Porous alumina based capacitive MEMS RH sensor

Abstract

The aim of a joint research and development project at the BME and HWU is to produce a cheap, reliable, low-power and CMOS-MEMS process compatible capacitive type relative humidity (RH) sensor that can be incorporated into a state-of-the-art, wireless sensor network. In this paper we discuss the preparation of our new capacitive structure based on post-CMOS MEMS processes and the methods which were used to characterize the thin film porous alumina sensing layer. The average sensitivity is approx. 15 pFIRH% which is more than a magnitude higher than the values found in the literature. The sensor is equipped with integrated resistive heating, which can be used for maintenance to reduce drift, or for keeping the sensing layer at elevated temperature, as an alternative method for temperature-dependence cancellation.

Full text

Page 1

9-11 April 2008
©EDA Publishing/DTIP 2008 ISBN: 978-2-35500-006-5
Porous Alumina Based Capacitive MEMS RH Sensor

László Juhász1, András Vass-Várnai1, Veronika Timár-Horváth1, Marc P.Y. Desmulliez2, Resh S. Dhariwal2
1Budapest University of Technology and Economics (BME), Dept. of Electron Devices, Budapest, Hungary
2MicroSystems Engineering Centre (MISEC), Heriot-Watt University (HWU), Edinburgh, United Kingdom

The aim of a joint research and development project at the
BME and HWU is to produce a cheap, reliable, low-power and
CMOS-MEMS process compatible capacitive type relative
humidity (RH) sensor that can be incorporated into a state-of-
the-art, wireless sensor network. In this paper we discuss the
preparation of our new capacitive structure based on post-
CMOS MEMS processes and the methods which were used to
characterize the thin film porous alumina sensing layer. The
average sensitivity is approx. 15 pF/RH% which is more than a
magnitude higher than the values found in the literature. The
sensor is equipped with integrated resistive heating, which can
be used for maintenance to reduce drift, or for keeping the
sensing layer at elevated temperature, as an alternative method
for temperature-dependence cancellation.

I. INTRODUCTION
Nowadays, humidity is an important factor to consider in
certain industrial and agricultural processes, in the
preservation of objects of our art heritage and in other
numerous areas of our lives. There are numerous old and
new methods and devices to measure quantities that are
related to the humidity, namely the absolute humidity, the
relative humidity (RH) and the dew point [1].
Capacitive sensors are based on the changes of the
dielectric property of thick or thin films upon water vapor
uptake which depends on the relative humidity content of the
surrounding media. Due to the polar structure of the H2O
molecule, water exhibits a very high permittivity (κw=80) at
room temperature. The permittivity of dielectric films shows
a huge increase upon adsorption of water. In a porous
dielectric the air in the voids is replaced by adsorbed vapor
as the ambient humidity level increases therefore it is worth
using porous dielectrics instead of compact layers.
Capacitive sensors – as well as other absorption-based
humidity sensors – typically show a non-linear behavior as
function of RH. This behavior can be described by the
following phenomenological equation:

n
d
w
d
w
C
C






=
κ
κ
(1)
where κd and κw, are the permittivity coefficients of the
dielectric material at the dry and wet states, respectively, Cw
and Cw are the corresponding capacitances and n is a factor
related to (the morphology of) the dielectric film [1].
Because of the immense surface-to-volume ratio and the
abundant void fraction, very high sensitivities can be
obtained with porous ceramics. One of the ceramics to be
used is porous Al2O3 which has proven to be stable at
elevated temperatures and at high humidity levels. A well
established method for porous Al2O3 preparation is the
electrochemical oxidation of Al films under anodic bias. It
was demonstrated in [2], that it is possible to produce highly
sensitive, micro-sized AAO (anodic aluminum oxide)
sensors on silicon chips using thin film aluminum as starting
material.
The quality of the Al2O3 layers and consequently the sensor
behavior (sensitivity, humidity range, response time, stability
etc.) strongly depend on the thickness of the porous layer and
the density and size of the pores [2] [3]. The structure of
porous alumina prepared by the aforementioned method can
be influenced by the technological parameters such as
concentration and temperature of the electrolyte and the
current density (or voltage) in the anodizing cell [2] [4] [5] [6].
Theoretically, the best solution to cover the total humidity
range (0–100%) is to produce all size of pores: micro-,
meso- and macropores1. According to a recent study [5] a
highly reproducible, wide-range humidity sensor can be
achieved using nanodimensional pores of a narrow size
distribution.
The response time of ceramic humidity sensors is
generally limited by diffusion [7]. Moreover, ceramics are
highly sensitive to contaminants such as dust and smoke.
These sensors require maintenance which consists in having
the condensed vapor evaporated by heating up them from
time to time. The temporary heating could also be a solution
to reduce the drift due to the formation of chemisorbed OH-
groups [5] [8].
II. EXPERIMENTAL
Boron doped (p-type) 2-inch silicon wafer was used as
substrate. Wafers were cleaned in RBS detergent, refreshed
in dilute (1:20) HF, boiled in dionised water and, after
drying, oxidized in 100 l/hour dry O2 flow at 1100 °C for
40 minutes, resulting a 100 nm thick SiO2 layer serving as
insulation [2] [9] and as adhesive layer [2]. E-beam vacuum
vapor deposition was carried out to create a titanium
(100 nm) and aluminum (99.9% purity, 500 nm) thin film
layers structure. The titanium served as the lower electrode
of the vertical capacitive structure. Using the first mask for
photolithography (with AZ-type thin resist) the aluminum
was selectively anodized in aqueous solution of 10 wt%
sulfuric acid [5] opposed to the previous experiments [2]
where 23.1 wt% solution had been used. The anodic current

