Page 145 - Wire Bonding in Microelectronics

W ir e Bond Testing 123

WIRE Relation Between

Type of Composition (x) and r on the NDP Force
Production Elongation Bond Pull Test Recommendation

Normal Al <3% (0.25 ≤σ> 0.15) x 0.9 (x) − 3σ)

High Rel. Al <3% σ ≤ 0.15 x 0.9 (x) − 3σ)

All Al 0.5 to 20% σ ≤ 0.25 x [(x − 3σ)]/2

All Al >20% σ ≤ 0.25 x [(x − 3σ)]/3
All Au Use same elongation and σ rules as aluminum, except
that the elastic limit is less predictable from one
manufacturer to the next.

TABLE 4B-2 Summary of NDP Force Recommendations Relation

–
is recommended for cases where σ > 0.25 x, since this indicates that
some aspects of the bonding procedure are out of control and either a
low, meaningless NDP force would have to be used or too many
bonds would be stressed beyond their elastic limits and/or broken.
Table 4B-1 gives a summary of the NDPT recommendations for wire
with various elongations.
4B.3 Assessment of Any NDP Test-Induced
Metallurgical Defects
During the NDP test, with the NDP force limits derived above, the
wire is only subjected to approximately one metallurgical stress-fatigue
cycle. Bulk Al and Au will normally withstand hundreds of thou-
sands of such cycles when the stress is kept below the elastic limit. The
stress during the NDP test is primarily along the wire; thus, there are
essentially no outer-fiber-strains (from bending) in the bond heel area
to enhance the probability of unannealable crack formation.
Under these conditions, any stress-fatigue developed below the
elastic limits of the bond-loop system during the NDP test should be
small. Also, almost all devices whose reliability is critical enough to
require NDP tests (usually space applications) will routinely undergo
thermal screens, such as burn-in (~125°C for 168 h or equivalent), or
such screens could be added if desired. These screens should anneal
any threshold level of NDP-test-induced fatigue occurring below the
elastic limits, and they can also anneal some, if not all, of the stress-
fatigue which might occur above the elastic limit, assuming no crack
has formed. Thus, only a small fraction of the NDP-tested bonds, whose
breaking strength is in the inelastic stress range of Table 4A-2.1, would
retain a significant number of test-induced metallurgical defects after a
typical burn-in or other annealing period. Even for a case where a
small non-annealable crack remains in the bond heel, it would not