Sei sulla pagina 1di 9

Electrical Breakdown across Micron Scale Gaps in MEMS Structures

Fabian W. Strong 1, 2, Jack L. Skinner 2, Paul M. Dentinger 2, Norman C. Tien 1


1
University of California, Davis, Department of Electrical and Computer Engineering
2
Sandia National Laboratories, Livermore, CA

ABSTRACT

Large voltage differences between closely spaced MEMS structures can cause electrical breakdown and destruction
of devices 1-2. In this study, a variety of planar thin film electrode configurations were tested to characterize breakdown
response. All devices were fabricated using standard surface micromachining methods and materials, therefore our test
results provide guidelines directly applicable to thin film structures used in MEMS devices.
We observed that planar polysilicon structures exhibit breakdown responses similar to published results for larger
metal electrode configurations 3-6. Our tests were performed in air at atmospheric pressure, with air gaps ranging from
0.5 µm to 10 µm. Our results show a sharp rise in breakdown level following increases in gap width up to about 3 µm, a
plateau region between 3 µm and 7 µm, and breakdown in gaps over 7 µm following the Paschen curve. This profile
indicates an avalanche breakdown process in large gaps, with a transition region to small gaps in which electrode
vaporization due to field emission current is the dominant breakdown process.
This study also provides information on using multiple-gap configurations, with electrically floating regions located
near the energized electrodes, and the added benefit this method may provide for switching high voltage with MEMS
devices. In multiple-gap configurations, we noted a transition between direct tip to tip breakdown across electrode gaps
of 40 µm, and a preferential breakdown path through the electrically floating contact head region for electrode gaps over
100 µm.

Keywords: Electrical breakdown, Paschen curve, MEMS microswitch, surface micromachining

1. INTRODUCTION

1.1 Motivation
This research is motivated by the desire to build a MEMS microswitch capable of standing off high voltage levels,
yet needing only a small actuation distance, achievable using standard MEMS actuators.
Solid state switches can hold off high voltages and take up very little space, but they can leak current at high
voltages giving rise to constant power losses. Standard mechanical relays provide high isolation, but they are
comparably very large. A MEMS microswitch, with high off state isolation at high voltages due to a physical air gap,
can be contained in a very small package. MEMS fabrication processes allow mass manufacturing at reduced cost, when
compared to standard macro-scale relays, as well as easier integration into multiple-switch packages. We are currently
exploring methods for fabricating a MEMS based microswitch for high voltage switching. This paper presents the
results of the initial testing phase, in which various air gap distances and contact configurations were investigated to
determine useful designs.
Mechanical switches are prone to failure through electrical breakdown across physical contact gaps in an over-
voltage condition. Widening the gap between electrodes would raise the breakdown voltage level, but MEMS actuators
can only provide limited actuation distances while still providing sufficient force for metal electrical contacts to reach
the plastic deformation stage necessary for low contact resistance. A longer contact throw would also require more
actuation power and create higher current spikes. It is our desire to minimize the power requirements of our switch. For
these reasons, we investigated the breakdown response of planar polysilicon MEMS structures in a variety of
configurations to determine beneficial designs for high voltage switch contacts, allowing high voltage switching with
short actuation distances.
Our results also provide new design guidelines for placement and distribution of closely spaced MEMS structures
by stating empirical breakdown response data specifically suited to surface micromachined thin film structures
fabricated from standard MEMS materials (polysilicon and gold).
Previous electrical breakdown tests covering this range of gap dimensions have typically been performed using
substantially larger electrodes in combinations of pin and plane configurations with a variety of metal electrode
materials 4-5, or using metal traces on glass slides 6.
While there are a small number of published studies on electrical breakdown for freestanding planar devices made
with standard MEMS materials, they are somewhat limited in their variety of gap distances and configurations, and do
not focus on the same aspects of breakdown that we determined. Walraven et. al. focused on the transient responses of
electrostatic discharge (ESD) between the fingers of electrostatic comb drive actuators and the substrate 1, while Ono et.
al. focused on pre-breakdown micro-discharges, and detected breakdown effects by light emission rather than current 2.
We have confirmed that the breakdown response for planar MEMS structures is similar to the results of previous
work that used larger electrodes of different shapes and materials, across the full gap range where a transition in
breakdown process occurs (0.5 to 10 µm).

