IEEE SENSORS JOURNAL, VOL. 13, NO.

11, NOVEMBER 2013 4487

A Capacitive Humidity Sensor Suitable

for CMOS Integration
Nooshin Saeidi, Member, IEEE, Jörg Strutwolf, Amandine Maréchal,
Andreas Demosthenous, Senior Member, IEEE, and Nick Donaldson

Abstract— This paper describes the design, fabrication, and climate control for domestic or industrial applications), process
performance of a thin film humidity sensor fabricated in standard control (e.g. for high quality chemical products), pharmaceu-
CMOS process, hence it may be combined with an integrated tical and biomedical applications (e.g. respiration monitoring
circuit. The sensor is based on a capacitance between interdigi-
tated electrodes in the top metal layer and water adsorption in systems). This wide-ranging usage has inspired much research
the polyimide layer. The design is optimized by analytical and and industrial efforts to develop humidity sensors based on
then finite element models which show that, within the constraint different sensing mechanisms. Major mechanisms to sense and
of the CMOS structure, the sensitivity can be no greater than measure relative humidity can be classified in three categories:
one third of the sensitivity of the polyimide alone. Experimental optical, mechanical and electrical (capacitive or resistive).
sensors were fabricated in-house before an improved design
was fabricated in a commercial foundry. The different behavior Among the commercial humidity sensors the majority are of
of these sensors, despite their similar designs, leads to an the capacitive type. This technique offers low power consump-
investigation into the effects of fabrication process on the sensor tion, high output signal amplitude and wide operating range.
linearity. Characterizing the polyimide film by contact angle, Also, this type of humidity sensors are less influenced by
AFM and FTIR revealed that the difference in linearity of the temperature and require less complicated readout electronics
response between the two sensors resulted from different etching
techniques employed to pattern the film. compared to the resistive type. However, the sensitivity is not
high with most dielectrics. Capacitive humidity sensors can be
Index Terms— Thin film sensors, interdigitated electrodes, realized with a moisture sensing film (dielectric layer) either
CMOS integration, humidity measurement, polyimide film,
plasma etching. sandwiched between two parallel plates or deposited on top
of interdigitated electrodes (IDEs). Dokmeci et al. [1] demon-
strated a capacitive humidity sensor which utilized a thin
I. I NTRODUCTION
(120 nm) polyimide layer as the moisture sensing film, placed

M OISTURE is a major concern for electronic devices

implanted in the body or operated in humid envi-
ronments. Moisture that has penetrated the device package
between two parallel electrodes. The sensor with a capaci-
tance of 275 pF resulted in sensitivity of 0.86 pF/%RH [1].
Kang et al. [2] reported a parallel plate capacitive sensor
eventually causes condensation on the active area of the with patterned polyimide layer (2 μm thick) to form multiple
device, corrosion of the structures, performance deterioration columns having diameters of a few microns which allow
and device failure. The most direct way to ensure that the moisture to diffuse into them circumferentially to decrease
package is functional and remains dry inside, is to measure the response time. The sensor with a capacitance of 13 pF
the internal humidity. exhibited a sensitivity of 30.0 fF/%RH [2]. Humidity sensors
Humidity measurement is essential for a wide range of utilizing thin polyimide sensing layers and interdigitated elec-
applications including environmental monitoring (e.g. indoor trode configurations have been developed by several groups
Manuscript received January 27, 2013; revised May 26, 2013; accepted [3]–[5]. The sensitivity range was between 5 and 22 fF/%RH
June 10, 2013. Date of publication June 19, 2013; date of current version for sensors with capacitances between 5 and 8 pF.
October 2, 2013. This work was supported by the U.K. Engineering and We report on the design of a humidity sensor with interdig-
Physical Science Research Council under Grant EP/F009593/1, the European
Commission’s FP7 project NEUWalk under Grant 258654, and the Tyndall itated electrodes and polyimide as the sensing material, made
National Institute. The associate editor coordinating the review of this paper in a standard CMOS process so that it may be incorporated on
and approving it for publication was Prof. Massood Zandi Atashbar. a custom integrated circuit (Fig. 1). Combining the sensor with
N. Saeidi and N. Donaldson are with the Department of Medical Physics
and Bioengineering, University College London, London WC1E 6BT, U.K. its read-out circuit improves performance by avoiding external
(e-mail: [Link]@[Link]; [Link]@[Link]). interconnections. Once integrated with electronic devices, the
J. Strutwolf is with the Institute of Organic Chemistry, Department of sensor allows for real time monitoring of moisture inside the
Chemistry, University of Tübingen, Tübingen D-72076, Germany (e-mail:
[Link]@[Link]). package while the device is operating.
A. Maréchal is with the Institute of Structural and Molecular Biol- The interdigitated configuration was particularly selected
ogy, University College London, London WC1E 6BT, U.K. (e-mail: in order to utilize the topmost metal layer in standard IC
[Link]@[Link]).
A. Demosthenous is with the Department of Electronic and Electrical fabrication as well as the polyimide layer which is available
Engineering, University College London, London WC1E 7JE, U.K. (e-mail: in many fabrication processes as stress relief or passivation
[Link]@[Link]). layer.
Color versions of one or more of the figures in this paper are available
online at [Link] The following sections of this paper discuss the design,
Digital Object Identifier 10.1109/JSEN.2013.2270105 fabrication and characterization of an interdigitated capacitive
1530-437X © 2013 IEEE
4488 IEEE SENSORS JOURNAL, VOL. 13, NO. 11, NOVEMBER 2013

Fig. 1. Integration of the sensor with CMOS devices in a single process.

