Home Search Collections Journals About Contact us My IOPscience

Electrical breakdown phenomena for devices with micron separations

This content has been downloaded from IOPscience. Please scroll down to see the full text.

2006 J. Micromech. Microeng. 16 1366

(http://iopscience.iop.org/0960-1317/16/7/034)

View the table of contents for this issue, or go to the journal homepage for more

Download details:

IP Address: 153.96.96.22
This content was downloaded on 06/10/2015 at 08:41

Please note that terms and conditions apply.

INSTITUTE OF PHYSICS PUBLISHING JOURNAL OF MICROMECHANICS AND MICROENGINEERING
J. Micromech. Microeng. 16 (2006) 1366–1373 doi:10.1088/0960-1317/16/7/034

Electrical breakdown phenomena for

devices with micron separations
Ching-Heng Chen, J Andrew Yeh and Pei-Jen Wang
Department of Power Mechanical Engineering, National Tsing Hua University, 101,
Section II, Kuangfu Road, Hsinchu, Taiwan 30043, Republic of China
E-mail: [email protected]

Received 5 March 2006, in final form 3 May 2006

Published 30 May 2006
Online at stacks.iop.org/JMM/16/1366

Abstract
The electric breakdown of gaseous dielectrics predicted by Paschen’s law
has been successfully employed for the design of insulation among metallic
conductors separated by millimeter scale gaps or larger. However, today
most electrostatic actuated MEMS devices have been designed and
fabricated with micron-scale gaps that lead to a high risk of failure.
Paschen’s curves measured under macro or meso gaps hereby must be
re-examined and carefully evaluated for providing the guidelines of
insulation design in MEMS devices. In this paper, geometric configurations
commonly employed for MEMS electrodes, typically planar and
inter-digitated combs, have been designed and analyzed with commercial
software packages. For verification purposes, breakdown voltages were
measured on test samples fabricated with metallic, p-type and n-type silicon
substrates in standard MEMS processes. The samples were prepared by
etching insulation gaps measured from 2 to 21 µm. In conclusion, Paschen’s
curves for describing the electric breakdown phenomena of gaseous
dielectrics were revisited and compared to the published data so that design
guidelines of electrical insulation in MEMS actuators are established.

Introduction is common to pursue maximum usable forces by shrinking

insulation gaps until the onset of dielectric breakdown that
Nowadays as the speed and density of microelectronic chips leads to potential failure of devices. To improve the product
are growing as predicted by Moore’s law, the width of reliability and shorten the design cycle, it is necessary to
insulator between conductors has accordingly reduced. In explore and eventually theorize the basic phenomena of
addition, some innovative MEMS devices, such as digital electric breakdown in MEMS devices insulated by dielectric
micro-mirror devices insulated via dielectric gases, have been gases. In the late 19th century, Paschen [3] was the first
operated under a much intense electric field. Consequently, scientist to study the electric breakdown of dielectric gases
the fundamental physics for governing electric breakdown between metallic electrodes, and then formulate the so-called
in either IC or MEMS devices has drawn much attention Paschen law which has been so effective in the prediction of
in the past few years. Especially in MEMS technology, it electrical breakdown of dielectric gasses. Recently, Xiong
is well acknowledged that forces generated via electrically et al [4] reported that two major contributing factors, namely
charged electrodes in micron or even smaller separation are the roughness [2] and the ionization process of dielectrics, have
most efficient and effective mechanism for actuating MEMS dominant effects on the apparent electric breakdown of
devices. Driven by various application requirements, it is dielectric gases in the meso- and micro-scale.
most desirable to fabricate micro-electrodes with small gap From the literature, it is noted that verifications on
and large surface area because a strong electric field can be Paschen’s law for predictions of safety operational voltage
easily maintained under low applied voltage plus dimensional under millimeter scale gaps or larger have been extensively
manipulations [1]. In Madou’s book [2], Paschen’s law has studied in the past. In summary, Paschen’s law is based
been employed as the fundamentals of breakdown design in upon the Townsend breakdown mechanism in dielectric gases
electrostatic MEMS actuators. In practical applications, it dominated by two processes simultaneously, namely the

