion and makes a
better metal weld. This explanation appears to be reasonable; how-
ever, there are many unanswered questions, and much more must be
learned about the HF bonding mechanism. The explanation used by
Shirai is based on a strain-rate model for single crystal LiF, Ge, and Si
[2-41]. These are brittle ionic and covalent materials. Wire bonding is
done with soft, polycrystalline, face-centered cubic metals (Au and Al)
that respond to stress by deforming easily, although they would