Page 167 - Wire Bonding in Microelectronics

144 Cha pte r F i v e

Temp-cycle
cracking of
bond heel

Kirkendall
voids

FIGURE 5-10 An SEM micrograph of TS Au crescent, wedge-stitch-bond to
1 µm of aluminum, then thermal stressed for 6 h at 155°C, and temp-cycled
20 times between −65 and I55°C. This reveals two major intermetallic failure
modes. Kirkendall voids are seen around the perimeter, and a brittle fracture
crack formed at the thin heel as a result of the temp-cycling of intermetallic
diffused up from the Al pad.

be aware of this problem in devices that are likely to be temperature-
cycled, such as under-the-hood automotive electronics.
Philofsky [5-2] published metallurgical design limits for avoiding
bond failures due to the formation of intermetallic compounds. A
condensed version of his diagnostic tables is given in Table 5-3.
Thinner metallization has been cited to limit Kirkendall voiding
by restricting the availability of one of the intermetallic components
[5-2, 5-10]. Similar observations have been made more for Al wire
bonds on 1-µm plated Au films in which resistance drift failures
occurred, but when the same composition films were thinner, 0.25 µm
(10 µin), they were reliable [5-16].

5.1.4 Reversing the Au-Al Metallurgical Interfaces
Often, one may have reliability data for, say, a ball bond of one metal,
bonded to a thin-film pad of the other; however, some new bonding
situations may call for the reverse, in which the wire is of the former
pad metallurgy and the pad, of the former wire. Intuitively, one might
think that it would make no difference metallurgically in the bond-
ability or the reliability of the bond. Intuition is often wrong, and
when one examines the metallurgical conditions that exist during
and after bonding, it becomes apparent that significant differences
may occur. Basically it is equivalent to changing the bottom two left
and right hand blocks in Fig. 5-3, resulting in entirely different