1 Micropores have widths smaller than 2 nm. Mesopores have widths between
2 and 50 nm. Macropores have widths larger than 50 nm. [10]

Page 2

9-11 April 2008
©EDA Publishing/DTIP 2008 ISBN: 978-2-35500-006-5
density was kept at a constant value (8 mA/cm2) [4]. After
photoresist removal the remaining aluminum was etched
away using a selective wet etchant (16 H3PO4 : 1 HNO3).
Titanium and copper were evaporated on the top of the wafer
and windows were opened in thick AZ-type photoresist to
enable the formation of the upper grid-electrode and contact
surfaces using gold electroplating. After stripping the
photoresist, wet copper etchant (10% Na2S2O8) was used. The
third photolithographic step protected the meanders of the
resistive heating and the surrounding surfaces of the contacts
during the titanium wet etching (10 NH4OH: 1 H2O2). The
ready-made wafer was cut to dices and the selected dices were
fixed onto TO-type headers with silicon-based glue. The gold
contact pads of each chip were connected to the leads with
thermocompression using 50 µm gold wire. The completed
capacitive sensors structure can be seen on Fig. 1.
During the selective anodization of the aluminum layer
two additional effects have to be taken into consideration.
One is the distortion of the planned structure due to under-
etching on the perimeter of the photoresist windows; the
other is the possible formation of small dots on seemingly
random sites due to inadequate resist thickness or quality.
The packaged sensors were tested in controlled humidity
environment to acquire RH–capacitance characteristics, but
additional methods were also applied to examine the
previously prepared porous alumina layers to explain the
results.
III. RESULTS AND DISCUSSION
To test the devices, two types of environment tests were
used: fixed-point humidity environment over five different
saturated salt solutions at 25 °C [11], and a humidity
chamber (ESPEC SH-241). Capacitance measurements were
performed using HAMEG LC meter at 16 kHz. Typical
sensing characteristics are shown in Fig. 2.
Both the characteristics measured over saturated salt
solutions and in the humidity chamber show the same
sensitivity of our structure and match the previous results
obtained with another test-structure [2], with the exception
of the sensitivity. The average sensitivity is three times
higher: approx. 15 pF/RH%. This is much better than the
typical 0.2-0.5 pF/RH% as found in [12].
The sensing characteristics are non-linear, as desired, both
a large hysteresis loop (Fig. 3.) and an enhancement in the
sensitivity over approx. 80% RH can be observed. The
characteristics are also temperature-dependent. In the next
sections these properties are investigated.
Au
Ti
Cu
Al O
SiO
23
2
Si,
1 mm
(layer thicknesses are not to scale)
line of the planned backside
etching

Fig. 1. Top view and cross-section of the capacitive RH sensor.
0
200
400
600
800
1000
1200
1400
1600
1800
02040 6080100
Relative humdity (%)
New 1.
New 2.
New 3.
Old 1.
Old 2.
Old 3.
Capacitance (pF)

Fig. 2. Typical sensing characteristics of the sensors
compared to the old test-structure.
A. Hysteresis loop and linearity
At relatively low RH levels very thin layers develop from
water molecules on the pore walls and heat of condensation
releases. This phenomenon can be explained by the BET
theory of adsorption on microporous materials [13]. The
BET theory assumes that the solid material has fixed number
of adsorption sites and the adsorbed layers can be several
molecules thick. It is also put forward, that the heat of
adsorption in all layers beyond the first is equal to the latent
heat of condensation:

00
1
c
1
)(p
p
v
c
cvppv
p
mm
−
+=
−
(2)
where p is pressure of the gas, p0 is the saturated vapor
pressure, v is the quantity of the adsorbed gas, vm is the
monolayer capacity of the surface (the quantity of one
adsorbed layer), c is a temperature-dependent value:

RT
EE
L
ec
−
=
1
(3)
where E1 is the adsorption heat of the first layer and EL is the
heat of condensation. A further assumption is that the vapor
condenses on the adsorbent as a liquid when its pressure
reaches the saturated vapor pressure, i.e., at p=p0 the number
of the adsorbed layers is infinite.
An equation similar to (2) might describe the adsorption
branch of the response curve shown in Fig. 3.
As the humidity level increases, the pores start to fill. This
filling can be explained by the capillary condensation theory.
In case of higher RH it has been shown by Kelvin that, at a
given pressure, vapor condenses into those pores, which
have radii smaller than the Kelvin radius, given by following
equation:

θ
γ
2
rRT
cos)/ln(
0
⋅−=
V
pp
(4)
where γ is the surface tension of the liquid, V is the
molecular volume and R and T have their usual meanings.
The negative sign implies for θ<90 °, i.e. p is less than p0.

Page 3

9-11 April 2008
boundaries of the alumina layer. This phenomenon might
affect the original sensing mechanism as well.
The resulting characteristics (type IV isotherms), showing
a hysteresis loop are typical in case of adsorption on
mesoporous adsorbents [13]. To obtain some information of
the alumina layer, we have made sorption measurements,
following the IUPAC (International Union of Pure and
Applied Chemistry) recommendations [10].
A Hydrosorb TM 100 computer controlled volumetric
instrument (Quantachrome, USA) chamber was used for the
purpose of sorption measurements. Unfortunately the 2"
wafers with the alumina coating have significantly lower
surface area than the required 1.5 m2. Our only choice was
modeling the porous structure by some alumina powder.
Prior to the measurements, the same powder was treated at
20 °C and 200 °C in low vacuum (2.7 Pa) for two days. Both
measurements were performed at the constant temperature of
20 °C. The resulting isotherms are shown in Fig. 5.
Both curves can be considered type IV isotherms
according to IUPAC classification. According to the results,
it is obvious that the heat treatment played an important role.
The powder could take up significantly more water after it
was outgased at 200 °C than at 20 °C. There is an inflexion
at approx. 15% RH (p/p0=0.15) in both isotherms, showing
the formation of a monomolecular layer of water molecules
on the porous structure. From this point up to 70% RH
(p/p0=0.70) the isotherms can be regarded linear. Over 70%,
due to the capillary condensation, the rate of adsorption
increases in the case of both isotherms. This behavior is
similar to the RH–capacitance characteristics of the sensor.
The fact that after the powder was outgased at 200 °C and
the hysteresis did not close in the entire RH region, implies
that, beside the physical adsorption, chemisorption can
occur. Their different activation energy might explain the
sensors temperature-dependence.
©EDA Publishing/DTIP 2008 ISBN: 978-2-35500-006-5
390
440
490
540
590
0204060 80
100
Relative humidity (%)
AdsorptionDesorption
Capacitance (pF)

Fig. 3. Adsorption and desorption branches measured on
one of the test-structures.
The capillary condensation might cause hysteresis in the
characteristics [13]. As p increases, wider and wider pores
will be filled, as p decreases, the pores get empty. There are
two different contact angles on the two branches of the loop,
while the pores are filling, there is an advancing contact
angle, θA, and while they are emptying, there is a “receding”
contact angle, θR. Since θA is greater than θR, according to
the equation (4), the value of p for a given value of r is less
during desorption than adsorption causing the hysteresis
(Fig. 3.). This gives the explanation of the upper part of the
sensing curve (desorption). Over 80% RH level physical
adsorption will be taken over by capillary condensation in
the whole porous structure.
Wetting measurements were conducted on the alumina
surface formed on the oxidized silicon wafer to measure the
different contact-angles. We have approximated the surface
as one side of an infinitely wide pore. The contact-angles
were measured by the goniometric method using an OCA 40
Automatic Contact Angle Meter. Small droplets of water (8
µl) were dropped onto the surface. The advancing contact
angles were approx. 70°, the receding 38°. The latter was
taken up with the continuous withdrawal of the droplet. The
difference between the advancing and receding contact
angles might be the main reason of the hysteresis.
B. Temperature-dependence of the sensor
The other major side effect is the temperature-dependence
of the vapor adsorption phenomena in the porous structure,
which has been proven by measurements. A constant RH
was maintained in the humidity chamber and the temperature
was varied in the range of 5–95 °C. Prior to the
measurements the sensors were heated up to 100 °C and kept
at this temperature for 10 minutes resulting in dry pores. The
results are shown in Fig. 4.
A possible explanation of this phenomenon was given by
R. K. Nahar [7]. He suggested that, besides the capillary
condensation of water vapor, there is lateral moisture
diffusion into the pore walls as well, causing a virtual
widening of pores especially at higher temperatures,
following the exponential diffusion rules. The T–C curve’s
hysteresis might be caused by the difference between the
diffusion mechanism of the water molecules penetrating into
or exiting from the structure of the pore walls or the grain
C. Integrated heating
Humidity sensors based on porous ceramics like alumina
require maintenance from time to time, which can be assured
by heating up the sensing structure. This solution reduces the
drift caused by the formation of chemisorbed OH- groups.
The integrated heating elements could also be used to keep
the sensing layer at elevated temperature [8], which could be
another solution for treating the temperature-dependence
besides the electronic signal processing.