1.2 Breakdown processes


In 1889, Friedrich Paschen first expressed the electrical breakdown voltage in a gas-filled gap as a function of the
product of gas pressure and gap distance 7. Townsend later reported that avalanche breakdown is caused by ionization of
atoms within the gap due to acceleration of free electrons inside an electric field 8. When a free electron’s kinetic energy
has increased enough within its mean free path to allow an ionizing collision with another atom, the result is two free
electrons and a positive ion. These species are then accelerated until they cause new ionizing collisions, bringing the
population to four electrons, then eight electrons, and so forth, cascading in an exponential increase across the gap from
each initiating electron.
All previous empirical studies of breakdown across micron scale distances have noted a departure from the Paschen
curve for very small air gaps at atmospheric pressure 1-6. This is due to the fact that the Paschen curve is based solely on
an avalanche breakdown process that is dependent on impact ionizations occurring on the gas atoms within the air gap.
However, at gap distances of only a few times the mean free path of electrons (which is about 0.5 µm in atmospheric
air), there are not sufficient numbers of ionizing collisions to produce an avalanche breakdown.
With only a few collisions occurring at small gap distances, vaporization of electrode material due to field emission
current becomes the dominant breakdown method. This ‘vapor arc’ is the typical process for breakdown in a vacuum.

Standard Paschen Curve


Vacuum Curve
Modified Paschen Curve

Figure 1. Modified Paschen Curve


for air at atmospheric pressure
In the vapor arc process, a high electric field overcomes the work function of the electrode material and pulls
electrons from the cathode into the air gap. Rough surface texture on the electrodes, which is practically unavoidable in
fabrication, allows the field to be enhanced at any small asperity or sharp protrusion. The resulting intensified field
emission current, being highly localized and constricted, heats the material of the tiny emitting asperity to the point
where it vaporizes. This vapor quickly spreads across the gap and creates a highly conductive bridge, providing a path
for breakdown current.
To summarize: for large air gaps, breakdown response follows the traditional Paschen curve, and at very small air
gaps, it follows the vacuum curve. In between these extremes is a transition region where experimental data has shown a
flat response, typically called the plateau region. Experimental results have confirmed that actual breakdown responses
will follow a modified Paschen curve, as shown in figure 1. There are various different explanations proposed for the
plateau region, and they all may have an effect on breakdown, depending on the specific device configuration used.
Germer and Schaffert have proposed that there is a tendency for discharges to take a longer path, rather than arcing
directly across the minimum gap distance, allowing an avalanche process enough distance to build up to breakdown
levels, and that this effect provides the plateau seen in most empirical testing 3, 9. For gap distances between the Paschen
minimum point (~ 7 µm in atmospheric air) and the gap distance on the vacuum curve corresponding to the Paschen
minimum voltage (~ 3 µm gap distance), a longer indirect path would allow an avalanche to progress to breakdown at
the minimum Paschen voltage. However, an avalanche path would need to follow electrical field lines because they
determine the acceleration direction of electrons. For flat and parallel electrodes (providing a homogeneous field
between them), this breakdown method would have to occur at the edges of the electrodes where the field lines can
curve into longer paths, limiting the electrode area where this type of breakdown can occur.
Although there is insufficient direct-path gap distance in the
plateau region to create an avalanche breakdown, there are still
some limited number of ionizing impacts occurring within the
gap, which in turn contribute some space charge. This charge
could enhance field emission from electrode surfaces and lower
the breakdown level from the values that the vacuum curve
indicates.
Additionally, shape enhancement factors can have a
significant effect on field strength and distribution, so they are 50 um 50 um
also quite likely to affect breakdown processes at these small
distances. Basic electrode shape at the micro-scale, and surface
A B
texture in the nano-scale, would both contribute to shape
enhancement effects.
There are obviously many contributing factors to the
breakdown response across small air gaps, and the impact of
each will depend on the configuration, material, and even the
fabrication process for the specific electrodes being used. For
this reason it is suggested that any general models of breakdown
be used only as guidelines, and that some empirical testing be
performed for the specific electrode configuration and 50 um
fabrication process being used.
C
2. EXPERIMENTAL

