Three technologies for a smart miniaturized gas-sensor: SOI CMOS, micromachining, and CNTs - challenges and performance

Abstract

In this paper we propose a new type of solid-state gas sensor by combining three recent advances, namely silicon-on-insulator CMOS technology, through wafer etching and growth of gas-sensitive carbon nanotubes. We have developed novel tungsten-based CMOS micro-hotplates that offer ultra low power consumption (less than 10 mW at 250degC), on-chip CNT deposition at temperatures up to 700degC, and full integration of CMOS circuitry. Moreover, the tungsten micro-hotplates possess better stability than other CMOS materials such as polysilicon. The multi-walled CNT resistive gas sensors showed a good response to PPB levels of NO₂ in air but required additional heating to provide reasonable baseline recovery times. We believe that our approach is attractive for the mass production of low-cost, low-power gas sensors in silicon foundries.

Full text

Page 1

Three technologies for a smart miniaturized gas-sensor: SOI CMOS,
micromachining, and CNTs – challenges and performance
F. Udrea1*, S. Maeng2*, J.W. Gardner3, J. Park2, M.S. Haque1, S.Z. Ali1, Y. Choi2, P.K. Guha1, S.M.C.
Vieira1, H.Y. Kim2, S.H. Kim2, K.C. Kim2, S.E. Moon2, K.H. Park2, W.I. Milne1, S.Y. Oh2
1Department of Engineering, University of Cambridge, Cambridge, UK
2 IT Convergence & Components Laboratory, Electronics and Telecommunications Research Institute (ETRI), Daejeon, Korea
3School of Engineering, University of Warwick, Coventry, UK
*Corresponding authors: fu@eng.cam.ac.uk, tel 44 1223 748319 and slm221@etri.re.kr, tel: 82 42 860 1771

Abstract
In this paper we propose a new type of solid-state gas
sensor by combining three recent advances, namely
silicon-on-insulator CMOS technology, through wafer
etching and growth of gas-sensitive carbon nanotubes. We
have developed novel tungsten-based CMOS micro-
hotplates that offer ultra low power consumption (less than
10 mW at 250°C), on-chip CNT deposition at
temperatures up to 700°C, and full integration of CMOS
circuitry. Moreover, the tungsten micro-hotplates possess
better stability than other CMOS materials such as
polysilicon. The multi-walled CNT resistive gas sensors
showed a good response to PPB levels of NO2 in air but
required additional heating to provide reasonable baseline
recovery times. We believe that our approach is attractive
for the mass production of low-cost, low-power gas
sensors in silicon foundries.

Introduction
There is renewed scientific and commercial interest in
solid-state gas sensors and in the related field of electronic
noses because of recent advances in silicon
microtechnology [1-5]. However, the problems of batch-
to-batch reproducibility of gas-sensitive thick films and
high power consumption have negated their use as
accurate monitors of hazardous gases, whilst other sensor
types (e.g. pellistor and electrochemical) are either too
insensitive, power hungry or too expensive for the mass
markets (e.g. automotive, PDAs, mobile phones). In
wireless environment, DC power levels well below 100
mW (1 mW in pulse operation) are required at a suitably
low cost. Recently reports on Carbon Nanotubes (CNTs)
have created a lot of interest in gas sensing applications [6-
10]. Carbon Nanotubes have unique electrical, mechanical
and optical properties and, unlike bulk material, CNTs,
have a high surface to volume ratio that results in good
sensitivity even for small volumes. CNTs are sensitive to
different toxic and VOC gases and the important aspect of
the CNT sensor is that it could be operated at lower
temperatures for detecting very low concentrations of NO2
or other toxic gases as well.
We present here an ultra-low power smart gas-
sensor with innovative CMOS micro-hot plate design and
an integrated, fully compatible gas-sensitive CNT layer.
We have achieved this through the challenging integration
of three major technologies:
(i) High temperature thin SOI CMOS process using
tungsten metallization: The use of CMOS offers full
integration of drive, processing and read-out circuitry.
Thin SOI offers additional electro-thermal isolation and
the option of using the buried oxide as an etch stop during
Deep Reactive Ion Etching (DRIE). The use of tungsten
metallization allows junction temperatures up to 250o C
with no risk of electro-migration and negligible drift in
time. Tungsten is used here both as an interconnect in the
CMOS circuitry and as a layer for the micro-heater
embedded in the microsensor.
(ii) Membrane technology using DRIE: The use of
ultra-thin CMOS compatible membranes gives very low
thermal losses and fast response times. Special attention
was paid to back to front alignment, CMOS compatibility
(removal of charge during
suppressing over-etching effects and ensuring uniform
etch across the wafer and from batch to batch.
(iii) CNT growth: High quality MW and SW CNTs
were locally grown, self-aligned onto the pre-formed
sensing metal electrodes. For this the embedded tungsten
micro-heater was powered up to temperatures in excess of
700 oC. By optimizing the heater shape, it was possible to
optimize the heat flow within the heater region and hence
improve its temperature uniformity with variation of only
1%. Since the high temperature was confined to the heater
region during the CNT growth the electronic circuitry was
unaffected.

