Microsystem Technologies 3 (1997) 183 —190 ( Springer-Verlag 1997

Silicon hotplates for metal oxide gas sensor elements

G. Sberveglieri, W. Hellmich, G. Müller

183

Abstract Low-power-consumption metal oxide gas sensors have been investigated in the past [Hoefer (1994), Perego
and gas sensor arrays can be produced by combining (1995), Dori (1996), Hellmich (1996)]. In this paper we will
micromachining and thin-film technologies. In the present focus on the results obtained within the ESPRIT Project
paper the state-of-the art in this field is reviewed. In the first ’’MMMGAS’’. These film technologies have been combined
part the problem of thermal losses from a heated metal oxide with silicon micromachining techniques to arrive at miniatur-
film is addressed and the necessity of miniaturisation of gas ized gas sensor elements. In addition accelerated life time tests
sensing devices is pointed out. The thermal properties of were performed to establish the thermal stability of such
realized silicon hotplates are compared and analysed with microstructures.
respect to thermal equivalent circuit models. In addition, the The gas sensor elements developed within this project are
results of accelerated thermal ageing tests are presented. These small consuming only some 100 mW of heating power. This
latter results demonstrate that micromachined heater elements figure is likely to be reduced by one order of magnitude in the
are likely to exhibit device lifetimes of the order of 30 years near future. A small power consumption not only makes
when operated at membrane temperatures around 400 °C. In battery operation feasible but also reduces the costs of the
the second part of the paper attention is drawn to novel sensor housing. For a heating power of 1 W a cylindrical
methods of gas detection which are enabled by employing sensor housing with 1 cm diameter and 1 cm height will have
stacks of different thin film materials on micromachined a surface temperature of around 70 °C when operated in
hotplates. In this context, recent results on temperature-and a room temperature environment. This requires temperature
field-effect modulated gas detection experiments are discussed. stable plastic or metal housings. For a reduced power
consumption of 100 mW, a housing temperature of only 30 °C
1 is reached. As a consequence the size of the housing can be
Introduction reduced and cheap plastic materials can be used. Since
In recent times the market demand for low cost gas sensors has housing is an important cost driver, the final price of the whole
been growing due to increased environmental standards. Such sensor will be reduced by reducing its power consumption.
sensors exhibit a high selectivity, a low power consumption The combination of thin film technology with silicon
and mechanical robustness. Silicon micromachining allows for micromachining offers novel possibilities of selectivity control.
small heated surface areas and therefore for a reduced power The thermal time constants of micromachined sensors are in
consumption. Selectivity can be achieved by building arrays the range of 1 to 10 msec (Table 1). As a consequence fast
consisting of several micromachined sensor elements [Endres temperature cycling experiments can be performed. Further-
et al. (1996), Hellmich et al. (1996)]. For arriving at battery- more the sensitivity of the gas sensitive layer can be
operated sensor elements several micromachining approaches dramatically influenced by an electrical field applied perpen-
dicularly to the direction of the current flow. This biasing field
can be generated between gas sensitive layer and heater,
without providing for additional electrodes!
Received: 16 May 1997/Accepted: 22 May 1997

G. Sberveglieri, W. Hellmich 2
Università degli Studi di Brescia, Dipartimento di Chimica e Fisica per Micromechanical solutions
l’ingegneria e per i Materiali, Gas Sensor Laboratory, Via Valotti 9, Commercial metal oxide gas sensors (see for instance Fig. 4)
I-25133 Brescia, Italy usually consist of a ceramic substrate heated to temperatures
ranging from 400 °C to 600 °C. The substrate is suspended
G. Müller
Daimler-Benz AG, Research & Development, F2 M/M
by wires in the housing. The thin wires provide for thermal
Postfach 80 04 65, D-81663 München, Germany isolation between the hot substrate and the cold environment
(Chap. 5). Depending on the ratio of the substrate and
Correspondence to: W. Hellmich housing sizes the power consumption of commercial gas
This work was supported in part under the ESPRIT contracts no. 6374
sensors is in the order of 1 Watt. Because of the high substrate
‘‘MMMGAS’’ and ‘‘SMOG’’. The authors should also like to acknow- temperatures the bonding of the wires to the substrate has to
ledge interest and support by Drs. B. Thomas and S. Paasche and be temperature stable. So it is more expensive than standard
Mr. G. Marca. bonding techniques.
 Table 1. Properties of micromachined hot plate structures as developed by different research groups. For comparison the thermal properties of
a miniaturized sintered pellet structure are given