2.1 Device design


All devices were fabricated from thin films deposited on a
crystalline silicon wafer (100), with initial isolation layers of
thermal oxide (1 µm) and silicon nitride (0.7 µm). A sacrificial
40 um
oxide layer (2 µm) was used for patterning of freestanding
structures, and a layer of low stress nitride (0.7 µm) was used for
further isolation regions. All electrodes are patterned from a 2 D
µm thick layer of polysilicon. Some devices were sputtered with Figure 2. Test structure configurations:
gold (0.25 µm) on top surfaces and selected sidewall areas. A) Single gap, 5 um radius electrodes,
Three styles of devices were investigated in this study; B) Single gap, 25 um radius electrodes,
single gaps, in-line double gaps, and double-gap switches with C) In-line double gap, 10 um radius electrodes,
offset & overlapping contact heads (see figure 2). D) Switch layout, 10 um radius electrodes with
offset and overlapping contact head.
The single air gap electrode configurations, ranging from 0.5 µm to 10.0 µm in gap distance, used planar electrode
widths of 10 µm to 50 µm (in 10 µm increments). The electrodes for all devices in this study had rounded ends, with
diameters matching the electrode widths.
The double gap test structures also used planar electrode widths from 10 µm to 50 µm, had two identical air gaps
(4, 6, or 8 µm) and an electrically floating region of the same width as the energized electrodes, using an in-line
configuration.
The third type of test structure also contained two energized electrodes and a floating region, but used an offset and
overlapping design. This configuration is intended to be used for microswitch contacts.
In the single and in-line double gap structures, all parts of the device are directly anchored to the substrate. In the
switch designs, the energized electrodes are anchored to the substrate, but the electrically floating contact head regions
have been mechanically released from the substrate and can be actuated towards the electrodes, physically making
electrical contact, and creating a continuous conduction path between electrodes in a switchable manner. Electrical
isolation from the actuator mechanism is provided by a block of silicon nitride between the contact head and the
actuator assembly.
High Voltage Circuit

A
Electrode 1 B Electrode 2
C

Electrically Floating Region


(Contact Head)

Figure 3. Switch contact configuration consisting of two energized electrodes and one
electrically floating region to be used as a contact head for switching. Critical dimensions are;
A) Electrode gap, B) Overlap distance, and C) Actuation gap.
All switch devices use 20 um wide electrodes and contact heads.
Electrode and contact head ends are all completely rounded (10 um radius).

50 um 50 um

(A) (B)

50 um 50 um

(C) (D)

Figure 4. Switch configurations used in this study (see table 1 for dimensions).

Table 1. Switch configuration critical dimensions.


Switch Electrode Gap (um) Overlap (um) Actuation gaps used (um)
A 40 20 2.0 to 8.0 (0.5 increments)
B 40 5 4, 5
C 40 50 3, 4
D 100 20 2, 3, 4, 5
2.2 Test procedures
The breakdown tests for this study were performed using a Micromanipulator probe station allowing visual
monitoring, with a Hamamatsu C4900 power supply as the high voltage source, a Tektronix PS2520G for the 15V
supply, and a Hewlett Packard E3631A for the 0V to 5V reference signal.
The current was determined by monitoring the voltage across a series 1MΩ resistor using a Hewlett Packard
34401A digital multimeter (see figure 5).
For each test, voltage was applied to the device under test (DUT) at a microdevice probe station, and manually
ramped up (at approximately 5 V/sec) until a sharp rise in current was noted. In all cases, the sharp rise in current was
accompanied by a visible breakdown result, ranging from minor blemishes to full electrode destruction. Multiple tests
(6 to 8) were performed at each gap distance. Error bars in figure 6 indicate highest and lowest breakdown voltage
readings, with average values plotted as data points.