Micro-Hotplate Design
The device schematic cross-section and photographs of the
manufactured smart sensors are shown in Figures 1-3. Two
major designs have been employed. The first is based on
silicon/tungsten resistive heaters and the second is based
on a novel FET heater, the latter having the advantage of
the temperature MOS gate control. All the layers used for
the microheater, as well as the sensing electrodes, are
formed during the CMOS sequence, with no additional
post-processing steps required. A thermal sensor in the
form of an SOI thermo-diode or a silicon resistive
temperature detector (RTD) was integrated directly below
the heater to monitor accurately the temperature during
operation. The membrane was formed using a low
DRIE de-clamping),
8311-4244-0439-X/07/$25.00 © 2007 IEEE

Page 2

frequency DRIE technique that ensured good uniformity
and no under-notching effects. The membrane deflection
was measured across the CMOS SOI wafers and from
wafer to wafer both at room temperature and elevated
temperatures (when the micro-heater was powered up).
The deflection/diameter ratio was around 1% and the
deflection increase with temperature was only 1nm/oC (Fig.
4). These were indications of low residual and thermally-
induced stress. The power consumption was below 15 mW
at 250 oC for a 500 µm (diameter) membrane and below 8
mW for a 300 µm membrane.

Silicon/Metal
heater
CNTs
Sensing electrode
Handle silicon
Interdielectric
layers
Buried oxide
Passivation

Fig. 1 Cross-section of resistive heater based resistive gas sensor with
local growth of vertical CNTs above. CNTs can be Spaghetti like
structures shown here.

M3: interdigitated electrodes

M2: thermometer
M1: interconnects
CNTs
p- channel
MOSFET
heater
Handle silicon
Buried oxide
Interdielectric
layers
Trench oxide
Passivation
n
p+
p+
Fig. 2 Cross-section of FET heater based resistive gas sensor with
integrated tungsten heater structure. CNTs are formed by local growth,
above the third metal (which forms the detection electrodes)

CNT Growth

On-chip growth of carbon nanotubes [16] has been
achieved by using the microhotplates as the thermal source
for depositing CNTs locally on the interdigitated
electrodes (IDEs). The SEMs of the locally grown multi-
wall Carbon Nanotubes (MWCNTs) are shown in Figs. 5
and 6 with Raman spectroscopy depicted in Fig. 7.



Fig. 3 Fabricated FET heater based micro-hotplate with integrated CMOS
SOI electronics. Back-to-front alignment is within 1 µm which ensures
high reproducibility chip to chip and batch to batch.

Fig. 4 Interferometer image showing the deflection (not to scale) of a
fabricated microhotplate. The deflection is below 3 µm for a 300 µm
diameter at room temperature. The deflection grows with a very low rate
of ~ 1 nm/oC at high temperatures. This is an indication of both low
residual stress and low thermally induced stress in the layers.

The resistance of the sensing layer and the microhotplates
were stable even after 600 hours of operation showing
very good reliability of CNT based sensing devices. As
expected, the high thermally conductive layer of CNTs has
no impact on the power consumption, while bulk CNT
growth can increase the power consumption quite
significantly, e.g. by 20% (Fig. 8).


Fig.5 The SEM of a local growth on CMOS microhotplates using
tungsten heater. Multi-chips were grown at the same time, by powering
up the tungsten micro-heaters in parallel.
Alignment
Marks
(metal 3)
Heater
2.5µm
300µm
Locally grown
CNTs
Analog
Readout
Circuit
Drive
Circuitry
Membrane
FET Heater
+ Sensing
Electrodes
100µm
832

Page 3

Fig. 6 Photograph of a CNT based gas sensor with the dark area
showing the CNTs. The inset picture is a zoom of the SEM showing
‘spaghetti -like’ CNTs.