0960-1317/06/071366+08$30.00 © 2006 IOP Publishing Ltd Printed in the UK 1366

 Electrical breakdown phenomena for devices with micron separations

ionization process of gas molecules via electron collisions and Table 1. Configurations and dimensions of electrodes fabricated on
the secondary electron emission process via ion bombardments the SOI and glass wafers.
on cathode surface. It should be noted that the existence of Length Width Gap Height
breakdown voltage depends on a critical number of electron Configurations\dimensions (µm) (µm) (µm) (µm)
multiplications throughout the ionization process induced by Silicon planar electrodes 100 20 2–21 40
collisions between electrons and gas molecules. According Silicon comb electrodes 26 4 2–21 40
to Nasser’s book [5], apparent breakdown voltage for generic Aluminum planar electrodes 100 20 2–21 0.3
dielectric gases is successfully governed by Paschen’s curves Aluminum comb electrodes 26 4 2–21 0.3
which are formulated as a product of gas pressure and
gap distance in Vs = f (pd). From the basic terms in Table 2. Manufacturer specifications of 4 inch SOI wafers.
the function, the breakdown voltage of common dielectric
gases under a uniform electric field is explicitly defined in Resistivity Silicon layer Oxide layer
SOI wafer Dopant (
cm) (µm) (µm)
terms of pressure p and gap distance d. In 1974, Dakin
et al [6] published more measured data on the breakdown n-type Antimony 0.008–0.02 40 2
voltage of various dielectric gases, and they reinforced the p-type Boron 0.01–0.05 40 4
effectiveness of Paschen’s law under uniform electric field
conditions. It is interesting to note that most published
data were obtained under reduced gas pressure in order to time, Wilson et al [16] published measured similar breakdown
circumvent the measurement difficulties with large electrodes voltages in vacuum caused by micro-discharges measured on
at minute gap. However, micro-electrodes in MEMS devices planar electrodes.
are typically separated by several micron gaps fabricated In this paper, micron-scale electrodes made of doped-
on silicon wafers which are encapsulated with air under silicon or sputtered metals were fabricated via standard MEMS
atmospheric conditions. When the gap distance is a couple processes on specific wafers. The insulation gaps between the
of microns, interesting results were reported in the 1950s by electrodes were designed from 2 to 21 µm with increments of
Germer and Kisluik [7, 8] and later concluded in Schaffert’s 1 µm. Furthermore, breakdown voltages under atmospheric
book [9] stating that the Townsend avalanche mechanism is conditions were measured and compared to the literatures with
less dominant under some special conditions. the help of automatic voltage–current test apparatus. On
More recently, Dhariwal et al [10] first conducted micron- the other hand, computer simulations for electrostatic field
gap experiments with macro-electrodes made of various intensity were conducted on Maxwell 2D, copyrighted by
metals to verify breakdown voltages of dielectric gases under Ansoft Inc., Pittsburg, PA, for illustrating the effects of surface
atmospheric conditions. They concluded that under nitrogen roughness on local field concentration near the electrodes.
and air atmospheric pressure Paschen’s law is not applicable
to electrodes separated with gaps smaller than 4 µm. Later Electrode fabrication
in 2002, Slade and Taylor [11] compared the data obtained
by Dhariwal et al to the data obtained by Lee et al [12]; and As a first step, non-metallic electrodes were fabricated with
they concluded that the Townsend avalanche mechanism is less 40 µm thickness on silicon layers of silicon-on-insulator
dominant in electric breakdown for gaps smaller than 6 µm. wafers (SOI) available from the stocks. Two types of doped-
In 2005, Radmilovic-Radjenovic et al [13] simulated field silicon wafers, namely n-type and p-type, are available for
emissions via particle-in-cell plus Monte Carlo method for these experiments. n-type silicon wafers are typically doped
ion-enhanced field emissions on macro-scale cathode. They by antimony while p-type silicon wafers are doped by boron.
reported that simulated results agree well with experimental Both of the doped wafers have been measured with electric
data when the gap is smaller than 5 µm. resistivity from 0.008 to 0.02
cm. After the lithographic
In contrast to macro-electrodes, Ono et al [14] concluded masks of plate and inter-digitated comb configurations were
that micro-discharges do occur in micro-gaps under electric made, the electrodes were fabricated via inductively coupled
field below the breakdown threshold because of uncertainty plasma (ICP) etching process available in MEMS fabrication
in the ionization process. Experimental data for electric labs on campus. As for the metallic electrodes, they were
breakdown deviate from Paschen’s curves for metal-to-silicon fabricated by evaporating 300 nm thick aluminum films on
cases on micro-electrodes which exhibit numerous micro- glass wafer substrates and then etched via metal RIE process
discharges. But for silicon-to-silicon cases, Paschen’s law in the labs. The insulation gaps for both the non-metallic and
is still valid. It is interesting to note that they reported that metallic electrodes were chosen to be from 2 to 21 µm with an
the micro-discharges evaporate the anode electrode which increment of 1 µm. Finally, all test samples were confirmed
presumably leads to increase in pressure and decrease in for dimensional tolerances of ±0.3 µm on gap width due to
breakdown threshold during the pump experiments. In 2003, the inherent process uncertainty.
Wallash et al [15] confirmed the discrepancy in breakdown Tables 1 and 2 summarize the electrode configurations
voltages based upon Paschen’s law for cases of metallic micro- and pertinent information employed for this study. Then,
electrodes separated with micron-scale gaps. In addition, planar and inter-digitated comb electrodes etched on the
breakdown voltage exhibits a plateau region near 3 to 2 µm 40 µm thick SOI wafers are shown in figures 1(a) and (b)
and then drops linearly as the gap keeps decreasing. The under microscopic pictures. For the same samples, figures 1(c)
breakdown mechanism is field emissions and is formulated and (d) show photographs of the electrodes after the
into Fowler–Nordheim lines with a constant slope. At the same breakdown measurement. As further comparison tests,