HAMAMATSU C4900 HP E3631A


POWER SUPPLY POWER SUPPLY
+ +
15V Supply in - - 15V Supply out
HP E3631A
+
0 to 5V Reference in + POWER SUPPLY
- - 0 to 5V Reference out

+
0 to -1250V Output
-
Rref DUT
HP 34401A 1MΩ
DMM Vmeas + PROBE
- STATION

Figure 5. Test setup for breakdown tests.

3. TEST RESULTS

3.1 Single gap structures Single Gaps (Bare Si)


Tests of our single gap structures showed a response 420
similar to those published for the same air gap distances
using large metal electrodes. The smallest air gaps 410
B re a k d o w n V o lta g e (V )

followed a steep rise from 0.5 µm to 3.0 µm gaps, the


Vacuum
results leveled out over a plateau region from 3 µm to 7 400
Response Plateau
µm, followed by a more gradual rise along the Paschen 390
curve for gaps over 7 µm wide (figure 6).
Paschen
It was also noted during testing that pre-breakdown 380
Response
field emission current at moderate voltages was
370
significantly higher for gaps smaller than 3 µm, and
increased in relation to shortness of the gap. This effect 360
was more pronounced in thinner electrodes (10 µm wide)
than thick electrodes (50 µm wide), most likely due to 350

shape related field enhancement 9. Further tests are


0 1 2 3 4 5 6 7 8 9 10
Gap Distance (um)
planned to measure a more accurate pre-breakdown
current response.
Figure 6. Single gap breakdown response
Breakdown indications ranged from small blemishes showing three regions of interest. Error bars
on electrode tips, to melted tips, or even complete melting indicate highest and lowest readings, and
of polysilicon electrode structures and signal lines. In the data points are averages of all readings.
case of gold sputtered devices, the typical result was
melting of the gold coatings, sometimes accompanied Single vs. Double Gaps
by destruction of the underlying polysilicon electrode 900

structures. 800

Breakdown Voltage (V)


700
3.2 Double gap structures 600
Devices with in-line double gaps (on either side
500
of a floating region) showed a much higher standoff
than devices with corresponding single gap distances 400

(figure 7). 300

200

3.3 Switch structures 100

Breakdown results for various devices having 0

electrically floating contact heads in addition to two 0 1 2 3 4 5


Gap Distance (um)
6 7 8 9 10

energized electrodes, in the switch configuration, were Double Gaps Single Gaps
compared. The measured breakdown level for switch
devices with a 40 µm electrode gap was reasonably Figure 7. Comparison of single and in-line double gaps.
consistent at about 600V, regardless of the overlap
distance (between the electrodes and the ends of the Switch Configurations
electrically floating region), and with no significant
800

dependence on the actuation gap (figure 8). See 700

figures 2, 3 & 4 for layouts of the device Breakdown Voltage (V)


600

configurations tested. 500

In comparison, switch devices with a 100 µm 400

electrode gap showed a 700V breakdown level, with a 300

slight dependence on the actuation gap. Test results 200

indicate that a longer electrode gap provides improved 100

standoff, but not quite equal to Paschen curve 0


predictions (figure 9). 0 1 2 3 4 5 6
Actuation Gap Distance (um)
100um gap, 20 um ovlp 40um gap, 20um ovlp
40um gap, 5um ovlp 40 um gap, 100um ovlp