Heater: Heating power Temperature Thermal time Type Reference

lateral extension (mW) (°C) constant
(lm) (ms)

900 130 500 10 membrane Fig. 1, Fig. 3;

200 67 500 0.6 spider Suehle et al. (1993)
200 \50 425 2 spider DiMeo Jr. et al. (1996)
100 \50 550 1 spider Semancik et al. (1995)
100 \10 300...600 ? spider Dücso8 et al. (1996)
ca. 100 20 350 4 spider Fung et al. (1995)
184 150 60 500 \200ms pellet Takada et al. (1996)

On the contrary, micromechanics is a way to reduce the obtain a good thermal isolation the silicon is etched off
heated surface area and thereby the power consumption of the underneath the heater. A thin membrane remains, carrying the
gas sensors. Table 1 presents a comparison between the power heater element. The membrane consists of Si N and SiO .
3 4 2
consumption and the thermal time constants as obtained by Both materials display a low thermal conductivity and provide
employing different micromecanical approaches. By an for thermal isolation (Sect. 3.2). On top of the heater element
appropriate design one can achieve, that the power consump- the gas sensitive layer is deposited employing the two-step
tion is proportional to the linear extent of the heated area (Sect. RGTO-process [Sberveglieri et al. (1990)]. This type of
3.5) and that the time constants are indirectly proportional to membrane structure has been used to obtain the results of this
this extent. As a consequence smaller heated areas lead to paper.
reduced levels of power consumption and faster response The advantage of the above-described membrane structure
times. In the last line of Table 1 a conventional hot wire sensor is its comparatively low power consumption even when using
with reduced size is presented for comparison. relatively large heated surface areas. This is because the sensor
Micromechanics furthermore integrates the transition can be designed in such a way that any cold parts neither of the
between the hot sensitive layer and the cold environment into frame nor of the housing are closer to the heated surface than
the sensor chip. The sensor chips therefore can be bonded and the wafer thickness.
assembled like normal silicon chips by standard equipment. The disadvantage lies in the processing from the backside of
There are mainly two kinds of silicon hotplates, the membrane- the wafer. This requires additional equipment for the
type and the spider-type ones. lithography on the backside of the wafer.

2.1 2.2
Membrane-Type Spider-Type
One example of a micromachined gas sensor element is shown Another example of a micromachined gas sensor is displayed
in Fig. 1 [Hellmich et al. 1995]. A silicon frame carries a thin in Fig. 2 [Suehle, J. et al. (1993)]. In this device spider-
membrane. The center of the membrane is heated to some like support beams carry a hotplate with the sensitive layer.
400 °C by a platinum heater which is also used as a temperature Under the hotplate the substrate is etched off to provide for
sensor, for convenience. The functions of heating and thermal isolation. In this case the etching is performed from
temperature sensing, however, can also be separated by the front side, so advantageously no photolithography from the
integrating another resistor for temperature measurements. To backside is required and all lithographic steps can be done
using standard semiconductor processing equipment. Only at
the end a KOH etching step is required, which cannot be
performed in a standard silicon foundary. Very probably this is