1367
C-H Chen et al

(a) (b)

(c) (d )

Figure 1. Photographs of silicon electrodes under a microscope. (a) Planar electrodes with 2 µm gap before test; (b) comb electrodes with
20 µm gap before test; (c) breakdown voltage at 406 V with the left electrode being burnt; (d) breakdown voltage at 396 V with one of the
fingers on the left electrode being burnt.

figures 2(a) and (b) show photographs with dimensions electrodes is too cumbersome to be accurately measured by
different from the samples in figure 1 plus the electrodes AFM scanning.
being made of etched-aluminum film on glass wafers. After From our experiences, the breakdown voltages are very
the breakdown experiments were done, figures 2(c) and (d) sensitive to contamination on wafers because the electrodes
are photographs of the same electrodes with some parts being were fabricated via ICP or metal deposition processes. After
melted and vaporized. It should be noted that due to the field of the wafers were carefully cleaned with appropriate procedures,
view on the microscope, the contact pads for electric probes are the test samples were placed on a probe station built in-
not shown in the pictures. The same experimental procedures house located in a dark chamber confined with a class 1000
were conducted for all samples prepared for measurement in cleanness environment. Meanwhile, the ambient temperature
a controlled dust-free environment. was controlled at 20 ◦ C under laminar air flow. A set of
stage manipulators with probes was employed for providing
electrical contact on the test electrodes. All the test samples
Simulations and experiments were grounded to remove residual charges prior to the
experimental procedures. In the past, the definition of the
For exploration of electric field distributions between the breakdown phenomenon is to measure the applied voltage
electrodes with various configurations, numerical simulations while observing the on-set of a self-sustaining discharge under
have been conducted for locating the position where maximal low pressure conditions. However, if it is under atmospheric
field intensity occurs between the electrodes excited with the conditions while the electric field is non-uniform between
experimental conditions. Since computer-aided engineering the electrodes, a corona discharge (also self-sustaining) is
programs for analyzing electrostatic fields have been well observed so that abnormal breakdown could be recorded
recognized for their accuracy and efficiency, Maxwell 2D [17]. In this study, valid breakdown voltage is defined as
copyrighted by Ansoft Inc., Pittsburg, PA, was employed for the conditions that a substantial electric current occurs while
conducting the numerical simulations. After the dimensions an abrupt voltage drop is measured across the gap which is
of the electrodes, boundary conditions and excitation sources electrically similar to the short-circuit condition.
are entered into the program, electrostatic field distributions Figure 3 illustrates the schematic drawing of the
for the planar and inter-digitated comb configurations were test apparatus employed for the breakdown measurements
calculated and then converted to maximum field intensity consisting of a programmable power supply made by Stanford
values. Though the simulations can be conducted with proper Research Systems Inc., Model PS350, being employed for
pre-processes, the effects of surface roughness were assumed the voltage source. In addition, a precision multi-meter
to be negligible because the sidewall surface roughness on the made by Keithley Inc., Model 485, is used for recording

