a. Gate Metal Sinking refers to a mechanism, which
the channel depth decreases when the gate metal diffuses down into the GaAs,
effectively moving the Schottky barrier deeper into the part. The effect is
accompanied by a decrease in Idss, gm, and pinch-off
voltage. The gate sinking mechanism occurs when the gate metal diffuses into the
GaAs. Defects in the FETs barrier can lead to local diffusion of gold through
the barrier, resulting in premature failure due to localized gate sinking. This
usually occurs at the edges of gates where shadowing has resulted in thinner
barrier metal deposition.
b. Electromigration is the movement of metal atoms in
a conductor carrying a high current caused by momentum exchange between moving
electrons and the atoms. This is an important mechanism in high-frequency power
transistors where high current densities and high temperature can occur
simultaneously. Electromigration results in voids forming in the metal and in
the accumulation of metal atoms. When the voids increase enough, they may sever
a metal run causing part failure. Likewise, when the metal accumulation becomes
large enough in the lateral or vertical (under metal air bridges) direction,
they can cause shorting between conductors. This effect varies significantly for
different metals; for example, the lighter metals such as aluminum are more
susceptible. The rate of electromigration varies as approximately the square of
the current density and has an activation energy, which varies from 0.5 eV to
greater than 1 eV, depending on the metal and its structure. Electromigration
was an early failure mechanism in GaAs FETs. Electromigration in the
drain and source metallization have lead to failures when metal accumulates,
causing shorting to air bridges or other adjacent structures. Covering the metal
with Silicon Dioxide (SiO2) or silicon nitride (Si3N4) will retard
Electromigration.
c. GaAs FETs and HEMTs are susceptible to several
different forms of surface degradation, which affect their ultimate life.
Described are failure modes:
1. Unpassivated parts are subject to surface
oxidation which releases free arsenic. The result is a reduction in breakdown
voltage, and an increase in low voltage gate current. Because of the breakdown
voltage reduction, the power output is decreased and eventually catastrophic
failures occur due to gate-drain breakdown.
2. Parts passivated with SiO2 experience
erosion of the GaAs surface, due to an interaction of the GaAs with the
SiO2. This results in a narrowing of the channel and an increase in
the source and drain resistance. Mobile ion contamination of the oxide or
nitride passivation can result in instability. The effect is an increase in
source resistance, due to an increase in the surface depletion layer, which
narrows the channel depth. The original performance can be recovered by a
175oC/24-hour bake. Passivation of the surface above the active layer
is necessary to prevent instabilities.SiO2 is not an adequate
passivation and sputtered silicon nitride may not be suitable because of mobile
ion content.
d. The ohmic contacts to GaAs parts can increase in
resistance with time at elevated temperature, eventually leading to wear-out
failure. The most common metal system for the contacts is Gold Germanium/Nickel
(AuGe/Ni). Typically, their activation energy is 1.4 to 1.8 eV with a
Mean-Time-Between-Failure (MTBF) of greater than 108 hours at
100oC. This may provide adequate contact life at a operating
temperature below 100° C, but rapid degradation can result at temperatures above
150°C. The mechanism involved with this degradation is the continued alloying
and intermixing of the contact metals. In particular, nonstochiometric regions
are formed when Ga diffuses through the AuGe into the Au layer, while Au
diffuses inward forming high resistive alloys, which causes the contact
resistance to increase.
e. Source-drain burnout is a common failure mode of
GaAs FETs, accounting for 30 to 50% failures. This catastrophic failure mode
results in substantial melting in the active region due to thermal runaway.
There are two types of burnout of concern: instantaneous and long term, each
caused by different mechanisms. Instantaneous burnout, including ESD failures,
will occur when the applied source-drain voltage exceeds the breakdown voltage.
Long-term burnout occurs during DC aging, resulting in catastrophic failure
after the device has been stressed for some time. This burnout is associated
with changes at the surface of the device near the drain contact. Before
breakdown, light emission is observed near the drain contact of n+ ledge,
associated with microplasmas or localized breakdown sites.
f. Many FETs use Al gate metallization and AuGeNi
ohmic contacts. Gold metallization is used for the wire bonding pads because it
is compatible with the ohmic contact materials and the gold wires used to
package the devices. Consequently, a transition is required between the Al gate
metal and the Au pad metal. Early devices had an Au-Al contact between the gate
and pad. When stressed at elevated temperatures, Au-Al intermetallic phase
formation occurred and device failure resulted due to Kirkendall voiding. This
same “purple plague” mechanism has been a traditional failure mechanism in
silicon devices in which Au wires are used to connect to Al metallization. Later
GaAs FETs employed a barrier layer between the Au and Al to prevent
interdiffusion. Although this barrier has been effective in reducing the effect,
failures were still observed at high temperatures. Gold apparently migrated
along the surface to interact with the Al gate fingers.