Fig. 1. Cross-section through a micromachined membrane-type gas Fig. 2. Micro hotplate gas sensor as developed by Suehle, J. et al.
sensor (1993)
the reason for the popularity of this latter approach [Suehle a length l can be described by:
et al. (1993), DiMeo Jr. et al. (1996), Semancik et al. (1995),
Fung et al. (1995)]. 2njl ·*T
Pcy\ (2)
Recently a slightly modified approach has been reported ln(ra/ri )
by Dücso8 et al. (1996). Instead of etching the silicon substrate
by a concentrated (some 30%) KOH solution, the silicon is This cylindrical model will be used for estimating the heat
made porous by HF in a first step and then the porous silicon is loss across the membrane of membrane-type heater elements.
etched by a diluted (1% ... 2%) KOH solution. This approximation necessarily will become better, for
The spider-type device exhibits a high stability against decreasing heater size and increasing distances between heater
mechanical shocks. The air can flow around the small hotplate and silicon frame. Because of corner effects it is better to use
and does not need to flow around the whole chip as in the case a cylindrical rather than a planar approximation in this case.
of the membrane-type solution. 185
One disadvantage of this concept is the limited distance 3.3
between the hot and the cold parts which is necessarily smaller Spherical geometry
than the wafer thickness. Another disadvantage is, that the gas This is the approximation that best describes the heat
sensitive layer must be deposited after the etching of the conductivity of the ambient air in most cases. The reason is
substrate, because this layer cannot withstand the etching that ‘‘normally’’ the equi-temperature surfaces in the far field
solution. The heater of the spider type has been made of poly-Si of any body in three-dimensional space are spheres. This
passivated by SiO . Consequently it can be assumed that the means that the extension of the body must be small compared
2
life-times are shorter compared to a Pt-heater devices (Sect. 4), to the extension of equi-temperature surfaces. The word
although life-time data were not reported yet. ‘‘normally’’ hides the other condition: The environment with
a isotropic heat conductivity has to extend far beyond the
3 equitemperature surface. The heat conductivity may even
Heating power consumption depend on temperature, which is the case for the heat
There are simple models by which we can estimate the different conductivity of air.
contributions to the power consumption of a gas sensor. These So both, the hot gas sensor and the cold ambient are
models are helpful for providing rough estimates in that they approximated as concentric spheres with radii ri and ra,
provide helpful information on the influence of various respectively. For the radius of the inner sphere ri a character-
geometrical parameters. More detailed results can be obtained istic length of the heated area is used. The difference between
from sophisticated methods like FEM-simulations. In Sect. 3.5 ra and ri should be comparable to the smallest distance between
and Sect. 3.6 those models are applied to both micromachined hot gas sensitive layer and cold surrounding elements like the
and commercial gas sensors. sensor housing. The outer radius ra on the other hand has only
a minor influence on the transported heat Psph in equation (3),
as long as ra is large compared to ri.
3.1 Equation (3) describes exactly the heat Psph transported
Planar Geometry between two concentric spheres with radii ri and ra and the
Consider two parallel planes with an area A separated by temperature difference *T. The medium between the two
a medium with a heat conductivity k and a distance d small spheres is assumed to have the thermal conductivity k.
compared to their lateral extension a (a2\A). Stationery heat
conduction within such a configuration can be treated in 4nj ·*T
Psph\ +4njri · *T (3)
analogy to Ohm’s law. In such a case a power flow Ppl will be (1/ri[1/ra)
necessary to sustain a temperature difference *T between the
two plates : If one wants to optimize the power consumption, the
most powerful screw is the size of the heated area represented
A a2 by the inner radius ri. For a low power consumption ri
Ppl\*T · j · \*T · j · (1) has to be as small as possible.
d d

The planar geometry model provides a good approximation 3.4

for the heat losses across the support beams of spider-like hot Radiation
plates (Fig. 2) or the bond wires of commercial gas sensors The heated area also looses heat by radiation. This is described
(Fig. 4). It also describes the heat loss across the membrane of by Stefan Botzmann’s law (4):
membrane-type heater elements, provided the distance
between heated area and silicon frame is small compared to the Phl\peAT4 (4)
heater size.
Phl is the radiation power, r\5.67 · 10[18Wm[2K[4 is the
radiation constant, A is the radiating area, T is the absolute
3.2 temperature and e is the emissivity. Phl is normally small
Cylindrical geometry compared to the heat flow through the ambient air. At
Heat conduction Pcy through a medium with a heat conductiv- temperatures of the order of 400°C the radiative power loss
ity k between two concentric cylinders with radii ra and ri and amounts to a few per cent of the total heating power only.
 3.5
Power consumption of MMMGas heater elements
The total power consumption P of a gas sensor can be split into
three parts: The heat flow transported away by the ambient air
Pair , the heat carried away via solid-state heat conduction
across the membrane or the support beams Psolid and finally
the flow of thermal radiation Phl from the heated surface.
Pair can be split again into two parts: One part is the diffusion
of heat through stationary air Pk and the other is convection
Pc, that refers to heat transported via air in macroscopic
motion. It is difficult to measure Pc, but FEM - simulations
186 showed, that Pc is expected to be in the order of 5% of Pk for
the microstructures discussed here.

