High-Sensitivity Capacitive Humidity Sensor Using 3-Layer

Patterned Polyimide Sensing Film

J. Laconte, V. Wilmart, D. Flandre, J.-P. Raskin
Research center in micro and nanoscopic electronics devices and materials (CeRMIN)
Universite catholique de Louvain
Microelectronics Laboratory, Microwave Laboratory
3, Place du Levant
B-1348 Louvain-La-Neuve (Belgium)
jlaconte@dice.ucl.ac.be; flandre@dice.ucl.ac.be; raskin@emic.ucl.ac.be
Abstract
This paper investigates novel isolation and patterning
schemes to increase the sensitivity of a capacitive humidity
sensor based on a polyimide sensitive layer. We obtained
an optimum sensitivity of 23% using aluminum
interdigitated electrodes with fingers of 1 p m width
separated by 1 pm, deposited on a first insulating
polyimide layer and covered by two more polyimide layers
whose upper one features a regular array, of holes to
increase the active surface area. A mathematical model
was developed to optimize the sensitivity regarding water
absorption factor, electrodes design and ratio between
active and parasitic capacitances. The process was
optimized to be fully compatible with usual CMOS-IC
processes in order tofinally be able to fabricate a humidity
smart sensor.
Keywords
Capacitive humidity sensors, polyimide, interdigitated
electrodes, CMOS-compatibility.
INTRODUCTION
In order to introduce the automatic control of humidity in
OUT daily life to make living and working environments
more comfortable, low cost mass-production humidity
sensors are required. Among several types of humidity
sensors already developed, the capacitive sensor is
preferred because of its linear sensitivity in a wide range of
relative humidity. The principle of this kind of sensor lies
in the increase of the apparent dielectric permittivity of a
sensitive layer due to moisture absorption [l].
Polyimide is a polymer particularly suitable as humidity
sensitive layer thanks to its full compatibility with
integrated circuit process, its great chemical stability, and
its high permeability to water. In addition, most of new
polyimides are also very easy to coat and pattern thanks to
their photosensibility properties. Good mechanical
properties such as high elasticity and low thermal dilatation
coefficient can also he mentioned [21.
In this study, we demonstrate how to modify the design and
fabrication parameters of the interdigitated electrode
fingers, the insulation layer between the electrodes and the
substrate, and the sensitive layer of a classical capacitive-
type humidity sensor in order to improve its humidity
sensitivity and its integration into a standard CMOS-IC
process.
Figure 1. Encapsulated humidity sensor on polyimide
using 2 polyimide sensitive layers.
DESIGN STRATEGY
Sensi t i ve layer
The major challenge of our design is to optimize the
sensitivity to humidity. Low sensitivity can be expected for
classical capacitive type sensors. The capacitance ratio
between wet and dry environment is equal to the dielectric
constants ratio (Eq. 1) in case of staked capacitors. We can
use this equation in first approximation.
Eq. 1
The dielectric constant of dry polyimide (&dr y) is equal to
3.2 [2]. Empirical Looyenga [I ] formula can be used to
calculate with =78 (Eq. 2):
0-7803-81 33-5/03/$17.00 02003 i EEE 372
1 I l 3
Eq. 2
where the y factor is the fractional volume of water
absorbed in the polyimide film. This factor depends on
polyimide type. To increase the capacitance ratio of Eq. 1,
we selected a polyimide with a high water absorption factor:
Pyralin PI2723 from HDMicrosystems ( y =4.2% at 100%
RH, Relative Humidity). With this polyimide, E varies from
3.2 (= at 0% RH) to 4 (=&,+e, at 100% RH). Sensitivity
to relative humidity can therefore be expressed by Eq. 3
where AC is the capacitance variation between dry (Cdr y)
and wet ( C , ) polyimide, E , , ~ and represent respectively
permittivity of dry and wet polyimide and ARH is the
interval of RH variation.
Ea. 3
A maximum sensitivity of 25% can be expected in this first
approximation of stacked capacitor. Practically, as it will
be seen further, this value is more critical in case of
interdigitated electrodes due to the parasitic capacitances.