1368
 Electrical breakdown phenomena for devices with micron separations

(a) (b)

(c) (d )

Figure 2. Photographs of aluminum electrodes under a microscope. (a) Planar electrodes with 3 µm gap before test; (b) comb electrodes
with 12 µm gap before test; (c) breakdown voltage at 400 V with the electrodes being melted; (d) breakdown voltage at 404 V with the
electrodes being melted.

spark under atmospheric conditions. Therefore, breakdown

voltages for silicon electrodes were recorded by a current surge
triggered at milliamp level whereas typical glow discharge was
observed. In contrast, aluminum electrodes were melted and
vaporized by spurious surge current when the breakdown was
recorded as shown in figures 2(c) and (d). After all the samples
were tested, a visual examination was conducted piece by piece
under an optical microscope for abnormal breakdown possibly
due to fabrication defects.

Results and discussions

Figure 3. Schematic drawing of the experimental apparatus for As mentioned in the previous section, the breakdown threshold
breakdown measurement. SRS PS 350 is the programmable voltage current has been set to milliamp level and is too insensitive for
source and Keithley 485 is the digital multimeter attached to a accurate measurements in this study. Therefore, the multi-
personal computer. meter was carefully readjusted to the sensitivity for measuring
the initial discharge current at 10 pico ampere trigger. During
the electric current. A data-acquisition program coded in the automatic test runs, the current readings were very small
Labview language, copyrighted by National Instrument Co., initially, namely near-zero compared to the threshold value,
was employed for controlling the voltage source and reading whereas the source voltages were increasing linearly. For
the status on the Keithley 485. Programmed at automatic example, when the breakdown was initiated, the current surged
runs, the tripped-current on PS350 was set at 1.8 mA up to several milliamp at 406 V for the 2 µm case, and the
so that the Keithley 485 would not be over-loaded with silicon electrodes on the left-hand side (cathode electrode) was
spurious currents. The PS350 was initially set to 50 V, burnt into black color as shown in figures 1(c) and (d) which
and then gradually increased linearly in voltage until electric contradicts Ono’s experiment [14] showing micro-discharges
breakdown occurred and was recorded by the Keithley 485. appearing in the vicinity of the anode. The same results were
According to many observations in the experimentation, observed on aluminum electrodes when the breakdown was
electric breakdown occurred as a glow discharge or a dark initiated. It is rather interesting to observe that the aluminum

1369
C-H Chen et al

600

500

Breakdown Voltage, V, (Volt)

400

300

200
experimental data (n-type Si planar-electrode)
Dakin's data [6]
Dhariwal's data (metal-electrode)[10]
100

0
0 5 10 15 20 25 30 35
Gap Width, d, (micron)

Figure 4. Plots of measured breakdown voltages versus gap width in an atmospheric environment. The triangular dots are the experimental
data measured by Dhariwal et al [10] on macro metallic electrodes; the square dots are experimental data which were deduced to small gap
width at atmospheric pressure obtained by Dakin et al [6] under low gas pressure and larger gaps.