P\Pair]Psolid]Phl (5)

The measured total power consumption P as a function of

the membrane temperature is plotted as a function of the
membrane temperature T in Fig. 3. At the preferred operation Fig. 3. Power consumption P and thermal response time q of
temperature of 400 °C the total power consumption amounts MMMGas sensor elements
to some 100 mW. The temperature of the sensor housing is
about 30 °C, i. e. 10 °C above the ambient temperature. The
total membrane area and its heated interior had lateral heating power amounting to some 100 mW. This is in very
extensions of 1700 and 900 lm, respectively. So the inner good agreement with the estimations on the basis of Eqs (2)
radius ri is of the order of 450 lm and the outer one ra about and (4) made above.
850 lm. Since the total power consumption of micromachined gas
For the calculation of the heat flow through the ambient air sensor elements can be attributed to more than 90% to the heat
Pair the spherical model (3) is used neglecting the effect of conductivity of the ambient air and only 5% to convection, it is
convection, i. e.: Pair+Pk. A difficulty is the temperature reasonable to compare the temperature dependence of the
dependence of the heat conductivity of the ambient air. measured power consumption and the heat conductivity. To
It amounts to 0.026 W/m/K and 0.051 W/m/K at room this end the ‘‘model’’ power Pm is defined by (6). It is
temperature and at 400°C, respectively [Enzyclopédie des gas proportional to the heat conductivity k(T), the excess
(1976)]. Since the temperature gradient is largest near the temperature (T[Tu) and a proportionality constant C.
heated area, it is more accurate to use the heat conductivity of Figure 3 reveals a very good agreement of Pm with the
400°C and to overestimate the power Pair slightly. For these measured power consumption P.
data (3) predicts, that 113 mW are carried away by the ambient
Pm\C · j(T) · (T[T ) (6)
air. This is slightly more than the measured total power !
consumption. In Fig. 3 also the thermal time constants of the MMMGas
The membrane consists of a Si3N4 and a SiO2 layer with sensor elements are given. The thermal time constants were
thicknesses of 100 nm and 700 nm, respectively. The inner and measured by switching the membrane temperature from
outer radii ri and ra are again 450 lm and 850 lm, respectively. Tbefore to Tafter. It turned out, that the thermal time constants
Because the outer radius ra is about twice as large as the inner depended only on the value Tafter . On an absolute scale the
radius ri, the cylindrical model (2) seems to be more thermal response times range from 20 ms to 10 ms at room
appropriate to calculate the heat Psolid carried away by the temperature and 500 °C, respectively.
membrane. Since Psolid will turn out to be small compared to These time constants are also compared to the heat
the total power consumption, it is not very important whether conductivity of air. Since a high heat conductivity shortens the
a planar or a cylindrical geometry is assumed. Formula (2) time constants, the time constants are expected to be
predicts a solid - state heat loss across the membrane of 2 mW, proportional to the inverse of the heat conductivity. So the
only. formula C@/j(T) has been fitted to the time constants. Figure 3
Equation (4) finally predicts a radiative heat flow of Phl shows, that indeed the temperature dependence of air accounts
of 1.8 mW at an emissivity e of 0.2. Experimentally the for most of the temperature dependence of the thermal time
power loss due to the ambient air Pair can be eliminated by constants.
measuring the power consumption of the gas sensor in
vacuum. Finally the heat flows over the membrane Psolid and the 3.6
radiation power Phl can be separated by their different Power consumption of commercial thick film sensors
temperature dependence. Psolid is proportional to the excess (3) and (4) can also be applied to commercial thick film gas
temperature *T and Phl is proportional to the fourth power of sensors such as those produced by Umweltsensortechnik
the absolute temperature T. Pfaller (1993) performed these GmbH. Figure 4 shows a cross-section through such a sensor
measurements and found out that both contributions amount element. The heated substrate has a square geometry with
to 3 mW and 2.5 mW, respectively at 400 °C with the total a linear dimension of 3 mm. The thickness of the substrate is
 187

Fig. 5. Thermal stability test of a micromachined gas sensor

Fig. 4. Cross section through a commercial gas sensor (UST)

