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

Подождите немного. Документ загружается.

preferentially between 750 and 850 8C whilst perovskite-type structures are formed

between 850 and 950 8C. A processing route bypassing the formation of pyrochlore

has been developed for PMN [23]; MgO is pre-reacted with the Nb

to form

MgNb

which is then reacted with PbO to form the perovskite phase.

A composition approximating to Pb

0.85

0.1

0.2

0.8

2000) is only

signiﬁcantly aﬀected by d.c. ﬁelds above 2 MV m

, as shown in Fig. 5.48. A

number of factors contribute to the stability:

1. the grain size (about 1 mm),

2. the relaxor characteristics,

3. a possible antiferroelectric structure between 735 and +65 8C and

4. T

at 70 8C due to the combined eﬀects of the zirconium and lanthanum

contents.

Factor 3 may account for the maintenance of the e

level at low temperatures

and the small eﬀect of d.c. ﬁelds up to 60 8C, and 1 and 2 contribute to a

322 DIELECTRICS AND INSULATORS

Fig. 5.47 Eﬀect of d.c. bias (MV m

) on the dielectric properties of PbFe

0.55

0.1

0.35

TEAMFLY

Team-Fly

smoothing out of the peaks which might otherwise occur at 740 and +70 8C.

Factor 2 has the marginal disadvantage of introducing frequency dependence of

and tan d below 0 8C. Tan d measured at 100 kHz rises to above 0.04 at

720 8C.

The composition sinters at 1100 8C which allows some reduction in the

palladium content of the Ag–Pd electrodes and reduces the volatility of PbO. The

resistivity is very high, leading to a leakage time constant greater than 10

sat

room temperature and about 10

s at 125 8C. There are very few signs of

degradation under high d.c. ﬁelds up to 150 8C.

The La

ion has r

¼132 pm and r

¼106 pm and appears to be able to

occupy either A or B sites, although a higher lead content may be needed to give

appreciable B-site occupancy. The high resistivity and resistance to degradation

may be due to the lowering of the concentration of vacant oxygen sites by the

replacement of Pb

by La

, accompanied by La

on B sites acting as

acceptors for electrons and compensating the La

ions on A sites.

5.7.3 Multilayer capacitors with base metal electrodes (BME)

The multilayer structure discussed in Section 5.4.3 is applicable to all ceramic

dielectrics. It enables the thinnest dielectric plates to be engineered into robust

units that are readily attached to the substrates carrying the electronic circuitry

by ‘surface mount technology’ (SMT). The structure also gives minimum

electrode inductance permitting use at high frequencies. Very signiﬁcant progress

has been made over the years in reducing the thickness of both dielectric and

electrode layers and so increasing volumetric eﬃciency. An appreciable part of

the cost of a multilayer capacitor lies in the palladium-containing electrodes.

HIGH-PERMITTIVITY CERAMICS 323

Fig. 5.48 Dependence of e

on T and d.c. bias (——, 0; - - - -, 2 MV m

; ---- ----, 4 MV m

)

for Pb

0.85

0.1

0.2

0.8

(PLZT). (After G.H. Maher (1983) 33rd Electron. Comp. Conf.

(IEEE) 173–7.)

With reductions in ceramic sintering temperature, the cost of electrodes has

been reduced by the use of Ag–Pd alloys, typically 70 at.% Ag/30 at.% Pd. The

reduction in palladium content is limited to about 15 at.% because at lower levels

silver migration becomes a problem. The drive to lower cost has led to eﬀorts to

avoid it entirely. One approach is based upon the ‘injected’ or ‘fugitive’ electrode

(see Section 5.4.3 and Fig. 5.12) and the other on the use of the base metals nickel

or copper for the electrodes (BME) combined with dielectrics resistant to the

reduction ﬁring necessary to retain the electrodes in the metallic state.

Although the fugitive electrode approach is viable it is presently little used, if

at all. In contrast BME technology presently accounts for about 50% of MLCC

production.

When BaTiO

is ﬁred under reducing conditions oxygen is lost from the lattice

with the formation of doubly ionized oxygen vacancies and electrons in the

conduction band (see Section 2.6.2, Eq. (2.48)). On cooling to room temperature

a large fraction of the electrons may remain in the conduction band because the

ionization energies of V

and V

are low compared to kT at room temperature

(see Fig. 2.15). Acceptor ions such as Mn

on Ti

sites also give rise to oxygen

vacancies but without the liberation of electrons. They have the eﬀect of shifting

the minimum in the conductivity–oxygen pressure relation to lower oxygen

pressures (Fig. 5.49) so that the concentration of electrons in the conduction

band is reduced and positive holes in the valence band become the majority

carriers at oxygen pressures above 0.01 Pa (Eq. (2.53)). On cooling to room

temperature the positive holes combine with the ionized acceptors since the

ionization energies for acceptors lie in the range 1–2 eV. High-resistivity BaTiO

therefore requires both a suﬃcient concentration of acceptors and a suﬃcient

pressure of oxygen during ﬁring.