The top metal layer in a CMOS process is used for fabrication of the sensor
electrodes which can overlie the active circuits. The moisture sensing film is
polyimide.

Fig. 2. Capacitance output signal vs. relative humidity for HS-cover (curve A)
humidity sensor. In terms of sensor fabrication, standard IC and HS-perforated (B) at a temperature of 37 °C. The rate of humidity change
fabrication processes generally use similar processing steps, was 1% RH per minute. Shown are the cycles for adsorption and desorption.
except the etching steps which are either wet or dry. In
this work, particular attention was paid to the effect of the
etching technique on the sensor performance. Performance sensors for adsorption and desorption, when relative humidity
characterization of a single chip which combines the sensor was changed between 15% to 95%, are shown in Fig. 2.
with a capacitance to frequency readout circuit is presented The sensors exhibited nonlinear response with a very low
elsewhere [6]. sensitivity below 50 %RH (areas marked by a and b in Fig. 2).
Above 65 %RH the sensitivities of both sensors were
significantly increased in a narrow range (regions marked by
II. FABRICATION
a’ and b’ in Fig. 2). The sensitivities of the sensors are 15
The first two designs of sensor were fabricated in an in- fF/%RH (region a, HS-cover) and 7.3 fF/%RH (region b,
house process on top of a 4" p-type silicon wafer used as HS-performed). In the regions a’ and b’ the sensitivities are
the substrate. A 1 μm thick layer of oxide was deposited by 0.22 pF/%RH (HS-cover) and 0.74 pF/%RH (HS-perforated).
PECVD (Plasma Enhanced Chemical Vapour Deposition) on The HS-perforated sensor shows negligible hysteresis (Fig. 2,
top of the substrate followed by deposition of 0.5 μm-thick curve B) which might be attributed to the larger contact
layer of aluminium (1% silicon). The metal was patterned to surface area compared to HS-cover. The latter exhibits
form the sensor electrodes. The electrode width and spacing pronounced hysteresis (Fig. 2, curve A).
were 3 μm and 8 μm respectively. Polyimide (HD Microsys- The openings in the dielectric film increase the air-film
tems PI2545) was deposited over the electrodes to act as the contact area and should therefore make moisture absorp-
dielectric material for the humidity sensor. The thickness of tion/desorption quicker. However, due to the lower dielectric
the polyimide layer deposited on the metal structures was constant of air compared to polyimide it was expected that
2.65 μm. To examine the effect of surface area in contact the perforated sensor would result in a lower sensitivity; the
with the surrounding ambient on sensor response, polyimide opposite was observed as shown in Fig. 2. These observations
between the electrodes was partially etched using oxygen will be discussed in Section IV.
plasma. Device HS-cover is a sensor with blanket polyimide
film on top and device HS-perforated is a sensor with 4 μm
III. D ESIGN
wide openings in the polyimide between the electrodes. The
overall size of each sensor was about 2 mm × 2 mm and the Before we discuss the characterizations of the sensor fab-
number of electrode fingers was 182. ricated in the commercial process (XFAB), we describe the
results of our investigation on the effects of stray capacitance
as well as the electrodes geometry on the performance of
A. Characterization
sensor, when realized in a typical CMOS process.
The fabricated sensor was placed in a climate test chamber
(Heraeus Votsch HC 7033) which provided the controlled
humidity and temperature levels for sensor characterization. A. Parallel Plate Capacitor
A commercial precision humidity sensor (Honeywell The relative permittivity of polyimide changes from 3.35
HIH-4000-001) was also placed inside the chamber, close when dry (0% RH) to 4.35 when wet (100% RH) [7]. As the
to the sensor chip, as the reference sensor to measure sensor is based on this effect, its sensitivity is bound to be
relative humidity. The capacitance output of the sensors low e.g. S polyimide = (ε100 − ε0 )/ε0 = 0.3. Assuming that
was measured at 1 kHz and with a bias voltage of 1 V. the proper response is linear, the fractional change per % RH
Because of the intended biomedical application of the sensors would be a mere 0.003 or 0.3%. However, the actual sensor
all measurements were carried out at body temperature will be less sensitive than the polyimide. This can be shown
(37 °C). The capacitance of HS-cover and HS-perforated for the models described below.
SAEIDI et al.: CAPACITIVE HUMIDITY SENSOR SUITABLE FOR CMOS INTEGRATION 4489

(a) (b)
Fig. 3. (a) CMOS sensor structure: the metal electrodes are deposited on the field oxide, passivated with impermeable silicon oxy-nitride then coated with
polyimide. (b) Parallel plate capacitor with two passivation layers and polyimide.