400
450
500
550
600
650
700
0 20 4060 80 100
Temperature (°C)
Capacitance (pF)

Fig. 4. The measured temperature–capacitance characteristics
at constant 35% RH.

Page 4

9-11 April 2008
©EDA Publishing/DTIP 2008 ISBN: 978-2-35500-006-5
The resistive heating is divided to eight segments (see
Fig. 1.) to ease the measurements and improve the yield in
the development phase. Preliminary investigations show that
it is possible to elevate the temperature above 80 °C at the
center of the capacitive structure. Although the first results
are promising it is necessary to optimize the parameters.
To decrease the power consumption and to improve the
response time of the heating additional processing steps are
planned to reduce the heated mass of silicon. Using backside
anisotropic etching for bulk micromachining it would be
possible to form a MEMS membrane below the capacitance
and the surrounding heating elements.
D. Surface investigations: SEM, AFM
For the direct observation of the alumina layer, SEM and
AFM imaging were used. SEM images were taken by the aid
of a LEO 1540XB CrossBeam® system. In Fig. 6. the
mesopores can be seen clearly, the average mesopore size is
about 9 nm. By enhancing the contrast of the images,
populations of smaller pores were also found. We assume
that they might be micropores with the radius less than 1 nm.
AFM measurements were performed to assess the
roughness of the surface. The radius of the AFM tip was
20 nm and the measurements were made in contact mode.
The biggest vertical deflection of the cantilever was 10 nm.
According to the results of the AFM investigations, the
alumina surface can be considered rather smooth.
E. Inner structure: Small Angle X-Ray Spectroscopy
In order to obtain more precise information of the surface
and also the inner structure of the alumina layer, Glazing
Incidence Small Angle X-ray Scattering (GISAXS)
experiments were carried out. A long strip of the surface was
illuminated this way and scattering was generated by the
inhomogeneities (electron density fluctuations) in the
surface. We assumed that in case of anodized alumina the
only inhomogeneities might be the pores.
A strip of absorbing lead was placed in front of the
detector to eliminate both the direct beam and the specular
reflections from the surface.

180
Adsorption, 20 °C
Desorption, 20 °C
Adsorption, 200 °C
Desorption, 200 °C
Absorbed volume (cc/g, STP)
0.00.2 0.40.6 0.8 1.0
0
20
40
60
80
100
120
140
160
Relative pressure (p/p )
0

Fig. 5. Water-vapor adsorption isotherms on alumina powder.

Fig 6. SEM image of the porous layer on the test-structure.
The GISAXS spectra (with the background subtracted)
were measured at 3 different sectors having angles with
respect to the horizontal plane: 22.5°-37.5°, 67.5°-82.5° and
127.5-142.5°.
A Lorentzian curve was fitted to the results:

22
1
) 0 (
r
+
)(
q
I
qI
=
(5)
where r is the average radius of the “pores” and q is the
wave vector. This results r=8.6 Å, i.e., a width of about
17 Å=1.7 nm.
Due to the low signal-to-noise ratio of the measurements,
these results should be considered only semi-quantitative,
still it is proven that the average pore sizes are in the
microporous range, i.e. below 2 nm, and the system is fairly
isotropic. This average pore size proved our estimation about
the presence of micropores in the alumina layer, not clearly
visible by SEM.
V. CONCLUSIONS
The new AAO-based sensors are proven to be highly
sensitive for ambient RH changes. The average sensitivity
was approx. 15 pF/RH%, which is much higher than
0.2-0.5 pF/RH% found in the literature. The reason for the
sensitivity improvement might be the increased average pore
size we expect from the altered electrolyte concentration.
Forthcoming SEM surface and cross-section investigations
could prove this assumption.
The shape of characteristics is not linear, hysteresis and
temperature-dependence is present due to physical laws and
chemisorption of water molecules.
To overcome the parasitic temperature-dependence of the
AAO RH sensors electronic signal processing or the
continuous operation of the integrated heating could be used.
The processing of sensors is built on steps used in post-
CMOS MEMS processing and has an additional electro-
chemical oxidation step to prepare the porous thin film
alumina sensing layer. This way the integration of read-out
circuitry and a transceiver would be possible.
The integrated microheater for the conditioning of the
sensing layer is working, but its parameters should be
optimized. Its efficiency and response time could be
significantly improved by reducing the heated mass with
anisotropic backside etching, which would create a bulk
micromachined MEMS membrane.

Page 5

9-11 April 2008
[4] Iván Losonci, Csaba Pető, Kálmán Tihanyi, Galvanotechnikai
zsebkönyv (Pocket–book of galvanotechnology – in Hungarian),
Műszaki Kiadó, Budapest, 1982, p. 395.
[5] Oomman K. Varghese et al., “Highly ordered nanoporous alumina
films: Effect of pore size and uniformity on sensing performance,”
J. Mater. Res., Vol. 17, No. 5., 2002, pp. 1162-1171.
[6] O. Jessensky, F. Müller, U. Gösele, “Self-organized formation of
hexagonal pore arrays in anodic alumina,” Applied Physics Letters,
Vol. 72., Num. 10., 1998, pp. 1173-1175.
[7] R. K. Nahar, “Study of the performance degradation of thin film
aluminum oxide sensor at high humidity,” Sensors and Actuators,
B 63., 2000, pp. 49-54.
[8] Ching-Liang Dai, “A capacitive humidity sensor integrated with
micro heater and ring oscillator circuit fabricated by CMOS-MEMS
technique,” Sensors and Actuators, B 122(2), 2006, pp. 375-380.
[9] Giorgio Sberveglieri, Roberto Murri, Nicola Pinto, “Characterization
of porous Al2O3-SiO2/Si sensor for low and medium humidity
ranges,” Sensors and Actuators, B 23. 1995, pp. 177-180.
[10] J. Rouquerol, D. Avnir, “Recommendations for the Characterization
of Porous Soilds,” Pure & Appl. Chem., Vol. 66, No. 8. 1994,
pp. 1739-1758.
[11] T. Lu, C. Chen, “Uncertainty evaluation of humidity sensors
calibrated by saturated salt solutions,” Measurement, 2006,
doi:10.1016/j.measurement.2006.09.012
[12] Denes K. Roveti, “Choosing a Humidity Sensor, A Review of Three
Technologies,” Sensors Magazine, July 2001
[13] Stephen Brunauer, P. H. Emmett, Edward Teller, “Adsorption of
Gases in Multimolecular Layers,” J. Am. Chem. Soc., 60, 1938,
p. 309. doi:10.1021/ja01269a023
©EDA Publishing/DTIP 2008 ISBN: 978-2-35500-006-5
ACKNOWLEDGMENT
The authors wish to explain their gratitude to Dr. Krisztina
László for the BET sorption measurements and the GISAXS
opportunity. We are thankful to Dr. Éva Kiss for the wetting
measurements, Dr. Attila Tóth for help in microscopic
investigations and Brian Moffat for help in sensor processing
at HWU.
The present research is a part of the development of a
smart sensor and read-out circuitry for environment
monitoring by employing new materials and semiconductor
technology. Present research is partly funded by the Network
of Excellence PATENT DfMM in FP6 of the European
Union.
REFERENCES
[1] Z. M. Rittersma, “Recent achievements in miniaturized humidity
sensors – a review of transduction techniques,” Sensors and
Actuators, A 96, 2002, pp. 196-210.
Veronika Timár-Horváth, László Juhász, András Vass-Várnai,
Gergely Perlaky, “Usage of porous Al2O3 for RH sensing,” In K.
Chakrabarty, B. Courtois, E. Jung, V. Kempe, R. Laur (Eds.), Procs.
DTIP 2007, EDA Publishing, 2007, pp. 372-376.
Goutam Banerjee, Kamalendu Sengupta, “Pore size optimisation of
humidity sensor - a probabilistic approach,” Sensors and Actuators,
B vol. 86., 2002, pp. 34-41.
[2]
[3]