These must be reduced to improve the sensitivity. On an
other hand, water absorption factor can be increased as
proposed by [4-5-61 using a plasma etching on the sensitive
layer surface. We studied an alternative solution, better
controlled and more reproducible. To increase the active
surface area in contact with the ambient humidity, we
imagined to coat a second polyimide layer patterned with
holes on top of the first one.
EIectrodes desi gn
[7] shows that a polyimide thickness h equal to (Wgap +
Wfingers) practically contains all electrical field lines
(Figure 2).. This result was confirmed by our finite elements
simulation (ATLAS tool). Figure 3 shows the simulated
capacitance variation vs. polyimide thickness reported to
the sum W,, +W,,,,$. Maximum capacitance is reached
when h 2 W,, +W,,,,s. This result is the same whether the
total width W,, +W,,,,s is equal to 1 or 3 pm.
Figure 2. Schematic of sensor and notations.
In addition, as calculated in [3], the thickness of polyimide
must be minimized to have faster sensor response.
I
01 (I
Nwpaprwfinperj
Figure 3. Atlas simulated capacitance variation vs.
polyimide thickness reported to (W- +W,,,,) far total
width W,,=Ww+W,,, equal t o 1 and 3 pm.
Polyimide thickness was therefore chosen equal to W,, +
W,,,, and W,, +W,,,,3 as small as possible according to
photolithography.
Rati o bet ween parasi ti c and act i ve capaci tances
The parasitic capacitances which appear between the
interdigitated electrodes outside the polyimide film affect
the output signal but are independent on humidity
(Figure 4). Decreasing these parasitic capacitances is
critical to improve the sensor sensitivity because of the
rather low variation of active capacitance with humidity.
A mathematical model was developed to calculate theratio
between parasitic and active capacitances in case of a
squared sensor with a constant total area, a given insulator
thickness, a polyimide thickness sufficient to contain all
field lines and a given Wj nRe,f l Rap ratio. From [7] we can
expect that the ratio W,adw,.,,,s has no impact on the active
capacitance if h >W,, +WfinR,,s, It was therefore fixed
equal to 1 to ease the design.
373
I I " i I
the electrical field lines passing through the polyimide layer
(of thickness fp,).
where hj i ngLrs =fingers height
Figure 4. Schematic of all capacitances measured
between fingers at 100 kHr with grounded substrate.
Working around 100 kHz with the substrate grounded (see
next part), the parasitic capacitance can be approximated by
2 capacitances in parallel (Eq. 4). At this frequency, the
substrate is in deep depletion and the interface between
depleted silicon and substrate is quasi grounded.
Cpnrasi t i c =C I +C 2
Eq. 4
CI (Eq. 5 ) is the capacitance related to the electrical
field lines passing through the insulator layer.
Eq. 5
where n =fingers number; L = fingers length and
sensor width, =permittivity of the insulator
layer; a= proportionality factor optimized by
simulation expressing the repartition of electrical
field lines between the layers and fiN =thickness of
insulator layer.
C, (Eq. 6) is the capacitance related to the electrical
field lines passing the silicon depletion region.
The graphic of Figure 5 depicts the results of this
mathematical model. It appears that the ratio between
parasitic and active capacitances increases as the finger
width decreases. 1 pm is the smallest width we can obtain
with our photolithography. W,, and Wfi.,,,s were therefore
chosen larger or equal to 1 pm.
Finge,n*ldth ,m*mlu,
Figure 5. Simulations results of estimated capacitance
vs. fingers width for a squared 500x500pm2 sensor,
W-/W,.w. =1; 4 pm polyimide as sensitive layer and
400 nm thick oxide as insulator.
It also appears from Eq. 7 that the active capacitance
increases with the number of fingers and the fingers length,
in other words, with the sensor size. But the size is limited
by the goal of miniaturization and also by fabrication
process restrictions. Too long thin lines may not adhere
enough on insulator layer.
Figure 6 gives information about the sensitivity expected
for this kind of interdigitated capacitive sensors. The two
Cz = [ a( L, t i m + ~ ~ ) - a ( ~ , t i n r ) ]
c>-. / . n
, . curves show capacitance value without humidity (eps =3.2)
and in presence of 100% humidity (eps =4). The decrease
of sensitivity with the increaseof the fingers width is clearly
demonstrated.
