new perspectives of gas sensor technology

Introduction Looking back its history, gas sensor technology was inaugurated when three kinds of pioneering gas sensors were put in prac-tice in Japan, i.e., oxide semiconductor gas sens

Contents lists available atScienceDirect Sensors and Actuators B: Chemical

j o u r n a l h o m e p a g e :w w w e l s e v i e r c o m / l o c a t e / s n b

New perspectives of gas sensor technology

Noboru Yamazoe, Kengo Shimanoe∗

Faculty of Engineering Sciences, Kyushu University, 6-1 Kasuga-koen, Kasuga, Fukuoka 816-8580, Japan

a r t i c l e i n f o

Article history:

Received 15 August 2008

Received in revised form 11 November 2008

Accepted 5 January 2009

Available online xxx

Paper presented at the International

Meet-ing of Chemical Sensors 2008 (IMCS-12), July

13–16, 2008, Columbus, OH, USA.

Keywords:

Semiconductor

Gas sensor

Oxide

Receptor

Depletion

MEMS

a b s t r a c t

Two recent topics important for advancing gas sensor technology are introduced Semiconductor gas sensors have been developed so far on empirical bases but now a fundamental theory has been made available for further developments The theory reveals the roles of physical properties of semiconductors and chemical properties of gases in the receptor function MEMS techniques have been applied to fabri-cation of micro-platforms for use in gas sensors The micro-platforms appear to provide gas sensors with new innovative function

© 2009 Elsevier B.V All rights reserved

1 Introduction

Looking back its history, gas sensor technology was inaugurated

when three kinds of pioneering gas sensors were put in

prac-tice in Japan, i.e., oxide semiconductor gas sensors for gas leakage

alarms[1], solid electrolyte oxygen sensors for car emission control

systems[2], and ceramic humidity sensors for automatic cooking

ovens[3] These sensors demonstrated dramatically how

impor-tant it was to monitor a specific gas species in situ, in real time

and continuously for ensuring safety from gas hazards, protecting

environments, or making home appliances intelligent or friendly

to users Importance of such emerging technology was well

recog-nized world wide when the first International Meeting on Chemical

Sensors (IMCS) was held at Fukuoka, Japan, in 1983, under the

lead-ership of the late Professor Seiyama[4] Research and development

was triggered off all over the world to seek new and/or better gas

sensors

Currently various kinds of gases including reducing ones

(methane, propane, carbon monoxide, ammonia, hydrogen sulfide,

etc.) and adsorptive ones (oxygen, nitrogen dioxide, ozone, etc.)

have been made detectable with gas sensors using

semiconduc-tors, electrolytes or catalytic combustion Yet there are various

new demands to gas sensors ranging from detecting VOCs (Volatile

∗ Corresponding author Tel.: +81 92 583 7876; fax: +81 92 583 7538.

E-mail address:simanoe@mm.kyushu-u.ac.jp (K Shimanoe).

Organic Compounds) at very low concentrations (ppb levels) to constructing sensor network systems Needless to say, even the established gas sensors are demanded for innovations towards bet-ter sensing performances, lower power consumption and more compact device structures To meet these demands, semiconduc-tor gas sensors are considered to be best suited because they have advantageous features such as simplicity in device structure and circuitry, high sensitivity, versatility and robustness It is pointed out that most of the sensors in this group have been developed empirically from a lack of theoretical understandings Fortunately,

we recently succeeded in deriving theoretical equations to describe the response of these sensors to adsorptive or reducing gases quan-titatively[5–7] We expect that the new theory will provide useful guidelines on how to elaborate selection and processing of sens-ing materials and additives used as well as device structure and fabrication techniques to be used

From such a standpoint, theory of semiconductor gas sensors is introduced here as a main topic Another topic, MEMS-assisted gas sensors, is selected because those are likely to give rise to a new breakthrough in gas sensor technology

2 Theory of semiconductor gas sensors

2.1 Overview of empirical information and problems

Semiconductor gas sensors detect a specific target gas from

a change in the electric resistance of a sensing body which is a 0925-4005/$ – see front matter © 2009 Elsevier B.V All rights reserved.