The actual structure, shown in Fig. 3(a) creates an electric

field between the electrodes in the inorganic passivation and
in the polyimide, but also in the underlying layers and in the
gas above. Given that the vertical thicknesses are defined by
the particular CMOS process, the problem for the designer
is to find the optimal values for the electrode widths (W E )
and the gap widths (WG ). We tackled this problem in three
stages.
Stage 1 is based on standard electrostatic equations [8]. We
consider a parallel plate capacitor (Fig. 3(b)) with constant
thickness layers of dielectric (ε2 ), the silicon oxynitride,
adjacent to each plate, and a variable thickness of the water- Fig. 4. Variation of sensitivity with d1 for the parallel plate capacitor in
Fig. 3(b) with additional stray capacitance as a parameter. ε2 = 3.9 (silicon
absorbent dielectric (ε1 ) in the middle. The electric flux dioxide), ε1,0 = 3.35 (dry polyimide), ε1,100 = 4.35 (wet polyimide) and
density D is constant in all the dielectric layers. If the potential d2 = 1 μm.
differences across the layers are V21 , V1 and V22 , then
V = V21 + V1 + V22 . (1)
Eq. (6) shows how the sensitivity varies with d1 for the
The electric flux density is the same for all layers and is structure in Fig. 3(b). However, the real structure, shown in
equal to the charge per unit area on the plates Fig 3(a) also has capacitance in parallel due to the electric
field in the silicon dioxide underlying the metal. This ‘stray’
ε2 V21 ε1 V1 ε2 V22 Q
D= = = = . (2) capacitance will reduce the sensitivity. If we assume that the
d21 d1 d22 A stray capacitance is a constant, C S , then eq. (4) becomes
Substitute from (2) into (1):
Aε0 Aε0
V = Dd1 Dd21 Dd22 Dd1 Ct ot al = C S +   = x Aε0 +  . (7)
ε1 + ε2 + ε2 = + (d21 + d22 )
D
ε1 ε1 d1
+ d2 d1
+ d2
Q d1 d2 (3) εr1 εr2 εr1 εr2
= A ε1 + ε1

where d2 = d21 +d22. Substituting in (3), the total capacitance So x is a convenient number (units: m−1 ). The sensitivity
is function changes to
Q A Aε0 ⎧ ⎫
C= =  =  .
⎨ x + ε d1 + dε2 ⎪
(4) 1
V d d d ⎪ ⎬
1
+ 2 ε1
1
ε2+ d2 εr1 εr2 S= 1,100 2
− 1. (8)
⎩ x + d1 d2 ⎪
⎪ 1
⎭
This is the same as two parallel plate capacitors in series. The ε1,0 + ε2
effect of moisture is to change εr1 from εr1,0 at 0%RH to
εr1,100 at 100%RH. So Consider first the x = 0 curve in Fig. 4 (i.e. no stray
Aε0 Aε0 capacitance). The sensitivity continues to rise against d1 and
C0 =   and C100 =  . (5) is asymptotic to 0.3, the sensitivity of the polyimide itself
d1 d2 d1 d2
εr1,0 + εr2 εr1,100 + εr2 (as mentioned above). However, increasing d1 is also decreas-
ing the total capacitance so that when there is a constant
Therefore, the sensitivity is
shunt stray capacitance, its contribution is becoming relatively
 d1 d2 
ε1,0 + ε2
C100 − C0 smaller. It is not surprising therefore, that as x increases from
S= = d1 d2
− 1. (6) zero, there are maxima in the sensitivity curves which move
ε1,100 + ε2
C0
to lower d1 as x increases.
4490 IEEE SENSORS JOURNAL, VOL. 13, NO. 11, NOVEMBER 2013

B. Analytical Model
For the CMOS structure of Fig. 3(a), the stray capacitance
will not be constant as the gap increases. In order to understand
how the sensitivity depends on W E and WG , in the next
stage, we calculated the sensitivity for an IDE structure, using
conformal mapping and with an allowance for the electrode
thickness.
Analytical expressions for the capacitance of IDEs are often
derived by using conformal mapping of the space region of
the IDE onto a parallel plate capacitor geometry for which the
capacitance can be readily calculated [9], [10]. We used the
model by Igreja and Dias [10] which allows one to take into
account multiple layers: the polyimide, the air layer overhead
and the substrate (silicon dioxide) below the electrodes. Given Fig. 5. Sensitivity versus gap width for the IDE structures, computed
the large number of electrodes, the different capacitances of using conformation mapping as described in the text. There is no passivation
the two outermost electrodes compared to the inner ones can layer between the electrodes and the polyimide. The elevated electrodes are
approximated by flat electrodes with an effective width W ’R (Eq. 11). The
be neglected. The capacitance of one interior electrode of curves are for electrode widths of 1.5, 2.0, 2.5, 3.0 and 3.5 μm (from top to
length L relative to the ground potential within the framework bottom). Maximal sensitivity is about 0.15.
of a three layer model is given by
CI = L(Cair + Cfilm + Csub ). (9)
where Cair , and Csub are the capacitances of quasi infinitely
thick air and substrate layers and Cfilm is the capacitance
of the polyimide film. For details of the calculation of the
capacitances see the paper by Igreja and Dias [10]. The
capacitance of the array of N electrodes is then given by
CI
C=N (10)
2
The expressions to calculate CI were implemented in For-
tran 90/95. The algorithms for evaluating special functions
were taken from the literature [11], [12]. The thickness of the
electrodes, t E , was taken into account by the approximation Fig. 6. Calculated capacitance for a sensor area of 2 mm × 2 mm. The
electrode width and spacing are 1.5 μm and 3.5 μm. The relative permitivities
[9]   of the sensing film and the substrate are 3.5 and 3.9 respectively.
 tE 4π W E
WR = W E + 1 + ln (11)
π tE
which replaces the real width of the electrode, W E , by an C. Finite Element Model