"72 + Ld2
Eq. 6
where xd is the depleted region thickness (equal to
X h )
The active capacitance (Eq. 7) can be calculated as the
capacitance between two parallel aluminum lines related to
374
Figure 6. Simulation results of total estimated
Capacitance at 0% RH (eps13.2) and 100% RH (eps=4)
vs. fingers width for a squared 500x500pm2 sensor,
Wfl,.,= 1; 4 pm polyimide as sensitive layer and
400 nm thick oxide as insulator.
Regarding the lower insulating layer, thick thermal Si02
( 4 0 0 ~ 1 ) is usually used to decrease the parasitic
capacitance to the substrate. We investigated an other
solution, replacing oxide by polyimide which has lower
permittivity and which can be deposited in higher thickness.
Simulations were made using a 3 pm thick polyimide as
insulator to compare the sensitivity we can expect in both
case.
Figure 7. Simulation of the capacitance variation
between wet (eps=4) and dry (epsz3.2) polyimide VI.
fingers width using 400 nm oxide (- -) and 3 pm
polyimide (-) as insulator for a squared 50Ox500pm2
sensor, WWp,,-.= l and 4 pm polyimide as sensitive
layer.
Figure 7 shows the AC difference when oxide or polyimide
is used as insulator. A higher range between wet and dry
capacitance can be observed using polyimide to isolate the
electrodes from the substrate. A higher sensitivity could be
therefore expected in this case.
Remove substrate by silicon etching in post process to
isolate the sensor on membrane was also investigated but
not conclusive to date [8].
EXPERIMENTAL
The chosen polyimide (PI2723) is a negative photosensitive
polymer which permits thick layers from 2 to 6 pm
depending on the spin coating speed. Experiments were
made to optimize exposure, development and bake time in
order to improve resolution [Z].
Figure 8 describes the fabrication of the sensor type using 3
layers of polyimide: the lower to isolate electrodes from
substrate, the second as passivation layer and the upper with
hole pattern (Figure 8).
Firstly, a 3 pm thick polyimide was spin coated on substrate
(Figure S(a)) and polymerized by baking at 450C to obtain
the optimum electrical properties [2]. 1 pm thick aluminum
layer was then sputtered at 150C and patterned by plasma
(Figure S(b)). Thanks to a good selectivity between the
CCI&l2 plasma and polyimide, no special care bas to be
taken to etch aluminum. But the photoresist release was
critical due to the bad selectivity between photoresist and
polyimide when using usual chemistry. Tests showed that
boiling acetone removes
polyinude.
D SILICON
(4
(b)
photoresist without damaging
Figure 0. Schematic process flow using polyimide as
insulator.
Aluminum was also deposited on backside (Figure S(c)) to
bias the substrate for minimizing the depletion capacitance.
The next step was the deposition of a first 2pm thick
sensitive polyimide layer, its definition of sensitive area and
baking at 350C for polymerization (Figure 8(d)). A very
375
good adhesion was observed between the two first layers.
Finally, the last 2 pm thick polyimide layer was coated and
patterned by photolithography with holes to maximize the
exchange surface area (Figure 8(e) and 9). Aspect ratio
equal to 1 of pattemed holes in polyimide was observed.
Holes of 2 pm width separated by 30 pm were therefore
obtained after optimization of the couple exposure -
development time.
Figure 9. Encapsulated humidity sensor i n details.
A cross section in the middle of the interdigitated electrodes
array can be seen on Figure 10.
Figure 10. Cross section of 5 interdigitated electrode
fingers between 3 pm of insulating polyimide (on
silicon substrate) and 4 pm of sensitive polyimlde.
RESULTS AND DISCUSSIONS
As it can be seen in Figure 1 or 9, four metallic contacts
were used to apply AC current and measure the resulting
AC voltage independently on series resistances.
Measurements were performed at 100 kHz and 50 mV.
Frequency sweep measurements showed better
reproducibility and sensitivity at this chosen frequency
(Figure 11).
- f" ,,dth-l.lpm
0 Em I m I m S m * w r w a r u r I m 4 n r a
FWW imzi
Figure 11. Measurements of AC variation vs. frequency.
Calibration structures (short and open circuits) have been
used to extract all external access and parasitic capacitances
from the measurements. Measurements were performed at
constant temperature (25C) in the climatic chamber to
avoid any temperature influence. Humidity variation at
ambient temperature represents a more expected case in
current life.