0.8 mm. So the inner radius ri of (3) can be chosen to be

1.5 mm. The metallic housing is cylindrical and has an inner
diameter of 7.7 mm and a height of 5 mm. So the substrate is
seperated from the housing by some 2.3 to 2.4 mm. As
a consequence the outer radius ra is chosen to be 3.8 mm.
For a sensor temperature of 530 °C, at which the heat
conductivity of air amounts to k\0.058 W/m/K, and a hous-
ing temperature of 70 °C (3) the power consumption
due to heat conduction in air is predicted to be 0.83 W. The
power emitted by radiation Phl is 83 mW (4). The measured
heating power, on the other hand, amounts to some 1.05 W. If
one takes into account, that also the contacts of the housing
will remain cold but are much closer to the heated substrate Fig. 6. Life time and measurement time of micromachined gas
than the rest of the housing, it is clear, that we underestimate sensors
the power consumption of this commercial gas sensor by some
150 mW. Because of the higher operating temperature the
radiation becomes more important but the total power The results of an accelerated life-time test on micro-
consumption is still dominated by the heat conductivity of the machined gas sensors as developed within the MMMGas
surrounding air. project are shown in Fig. 5. The temperature is ramped in
25 °C steps. Shortly after a temperature step also the temper-
ature of the sensor housing is increased slightly by less than
4
one degree. The thermal time constants of the sensor housing
Thermal stability
amount to 88 s. This is the reason for the faster drop of the
Before working with micromachined sensors it is important to
heating power directly after the temperature jumps. To
establish the thermal stability of the heater elements. To this
suppress the effect of the rising housing temperature the
end, a number of membrane-type sensors have been slowly
relative drop of the heating power (dP/dt)/P shortly before the
destroyed by increasing the temperature in a step by step
following temperature jump is calculated and displayed in
manner. The temperature was inferred by measuring the heater
Fig. 6. These data show, that the degradation of the heaters is
resistivity. So both, the measured temperature and the
a temperature activated process with an activation energy of
resistivity are kept constant during each temperature step
1.95 eV:
whereby the heating power is being adjusted. A degradation of

A B
the heater increases its resistivity and the measured temper- dP/dt 1.95eV
ature will increase, although the actual temperature decreases. \9 · 104 [1/s] · exp (7)
P kB T
So the power control reduces the heating power to keep
resistivity and measured temperature constant. Consequently By defining a maximum allowable relative power change one
the decrease of the heating power is a measure for the is able to calculate a life-time of a micromachined sensor
degradation. structure. For normal sensor operation it seems reasonable to
 tolerate a relative heating power change of 1%. At 400 °C
sensor temperature a 1% heating power change leads roughly
to 1% or 4 °C temperature change. These results are labeled
‘‘measurement time’’ in Fig. 6. So measurement times of 6 and
28 hours are obtained at 600 °C and 550 °C, respectively. This is
sufficient for research purposes. At more realistic operating
temperatures of 400 °C measurement times of 104 hours or 1.5
years are to be expected.
If the resistivity change due to degradation is compensated,
a higher heating power change of — lets say — 20% can be
tolerated. The degradation can be compensated by measuring
188 the resistivity at ambient temperature from time to time. By
degradation this resistivity is increased, too. The data for
a 20% heating power change are labeled ‘‘lifetime’’ in Fig. 6. So
‘‘lifetimes’’ of 120 and 560 hours are obtained for 600 °C and
550 °C operation. This is enough for characterizing sensitive Fig. 7. Variation of heater power input and sensing layer conductance
layers up to some 600 °C. For realistic operation temperatures in a temperature cycling experiment
of around 400 °C the lifetime is some 30 years. During sensor
operation a warning may be displayed after a 10% heater
resistivity change, although the sensor will keep on working for
some time.
Extrapolating measurements taken within hours to years
may lead to false results. As (4) contains only one exponential,
a single thermally activated degradation mechanism has been
assumed. To support this thesis, long term measurements at
lower temperatures have to be performed.