effective width, WR . Note, however, that in Eq (11) the two-
dimensional structure of the electrode (i.e. width and height) is In the third stage, having settled on W E = 1.5 μm, we
approximated by a 1-D structure (i.e. an effective width). This modeled the CMOS structure using COMSOL Multiphysics
approximation does not take into account the effect that for FEM software (ver. 3.5a). Using the Electrostatics Mode, the
a constant film thickness, the volume of the sensing film will value for electrostatic energy density, Wes , can be calculated
change, depending on the electrode height and width. Based by integration over the whole computational domain 
on this model the sensitivity calculated for 5 different IDE 
structures as shown in Fig 5. Wes = D · E d (12)

Theoretical studies reported in literature [10], [13] suggest
that the thickness of the sensing film impacts the capacitance where D and E are the electric flux density and the field
of the sensor. Fig. 6 shows calculated capacitance for a intensity, respectively. Using the relation Wes = 0.5CV , the
constant sensor area, electrode width and spacing while the capacitance is then given by
height of the sensing film varies. With increasing height of the
Wes
film the capacitance increases until a limit is reached where the C = 2N L (13)
capacitance remains constant. In Fig. 6 the capacitance shows V 2
saturation when the height is greater than 4 μm. In this case the where N denotes the number of electrodes of length L and
film is thick enough so that the electric field is confined within V is the potential difference applied between two adjacent
the dielectric layer. This thickness is greater than the actual electrodes. We assumed a high value for N (N > 80), so that
thickness of the polyimide in the CMOS process (2.8 μm) the slightly different contributions of the edge electrodes is
but the figure shows that the loss in capacitance is only negligible. Furthermore the length L (2 mm) of the electrode
about 3%. is much larger than the width of the electrode, so that a
SAEIDI et al.: CAPACITIVE HUMIDITY SENSOR SUITABLE FOR CMOS INTEGRATION 4491

Fig. 7. Sensitivity versus gap width, calculated using COMSOL for the
geometry shown inset (the unit cell includes two half widths of adjacent
electrodes): dimensions in μm. Electrode width is 1.5 μm. Substrate and Fig. 8. Capacitance values for the constant sensor area with different
passivation permittivity is 3.9. Polyimide has a permittivity of 3.35 (dry) and geometry factors (α). Symbols are the result from FEM simulations, line
4.35 (wet). Gas above polyimide has a permittivity of 1.0. Total area covered is the result from an analytical model [10]. The inset shows the change of
is 2 mm × 2 mm. Maximal sensitivity occurs with WG = 4 μm. the number of electrodes, N,and the metallization ratio, η, with the electrode
width, W E .

2-dimensional geometry could be applied. The sensor struc- This as well as other combinations of W E with WG show,
ture (W E , thickness of passivation layer, electrode thickness, that the electrode geometry 0.5WG ≤ W E ≤ WG leads to the
height of the polyimide film, sensor size) used in the FEM optimal compromise between metallization and the number of
simulations is according to the structure of the commercially electrodes on a given area, for maximal capacitance.
fabricated sensor. The variation of the sensitivity with the For the example plotted in Fig. 8, FEM calculations in
change of the gap is shown in Fig. 7. COMSOL produced the points marked which are in good
agreement with the conformal mapping calculations.
D. Analytical and Finite Element Models for Calculating
Geometry Dependence of the Capacitance E. Discussion of Design
The advantage of the FEM approach using COMSOL is The CMOS structure imposes a water-impermeable passiva-
all the structure can be represented, including, here, the tion layer between the electrodes and the absorbent polyimide
impermeable passivation layer; thus Fig. 7 should be the most which reduces sensitivity, and it is further reduced by constant
realistic calculation. stray capacitance which is in parallel to the sensor capac-
Igreja and Dias [10] show the dependence of the sensor itance. In the very approximate model of the parallel-plate
capacitance on the metallization ratio η = W E /(W E + WG ). capacitor (Fig. 3b), these two effects caused a maximum in
However, in their approach the number of the interdigitated the sensitivity for a particular geometry (Fig. 4). For the more
electrode fingers is constant while η is varied, so the size of accurate FEM model, this maximum remained for variation in
the sensor is not fixed, whereas in practice, the sensor area the electrode gap.
is constrained and the minimal gap and electrode width are Fig. 5, the results from conformal mapping, suggested that
limited by the fabrication process design rules. the maximal sensitivity is about 0.15. The optimal gap and
Therefore, the capacitance was calculated versus α, the electrode widths are not very critical but are close to those
product of the number of electrodes (N) and the metallization used in our commercially fabricated sensor: 3.5 and 1.5 μm
ratio. respectively (see Section G). There is some reduction in
capacitance due to the given polyimide thickness but it is
WE small, about 3% (Fig. 6).
α=N = Nη (14)
WE + WG Comparing the sensitivities of the polyimide itself (0.30),
As an example, if there is a constant gap width of the model without passivation (0.15, Fig. 5) and the COMSOL
WG = 5 μm and 1<W E <20 μm, then 334>N>83. The α vs. model (0.10, Fig 7), we see that most of the reduction is due to
capacitance plot is shown in Fig. 8. The reversal of the curve stray shunt capacitance and less due to the passivation. Given
at α =100 is explained as follows. Keeping WG constant and the IDE structure, with impermeable dielectric between and
increasing W E will increase the metallization η (c.f. Eq.(14)) below the electrodes, the reduction from 0.30 to 0.10 (30%)
and will decreasing the number of electrodes on the sensor, seems quite reasonable.
as shown in the inset of Fig. 8. Therefore α goes through a
maximum. The maximum α is reached when W E = WG = F. Design Limitations
5 μm. At this point, η is 0.5 and N = 200. However, a Usually, in a standard fabrication process the thickness of
slightly higher capacitance is reached, when α = 89, for the sensing film is fixed. Therefore, design optimization is
the parameter combination N = 267 and η = 0.33. This limited to improving the geometrical configuration of the elec-
combination is realized by WG = 5 μm and W E = 2.5 μm. trodes with regards to a fixed height of the sensing film. For
4492 IEEE SENSORS JOURNAL, VOL. 13, NO. 11, NOVEMBER 2013