500
Breakdown Voltage, V, (Volt)

450

400

350

300
Planar electrodes
Comb-electrodes
250
0 5 10 15 20 25
Gap Width, d, (micron)

Figure 5. Plots for measured breakdown voltages versus gap width for n-type silicon electrodes in the planar shape and inter-digitated
comb. The two solid lines are the best-fitted polynomials based upon Paschen’s law with the triangular dots for the planar and square dots
for the inter-digitated comb.

electrodes were melted and vaporized by the electric energy on the various electrode configurations with doped-silicon SOI
as shown in figures 2(c) and (d). wafers. In the figures, the scatterings of data at a specific gap
After the experimental data were analyzed and fitted with width are due to the processing uncertainties inherent to the
Paschen’s law, the plots of breakdown voltages (V ) versus gap specific processes on each machine in the labs. Based upon
width (d) are shown in figure 4 for silicon electrodes. Based the plotted curves, it is noted that the cases of electrodes in the
upon the plotted data, it is postulated that the condition of inter-digitated comb have lower breakdown voltages compared
less available ionized gas molecules in the gap would lead to to the cases of electrodes in the planar shape. In addition,
higher breakdown voltages if the gap width is smaller than significant discrepancies also exist on the experimental data
5 µm. The statement is true for both the planar and the inter- between two wafer materials. Since MEMS actuators are
digitated comb configuration same as the predictions based mostly insulated by dielectric gases as well as gap width being
upon Paschen’s law. Figure 4 also shows the measured data smaller than 5 µm, it would be interesting to further investigate
obtained by Dakin et al [6] for larger gap width under low gas the cases of metallic electrodes with the two given electrode
pressure as well as the measured data obtained by Dhariwal configurations.
et al [10] for small gap width on macro metallic electrodes Due to the constraints on deposition metals in the lab,
under atmospheric conditions. It is evident that Dhariwal’s only aluminum was employed for the breakdown experiments.
data have shown the effects of field emissions and Dakin’s Again, the plots of breakdown voltages versus gap width for
deduced data agree to our data measured on silicon electrodes. aluminum electrodes fabricated with the two configurations
Figures 5 and 6 illustrate the plots of experimental data are shown in figure 7. From the plots, it could be concluded
together with the best-fitted Paschen curves for comparisons that field emissions gradually become important as the gap

1370
 Electrical breakdown phenomena for devices with micron separations

550

Breakdown Voltage, V, (Volt)

500

450

400

350
Planar p-type Si electrodes
Comb p-type Si electrodes
300
0 5 10 15 20 25
Gap Width, d, (micron)

Figure 6. Same as in figure 5 except the electrode materials being p-type, and the best-fitted polynomials are based upon Paschen’s law.

550
Breakdown Voltage, V, (Volt)

500

450

400

350
Planar Al electrodes
Comb Al electrodes
300

250
0 5 10 15 20 25
Gap Width, d, (micron)

Figure 7. Same as in figure 5 except the electrode materials being aluminum and the best-fitted polynomials are not based upon Paschen’s
law.

width is reduced to several microns such that electrostatic positive-charged ions. As illustrated in the figures, 30–50 V
field is intensified accordingly. On the other hand, it is also difference in breakdown voltage has been clearly observed in
noted that a plateau region exists between 3 and 7 µm which the two electrode configurations. It would be important to
implies that two mechanisms, namely Townsend discharge verify the observations via conducting numerical simulations
and field emissions, are competing with each other. At the with the experimental conditions being used as the boundary
edge of the plateau region, the breakdown voltage starts to conditions. After the maximal field intensity is calculated,
decrease gradually when the gap width is smaller than 3 µm figure 8(a) shows that the maximal field intensity is
while electrons generated by field emissions have commanded approximately 6 × 107 V m−1 when a voltage of 338 V
the breakdown phenomenon. is applied to planar electrodes at 5 µm gap width. The
If we compared the data plots between the p-type and corresponding maximal electrostatic field intensity value in
n-type wafers, the p-type wafers are 50 V higher than the the inter-digitated comb is calculated when 302 V is applied
n-type wafers. Based upon semiconductor physics, it is as shown in figure 8(b) so that the initiation of breakdown
comprehended that electrons move in the conduction band condition is established. The simulated results show that the
for n-type materials and holes move in the valence bond for maximal electric field intensity in inter-digitated comb is 12%
p-type materials. When voltage is applied, electrons acquire higher than the values in planar electrodes when same voltage
energy and activate to band free level. The process in p-type is applied. This result confirms the differences in breakdown
materials would take 1.12 eV as compared to electrons voltages between the two configurations being from 11% to
already in free level in n-type materials. Therefore, the 14% as shown in figures 5–7. Despite uncertainties in the
yields of secondary electrons for p-type materials are smaller experimental data, the simulated results agree well with the
than the yields for n-type materials. In order to replenish experimental curves. This concludes that lower breakdown
secondary electrons on the cathode surface, higher voltages voltages for inter-digitated combs are mainly due to electric
would be required for generating electron avalanches and field concentration effects at large curvature.

