Fatigue Failure of Encapsulated Gold-Beam Lead and TAB Devices

A thermal stress model for gold-beam lead and tapeappliqued beam (TAB) devices TC bonded to a substrate is presented. The bonded appliqued beam-centerline is assumed to be geometrically similar to that of a bonded gold-beam lead. The devices are assumed to be encapsulated with room-temperature vulcanizing rubber (RTV). For gold-beam lead devices and TAB devices with large chips and long beams (LL
c
\geq
0.0014 in
²
, where L is free beam length and L
c
is chip length), the analysis predicts that the RTV remains intact and drives the chip up and down during temperature cycling. For small TAB chips with short beams (LL
c
\leq
0.0004 in
²
), however, the RTV may rupture and would then not control the chip motion. The protective capability of RTV is uncertain for such a structure. An analysis of gold beam fatigue relates the cycles to median failure (N
f
) to magnitude of a temperature cycle (
\Delta
T). The analytical results are compared with temperature cycling data for a gold-beam leaded device at
\Delta
Ts of 190, 100, 75, 50, and 35°C. Although there are differences between the experimental and analytical results, the trends are close and the degree of agreement is encouraging. Additional comparisons are made to establish the influence of the beam dimensions, gold properties and bonded geometry on the fatigue life. The N
f
versus AT analysis combines estimates of cyclic strain (
\beta\varepsilon
) as a function of AT with experimentally obtained fatigue data (N
f
as a function of
\Delta\varepsilon
). The
\Delta\varepsilon
versus AT estimates are derived from elasticity and plasticity stress analyses. Since fatigue data on electrodeposited gold is apparently nonexistent, N
f
versus
\Delta\varepsilon
data on bulk gold is used. Equations of Weibull are used to scale the data to specimens with beam-lead dimensions. It is concluded that the model provides an adequate prediction of gold-beam lead device behavior over a wide range of temperature cycling conditions and beam geometries predictable from models of beams bent during bonding. Uncertainties exist over the relative severity of the high temperature portion of the thermal cycle versus the low temperature regime. The uncertainty is related to the effect of mean strain on fatigue life as well as to the annealing of gold. The low-cycle fatigue range is reasonably expressed by the Coffin-Manson law (i.e., N
f
= C(
\Delta
T)
²
) subject to the conditions that the plastic strain is large and is valid for a limited temperature cycle range. The elastic range extends for a
\Delta
T of roughly 50°C and normal fatigue behavior exists for
\Delta
T less than 50°C.