The sensor exhibited capacitance with almost linear relation

to relative humidity but, compared to the in-house fabricated
sensor, with a lower rate of change (0.03 pF/%RH) and a
sensitivity S of 0.078, as shown in Fig. 9. The sensitivity
is slightly higher than the value predicted by the COMSOL
model (c.f. Fig. 7). This could be attributed to a more sensitive
polyimide employed in the XFAB process.
The different behavior of HS-cover/perforated and
HS-1.5-3.5 sensors despite using similar material (polyimide)
for the moisture sensitive layer in their fabrication, led us to
investigate the possible influence of the fabrication process
on polyimide film properties which will be discussed in the
Fig. 9. Capacitance output signal vs. relative humidity for HS-1.5-3.5 at a
temperature of 37 °C. Rate of humidity change was 0.5% RH per minute.
following sections.

IV. T HE E FFECTS OF FABRICATION P ROCESS

the XFAB process selected in the present work, the maximum In the fabrication of HS-cover/perforated sensors, oxygen
thickness of polyimide was 2.8 μm. The minimum width of plasma was employed to etch polyimide from unwanted areas.
metallization (i.e. minimum electrode width) was 1.2 μm. For HS-1.5-3.5, which was fabricated in a commercial process,
With regard to Fig. 5 a small electrode width is preferred in wet etching was employed. The main reason behind using
order to obtain a high sensitivity from the sensor. Therefore, plasma treatment in fabricating HS-cover/perforated sensors
the electrode width was set to 1.5 μm. The minimum spacing was the requirement for etching narrow openings (4 μm wide)
between metallization (i.e. minimum gap between electrodes) in polyimide film for HS-perforated sensor.
was 1 μm. However, taking into account the passivation layer Plasma treatment is a standard technique for dry etching in
on the side walls of the electrode the thickness of the poly- semiconductor processing and is widely used for defining very
imide film between the electrodes would be only 0.2 μm. small features where wet etching is not efficient. Improving
The FEM simulations (Fig. 7) shows a low sensitivity in this the adhesion properties of polyimide is a known effect which
situation. The maximum sensitivity is reached when increasing results from plasma treatment [14]. Change in the dielectric
the gap to about 3.5 to 4.5 μm. We set the gap between constant or chemical properties of polyimide film exposed to
electrodes to 3.5 μm. different plasmas has also been demonstrated by the others
The substrate on which the sensor is fabricated would also [15], [16].
contribute to the base capacitance of the sensor (Eq. (9)). In the present work atomic force microscopy (AFM) was
For the commercially fabricated sensor the contribution of the employed to investigate the change in morphology of the
substrate is not fully specified. Several layers exist in the core polyimide surface on the plasma etched and wet etched
module of this standard CMOS fabrication process. Among sensors. The results showed a significant difference in surface
these layers are an initial oxide layer, other metal layers than roughness between HS-perforated (plasma treated) and HS-
the one used for sensor electrodes and intermetal dielectric 1.3-3.5 (wet etched) sensors (Fig. 10). The average surface
layers. The information regarding the type (e. g. the dielectric roughness for HS-perforated sensor was 4.439 nm while for
constant) of these underlying layers are not specified. HS-1.3-3.5 sensor it was 0.372 nm.
A humidity sensor design, 2 mm × 2 mm in size, which The increased roughness of the polyimide surface may
utilizes the topmost metal layer of a CMOS process for explain the sharp rise in the sensitivity of HS-cover/perforated
interdigitated electrodes was transferred to the commercial sensors and their non-linear response. It is expected that
process for fabrication. The acronym HS-1.5-3.5 is used for moisture would condense into pores of the rough polyimide
this sensor. surface [17]. With increasing the humidity level a thin water
A design with partial etched tranches between the electrodes film would then be formed from the entrapped moisture.
was also fabricated in the commercial process. According to A water film will considerably increase the output capacitance.
the design rules of the process, the minimum width for poly- To eliminate the possible effects of the two different
imide etch is 15 μm. Since the perforation of the polyimide processes (in-house and commercial), further studies were
layer is between the electrodes, the gaps between the sensor performed on samples prepared in the same fabrication process
electrodes have to have a width of more than 15 μm. The result but using different etching methods. Two silicon wafers were
of model calculations show low capacitance for such a big gap coated with 3.3 μm thick PI2545 polyimide, one exposed to
between the electrode fingers (c.f. Fig. 5). Accordingly, the oxygen plasma for 180 seconds and the other was wet etched.
sensors made in the commercial process (20 μm gap) showed To ensure that the photoresist is fully removed during dry
poor performance in humidity tests (results not shown here). etching of the polyimide, a small amount of the underlying
polyimide is also etched. Thus the thickness of the dry etched
polyimide was slightly smaller than the wet etched sample.
G. Characterization AFM measurements on blanket layers of polyimide confirmed
The same setup which was used for the in-house fabricated that plasma treatment considerably increases surface roughness
sensors was employed to characterize the HS-1.5-3.5 sensor. of the film as shown in Fig. 11.
SAEIDI et al.: CAPACITIVE HUMIDITY SENSOR SUITABLE FOR CMOS INTEGRATION 4493