1371
C-H Chen et al

(a)

(b)

Figure 8. Contour plots of electric field intensity calculated via electrostatic field simulations showing the difference in breakdown voltage.
Case (a) with field intensity being 6×107 V m−1 at 338 V under 5 µm gap for planar silicon electrodes. Case (b) with field intensity of
6×107 V m−1 at 302 V under 5 µm gap for silicon comb electrodes.

Though the effects of surface roughness on breakdown similar to the results reported by Wallash et al [15] and Ono
voltage were initially assumed to be insignificant, the et al [14]. Later, Longwitz et al [20] discovered platinum and
experimental breakdown voltages for the two materials have gold cathodes are sputtered on the electrode. Slade et al [11]
indicated that the effects were actually significant. According concluded the phenomenon and stated that the electric field
to the published data, the RMS roughness values of ICP etched at a whisker on the cathode surface can be high enough to
silicon are from 20 to 30 nm whereas the peak-to-valley value produce field emission electron beam at the whisker for gap
is from 200 to 300 nm [18]. In addition, Dhariwal et al width less than 4 µm. When this current density exceeds
reported that their electrodes were polished with diamond a threshold, the whisker explodes and forms the released
paste to a state of average grain size at 250 nm [10]. This metal vapor to triggered electric breakdown. For the cases
could confirm the agreement on the data between the current of silicon electrodes, electric conductivity of doped silicon is
study and Dhariwal’s experiments when the gap width is larger smaller than metals because of low mobility of electron in
than 5 µm. On the other hand, the RMS roughness value on the silicon crystal. Therefore, the current density at whiskers
the aluminum sidewall in this study could be approximately on the cathode surface cannot exceed the critical power to
5 nm according to Guerry’s measurements [19]. Therefore, vaporize silicon. At most, even though the current density at
this could be the rationale for higher electric breakdown the whiskers is large so that vaporization of the whisker does
voltages being measured on the aluminum electrodes in this occur, it is still not the case of catastrophic breakdown for
study. metallic electrodes. This could be the rationale that breakdown
It should be noted that figure 2 shows that the aluminum voltages of doped-silicon electrodes still follow the Paschen’s
electrodes are melted after breakdown has occurred, which is curves.