5
Advanced modes of sensor operation

5.1
Temperature cycling
The technique based on cycling the device temperature
between high (400 °C) and low temperatures (room temper- Fig. 8. Variation of the sensing layer conductance upon operation in
ature to 200 °C) is finding wider applications for obtaining CO and synthetic air of various humidity backgrounds. Test 1 and
higher selectivity for CO detection [Figaro (1991); Hellmich et test 2 distinguish two experiments performed using different values of
al. (1995)]. In this context an attractive feature of microm- the low temperature
achined silicon devices is the small heat capacity associated
with extremely thin membranes. As has already been shown
above, the thermal response of such membranes is of the order is reduced and a constant alarm level can be used between 20
of a few msec. For this reason, the micromachined heater and 80% relative humidity.
elements lend themselves very easily towards temperature-
modulated modes of operation. 5.2
In Fig. 7 we report the results of a temperature cycling Sensitivity control by field effect
experiment which had been performed using one of the A further attractive feature of using SnO2-thin films in
microstructures shown in Fig. 1. In this experiment, the sensor combination with micromachined heater elements is that the
was operated in a background of 100 ppm CO admixed to gas-sensitive metal oxide films can be arranged in field-effect
humidified synthetic air with the membrane temperature being configurations. It is necessary to use thin-film gas sensors,
periodically cycled between 400 and 150 °C. We observed that, because the electrical field will not penetrate thick film gas
under these conditions, the response to CO is enhanced by sensors. This possibility is sketched in Fig. 9. Basic experi-
about a factor of 4 and that the average power consumption is ments, during which only the bias voltage has been changed,
reduced to about 40% as compared to operation under showed, that the whole sensitive layer can be electron-depleted,
isothermal conditions. but only for a short time in the range of 10 to 30 minutes. As
Figure 8 summarizes the results of similar CO detection a consequence, to effectively influence electron density and
experiments. In this figure the conductance of the sensitive film thus the sensitivity of the gas sensitive layer, the bias voltage
at the end of the ‘‘cold’’ period is plotted as a function of the has to be cycled. A typical measurement with a cycled bias
humidity background. The data of Fig. 8 demonstrate that the voltage is shown in Fig. 10.
cross sensitivity to humidity can be clearly reduced by raising Fig. 11 summarizes the sensitivity dependence to CO,
the ‘‘low’’ temperature towards 200 °C. Under these latter humidity and NO2 on the electron density. The columns for
conditions the adsorption of water vapor on the sensor surface electron accumulation, for electron depletion and for
 layer does not depend on the composition of the ambient air,
Ggas/Gair\1 for such resistors. It can be clearly seen that
electron depletion reduces the sensitivity and may even change
the normal behavior of the sensitive layer. For the adsorption
of H2O and NO2 electrons are necessary to build the adsorbates
OH[ and NO~, respectively. So it is clear, that electron
3
depletion leads to a reduced sensitivity to these gases.
Under uninfluenced electron density or electron accumula-
tion CO is oxidized with oxygen ions on the SnO2- surface to
CO2. This mechanism causes electrons to be injected into the
bulk of SnO2, leading to a higher conductivity. Electron
depletion causes oxygen to desorb. If a CO molecule does not 189
find as many oxygen atoms on the SnO2-surface to be oxidized
to CO2 , the conductivity of the SnO2-layer will not decrease any
more. The decrease of the conductivity for electron depletion
may be attributed to the adsorption of CO at the SnO2 - surface.
Fig. 9. Micromachined gas sensor with electrical circuit for bias
voltage measurements
6
Conclusions
Gas-sensitive metal oxide semiconductor films have been
deposited onto micromachined substrate heater elements. In
particular, the following results have been obtained:
—We have shown that the micromachining approach allows
the heating power consumption to be reduced by about one
order of magnitude with respect of commercial devices. Such
a reduction is essential for realizing battery - operated sensor
arrays [Endres et al. (1995)]:
—the small thermal inertia allows for the successful
implementation of temperature-cycling methods in the gas
detection process;
—the thin film topology of the micromachined gas sensor
elements allows for novel methods of selectivity control such as
field-effect modulation.
Sensors of this type are expected to experience a wide
commercial distribution in the detection of environmental
Fig. 10. NO2 - measurement with cycled bias voltage: electron
depletion reduces the sensitivity to NO2 pollutants such as CO, NO2 and O3 as well as flammable gases
such as methane and propane.