Fig. 10. AFM images of a) HS-perforated sensor, plasma etched (average

surface roughness: 4.439 nm) and b) of HS-1.3-3.5 sensor, wet etched (average
surface roughness: 0.372 nm).
Fig. 11. AFM images of the polyimide samples. a) The wet etched sample
Plasma exposure may also have an impact on the surface (average surface roughness: 1.981 nm). b) The dry etched sample (average
surface roughness: 4.046 nm).
physicochemical properties. Wet etched and plasma etched
samples were therefore investigated by Fourier transform than these of a typical polyimide PMDA-OPA. This difference
infrared spectroscopy in attenuated total reflectance mode spectrum also put in evidence a shift in the frequencies of
(ATR-FTIR) [12] to obtain information on the chemical groups the symmetric and asymmetric carbonyl stretches of the dry
exposed at their surface. Fig. 12 shows the spectra recorded for etched sample, down from 1780 to 1771 cm−1 and up from
the wet (dashed trace) and dry (dotted trace) etched polyimide 1696 to 1710 cm−1 , respectively. This frequency shift most
films. They are presented in the 1900-1000 cm−1 mid-IR spec- likely arises from a change in the carbonyl close environment
tral range as no IR absorption was observed for these samples and could result in a modification of the surface material
at higher frequencies. They were normalised on the intensity hydrophilicity.
of the 1777 cm−1 band for direct comparison. As expected, Both, an increase in surface roughness [18], [19] and a
the two IR spectra are very similar, showing that no dramatic change in the physicochemical properties of functional surface
alteration of the surface occurred on oxygen plasma exposure. groups will increase the wettability of the surface and this
Nevertheless, many bands of the dry etched spectrum are of was investigated by contact angles measurements. The contact
higher intensity than in the wet etched spectrum. This might angle of deionized (DI) water drops on the surface of wet
represent an increased amount of material per surface area etched and plasma etched polyimide films was measured as
unit, in agreement with the AFM results which showed a presented in Fig. 13. The results show a 14° decrease in contact
rougher surface of the dry etched polyimide sample. This is angle for the film which was exposed to oxygen plasma for
confirmed by the calculation of the ‘dry minus wet’ difference 180 seconds, indicating an increase in the wettability of the
spectrum (Fig. 12, solid trace) which revealed no IR band other surface.
4494 IEEE SENSORS JOURNAL, VOL. 13, NO. 11, NOVEMBER 2013

the sensor was simulated. The most realistic model based

on FEM, yields a sensitivity which is in best agreement
with the sensitivity obtained by the commercially fabricated
sensor. The capacitance of the sensor was optimized by model
calculations taking into account the geometrical configuration
(i.e. thickness of the sensing film, electrodes width and the
gap between them) while keeping the sensor area constant,
by introducing a new geometrical parameter α. An optimized
capacitance improves both the sensitivity and the readout of
the sensor.
The surface morphology and physicochemical properties of
the polyimide are crucial for the performance of the sensor. In
comparison to wet etching, plasma exposure leads to a rougher
surface with small gaps and pores. Furthermore, the surface
chemistry is influenced by the plasma exposure, resulting in an
increase of the number of hydrophilic groups exposed at the
Fig. 12. FTIR spectra of dry etched and wet etched polyimide samples. Dry surface. Both processes facilitate moisture condensation on the
etched (blue trace) and wet etched (red trace) sample surfaces were applied
onto a silicon ATR prism for the recording of their IR absorption spectrum. surface and contribute to a non-linear response of the sensor
The two spectra were normalised on the intensity of the 1777 cm−1 band. The capacitance with a linear change of the relative humidity.
black trace shows the ‘Dry minus Wet’ difference spectrum calculated from In contrast the surface treatment by wet etching resulted in
the normalised spectra; its intensity was multiplied by a factor of 4 between
1900 and 1600 cm−1 . a less rough and less hydrophilic surface. However, the sensor
response showed a linear behaviour with increasing relative
humidity. Therefore, plasma etching is not recommended for
surface treatment.