doi: 10.1016/j.snb.2009.01.023

Fig 1 Three basic factors controlling semiconductor gas sensors.

porous assembly of tiny crystals (particles) of oxide

semiconduc-tors such as SnO2 Gas sensing properties have been accounted

for by three basic factors, namely, receptor function, transducer

function and utility factor, as schematically shown inFig 1 [6–10]

The first one is concerned with how each crystal responds to

the stimulant gas in problem, while the second does how the

response of each crystal is transduced into device resistance The

third one describes how the device response (resistance change)

is attenuated in an actual porous sensing body due to a

con-sumption of the stimulant gas during its diffusion inside Among

these factors, only the last one has been clarified theoretically

[9,10] As for the receptor function, it is known that oxygen is

adsorbed on the crystals in air, presumably as O−, over the

tem-perature range of interest[11]to form a depletion layer in them

Upon contact to a stimulant (H2), the oxygen adsorbates are

con-sumed more or less, causing the depletion layer to decrease The

transducer function can also be easily understood qualitatively

based on the double Schottky barrier model so far assumed

popu-larly

It is remarked, however, that the double Schottky barrier

model is nothing but estimated as an extension from other

semi-conductor devices In fact, some sensor devices fabricated with

wet-coating techniques have been found to exhibit

temperature-independent resistance in air in disagreement with the model,

and tunneling transport of electrons between adjacent crystals

has been strongly suggested instead[8] The scheme of receptor

function above is also too qualitative, failing to account for many

pieces of experimental information including grain size effects on

sensitivity Definitely quantitative approaches to the receptor

func-tion are badly needed This is a main concern here in the first

topic

2.2 How to formulate receptor function

2.2.1 Scope of formulation

On each semiconductor crystal, adsorption and/or reactions of

gases take place to capture or release electrons, while a

corre-sponding redistribution of electrons takes place inside to achieve an

electrostatic equilibrium The surface chemical affair and the

sub-surface physical one are not independent but united together, and

this provides a base on which the receptor function is formulated

as follows

The chemical affair can be formulated easily In an oxygen

atmo-sphere, for instance, oxygen adsorption is expressed as follows:

Here KO2and PO2are adsorption constant and partial pressure of

oxygen, respectively, and [e]sand [O−] are surface densities of free

electrons and O−, respectively In the presence of a reducing gas

(H2), O−is consumed by the reaction:

By coupling with(R1), the following equation results in the steady state:

KO2PO2[e]s2= [O−]2+ cPH2[O−] (2)

PH2is partial pressure of H2, and c is a constant defined by c = k2/k−1,

where k2 and k−1are rate constants of(R2)and reverse reaction

of(R1), respectively Eqs.(1) and (2)combine between [e]s and [O−] in the oxygen atmosphere in the absence and presence of H2, respectively Since the electrostatic equilibrium condition gives rise

to another interrelation between [e]sand [O−] as stated later, the

two variables can be determined uniquely if PO2, PH2and physical parameters of semiconductor crystals are fixed

2.2.2 Electrostatic equilibrium for large crystals

Let us consider electrostatic equilibrium inside a semiconductor crystal In case the crystal is large enough, depletion is limited in the shallow region from the surface We can assume a flat surface for which the electrostatic equilibrium has been discussed well by using an energy band diagram as shown inFig 2 A location in the

crystal is expressed by a depth from the surface, x Under simplifying

conditions of complete ionization of donors, no tailing of electron distribution, and no surface states other than O−, Poisson’s equation

Fig 2 Potential energy diagram for depletion in a large semiconductor crystal Here

qV(x) is the potential energy of electron, qVssurface potential energy of electron, x distance from the surface, w depletion depth, Ecconduction band edge, Ev valence

band edge, and EF is the Fermi level.

can be solved easily to give the following equations[5]:

qV

kT =1

2

 x − w

LD

2

(3)

qVs

kT =1

2

 w

LD

2

(4)

[e]s= Nd exp

−qV

s

kT



(5)