Figure 8. Breakdown results for a variety of switch


configurations. Electrode gaps of 40 um or 100 um,
and contact head overlap of 5, 20, or 100 um.
Breakdown Voltage (V)

Electrodes Only
+ -
Switch Configuration
+ -
In-line Configuration + -

0 20 40 60 80 100 120 140


Electrode Gap (um)
Figure 9. Test data plotted as breakdown voltage vs. electrode gap,
with values shown in relation to Paschen curve.
4. DISCUSSION

4.1 Tip vs. sidewall breakdown


A difference was noted in the breakdown patterns between switch devices with 40 µm electrode gaps, and those
with larger electrode gaps. A transition point for preferred breakdown path occurs between 40 µm and 100 µm electrode
spacing.

40 um 40 um 40 um

A B

40 um 20 um

C D

100 um 40 um
Figure 11. Only sidewall damage was found in devices
with 100 um electrode gaps (no tip damage),
E F indicating the preferred breakdown path is through the
Figure 10. Tip vs. sidewall breakdown damage in floating region, rather than directly across electrode gap.
switch configurations with a 40 um electrode gap.
A) Undamaged device, B) Tip damage only,
C & D) Tip and minor sidewall damage
E & F) Extensive breakdown damage, including
typical damage at tip of electrode.
Visual inspection after breakdown of samples with a 40 um
40 µm electrode gap showed that switch devices which
experienced only slight breakdown damage all show
indications of breakdown originating at the very tip of the
electrodes. The devices with more extensive breakdown
effects all show damage to electrode tips in addition to
damage on the sidewalls between the electrodes and the
contact head (figure 10). Negative terminals are shown to
the left in all figures. Damage is worst on these terminals
due to the larger mass of the positive ions impacting
them, compared with the small mass of the electrons
bombarding the positive electrodes.
The observed damage pattern indicates that electrical
breakdown in these structures begins at the electrode tips
and propagates directly across the electrode gap,
spreading to the small actuation gaps at the sides only Figure 12. Breakdown path through floating zone
after the tips have vaporized and provided a highly on in-line double gap structures (distance between
conductive path between the electrodes and the floating energized electrodes is over 130 um).
region. Some of the 40 µm devices showed significant
damage to the electrode tips, with no damage to the floating region’s sidewalls, helping to confirm that breakdown is
initiated directly between the tips of the energized electrodes, and not through the contact head.
In contrast, the devices with a 100 µm electrode gap all show damage on the edges overlapping the floating region,
with no damage at the electrode tips (see figure 11). These results indicate that, at some electrode gap spacing between
40 µm and 100 µm, the preferential breakdown current path makes a transition from a direct route through the air
between the electrode tips to a path going across the actuation gaps and through the electrically floating contact head. It
was noted that the preferential breakdown path for the in-line double gap configurations was also through the floating
region (figure 12).
Tests of switch structures in which the contact head had been physically removed, leaving no floating region in
proximity to the gap, showed breakdown levels near the Paschen curve predictions (open squares in figure 9). The
contrast between devices with and without the floating region shows that providing the opportunity for the breakdown
through an alternate path lowers the standoff capability of the switch structures with wide electrode gaps.

4.2 Gold coated devices 10 um


The breakdown values for gold coated single gap structures all fell within the error ranges found in the
corresponding bare polysilicon tests. The results for gold coated devices with double gap in-line and switch
configurations were sometimes slightly lower than the bare polysilicon results, but were all within 10% of measured
values from identical bare devices. Gold was sputtered onto the polysilicon in 0.25 µm thick layers, so all corresponding
gap distances were actually 0.5 µm narrower in gold coated devices.
The same difference in preferred breakdown path between 40 µm and 100 µm electrode gaps was seen in the gold
coated devices (figure 13).