8.1 -
5.0 .
5.8
.
w m 80 71 80 00 lm
RH 1911
Figure 12. Capacitance variation vs. relative humidity
for 500x500 pm2 sensor, W,,, =1 pm, 3 pm thick
polyimide as insulator, 2 polyimide sensitive layers.
Figure 12 shows the linear capacitance variation with RH.
A comparison between measured values and simulations
can be seen in Figure 13. The graphic conf m the good
agreement between our analytical model and results
obtained with finite elements tools. The small shift between
measurements and simulations values is due to an
overestimation of parasitic capacitances in the theoretical
model.
376
I
0 6 I S 2.5
Fins- *Id* lmuDral
Figure 13. Comparison between analytical, finite
elements simulations and measurements values for
500x500 pm2 sensor, 3 pm thick polyimide layer as
insulator, 2 polyimide sensitlve layers, at ambient
humidity.
In Figure 14, a comparison is made between the 3 studied
configurations and shows the important sensitivity gain
using polyimide as insulator with the substrate until a
maximum of 23% with fingers 1 pm width.
2s I
& + 0 5 1 5 2 5 35
Fl"g.. xldm 1mism-1
Figure 14. Sensitivity to humidity vs. fingers width for
500x500 pmz sensor; oxide-400 nm, Poly1 =3 pm,
Pol y2~2 pm and Poly3=2 pm.
CONCLUSIONS
We demonstrated that a novel capacitive-type humidity
sensor, using polyimide as insulator between our
interdigitaded electrodes and the substrate and a second
patterned sensitive layer, increases the sensitivity to RH by
at least 7% in comparison with classical design. The
fabrication of this sensor is easy, low cost and fully
compatible with usual CMOS-IC processes on the contrary
to most designs based on planar capacitor while providing
state-of-the art sensitivity values [9-10]. We expect a
further increase of sensitivity when decreasing more the
parasitic capacitance and increasing water absorption
factor. Improvements are still in progress as well as a joint
integration of such sensor with its close CMOS electronics
on SO1 (Silicon on Insulator) substrate in order to finally
obtain a humidity smart sensor.
ACKNOWLEDGMENTS
We would like to thank the PCIM Laboratory, Universitk
catholique de Louvain, Belgium, for providing the climatic
chamber and in particular M. Sinaeve for its maintenance.
REFERENCES
[ I ] H. Shibata et al., A digital Hygrometer Using a
Polyimide Film Relative Humidity Sensor, IEEE
Transactions on Instrumentation and Measurement,
Vol. 45, No. 2, April 1996.
[ 2] HDMicrosystems, Pyralin PI2720 Processing
Guidelines.
[3] D. Denton et al., Moisture diffusion in polyimidefilms
in integrated circuits, J ournal of electronic materials,
Vol. 14, No. 2,1985.
C. Laville et al., Humidity sensors f or a pulmonary
function diagnostic microsystem, Sensors and
Actuators B, Vol. 76, pp. 304-309.2001.
[5] C. Laville et al., Comparison of Three Humidity
Sensors For a Pulmonary Function Diagnostic
Microsystem, IEEE Sensors J ournal, Vol. 2, No. 2,
April 2002.
[6] Y.L. Yang et al., Improvement of Polyimide
Capacitive Humidity Sensor by Reactive Ion Etching
and Novel Electrode Design, in Proc. of the
Intemational IEEE Sensors Conference on sensors
(Orlando), pp. 51 1-514, J une 2002.
P. Van Gerwen, Nanoscaled interdigitated electrode
arrays for biochemical sensors, Sensors and
Actuators B 49, pp. 73-80, 1998.
J . Laconte, et al., SOI CMOS compatible low-power
microheater optimization and fabrication for smart
gas sensor implementations, in Proc. of the
Intemational EEE Sensors Conference on sensors
(Orlando), pp. 1395.1400. J une 2002.
T. Suzuki, Performance of polyimide-based humidity
sensors, microelectronic engineering research
conference, 2001.
[I O] G. Hay et al., Capacitive-type humidity sensors
manufactured using offset lithography, in Proc. of
16th European Conference on Solid-State Transducers
(Prague), pp. 1067-1070, September 2002.
[4]
[7]
[8]
191
377