Moulson A.J., Herbert J.M. Electroceramics: Materials, Properties, Applications

Подождите немного. Документ загружается.

chromium metal to the other. The oxygen activity in the Cr/Cr

mix is known

from thermochemical data relating to the equilibrium reaction:

!2Cr þ 3=2O

(4:40Þ

A common form of disposable sensor is a long cardboard tube with the sensor

elements built into one end. The tube is plunged into the molten steel and the

measurements made within a few seconds.

Amperometric mode An advantage of the lambda sensor as described above is

its sensitivity close to the l ¼ 1 point where the oxygen activity in the exhaust

changes rapidly with burn conditions and consequently so too does the cell e.m.f.

The disadvantage is that under ‘lean-burn’ (oxygen-rich) conditions the log

dependence on oxygen activity renders the cell insufficiently sensitive to provide

effective control. This shortcoming has led to the introduction of sensors used

in the amperometric mode discussed below and illustrated schematically in

Fig. 4.40.

The requirement is to measure the concentration (of order 1 vol.%) of oxygen

in a gas. The oxygen diﬀuses down the capillary and by Fick’s Law the ﬂux, J,is

J ¼DAdc=l (4:41Þ

in which D is the oxygen gas diﬀusion coeﬃcient, l the capillary length and A its

cross-section; dc is the concentration diﬀerence between the capillary ends.

The oxygen ions are driven into the electrolyte by an applied e.m.f. and the

current (I), limited by the supply of oxygen ions, measured. It follows that

I ¼ 4eJ (there are four electrons carried by each oxygen molecule)

202 CERAMIC CONDUCTORS

Fig. 4.39 Schematic of an oxygen meter for monitoring the oxygen content of steel during

reﬁning (adapted from [15]).

TEAMFLY

Team-Fly

I ¼ 4eDAdc=l (4:42Þ

Assuming that the oxygen concentration at the surface of the electrolyte is zero,

and with D, A and l all constant, then

ðoxygenÞ

/ I (4:43Þ

where C

(oxygen)

is the oxygen concentration in the gas.

The capillary tube can be replaced by a porous ceramic appropriately limiting

gas ﬂow. Because of the constant urge to reduce component size and cost

multilayer technology is exploited in the fabrication of sensors.

The sensor illustrated in Fig. 4.41 for measuring NO in oxygen well illustrates

the complexities in design which multilayer technology can readily cope with. In

the ﬁrst stage the gas mixture enters the sensor where the oxygen is pumped

away. The potentiometric sensor monitors the oxygen level and controls the

pumping. The NO diﬀuses through the second barrier and dissociates at the

amperometric sensor electrode, the oxygen ions being driven through the sensor

into the air reference compartment.

CERAMICS-BASED CHEMICAL SENSORS 203

Fig. 4.40 Schematic illustrating the principle of the amperometric mode.

Fig. 4.41 A planar form of NO sensor the entire fabrication exploiting multilayer technology

(adapted from [16]).

As an alternative to the limited gas diﬀusion technique described above, high

oxygen activities can also be measured by driving a current though the wall of a

membrane separating the test and reference gases (e.g. Fig. 4.36). The driving

potential diﬀerence is reversed so that the current is measured under the applied

voltage reinforced by the cell e.m.f. for one half of the cycle and by the applied

voltage minus the cell e.m.f. for the next half cycle. It can be shown that the test

oxygen activity is directly related to the ratio of the currents ﬂowing under the

two diﬀerent voltage gradients. The technique is suited to measuring oxygen

concentrations over a wide range [17].

Sensors for chemical species other than oxygen

The following examples serve to illustrate the versatility of the experimental

approach provided suitable electrolytes can be found. The technology is reviewed

by D.J. Fray [15,18].

Sensing hydrogen in molten aluminium Because of the detrimental effect

dissolved hydrogen can have on the mechanical properties of many metals it is

monitored in metallurgical processing. For this the requirement is for a proton

conductor and suitable ceramics have been identified.

It has been known for many years [19] that ‘water’ can be incorporated into

silica in the form of hydroxyl groups which have a characteristic absorption in

the infrared. This was a problem encountered in the early days of ﬁbre optics

communications’ technology and solved by excluding water from the processing

atmospheres. This ‘water’ can be readily introduced by high temperature

annealing in water vapour and removed by heating in a water-free atmosphere.