5. CONCLUSIONS AND FUTURE WORK

We have shown that electrical breakdown across air gaps in planar


MEMS devices reacts in a manner similar to previously reported
40 um empirical tests performed with larger three-dimensional metal
electrodes. The breakdown voltage response exhibits a sharp rise with
(A) gap distance up to 3 µm, then a plateau region up to the Paschen
minimum at 7 µm, and follows the Paschen curve for gaps larger than 7
µm.
The addition of a separate free-standing structure in close proximity
40 um to the energized electrodes of an air gap, such as a switch contact head
that forms an electrically floating region, can provide an alternate path
for breakdown. This alternate path does not show a significant deviation
of breakdown levels from the Paschen curve when the electrode gap is
40 µm or less, indicating direct breakdown initiation, but there is a
definite lowering of standoff capability from Paschen curve predictions
20 um 20 um
for devices with electrode gaps of 100 µm or greater.
The actuation gap for all switch configuration devices tested in this
(B) study was limited to a maximum of 8.0 µm. Future studies should
examine the effect of increasing this actuation gap, in conjunction with
Figure 13. Breakdown effects on gold a much wider range of electrode gaps, to further characterize the levels
coated devices. A) 40 um electrode gap at which an indirect breakdown path through the contact head is taken.
breaks down between tips, bypassing The configurations using a double-gap and electrically floating
the floating region. B) 100 um electrode
gap breaks down through contact head
contact head show great potential for increasing the standoff voltage in
as preferential path. MEMS microswitches. We have measured standoff voltages in this
configuration that were much higher than any levels possible for a
switch traversing the same actuation gap with a single-contact
configuration.
ACKNOWLEDGEMENTS

The authors would like to thank Dan McCormick for fabrication assistance performed at the Berkeley Sensor and
Actuator Center. We also thank Greg Cardinale, Steve Gianoulakis, Alec Talin, and Frank Jones, of Sandia, CA, for
advice, assistance, testing facilities, and many fruitful discussions. Finally, we would like to thank the Sandia MESA
Institute for providing funding assistance throughout the project. Sandia is a multiprogram laboratory operated by
Sandia Corporation, a Lockheed Martin Company, for the United States Department of Energy’s National Nuclear
Security Administration under contract DE-AC04-94AL85000.

REFERENCES

1. Walraven, J., Soden, J., Tanner, D., Tangyunyong, P., Cole, E., Anderson, R., Irwin, L., “Electrostatic
Discharge/Electrical Overstress Susceptibility in MEMS: A New Failure Mode”, Proc. Of SPIE vol. 4180, 2000,
pp. 30-39
2. Ono, T., Sim, D., Esashi, M., “Micro-discharge and Electric Breakdown in a Micro-gap”, J. Micromech.
Microeng., vol. 10, 2000, pp. 445-451
3. Germer, L., “Electrical Breakdown between Close Electrodes in Air”, Journal of Applied Physics 30 (1), 1959, pp.
46-51
4. Slade, P., Taylor, E., “Electrical Breakdown in Atmospheric Air Between Closely Spaced (0.2um-40um)
Electrical Contacts”, IEEE Transactions on Components and Packaging Technologies, 25 (3), 2001, pp. 390-396
5. Wallash, A., Levit, L., “Electrical breakdown and ESD Phenomena for devices with nanometer-to-micron gaps”,
Proc. Of SPIE, vol. 4980, 2003, pp. 87-96
6. Dhariwal, R., Torres, J., Desmulliez, M., “Electric Field Breakdown at Micrometer Separations in Air and
Nitrogen at Atmospheric Pressure”, IEE Proc. Sci. Meas. Tech., vol. 147, 2000, pp. 261-265
7. Paschen, F., Weid. Ann., 37 (69), 1889
8. Townsend, J., Electricity in Gases, Oxford University Press, 1915
9. Schaffert, R., “Charge Transport Mechanisms in the Transfer of Latent Electrostatic Images to Dielectric
Surfaces”, IBM Journal of Research and Development, 6 (2), 1962, pp.192-199

50 um 50 um

50 um 50 um

Potrebbero piacerti anche