[O−]=−Qsc

Here, q is the elementary charge of electron, kT thermal energy,

V and Vs electric potentials at depth x and surface, respectively,

w depletion depth, Nddonor density, LDDebye length defined by

LD= (q2Nd/εkT)−1/2(ε, permittivity), and Qscis the surface charge

density [e]sand [O−] are seen to be correlated implicitly through

w by Eqs.(4)through(6) By coupling these equations with(1)or

(2), [e]sand [O−] can be solved for given PO2and PH2

2.2.3 Electrostatic equilibrium for small crystals

In case the crystal is small, we encounter two kinds of

uncon-ventional phenomena First, depletion extends to cover the whole

crystal with increasing PO2, and what would happen thereafter?

Second, the solutions of Poisson’s equation become dependent on

the shape of crystal These phenomena have important meanings

for actual semiconductor gas sensors, as discussed later

The first phenomenon is illustrated schematically in Fig 3,

where three thin plates with thicknesses of l, l/2 and l/4 are exposed

to stepwise increasing partial pressures of oxygen, PO2(I), PO2(II) and

PO2(III) When depletion extends to l/4 from both surfaces for the

thickest plate at PO2(I), depletion just covers the whole region for

the l/2 thick plate, while the thinnest plate is put into a new type of

depletion When depletion covers the whole region of the thickest

plate at PO2(II), the thinner plates are both in the new type

deple-tion The new type depletion is seen to show up more easily (at

lower PO2) as the thickness is reduced or PO2is increased The new

type depletion and the conventional type one are called here

vol-ume depletion and regional depletion, respectively, and the border

state is called boundary depletion

Nature of volume depletion is easily understood from the

dia-grams of potential energy and electron distribution shown inFig 4,

where the one-dimensional coordinate, x, is redefined as a distance

from a center of the crystal (thickness 2a) The crystal is in the flat

band state at PO2= 0 as assumed In the presence of oxygen (PO2(I)),

surface state (O−) is formed and depletion takes place from both

Fig 3 Stages of depletion in thin semiconductor plates with three different

thick-nesses placed under three different partial pressures of oxygen.

Fig 4 Potential energy diagram (a) and electron distribution diagram (b) for

deple-tion in a thin semiconductor plate.

sides of the plate to a depth at which an electronic equilibrium is reached between bulk and the surface state (O−) (regional deple-tion) Free electrons have been transferred from a shallow region as shown in the electron distribution diagram, where a tailing effect

of distribution is taken into account With increasing PO2, the sur-face state (O−) shifts to lower energy and so depletion depth w also increases until it reaches finally the center of the plate (w = a)

at PO2(II) (boundary depletion) For a further increase in PO2 to

PO2(III), however, there is no room to extendw further Instead the location of Fermi level shifts down by an adequate quantity

(pkT), while keeping the potential energy profile the same as that for PO2(II), to satisfy the new equilibrium (volume depletion) At this stage, free electrons are squeezed out of the whole plate, as seen from the electron distribution diagram This is a reason why

it is named volume depletion It is inferred that volume deple-tion corresponds to Region III reported by Rothschild and Komen [12,13]

The potential energy of electrons inside the plate is formulated for each type of depletion as follows[6]:

qVr(x)

kT =1

2

 x − (a − w)

LD

2

(7)

qVb(x)

kT = (1/2L2

qVv(x)

kT = (1/2L2

Fig 5 Potential energy diagram drawn relative to the flat band state.

Suffices r, b and v stand for regional, boundary and volume

deple-tion, respectively Once the potential energy is known, [e]scan be

derived from(5) On the simple abrupt model (Fig 2), [O−] is given

by(6) For more precise treatment, it should be modified to

HereA(a, w) is a quantity to be corrected for non-ideal behavior

such as the tailing effect Thus the two variables, [e]sand [O−], are

correlated implicitly by(9) and (10), and hence they are determined

uniquely by coupling these equations with(1)or(2)

The meaning of Fermi level shift, pkT, is worth being discussed.