More recently, H. Iwahara et al. [20] reported that some compounds having

the perovskite structure (see Section 2.7.3) become proton conductors if

hydrogen is introduced into the crystal, and the solubility of ‘water’ and proton

mobility in ‘perovskites’ are now actively researched topics [21]. The perovskites

which can be tailored to exhibit high protonic conductivity have compositions of

the type

1x

3x

in which M is a trivalent dopant and x the oxygen deﬁciency per formula unit.

CaZr

0.9

0.1

3x

is an example, the indium substituting for zirconium and the

compensating defect being oxygen ion vacancies. Other examples are

SrZr

0.95

0.05

3x

, BaCe

0.8

0.2

3x

, BaCe

0.9

0.1

3x

and SrCe

0.95

0.05

3x

When these compositionally modiﬁed perovskites, which are p-type semi-

conductors, are heated to high temperatures in a water vapour atmosphere

protons are incorporated into the structure, most probably is the form of

hydroxyl groups, with the proton able to ‘hop’ from oxygen to oxygen. The

204 CERAMIC CONDUCTORS

proton mobility will depend upon the strength of the hydrogen bond which will,

in turn, depend critically on structural details. The proton transport number can

be as high as 0.99. Arrhenius plots of the conductivities of the perovskites listed

above are shown in Fig. 4.42.

The exploitation of proton conductors is discussed by H. Iwahara [22]. Figure

4.43 is a schematic of a commercial sensor for measuring the hydrogen content of

molten aluminium [23]. Silicon nitride and carbon are used in the construction

because both are corrosion-resistant to molten aluminium. The electrolyte tube is

CaZr

0.9

0.1

3x

ceramic. The reference hydrogen activity is ﬁxed by a mixture of

1 vol.% hydrogen in argon fed to the platinised outer sensor surface via the

spring-loaded steel tube. The hydrogen gas collected in the inverted sensor tube is

in equilibrium with the hydrogen dissolved in the molten aluminium. The

measured e.m.f. is given by the Nernst equation

CERAMICS-BASED CHEMICAL SENSORS 205

Fig. 4.42 Electrical conductivities of modiﬁed ‘perovskite’ proton conductors (adapted from

[22]).

E ¼ RT=2F flnða@

Þg (4:44Þ

in which a@

and a

are respectively the hydrogen activities on the measuring

and reference sides of the cell. The factor 2 arises because two electrons are

associated with the hydrogen molecule.

Sensing sodium The b-aluminas (see Section 4.5.3) can be successfully

exploited as a sodium sensor. Sodium is widely used in the metallurgical

industry, for example for removing arsenic and antimony from zinc and lead, for

removing the same two elements and oxygen from copper, and phosphorus from

iron. It is also added to aluminium–silicon alloys to control microstructure and,

in consequence, mechanical properties.

The very high vapour pressure of sodium at high temperatures poses problems

for controlling the sodium content of a melt, necessitating its constant and

accurate monitoring.

There are no new principles involved as far as an electrolytic sensor is

concerned. The electrolyte membrane is b-alumina and the reference sodium

activity ﬁxed, at constant oxygen activity, by a mixture of sodium ferrite

(Na

, which can be written 5Na

O 8Fe

) and Fe

, just as

it is common practice to ﬁx oxygen activity by a mixture of a metal and its

oxide.

Sensing CO

ACO

sensor developed by G.M. Kale et al. [24] illustrates a

principle permitting the sensing of species other than that involving the mobile

ion in a solid electrolyte. For example sensing CO

employing a Na-ion

206 CERAMIC CONDUCTORS

Fig. 4.43 A protonic electrolyte used for measuring the hydrogen content of molten

aluminium (adapted from TYK Corporation (Japan) literature; see [23]).

conductor, in this case ‘Nasicon’*. The experimental arrangement is shown in

Fig. 4.44. The Nasicon tube is packed with Na

which serves as an ‘activity-

linking’ compound. Because of the equilibrium reaction

!Na

O þ CO

(4:45Þ

any change in CO

activity in the gas phase in contact with the solid Na

causes a change in Na

O activity. At the surface of the electrolyte the activity of

O in the Nasicon changes to the same value. This change in activity can be

sensed through the corresponding change in cell e.m.f.

At the platinum contact with the solid Na

the electrochemical reaction is:

! 2Na

þ 2e

þ CO

(4:46Þ

And at the Nasicon/gas interface,

2Na

þ 2e

! Na

O(4:47Þ

The oxygen partial pressures are the same on both sides of the cell so that the

overall cell reaction is given by Eq. (4.45).