10001200140016001800
300
400
500
600
700
800
900

Intensity (a.u.)
Frequency cm
-1
D Peak
G Peak

Fig. 7 Raman spectroscopy showing CNT formation on CMOS
microhotplates.

0
5
10
15
20
25
30
35
40
45
0100200
Temperature (C)
300400 500600
Power (mW)
Plain Heater
With Local CNT growth
With Bulk Grown CNTs
Fig. 8 The power consumption before and after CNT growth. The bulk
growth has resulted in additional power losses due to high conductivity of
the CNT layer. In contrast, local growth of CNTs did not impact the
power consumption, as the sensing area is fully isolated from the rest of
the chip though the membrane.
Results

The sensor was found to offer a response of 8 % to 100
ppb of NO2 and 20% at 20 ppb before falling off at lower
concentrations (Fig. 9). It showed improved recovery time
of the zero gas line at elevated temperatures (few seconds).
The best response was seen at room temperature but a
higher temperature was required to refresh the baseline
resistance. At higher temperatures the sensitivity is lower
but no refreshing is required at an elevated temperature of
270oC (10 mA of heater current) for baseline recovery so
there is slight trade off between sensitivity and
reversibility. The thermal response time was in the order
of few ms (Fig. 10) allowing fast thermal modulation of
the micro-heater and pulse drive for further reduction of
the power consumption. The CNTs were initially found to
have poor stability, but after “conditioning”, using the
embedded micro-heater and drive beyond a certain
stability point, the resistance of the CNT layer became
stable if operated below 300 oC. The smart CNT micro-
sensor also showed responses to methanol and ethanol (not
shown here). The ultra-low power consumption of the
hotplates and the growth of CNTs on multi-chips at the
same time, in parallel, show great potential for high
volume manufacturability.
The precise mechanism by which CNTs respond
to oxidizing and reducing gases is not fully understood.
Nevertheless it has been reported that it probably involves
the sorption of oxygen molecules from the atmosphere
onto the surface of the CNTS [6,8,14]. Then, for example,
the gas reacts with the sorbed oxygen leading to the
abstraction (or injection) of electrons into the conduction
band of the carbon nanotubes.



Fig. 9. Response of CNTs to 20, 2, 0.5, 0.1 ppm of NO2. The response
time was in the order of minutes, and the embedded micro-heater was
used to speed up the recovery time (facilitate NO2 desorption)
SEM of
CNTS
Concentration (ppm)
150µm
833

Page 4

0
100
200
300
400
500
600
700
01020
Time (ms)
30 4050
Temperature (C)
Fig. 10 Transient response of the heaters. Due to the very low thermal
mass of the membrane, the thermal rise and fall times were very fast (few
ms). This allows pulse drive for further reducing the power consumption.

It is believed that defects within the carbon nanotubes play
an important role in their response to gaseous molecules
because these defects create reactive sites and can be
explained by theoretical investigations [11,12]. These
defects can induce dangling bonds in the tube edges, ends
and side-walls. It is also believed that the electrons are
localized on the individual tubes and that the junctions
between NTs can affect the mobility through the distance
between CNTs varying [13,14]. At room temperature NO2
molecule can dynamically co-exist with N2O4, NO3
molecules which do not influence the density of states and
have longer desorption time; therefore could be one of the
reasons for the longer time recovery in the gas sensing
process [15]. The baseline can be recovered quickly at
high temperatures around 250o C – 300o C offering good
stability of the CNT gas sensing devices at very low
powers.

Conclusions

In this paper we have demonstrated a CNT based smart
gas sensor. CMOS compatible micro-hotplates were
fabricated followed by a fully compatible local growth of
the CNT gas-sensing layer. A good sensitivity to NO2 gas
has been observed at room temperature and the embedded
micro-heater was used to greatly improve the recovery
time of the CNTs. The smart sensing device has ultra low
power consumption, excellent reproducibility and good
long-term stability, and we believe has strong potential as
a commercial sensor.

Acknowledgements

The work was partly supported by the Ministry of
Information and Communication, Republic of Korea,
under project no. 2005-s-605-02 and by the Rural
Development Administration, Republic of Korea, under
project no. 20070501034006. F. Udrea, W. I. Milne and
J.W. Gardner also acknowledge the award of the EPSRC
project: EP/F004931 and the continuous support of the UK
research council in this area.