Fig 5is a potential energy diagram redrawn relative to the flat band

state On going from boundary depletion to volume one, the

poten-tial energy shifts up by pkT The shift causes the surface potenpoten-tial

energy to increase and so causes [e]sto decrease correspondingly

On the other hand, oxygen adsorption equilibrium(1)indicates that

an increase in PO2should be met by a change in [O−]/[e]s ratio

In the stage of volume depletion, [O−] cannot increase so much

since most of conduction electrons available have been exhausted

already That is, oxygen adsorption is controlled by a supply of

elec-trons This means that the change in the above ratio is achieved

mainly by that in [e]s Thus it can be stated that the Fermi level

shift plays a role to connect between the surface chemical

equilib-rium and the subsurface electrostatic equilibequilib-rium in the stage of

volume depletion

Now we consider the second phenomenon, dependence on

the shape of crystals Fig 6 illustrates the coordinates

sys-tems conveniently selected for three shapes of crystals, i.e.,

one-dimensional coordinate (plate), three-dimensional spherical

coordinates (sphere) and three-dimensional columnar coordinates

(column) By choosing such coordinates systems, Poisson’s equation can be simplified as follows:

d2V

dx2 = −qNd

1

r2

 d

dr



r2dV

dr



= −qNd

1 r

 d

dr



rdV

dr



= −qNd

The solution of(11)has already been shown as(7)–(9)for plates Similarly(12) and (13) can be solved mathematically under the same boundary conditions[6] Thus [e]sand [O−] can also be deter-mined uniquely for spheres and columns

2.3 Response to oxygen and other stimulants

As so far mentioned, we can formulate receptor function theo-retically In order to visualize it, however, we need to transduce it

into device resistance (R) For this purpose, it is assumed that R is

proportional to [e]s, and that it follows the following equation: R

R0 = [e]s

Nd = exp−qV

s

kT



(14)

R0 is device resistance at PO2= 0 (flat band state) Although R is

influenced by many factors other than the receptor function, those

factors are cancelled out for R/R0, which is called reduced resistance hereafter

For devices consisting of large crystals, R/R0is correlated with

PO2and PH2by the following equations

In pure oxygen,

X = m exp



m2

2

 , RR

0 = exp



m2

2



(15)

In oxygen mixed with H2,

X =

1+Y m

1 /2

exp



m2

2

 , RR

0 = exp



m2

2



(16)

X is reduced adsorptive strength of oxygen defined as

X = (KO2PO2)1/2/LD, Y is reduced reactivity of H2 defined as

Y = cPH2/NdL D(see(2)for c) and m is reduced depth of depletion

defined as m = w/LD The correlations, obtained by numerical calculations, explain the power laws governing the response to O2 and H2, but no other important information relevant to sensing properties can be drawn

Fig 6 Coordinates systems selected for plates, spheres and columns.

Fig 7 Correlations between reduced resistance (R/R0 ) and reduced adsorptive

strength of oxygen (X) for semiconductor spheres different in reduced size (n) as

drawn on logarithmic scales.

Situation becomes utterly different for devices using small

crys-tals When the crystals are spheres of radius a, for example, the

following equations are derived for the response to oxygen

For regional depletion,

X =n

3

 

1− (n − m)

n

3

− AS(n, m)

 exp



m2

6



1+ 2(n − m) n R

R0 = exp



m2

6



1+ 2(n − m) n

(17)

For volume depletion,

R

R0 =3

n



X +3

n



AS(n, n) exp



n2

6



=3

a



Here n is reduced radius defined as n = a/LDand m is reduced depth

of depletion AS(n, m) is a correction term for non-ideal behavior

such as the tailing effect Cs(n) is a constant fairly close to unity

for small n (<4) The correlation between R/R0 and X shown by

(17)is implicit and non-linear, while R/R0is a linear function of

X or PO2as indicated by(18) The correlations given by these

equa-tions are drawn on logarithmic scales inFig 7, where n is varied

between 1 and 10 For large n (=10), regional depletion prevails over

a wide range of X up to about 108, while its range deteriorates rather

sharply, being replaced by volume depletion, with decreasing n, and

it becomes practically invisible at n = 2 The meanings of the

cor-relations are made clearer inFig 8, where the same correlations

are drawn for a small range of X up to 150 on linear scales Linear

correlations are obtained between R/R0and X for volume

deple-tion, with their slopes (sensitivity) being equal to 3/n, although

regional depletion is also visible for n > 3 Evidently this accounts

for the grain size effect on the response to oxygen Similar

discus-sion can be made on the crystals of other shapes For brevity, only

the correlations for volume depletion are compared below:

R

R0 =1

a



(KO2PO2)1/2+ Cp(n) (plate) (19)

R

R0 =2

a



(KO2PO2)1/2+ Cc(n) (column) (20)

R

R0 =3 a

 (KO2PO2)1/2+ Cs(n) (sphere) (18)

It is understood that the linearity constant (sensitivity) increases

to two or three times, on going from plates to columns or spheres This is nothing but indication of a shape effect on sensitivity It

is worth stating that the linearity constants are in coincidence with the surface area/volume ratios of the crystals of respective shapes

The response to H2in air can be treated similarly Under the condition that the rate of reaction(R2)is much faster than the rate

of desorption of O−, reduced resistance for a spherical crystal-based device is given by

Rg

R0 = Nd1/2(ac/3KO2PO2(air))−1/2PH2−1/2 (21)

Here Rgis resistance under exposure to H2, PO2(air) is PO2in air, and

c has already been defined for(2) Conventionally defined response,

Ra/Rg(Ra, resistance in air), is expressed as

Ra

Rg = Sc

aNd

1 /2

S is shape factor and is allotted values of 1, 2 and 3 for plates, columns and spheres, respectively Ra/Rgis linear to the square root

of PH2(power law) Its slope (sensitivity) is determined by the sev-eral physical and chemical parameters included in the parentheses

To achieve high sensitivity, spherical crystals (S = 3) are best suited,

and a and Ndshould be reduced as much as possible It is also under-stood that selectivity among reducing gases is determined by their

reactivity (c) under these conditions.

The response to adsorptive gases can also be formulated, as exemplified here for the case of NO2 Its adsorption is expressed as

KNO2and PNO2are adsorption constant and partial pressure of NO2, respectively, and [NO2 −] is surface density of NO2− NO2molecules

compete with O2 to capture electrons so that formulation of the response to NO2is fairly complex for the stage of regional deple-tion For volume depletion, in contrast, it is quite simple, reduced

Fig 8 Redrawing of the same correlations inFig 7 on linear scales.

resistance being expressed as

R

R0 =S

a



KNO2PNO2+S

a

 (KO2PO2)1/2+ C(n) (24)

S is shape factor, being 1, 2 and 3 for plates, columns and spheres,

respectively C(n) is a small constant close to unity for small n.

Adsorption of NO2and O2is seen to contribute to reduced

resis-tance linearly as well as additively Conventionally defined response

is given as

Rg

Ra =S

a

 R

a

R0

−1

Rais device resistance in air in the absence of NO2 It is noted that

Ra/R0 cannot be derived from(24)since the condition of volume

depletion is not always satisfied in the absence of NO2 The response

is seen to be inversely proportional to a.

2.4 Applicability to experimental data

The theory developed above has been shown to be well

con-sistent with experimental data, though available data are still

limited at present Here two examples are introduced.Fig 9 is

a reproduction of the data showing grain size effects observed

with SnO2-based devices[14] In the region of smaller grain sizes,

device resistances in air (Ra) as well as the responses to H2or CO

(Ra/Rg) tend to increase sharply with decreasing size Previously

we estimated that such behavior could be associated with

deple-tion of the whole region of each constituent grain In the light of

the present theory, however, this estimation should be corrected

largely, though not totally wrong First of all, it is shown that Ra

should be independent of grain sizes if volume depletion is assumed

to prevail The sharp increase of Ra observed can be attributed

instead to an increase in the number of grains without donors

(insulating grains) This can happen when the grain size (diameter)

is made smaller than an average separation between neighboring

donors under the condition of a fixed donor density (Nd) From the

grain size value at which Rabegins to increase, Ndis estimated to

be 5.6× 1018cm−3 This in turn leads to LD= 2.4 nm at 600 K,

show-ing that reduced sizes, n = a/LD, of the grains tested are between 5.4

(largest) and 0.9 (smallest) This range of n is consistent with the

assumption of volume depletion From(22), the response to a fixed

Fig 9 Dependence of device resistance in air (Ra ) and response to H 2 or CO in air