4.6.2 Gas-sensors based on electronically conducting

ceramics

The science of gas-sensors based on changes in electronic conductivity of a

semiconducting ceramic is a complex matter depending upon a combination of

surface chemistry and electron transport in mainly transition metal oxides. As is

usual in exploiting electroceramics, the applications’ technology runs ahead of

CERAMICS-BASED CHEMICAL SENSORS 207

Fig. 4.44 Schematic of a CO

sensor incorporating sodium carbonate as a ‘linking’

compound. N.B. Because of the small size of the sensor (the outside diameter of the tube is

only approx. 4 mm) equilibration between the atmosphere and sodium carbonate is rapid at

400 8C.

* Nasicon is a generic term for sodium ion conducting oxides containing elements such as Zr, Si, and P.

The values of their conductivities are similar to that of b-alumina.

the basic understanding. The following discussion provides the base for further

study for which the literature sources [25–29] are recommended.

Many semiconducting oxides, for example doped SnO

, ZnO, and TiO

, show

changes in electrical resistivity in the presence of small concentrations of

ﬂammable gases such as propane, and toxic gases such as carbon monoxide. This

phenomenon has led to the development of a wide variety of gas-sensors which

are ﬁnding a rapidly expanding range of applications.

Because of a combination of unique physical properties, namely high

transparency in the visible region of the spectrum, semiconducting characteristics

and good thermal, mechanical and chemical stabilities, tin oxide is exploited for

resistors (see Section 4.2.1), furnace electrodes (see Section 4.1.4) and in

optoelectronic devices (see Chapter 8) as well as dominating in gas-sensor

technology. Unless otherwise stated the discussion is based upon n-type

semiconducting SnO

and the reader is referred to Section 4.1.4 for introductory

remarks relating to its solid state science.

The sensors detect gases because of a change in electrical resistance which

accompanies the reaction between adsorbed species such as O



and O

2

and

the ambient gases being sensed, the ‘target’ gases. Figure 4.45 illustrates how the

electron distribution in a semiconductor is modiﬁed as oxygen interacts with the

surface leading to the adsorption of the most likely species, O



(Note. The Debye length (L

), although not introduced into the present

simpliﬁed discussion, is a parameter frequently referred to in the gas-sensor

literature. It was originally introduced into ionic solution theory and later

applied to semiconductor theory where it is especially applicable to semicon-

ductor/metal and semiconductor/semiconductor junctions. It is a measure of the

distance beyond which the disturbance at the junction has eﬀectively no inﬂuence

on the electron distribution and therefore closely related to ‘d’ (see Eq. (4.49)). It

is a material parameter given by L

¼(ekT=e

1=2

where c

is the undisturbed

electron concentration, essentially the extrinsic electron concentration in the case

of doped n-type tin oxide, and the other symbols have their usual meaning.)

208 CERAMIC CONDUCTORS

Fig. 4.45 Schematic illustrating the electron-depleted layer at a surface of SnO

carrying

chemisorbed oxygen.

With the chemisorption of, for example, oxygen a surface density of electron

acceptor states leads to the establishment of a Schottky barrier. The process is

essentially the same as that which occurs in the case of the PTC thermistor (see

Figs 4.21 and 4.10). The electron potential barrier height (’) is given by

’ ¼ e

=2eN

(4:48Þ

where N

and N

are respectively the surface density of ‘acceptors’ and the

volume density of ‘donors’.

Under the same simplifying assumptions as for the PTC thermistor (where the

surface states are shared between two grains) the depleted layer thickness d is

given by

d ¼ð2e’=e

1=2

¼ N

(4:49Þ

Summarizing the basics of the gas-sensing mechanism, oxygen molecules from

the surroundings become chemisorbed (that is chemically bonded) to the surface

atoms, the process being described by the reaction:

ðgÞþ2e

¼ 2O



ðadÞ (4:50Þ

As the adsorption proceeds a positive space charge develops in the surface layer

of the oxide as electrons transfer from donor defects to the adsorbed oxygen. The

electrostatic ﬁeld generated in the surface regions of the crystal tends to oppose

the charge transfer process and eventually it stops. The thickness of the charge-

depleted layer is determined by the availability of electrons able to transfer to the

surface states and by the surface concentration of adsorbed oxygen. Measure-

ments indicate the thickness of depletion layers to lie in the range of a few

nanometres to approximately 500 nm. Reaction of reducing gases at the surface

removes some of the adsorbed oxygen and releases electrons which are then

available for conduction.

The above description is on the basis of the semiconductor being n-type. In the

case of a p-type semiconductor (e.g. Cr

2x

;0:01 5 x 5 0:45 [29]) then a

decrease in resistance would accompany an increase in the amount of adsorbed

oxygen.