References
Cavicchi RE; Suehle JS; Kreider KG; Gaitan M; Chaparala P (1995)
Optimized Temperature Pulse Sequences for the Enhancement of
Chemically-Specific Response Patterns from Micro-Hotplate Gas
Sensors, Transducers ’95, Eurosensors IX, proceedings, Stockholm,
Sweden, 823—826
DiMeo Jr F; Semancik S; Cavicchi, RE; Poirier GE (1996) Hetero-
epitaxy of Conducting Oxides on Silicon Using Oxide Buffer
Layers, The 6th International Meeting on Chemical Sensors,
proceedings, Gaithersburg
Dori L; Cardinali GC CNR - LAMEL, Bologna, Italy, private
communications.
Dücso8 Cs; Vázsonyi Ë; Ädám, M; Bársony I; Gardeniers JGE; van den
Berg A (1996) Porous silicon bulk micromachining for thermally
isolated membrane formation, Eurosensors X, proceedings,
Leuven, Belgium, 227—230
Fig. 11. Variation of the CO, humidity and NO2 sensitivity with gate Endres HE; Goettler W; Jander HD; Drost S; Sandmaier H;
bias voltage Sberveglieri G (1995) Improvement in signal evaluation methods
for semiconductor gas sensors, Sensors and Actuators B, 27,
uninfluenced electron density are black, white and grey, 267—270
Endres H-E; Göttler W; Hartinger R; Drost S (1996) A thin - film SnO2
respectively. The sensitivity in this plot is the deviation of the sensor system for simultaneous detection of CO and NO2 with
conductivity with test gas divided by the conductivity in wet neural signal evaluation, conference ‘‘Sensoren und ihre Anwen-
synthetic air Ggas/Gair from 1. As the resistivity of an insensitive dungen’’, Bad Nauheim
 Enzyclopédie des gas, L’ air liquide, (1976) Elsevier, Amsterdam Kunt TA; Mc Avoy TJ; Cavicchi RE; Semancik S; Ratton L;
Figaro (1991) Technical Catalogue, Gas Sensor TGS 203 Fuja T (1996) Application of Data-Based Dynamic Modeling
Fung S; Tang Z; Chan P; Sin J; Cheung P (1995) Thermal Analysis and Techniques to Temperature-Programmed Microhotplate Gas
Design of a Micro-Hotplate for Integrated Gas Sensor Applications, Sensors, The 6th International Meeting on Chemical Sensors,
Transducers ’95, Eurosensors IX, proceedings, Stockholm, Sweden, Gaithersburg
818—821 Pfaller K (1993) Entwicklung und Vermessung miniaturisierter
Hellmich W; Krenkow A; Bosch - v. Braunmühl, G Müller; C Perego; Gassensorelemente, diploma thesis, Fachhochschule München
G Faglia; G Sberveglieri (1995) A selective and highly sensitive CO Sberveglieri G; Faglia G; Groppelli S; Nelli P; Camanzi A (1990)
sensor based on RGTO - SnO2 films deposited onto Si micro- A new technique for growing large surface area SnO2 thin fims
machined structure, Proc. Second Asia Conference on Chemical (R. G. T. O. technique), Semicond. Sci. Technol., 5: 1231ff
Sensors (EACCS’95) Xi’an, China, 137—141 Semancik S; Cavicchi RE; Kreider KG; Suehle JS; Chaparala P (1995)
Hellmich W; Müller G; Bosch - v. Braunmühl Ch; Doll T; Eisele I Selected-Area Deposition of Multiple Active Films for Conduc-
(1996) Field - Effect - Induced Gas Sensitivity Changes in Metal tometric Microsensor Arrays, Transducers ’95, Eurosensors IX,
190 Oxides, Eurosensors 96, proceedings, Leuven, Belgum, 505—508 proceedings, Stockholm, Sweden 831—834
Perego C; Sberveglieri G; in D’ Amico A; Sberveglieri G (ed.s) (1995) Suehle JS; Cavicchi RE; Gaitan M; Semancik S (1993) Tin Oxide
Sensors for Domestic Applications, pp. 217—225. Singapore: World Gas Sensor Fabricated Using CMOS Micro - Hotplates and In - Situ
Scientific Processing, IEEE Electron Device Letters, Vol. 14, NO. 3:
Hoefer U; Kuehner G; Schweizer W; Sulz G; Steiner K (1994) Sensors 118—120
and Actuators B 22: 115—119. Takada T; Shiozawa H; Kobayashi H; Tanjou H; Matsubara Y (1996)
Hellmich W (1996) Feldeffektinduzierte Empfindlichkeitssteuerung Miniaturized Hot Wire Type Semiconductor Gas Sensor Operated
bei mikromechanisch hergestellten Halbleitergassensoren auf SnO with a Battery, The 6th International Meeting on Chemical Sensors,
2
- Basis, PhD - thesis, Universität der Bundeswehr München Gaithersburg