(Ra/Rg ) on grain sizes of SnO 2(diameter d)[16]

Fig 10 Response data to H2or CO as correlated with a−1/2.

concentration of H2or CO should be proportional to a−1/2:

Ra

Rg = (3cPgas/Nd)1/2a−1/2 (gas; H2 or CO) (26)

The response data are plotted against a−1/2 inFig 10 It is seen that three data on the larger grains fall on a straight line passing through origin in either cases of H2 (A) and CO (B) in agreement with(26), while the remaining data on the smaller grains deviate upward probably through improvements of utility factor due to the appearance of insulating grains The ratio of the slopes of straight

lines, A/B, gives the ratio of reaction rate constants (k2) of H2and

CO; k2of H2is analyzed to be 14 times as large as that of CO Sim-ilar H2response data obtained with Al-doped SnO2[15]are also plotted in the same figure (C) The slope ratio, C/A, gives the ratio

of donor density (Nd), which is analyzed to be 1/29 In this way,

the influences of a, k2and Ndon the response can be accounted for quantitatively

Sensing behavior to NO2in air can be explained satisfactorily as well Sensitivity to NO2has been shown to be enhanced greatly with decreasing grain size or lamella thickness for WO3-based devices,

as shown inFig 11 [16] This is consistent quite well with(25), which predicts sensitivity inversely proportional to grain size or lamella thickness As another typical feature, response behavior to

NO2is strongly dependent on operating temperature As shown in Fig 12, the response is linear to concentrations of NO2 down to small concentration levels at a low temperature (200◦C) At higher temperatures (300 and 400◦C), however, besides a decrease in sen-sitivity, the response becomes non-linear in the region of smaller concentrations, and the non-linear region grows more conspicuous with increasing temperature Notably even in these cases linearity

is maintained for the larger concentrations of NO2 Such behav-ior is consistent with what is expected from the theory At a low

temperature where adsorption constant (KNO2) is large enough to fulfill the condition of volume depletion down to a small concen-tration of NO2, the linearity holds seemingly over the whole range

of NO2concentrations according to(25) With increasing

temper-ature, KNO2is lowered to cause regional depletion to appear in a range of smaller concentrations of NO2, while volume depletion dominates at larger concentrations This explains why the response exhibits non-linear behavior followed by linear one at the higher temperatures as observed as well as why linearity constant (sensi-tivity) of the linear region decreases with increasing temperature It

is worth noting that the dependence of response behavior on tem-perature inFig 12is very similar to that on reduced size (n) inFig 8

Fig 11 Response data to NO2 (1 ppm) in air at 200 and 300 ◦ C for WO 3 -based devices

as correlated with grain size (diameter d) or lamella thickness (l).

In fact this similarity is rationalized because a decrease in KNO2is

equivalent to an increase in n.

3 MEMS-assisted gas sensors

An important technology called MEMS

(Micro-Electro-Mechanical System) was born about two decades ago Since then,

it has been developed greatly for realizing various types of physical

Fig 12 Response of a lamellar WO3 -based device at three operating temperatures

as a function of NO 2 concentration.

Fig 13 Structure and heating and cooling characteristic of a micro-platform[17]

sensors and actuators Although gas sensors have no electro-mechanical parts, micro-fabrication techniques elaborated for MEMS can be transferred to gas sensors [17,18] In recent years, MEMS techniques have been gathering a strong focus in the field

of gas sensors, especially for the purpose of realizing a micro-platform, suspended in a cavity of silicon for thermal insulation, on which sensing materials are coated One of the greatest difficulties associated has been how to attain stably an operating temperature high enough to secure gas sensing (up to 450–500◦C) After various efforts, such micro-platforms have been realized An example is shown inFig 13 [19] The platform, a square of about 100␮m in width, is attached with a micro-heater and a pair of comb type electrodes On heating in a pulse mode of 100 ms on and 100 ms off, device temperature changes rapidly between 450◦C and room temperature, the high temperature being stabilized in as short as