The most common form of gas sensor is based on a porous, sintered ceramic as

shown schematically in Fig. 4.46. As adsorption occurs over the surface of the

semiconductor grains, the barriers to charge transport develop, especially at the

grain boundaries and at particle contact areas (the ‘neck’ regions).

The ideal sensor would have high ‘sensitivity’ and be ‘selective’. That is, it

would show a large resistance change for a small change in the target gas

concentration, and have the ability to discriminate between diﬀerent gases. It

would also show a reproducible response over the required life-time and be

economically viable.

CERAMICS-BASED CHEMICAL SENSORS 209

The sensitivity of a polycrystalline gas-sensing ceramic depends upon sensor

composition, ceramic particle size (since this determines reactive surface area,

and particularly its relationship to L

), sensor geometry, temperature (which

strongly aﬀects adsorption processes and oxide conductivity) and accessibility of

the sensor bulk material (i.e. permeability) to the gas or gases.

It would be inappropriate here to consider the eﬀects of these variables in

detail since optimising sensor performance is still essentially a matter of

‘technical development’ involving not only them but also the sensor housing and

associated circuitry. Optimisation must, of course, be achieved within the

inevitable cost constraints.

The sensor compositions usually include noble metals, particularly palladium

and platinum, since they are found to increase sensitivity through catalysing the

reaction between the oxygen adsorbate and reducing gases. In this role the

additive is said to be acting as a ‘chemical sensitizer’.

The noble metal might also act in a quite diﬀerent way, as an ‘electron

sensitizer’. Because of its very ﬁne particle size (510 nm) relative to the typical

SnO

grain size there are many metal/semiconductor contact points. Also,

because of the high speciﬁc surface area (50 m

1

) of the particulate metal it,

too, presents a large adsorbent surface available to oxygen species, the

adsorption process removing electrons from the metal. Because the work

function of the metal (5.12 eV) is larger than the electron aﬃnity of the

semiconductor (4.45 eV) electrons will ﬂow out of the semiconductor into the

metal, increasing the depletion layer thickness. This is very likely an over-

simpliﬁcation of what is a complex charge transfer process, and the role of Pt in

inﬂuencing the sensitivity to CO of Sb-doped SnO

is addressed in detail by L.

Morris and D.E. Williams [29].

210 CERAMIC CONDUCTORS

Fig. 4.46 (a) Schematic diagram of a section through a porous SnO

compact with a porous

electrode; (b) detail of the ‘neck’ region between the grains showing the eﬀect of adsorbed

oxygen.

‘Selectivity’ is mainly achieved by ‘trial and error’. It can be engineered

through speciﬁc dopants to the tin oxide, for example Fe

to sensitize for NO

or V

and Pd to sensitize for CO and other combustible gases, as illustrated

below. It can also be engineered by interposing selective diﬀusion barriers

between the sensor and the ambient.

‘Sensitivity’ is a strongly microstructure-dependent property, the depletion

layer thickness relative to the particle size of the sensing ceramic being the

particularly important parameter.

It seems likely from Fig. 4.46 that the greatest sensitivity would be achieved

when the particle size is close to twice the depletion layer thickness. Sensor

ceramics therefore have a particle size typically well into the sub-micron range,

and often in the range 10–100 nm. Figure 4.47 demonstrates the dependence of

sensitivity on grain size, the onset of the steep rise at around 10 nm being

associated with the grain size approximating to the depletion layer thickness.

The now classic ‘Taguchi’ sensor design, dating from the 1960s, is shown in

Fig. 4.48(a) and a present day format in Fig. 4.48(b).

The planar format sensor is most likely to be fabricated by screen-printing, the

sensing layer being typically 20–30 mm thick. The doped tin oxide can be

prepared chemically, for example by precipitating Sn(OH)

, together with

hydroxides of the dopants, by passing NH

through an aqueous solution of

SnCl

together with dopant chlorides. The grain size will depend on the

subsequent calcining temperature and time, typically 800 8C for 1 h.

The screen-printing paste is made up in the manner described in Section 4.2.2

and the patterns printed. The heater element and its leads would typically be

platinum and the interdigitated electrodes gold. As illustrated below (Fig. 4.49)

the heater might also be a composition based on ruthenium oxide (see Section

4.2.2). The contacting tabs are also screen-printed gold.

CERAMICS-BASED CHEMICAL SENSORS 211

Fig. 4.47 Dependence on grain size of the sensitivity (k) of a tin oxide sensor to 100 p.p.m.

hydrogen in air. Adapted from [28]. (k is deﬁned as the sensor resistance measured in air to

that measured in the hydrogen-containing air.)