-1 color=black>

Process-Induced Thermal Effect on Packaging Yield of RF MEMS Switches 1
Copyright © 2002 by ASME
Proceedings of IMECE2002
ASME International Mechanical Engineering Congress & Exposition
November 17-22, 2002, New Orleans, Louisiana
IMECE-EPP-34266
PROCESS-INDUCED THERMAL EFFECT ON PACKAGING YIELD OF RF MEMS
SWITCHES
Lei L. Mercado/Motorola
Shun-Meen Kuo/Motorola
Tien-Yu Tom Lee/Motorola
Russ Lee/Motorola
ABSTRACT
RF MEMS switches offer significant performance
advantages in high frequency RF applications. The switches are
actuated by electrostatic force when voltage was applied to the
electrodes. Such devices provide high isolation when open and
low contact resistance when closed. However, during the
packaging process, there are various possible failure modes that
may affect the switch yield and performance.
The RF MEMS switches were first placed in a package and
went through lid seal at 320ºC. The assembled packages were
then attached to a printed circuit board at 220ºC. During the
process, some switches failed due to electrical shorting. More
interestingly, more failures were observed at the lower
temperature of 220ºC rather than 320ºC. The failure mode was
associated with the shorting bar and the cantilever design.
Finite element simulations and simplified analytical
solutions were used to understand the mechanics driving the
behaviors. Simulation results have shown excellent agreement
with experimental observations and measurements. Various
solutions in package configurations were explored to overcome
the hurdles in MEMS packaging and achieve better yield and
performance.
INTRODUCTION
GaAs FET switches and diode switches have been widely
used in RF front-ends of cell phones to switch antenna bands
and transmitter/receivers. However, in multi-band cell phones,
GaAs FET switches do not have sufficient isolations to
minimize cross-interference and signal jamming from channels
in close proximity. Pin diodes, on the other hand, typical need
considerable amount of power to operate which significantly
lower the operation life of the battery. MEMS switches provide
high isolation when open, low insertion loss when closed and
can be operated at very low power consumption. Therefore, RF
MEMS switches are a very attractive solution to switch antenna
bands and transmit/receive switching for future multi-band, high
bandwidth cell phones. MEMS switches are usually categorized
by the contact methods: capacitive (metal-insulator-metal)[1-3]
and resistive (metal-to-metal)[3-6]. In this paper, electrostatic
switches with metal-to-metal contact are considered. In
electrostatic actuated switches, the switches are actuated by
electrostatic force when voltage was applied to the electrodes.
Figure 1 (a)-(c) show the schematic drawing of a generic RF
MEMS switch structure. The shorting bar is connected to the
electrostatic actuated cantilever beam and moves to open and
close the switch.
A
A
A
A
B
B
B
B
ES electrodes
µ
-cantilever
(a) Top View of the Switch
shorting bar
A-A
A-A
Substrate
SiON
(b) Side View of the Switch along Line A-A'
B-B
B-B
In
Out
(c) Side View of the Switch along Line B-B'
Figure 1 Schematic Drawing of RF MEMS Switch Structure
Proceedings of IMECE2002
ASME International Mechanical Engineering Congress & Exposition
November 1722, 2002, New Orleans, Louisiana
IMECE2002-39
266
EPP TOC 2
Copyright © 2002 by ASME
PROBLEM DEFINITION
Packaging processes present the major challenge for the
communication of MEMS products. MEMS switches with free
moving parts are very sensitive to the temperature excursion
during the packaging process. A mismatch of thermal expansion
coefficients between materials layers causes the switch to
deform. The structure distortion may render the switches with
degraded performance or failure. Many researches have touched
upon the subject of MEMS device packaging. However, most
were trying to vary the packaging conditions to avoid the
deformation of the switches [6-9] and were not addressing the
problem at the design stage. By doing so, special packaging
process and materials have to be developed and significantly
increased the cost of final product. Therefore, to ensure the
stability and performance of packaged switches remained in
acceptable specification and maintain the low cost packaging
solution, the interaction between MEMS structures and
temperatures must be taken into consideration during the design
phase.
The MEMS packaging process flow is shown in Fig. 2. The
individual control die with the RF-MEMS switch pattern was
first placed in a cavity package with die-attach materials. The
inputs and outputs of the switches were then wirebonded to the
corresponding pads in the package. The package is placed in a
plasma asher that ramps up to 200ºC to etch away the polyimide
under the switch beam structure. When all the polyimide was
etched away, the switch was then released and can move freely.
Since the switch is etch-released at high temperature in the
asher, when it was cooled down to room temperature, there
might be a slight upward warpage in the switch due to the
thermal mismatch between the gold and the dielectric material.
The MEMS package must be hermetic because the switch
performance is very sensitive to the environmental conditions
such as humidity, contaminants, etc. A lid is placed on top of the
package. There is a band of Au/Sn solder around the edge of the
lid. The solder was reflowed at 320ºC so that the lid was sealed.
The sealed package was electrically tested. They were then
attached to a printed circuit board at 220ºC. Electrical tests
were performed at the end of the board attach. To study the
temperature impact of the package process on RF-MEMS
devices, yield was monitored when the devices were going
through each stage. An interesting phenomenon was observed:
while most devices were still functional at the higher lid-seal
temperature, the switches became "stuck" at the lower board-
attach temperature.
Why would the devices survive lid seal at 320ºC, but fail
from electrical short at 220ºC? The switch had only gone
through one cycle, so fatigue failure is unlikely. In electrostatic
loading, as voltage increases on the beam, the beam is pulled
closer to the ground, resulting in more contact. As temperature
increases, should there not be more contact between the beam
and the ground, therefore more devices failing from short? This
paper provides an explanation to this phenomenon through
finite element simulation and a simplified analytical solution.
We can learn from this process issue and achieve more robust
switch designs.
Die attach to
Package
Etching
(200
°
C)
Wire
Bond
Seal
(320
°
C)
Testing
Testing
Board Mount
(220
°
C)
Figure 2 Process Flow of Making a RF MEMS Switch
Package
FINITE ELEMENT SIMULATION
A three-dimensional model has been constructed to
simulate the cantilever MEMS beam, as shown in Fig. 3. The
cross-sectional view of this model is shown in Fig. 4. The beam
is composed of a top electrode, which is a thin layer of gold,
deposited on a 2-micron thick SiON dielectric layer. On the
bottom of the beam, located 40 microns away from the beam
tip, a thin gold metal strip forms the shorting bar. When positive
charges were applied to the top electrode, the electrostatic
forces between the beam and the ground would pull the beam
towards the ground. If the shorting bar touches the signal trace
in the same plane as the ground electrode, then the switch is
turned on.
As shown in Fig. 4, the shorting bar thickness, its inset
distance from the cantilever beam tip, and the gap between the
bottom of the dielectric beam and the top of the ground can all
be design parameters determining the geometry and location of
the shorting bar. Here we will examine two cases with different
shorting bar thicknesses.
Case 1: 0.5
µ
m Thick Shorting Bar with 2
µ
m Gap
To understand what this switch went through during the
process, we studied the steps from etching to lid seal, then
board attach. The switch was released at the 200ºC asher
temperature, so 200ºC was used as the stress-free temperature.
Gold has a coefficient of thermal expansion (CTE) of 15
ppm/ºC, while SiON has a much lower CTE of 0.55 ppm/ºC.
The entire beam will bend during the temperature excursion
because of the large difference in the CTE between the gold and
the dielectric,. Suppose the beam was horizontal (flat) at 200ºC,
as temperature increases, the gold expand more than the SiON