References

(1) I. Simon, N. Barsan, M. Bauer, and U. Weimar, “Micromachined
metal oxide gas sensors: Opportunities to improve sensor
performance,” Sensors and Actuators B, vol. 73, pp. 1–26, (2001)
(2) J. Suehle, R.Cavicchi, M.Gaitan, S. Semancik, “Tin oxide gas
sensor fabricated using CMOS micro-hotplates and in-situ
processing,” IEEE Electron Device Letters, Vol 14 (3), pp.118-120,
(1995)
(3) F. Udrea, J.W. Gardner, D. Setiadi, J.A. Covington, T. Dogaru, C.C.
Lu, W.I. Milne, “Design and simulations of SOI CMOS micro-
hotplate gas sensors,” Sensors and Actuators B, Vol 78, pp. 180-
190, (2001)
(4) M. Graf, D. Barrettino, K. Kirstein, and A. Hierlemann, “CMOS
microhotplate sensor system for operating temperatures up to
500oC,” Sensors and Actuators B, Vol 117, pp. 346-352, (2006)
(5) M. Afridi, J. Suehle, M. Zaghloul, D. Berning, A. Hefner, R.
Cavacchi, S. Samancik, C. Montgomery, C. Taylor, “A monolithic
CMOS Microhotplate based gas sensor system,” IEEE Sensors
Journal Vol2 No 6, pp 644-655, December (2002)
(6) J. Kong, N. R. Franklin, C. Zhou, M. G. Chapline, S. Peng, K. Cho,
and H. Dai, “Nanotube molecular wires as chemical sensors”
Science 287, 622 , (2000).
(7) S. Peng, K. Cho, “Chemical control of nanotube electronics”,
Nanotechnology 11, 57, (2000).
(8) P.G. Collins, K. Bradley, M. Ishigami, A. Zettl, “Extreme oxygen
sensitivity of electronic properties of carbon nanotubes”, Science
287 (2000) 1801.
(9) M. Arab, F. Berger, F. Picaud, C. Ramseyer, J. Glory, M. Mayne-
L’Hermite, “Direct growth of the multi-walled carbon nanotubes as
a tool to detect ammonia at room temperature”, Chemical Physics
Letters, 433, 175 – 181, (2006).
(10) C. Cantalini, L. Valentini, I. Armento, L. Lozzi, J.M. Kenney,
S.Santucci, “Sensitivity to NO2 and cross-sensitivity analysis to
NH3, ethanol and humidity of carbon nanotubes thin film prepared
by PECVD”, Sensor Actuators B, 95, 195, (2003).
(11) J. Zhao, A. Buldum, J. Han, J.P. Lu, “Gas molecule adsorption in
carbon nanotubes and nanotube bundles”, Nanotechnology 13, 195
(2002).
(12) S. Santucci, S. Picozzi, F. Di Gregorio, L. Lozzi, C. Cantalini, L.
Valentini, J.M. Kenny and B. Delley, “NO2 and CO gas adsorption
on carbon nanotubes: experiment and theory”, Journal of Chemical
Physics, Vol. 119, No. 20, pp. 10904-10910, (2003).
(13) L. Valentini, V. Bavastrello, E. Stura, I. Armentano, C. Nicolini,
J.M. Kenny, “Sensors for inorganic vapor detection based on
carbon nanotubes and poly(o-anisidine) nanocomposite material”,
Chem. Phys. Lett. 383, 617, (2004).
(14) L. Valentini, C. Cantalini, L. Lozzi, S. Picozzi, I. Armentano, J.M.
Kenny, S. Santucci, “Effects of oxygen annealing on cross
sensitivity of carbon nanotubes thin films for gas sensing
applications ”, Sensor. Actuators B 100, 33 (2004) .
(15) S. Peng, K. Cho, P. Qi, H. Dai, “An initial study of CNT NO2 gas
sensor ”, Chemical Physics Letters 387, 271 – 276, (2004).
(16) M. S. Haque, S. Z. Ali, N. Rupensinghe, I. Haneef, P. Guha, F.
Udrea, W. I. Milne, J. Park, S. Maeng, “Novel on chip growth of
carbon nanotubes using high temperature microheaters”, NT-07, 8th
International Conference on the Science and application of Carbon
Nanotubes, 24-29 June, p 15, A05, (2007).

834