R EFERENCES

[1] M. Dokmeci and K. Najafi, “A high-sensitivity polyimide capacitive

relative humidity sensor for monitoring anodically bonded hermetic
micropackages,” J. Microelectromech. Syst., vol. 10, pp. 197–204,
Jun. 2001.
[2] U. S. Kang and K. D. Wise, “A high-speed capacitive humidity sensor
with on-chip thermal reset,” IEEE Trans. Electron Devices, vol. 47, no. 4,
pp. 702–710, Apr. 2000.
[3] L. Gu, Q. A. Huang, and M. Qin, “A novel capacitive-type humidity
sensor using CMOS fabrication technology,” Sens. Actuators B, Chem.,
vol. 99, pp. 491–498, May 2004.
[4] J. Laconte, V. Wilmart, D. Flandre, and J. P. Raskin, “High-sensitivity
capacitive humidity sensor using 3-layer patterned polyimide sensing
film,” in Proc. IEEE Sensors, vol. 1. Oct. 2003, pp. 372–377.
[5] C. Laville and C. Pellet, “Comparison of three humidity sensors for a
pulmonary function diagnosis microsystem,” IEEE Sensors J., vol. 2,
no. 2, pp. 96–101, Apr. 2002.
[6] D. Cirmirakis, A. Demosthenous, N. Saeidi, and N. Donaldson,
“Humidity-to-frequency sensor in CMOS technology with wireless
readout,” IEEE Sensors J., vol. 13, no. 3, pp. 900–908, Mar. 2013.
[7] G. Rabilloud, High-performance Polymers: Chemistry and Applications,
vol. 3, Paris, France: Editions Technip, 2000.
[8] J. M. Crowley, Fundamentals of Applied Electrostatics. New York, NY,
USA: Wiley, 1986.
Fig. 13. Contact angle measurements on the polyimide samples. a) A contact [9] S. S. Gevorgian, T. Martinsson, P. L. J. Linner, and E. L. Kollberg,
angle of 68.7° on the wet etched sample and b) 54.7° on the dry etched sample. “CAD models for multilayered substrate interdigital capacitors,” IEEE
Trans. Microw. Theory Tech., vol. 44, no. 6, pp. 896–904, Jun. 1996.
[10] R. Igreja and C. J. Dias, “Analytical evaluation of the interdigital
electrodes capacitance for a multilayered structure,” Sens. Actuators A,
V. C ONCLUSION Phys., vol. 112, pp. 291–301, May 2004.
We presented and discussed the design of a capacitive [11] M. Abramowitz and I. A. Stegun, Handbook of Mathematical Functions:
With Formulas, Graphs, and Mathematical Tables. New York, NY, USA:
humidity sensor based on interdigitated electrodes configu- Dover Publications, 1964.
ration and using a polyimide film as the moisture sensing [12] S. Zhang and J. M. Jin, Computation of Special Functions. New York,
layer. This design allows for fabricating the sensor in a stan- NY, USA: Wiley, 1996.
dard CMOS process. Therefore the integration of the sensor [13] P. Van Gerwen, W. Laureyn, W. Laureys, G. Huyberechts, M. O. De
Beeck, K. Baert, J. Suls, W. Sansen, P. Jacobs, L. Hermans, and
with electronic circuitries in a single chip becomes possible. R. Mertens, “Nanoscaled interdigitated electrode arrays for biochemical
Using models with increasing complexity, the sensitivity of sensors,” Sens. Actuators B, Chem., vol. 49, pp. 73–80, Jun. 1998.
SAEIDI et al.: CAPACITIVE HUMIDITY SENSOR SUITABLE FOR CMOS INTEGRATION 4495