30 ms after switching on, as shown Such an excellent characteristic

is granted by elaborating the platform design with regard to heater conductance, heat capacitance, insulation of heat, and so on Micro-platforms are expected to provide gas sensors with var-ious benefits First, power consumption can be reduced drastically, especially through a pulse-mode heating operation with a small duty ratio By saving power consumption, sensor devices can be made drivable with a battery, which is beneficial to cordless or portable gas sensors For example, realization of battery-driven gas sensors durable for 5 years without charge is planned for installation in houses and vehicles Second, sensor devices can

be miniaturized drastically Coupled with the small power con-sumption, this feature makes the devices easier to install in a small space and so more appropriate to apply for various new ubiquitous sensor systems Third, the excellent heating and cooling characteristics can provide gas sensors with new functions For example, it is possible to scan operating temperature stepwise in

a short time Probably response data at 10 different temperatures will be acquired within a few seconds, and the resulting response

vs temperature profile will be a useful tool to identify the gaseous component(s) in problem False alarming will thus be eliminated

Fig 14 Structure of adsorption–combustion type VOC sensor (a) and sensing capability to toluene gas achieved (b).

effectively The same characteristics will facilitate to introduce

use-ful analytical techniques such as condensation and extraction into

gas sensing This is exemplified well by the adsorption–combustion

type VOC sensor developed recently[20] A catalytic combustion

type sensor fabricated on twin micro-platforms (Fig 14) is

oper-ated in a mode of pulse heating (for 0.4 s) and off or low power

heating (for 9.6 s) VOCs are adsorbed and accumulated on the

catalyst layer during the off- or low power heating-period Those

adsorbates are subjected to catalytic combustion during the pulse

heating period If the sensor response is continuously monitored,

the concentration of VOCs surrounding the sensor can be estimated

from the response peak height It is recognized that the key of

this sensor is possessed by the condensation of VOCs through

adsorption By optimizing the temperature of adsorption, the

sensor can detect as low as 10 ppb toluene in air, as shown

As mentioned above, gas sensors are expected to be

inno-vated greatly by the use of micro-platforms However, there are

great tasks to be cleared before the innovated sensors are

com-mercialized One of such tasks is to establish micro-fabrication

techniques to deposit a well-qualified sensing layer efficiently on

micro-platforms Micro-characterization methods should also be

established

4 Conclusions

Gas sensors will ever continue to be requested for further

advancements in the future as a key to solve or control various

problems associated with gases In this regard, particularly

semi-conductor gas sensors which have pioneered gas sensor technology

have to be innovated further The sensors of this group have been

developed successfully mostly on empirical bases thanks to the

excellence of the sensing properties oxide semiconductors

intrinsi-cally possess, but further developments seem to be difficult without

theoretical supports Fortunately, a theory dealing with the

recep-tor function of these sensors was proposed recently The theory was

derived by combining together the chemical affairs taking place on

the surface of individual semiconductor crystals and the physical

affairs inside It reveals the important roles played by the chemical

parameters of gases (adsorption constants and reaction rate

con-stants) as well as the physical parameters of semiconductor crystals

(grain size, grain shape and donor density) The theory is expected

not only to facilitate fundamental understandings but also to give

rise to a breakthrough in the design and processing of materials

toward innovations of gas sensors

Micro-platforms fabricated with MEMS techniques are almost

ready to be used in gas sensors Excellent heating and cooling

char-acteristics of them are expected to provide gas sensors with new

innovative functions toward realization of various next-generation

sensors

References

[1] H Ihokura, DENKIKAGAKU 50 (1982) 99.

[2] T Takeuchi, K Saji, H Kondo, I Igarashi, The Electrochemical Society Extended Abstracts, 78-1, Pittsburgh, PA, October 15–20, 1978, 196 pp.