1372
 Electrical breakdown phenomena for devices with micron separations

Conclusions References

In this paper, electric breakdown voltages of dielectric gas [1] Tabib-Azar M 1998 Microactuators (Dordrecht: Kluwer)
for both planar and inter-digitated comb electrodes in MEMS pp 37–85
[2] Madou M J 2002 Fundamentals of Microfabrication: The
devices fabricated with various substrate materials have been Science of Miniaturization (Boca Raton, FL: CRC Press)
measured in systematical procedures. Both metals and doped p 551
silicon have been employed for the fabrication of two electrode [3] Paschen F 1889 Wied. Ann. 37 69–96
configurations. The gap width between the electrodes has been [4] Xiong J, Zhou Z, Ye X, Wang X and Feng Y 2003
made from 2 to 21 µm with increments of 1 µm. Based Research on electric field and electric breakdown problems
of a micro-colloid thruster Sensors Actuators A
upon the experimental observations and conjectures, field 108 134–7
emissions at atmospheric pressure would become significant if [5] Nasser E 1971 Fundamentals of Gaseous Ionization and
metallic electrodes are adopted. In addition, the experimental Plasma Electronics (New York: Wiley-Interscience)
results for the doped-silicon electrodes in planar configuration pp 219–48
agree well with the data based upon Paschen’s law in the [6] Dakin T W, Luxa G, Oppermann G, Vigreux J, Wind G and
Winkelnkemper H 1974 Breakdown of gases in uniform
literatures. During the breakdown process, glow discharge fields; paschen curves for nitrogen, air and sulfur
occurs on the cathode and later burns the electrode a black hexfluoride Electra 32 61–82
color as reported by other papers. Also, p-type silicon wafers [7] Boyle W S, Kisliuk P and Germer L H 1955 Electrical
exhibit higher breakdown voltages as compared to n-type breakdown in high vacuum J. Appl. Phys. 26 720
silicon wafers because p-type silicon has higher band-gap [8] Germer L H 1959 Electrical breakdown between close
electrodes in air J. Appl. Phys. 30 46
energy for free electrons. For the cases of inter-digitated [9] Schaffert R M 1971 Electrophotography (London: Focal
comb electrodes, the breakdown voltages are lower because Press) pp 516–17
of field concentration effects occurring at large curvature. The [10] Dhariwal R S, Torres J M and Desmulliez M P Y 2000 Electric
field concentration effects have been verified by numerical field breakdown at micro-meter separations in air and
simulations via calculating the maximal field intensity in the nitrogen at atmospheric pressure IEE Proc. Sci. Meas.
Technol. 147 261–5
gap domain. In addition, even for heavily doped n-type silicon, [11] Slade P G and Taylor E D 2002 Electrical breakdown in
the concentration of free electrons is still far lower than those atmospheric air between closely spaced electrical contacts
of metals. Therefore, the breakdown mechanism should be IEEE Proc. Electron. Control p 245
different from metals. [12] Lee R T, Chung H H and Chiou Y C 2001 Arc erosion
It is interesting to note that the experimental results based behavior of silver electric contacts in a single arc
discharge across a static gap IEE Proc. Sci. Meas. Technol.
upon the cases of aluminum electrodes deviate significantly 148 8
from Paschen’s curves when the gap width approaches 5 µm. [13] Radmilovic-Radjenovic M, Lee J K, Iza F and Park G Y 2005
As the gap becomes smaller, the breakdown voltage starts Particle-in-cell simulation of gas breakdown in microgaps
dropping instead of increasing as compared to the cases of J. Phys. D: Appl. Phys. 38 950–4
doped-silicon electrodes. Hence, it is reasonable to set the [14] Ono T, Sim D Y and Esashi M 2000 Micro-discharge and
electric breakdown in a micro-gap J. Micromech. Microeng.
safe operation area for p-type and n-type silicon electrodes at 10 445–51
marginal values below 330 and 300 V, respectively. In contrast, [15] Wallash A and Levit L 2003 Electrical breakdown and ESD
for the cases of aluminum electrodes, the gap width should be phenomena for devices with nanometer-to-micron gaps
larger than 5 µm to avoid field emissions in the gap. The safe Proc. SPIE 4980 87–96
operation area for aluminum electrodes is set below 330 V for [16] Wilson C G, Wendt A E and Gianchandani Y B 2003 High
voltage constraints for vacuum packaged microstructures
normal operation. IEEE J. Microelectromech. Syst. 12 445–51
[17] Raizer Y P 1991 Gas Discharge Physics (New York: Springer)
pp 327–8
Acknowledgments [18] Wang H C 2003 Micro Electromechanical Systems Technology
and Applications (Hsinchu, Taiwan: PIDC) pp 210–25
The authors would like to thank the Center for Nano- (in Chinese)
Science and Technology, University System of Taiwan, and [19] Guerry A, Gondran C F H and Miller K 2004 Sidewall
the Precision Instrument Development Center, both located roughness measurement: a comparison of in- and off-AFM
techniques ASMC ’04 (4–6 May 2004) pp 221–6
in Hsinchu. Also, this work was financially supported by [20] Longwitz R, van Lintel H and Renaud P 2003 Study of micro
National Science Council in Taiwan, ROC, under grant glow discharges as ion sources for ion mobility
NSC94-2216-E007-037. spectrometry J. Vac. Sci. Technol. 21 1570–3

1373