[14] D. Bhusari, H. Hayden, R. Tanikella, S. A. B. Allen, and P. A. Kohl, Amandine Maréchal received the [Link]. and [Link].
“Plasma treatment and surface analysis of polyimide films for electro- degrees in chemistry from the École Normale
less copper buildup process,” J. Electrochem. Soc., vol. 152, no. 10, Supérieure, Paris, France, and the Ph.D. degree from
pp. F162–F170, 2005. Commissariat à l’Énergie Atomique, CEA, Saclay,
[15] Y. Nakamura, Y. Suzuki, and Y. Watanabe, “Effect of oxygen plasma France, studying the reaction mechanism of proteins
etching on adhesion between polyimide films and metal,” Thin Solid of medical interest. She was with Peter Rich’s Lab-
Films, vol. 291, pp. 367–369, Dec. 1996. oratory, University College London, London, U.K.,
[16] S. H. Kim, H. S. Moon, S. G. Woo, and J. Ahn, “Study on the properties to learn infrared spectroscopy and apply it to the
of interlayer low dielectric polyimide during Cl-based plasma etching study of enzyme reaction mechanisms. Her current
of aluminum,” Jpn. J. Appl. Phys., vol. 38, pp. 7122–7125, Dec. 1999. research interests include the different components
of the respiratory chain, and in particular, the under-
[17] B. Bhushan, “Adhesion and stiction: Mechanisms, measurement tech-
standing of the proton coupled electron transfer in the mitochondrial terminal
niques, and methods for reduction,” J. Vacuum Sci. Technol. B, Micro-
oxidase. She is involved in medical projects investigating the potential of
electron. Nanometer Struct., vol. 21, no. 6, pp. 2262–2296, 2003.
infrared for medical diagnostics.
[18] M. Miwa, A. Nakajima, A. Fujishima, K. Hashimoto, and T. Watanabe,
“Effects of the surface roughness on sliding angles of water droplets
on superhydrophobic surfaces,” Langmuir, vol. 16, pp. 5754–5760,
Jun. 2000.
[19] R. N. Wenzel, “Surface roughness and contact angle,” J. Phys. Colloid
Chem., vol. 53, no. 9, pp. 1466–1467, 1949.
Andreas Demosthenous (S’94–M’99–SM’05)
received the [Link]. degree in electrical and
electronic engineering from the University of
Leicester, Leicester, U.K., the [Link]. degree
in telecommunications technology from Aston
University, Birmingham, U.K., and the Ph.D.
degree in electronic and electrical engineering from
University College London (UCL), London, U.K.,
in 1992, 1994, and 1998, respectively. He was a
Post-Doctoral Research Fellow with the Department
of Electronic and Electrical Engineering, UCL,
Nooshin Saeidi (S’11–M’12) received the [Link]. from 1998 to 2000. He became an Academic Faculty Member in 2000,
degree in hardware engineering from the Iran Uni- and is currently a Professor leading the Analog and Biomedical Electronics
versity of Science and Technology, Tehran, Iran, Research Group. He has numerous collaborations for interdisciplinary
and the Ph.D. degree in electronic and electrical research. He has authored or co-authored over 170 articles in journals and
engineering from University College London (UCL), international conference proceedings. His current research interests include
London, U.K. From 2008 to 2011, she was a Visiting analog and mixed-signal integrated circuits for biomedical, sensor, and
Ph.D. Student with the Tyndall National Institute, signal-processing applications.
Cork, Ireland. She is currently a Research Asso- Dr. Demosthenous is an Associate Editor of the IEEE T RANSACTIONS ON
ciate with the Implanted Devices Group, Depart- C IRCUITS AND S YSTEMS I: R EGULAR PAPERS , the IEEE T RANSACTIONS
ment of Medical Physics and Bioengineering, UCL. ON B IOMEDICAL C IRCUITS AND S YSTEMS , and the IEEE C IRCUITS AND
Her current research interests include CMOS/MEMS S YSTEMS N EWSLETTER. He is on the International Advisory Board of
sensors, micropackaging for CMOS and MEMS devices, TSV based 3-D Physiological Measurement, Institute of Physics. He was an Associate Editor
integration technologies, and aligned wafer bonding. of the IEEE T RANSACTIONS ON C IRCUITS AND S YSTEMS II: E XPRESS
B RIEFS from 2006 to 2007. He is a member of the Analog Signal Processing
Technical Committee and the Biomedical Circuits and Systems Technical
Committee of the IEEE Circuits and Systems Society. He is a member of
the Technical Programme Committee of various IEEE conferences, including
ESSCIRC, BioCAS, and VLSI-SoC. He is a fellow of the Institution of
Engineering and Technology and a Chartered Engineer.

Jörg Strutwolf received the Diploma and Ph.D.

degrees in the Theoretical Chemistry Group, Uni-
versity of Bielefeld, Bielefeld, Germany. He has spe- Nick Donaldson received the M.A. degree in engi-
cialized on the investigation of interfacial transport neering and electrical sciences from Cambridge Uni-
processes by theoretical and experimental methods. versity, Cambridge, U.K. From 1977 to 1992, he was
His current research interests include electrochem- with the Medical Research Council, Neurological
istry at the nanoscale, dynamics and reactivity of Prostheses Unit, London, U.K., under the direction
soft interfaces, the combination of microfluidics and of Prof. G. S. Brindley. His main field of research
electrochemistry, nanostructuring of interfaces for was the technology and use of implanted device for
sensor application, and development of mathemat- the restoration of useful leg function to paraplegics.
ical models for transport phenomena simulation. Since 1992, he has been the Head of the Implanted
His places of research included University College London, London, U.K., Devices Group, University College London, Lon-
University of Warwick, Coventry, U.K., Universitat Rovira i Virgili, Tarragona, don. He has been a Principal Investigator for many
Spain, and Tyndall National Institute, Cork, Ireland. He is currently a Visiting projects related to implanted devices and functional electrical stimulation.
Scientist with the University of Tübingen, Tübingen, Germany, and has He is a Professor with UCL. His current research interests include the
an active collaboration with Aarhus University, Århus, Denmark, on the development of implanted devices that use natural nerve signals as inputs,
simulation of electrochemical systems and with Curtin University, Perth, especially for preventing incontinence, stimulators of nerve roots, the use of
Australia, on electroanalysis at arrays of micro and nanoscaled liquid/liquid electrical stimulation for recreational exercise of paralyzed legs, and methods
interfaces. to encourage functional neurological recovery after injury.

View publication stats