[3] T Nitta, Z Terada, S Hayakawa, Humidity-sensitive electrical conduction of MgCr 2 O 4 –TiO 2 porous ceramics, J Am Cer Soc 63 (1980) 259–300 [4] T Seiyama, K Fueki, J Shiokawa, S Suzuki (Eds.), Proceedings of International Meeting on Chemical Sensors, Fukuoka, Japan, September 19–22, Kodansha Ltd., Tokyo, 1983.

[5] N Yamazoe, K Shimanoe, Theory of power laws for semiconductor gas sensors, Sens Actuators B: Chem 128 (2008) 566–573.

[6] N Yamazoe, K Shimanoe, Roles of shape and size of component crystals in semiconductor gas sensor (1) Response to oxygen, J Electrochem Soc 155 (4) (2008) J85–J92.

[7] N Yamazoe, K Shimanoe, Roles of shape and size of component crystals in semiconductor gas sensors (2) Response to NO 2 and H 2 , J Electrochem Soc.

155 (4) (2008) J93–J98.

[8] N Yamazoe, K Shimanoe, C Sawada, Thin Solid Films 515 (2007) 8302–8309 [9] G Sakai, N Matsunaga, K Shimanoe, N Yamazoe, Sens Actuators B: Chem 80 (2001) 125–131.

[10] N Matsunaga, G Sakai, K Shimanoe, N Yamazoe, Sens Actuators B: Chem 83 (2002) 216–221.

[11] N Yamazoe, J Fuchigami, M Kishikawa, T Seiyama, Interactions of tin oxide surface with O 2 , H 2 O and H 2 , Surf Sci 86 (1979) 335–344.

[12] A Rothschild, Y Komen, On the relationship between the grain size and gas-sensitivity of chemo-resistive metal-oxide gas sensors with nanosized grains,

J Electroceram 13 (2004) 697–701.

[13] A Rothschild, Y Komen, The effect of grain size on the sensitivity of nanocrys-talline metal-oxide gas sensors, J Appl Phys 95 (2004) 6374–6380 [14] C Xu, J Tamaki, N Miura, N Yamazoe, Grain size effects on gas sensitivity of porous SnO 2 -based elements, Sens Actuators B: Chem 3 (1991) 147–157 [15] C Xu, J Tamaki, N Miura, N Yamazoe, Promotion of tin oxide gas sensor by aluminium doping, Talanta 38 (10) (1991) 1169–1175.

[16] Y.-G Choi, G Sakai, K Shimanoe, N Miura, N Yamazoe, Wet process-prepared thick films of WO 3 for NO 2 sensing, Sens Actuators B: Chem 95 (2003) 258–265 [17] J.S Suehle, R.E Cavicchi, M Gaitan, S Semancik, Tin oxide gas sensor fabricated using CMOS micro-hotplates and in-situ processing, IEEE Electron Dev Lett 14 (1993) 118–120.

[18] G Müller, A Friedberger, P Kreisl, S Ahlers, O Schulz, T Becker, A MEMS toolkit for metal-oxide-based gas sensing systems, Thin Solid Films 436 (1) (2003) 34–45.

[19] K Yoshioka, T Tanihira, K Shinnishi, K Kaneyasu, Development of extremely small semiconductor gas sensor, Chem Sens 23 (Suppl B) (2007) 16–18 [20] M Egashira, T Sasahara, Adsorption–combustion type micro-gas sensor, Mater Integrat 21 (5) (2008) 91–96.

Biographies

Noboru Yamazoe has been a professor at Kyushu University since 1981 until he

retired in 2004 He received his BE degree in applied chemistry in 1963 and PhD

in engineering in 1969 from Kyushu University His research interests include the development and application of the functional inorganic materials.

Kengo Shimanoe has been a professor at Kyushu University since 2005 He received

his BE degree in applied chemistry in 1983 and ME degree in 1985 from Kagoshima University and Kyushu University, respectively He joined the advanced materials and technology laboratory in Nippon Steel Corp and studied the electronic charac-terization on semiconductor surface and the electrochemical reaction on materials

to 1995 He received PhD in engineering in 1993 from Kyushu University His cur-rent research interests include the development of gas sensors and other functional devices.