Technical BaTiO

normally contains suﬃcient acceptors in the form of Al, Fe

etc. to render it insulating after ﬁring at atmospheric pressure, but it requires

324 DIELECTRICS AND INSULATORS

Fig. 5.49 Eﬀect of Mn

on the conductivity–oxygen pressure relation for BaTiO

at 965 8C.

additional acceptors if it is to be sintered at lower oxygen pressures. The addition

of 0.2–0.5 mol.% Mn forms suﬃcient Mn

and Mn

acceptor ions on Ti

sites on ﬁring at oxygen pressures below 10

Pa to result in an insulating

dielectric at room temperature.

The oxidation reaction for nickel is

Ni þ

ðgÞ!NiO ð5:46Þ

and the standard free-energy change at 1300 8C(DG

at 1573 K) is approximately

7100 kJ. If the activities of Ni and NiO are set equal to unity, the reaction

constant K is

K ¼ p

1=2

ð5:47Þ

and since DG

1573

¼RT lnK ¼RT lnp

1=2

¼ exp



2DG

1573



ð5:48Þ

 2.5 10

7

bar ð10

2

PaÞ

The oxygen potential must be kept below this value at the ﬁring temperature

(about 1300 8C) if the nickel electrodes are to remain metallic. This is achieved by

ﬁring in nitrogen carrying a controlled H

O/H

mixture. An oxygen meter (see

Section 4.6.1) is used to monitor the oxygen potential in the ﬁring zone.

HIGH-PERMITTIVITY CERAMICS 325

Fig. 5.50 Resistivity at room temperature as a function of oxygen pressure during sintering:

curve A, undoped BaTi

0.85

0.15

; curve B, (Ba

0.85

0.15

1.01

–(Ti

0.9

0.1

. (After [24])

The titanate can also be stabilized against reduction by the substitution of some

calcium for barium, and by establishing the ratio AO/BO

41. A suitable

composition is [(Ba

0.85

0.15

)O]

1.01

(Ti

0.9

0.1

, and its stability against ﬁring

under reducing conditions is compared with that for Ba(Ti

0.85

0.15

in Fig. 5.50

[24].

The success of the formulation is probably due to the presence of Ca

ions on

B sites acting as acceptors (Section 5.7.1). Certainly an AO/BO

ratio greater than

unity has been found to be essential. CaZrO

is itself very resistant to reducing

conditions, and the presence of Ca and Zr ions will confer some measure of

increased resistance to reduction on the composition as a whole. The loss of lattice

oxygen resulting from the reduction ﬁring, unless compensated for, has the

expected adverse eﬀect upon degradation of BME MLCCs and the need to

control this has stimulated research into what is undoubtedly a very complex

matter.

The BME process cannot be applied to compositions containing lead because

of the ease of reduction to metallic lead.

5.7.4 Barrier-layer capacitors (Class IV)

Although barrier layer capacitors are of little or no commercial signiﬁcance, the

principles upon which their manufacture and operation are based, are important.

Most materials containing TiO

, whether as a single phase or in combination

with other oxides, become conductive on ﬁring in reducing atmospheres. The

ease of reduction is strongly aﬀected by the other ions present: acceptor ions tend

to inhibit reduction and donor ions tend to enhance it. In most cases a high

resistivity can be restored by annealing in air or oxygen. The mechanism of

conduction is discussed in Section 2.6.2.

Barrier-layer capacitors are based on the limited reoxidation of a reduced

composition. This results, in the simplest case, in a surface layer of high

resistivity and a central portion of conductive material so that the eﬀective

dielectric thickness is twice the thickness h

of a single reoxidized layer and there

is an apparent gain in permittivity over that of a fully oxidized unit by a factor of

h/2h

, where h is the overall dielectric thickness (Fig. 5.51). Alternatively each

326 DIELECTRICS AND INSULATORS

Fig. 5.51 Schematic diagram of a partially reoxidized dielectric.

conductive grain may be surrounded by an insulating barrier layer so that the

dielectric property is dispersed throughout the ceramic.

Reoxidized layer units

Units depending on a reoxidized surface layer are generally made by ﬁring

BaTiO

or SrTiO

discs, approximately 0.5 mm thick, under reducing conditions.

A silver electrode paint is applied to the surfaces of the disc and ﬁred on at about

800 8C. The silver paint contains a PbO–Bi

–B

glass frit to which is added a

small amount (about 1 cat.%) of acceptor ions, e.g. Cu. This leads to the

formation of a thin (about 10 mm) insulating layer separating the electrodes from

the semiconducting titanate and to an associated very high capacitance. Because

the major part of a voltage applied to the capacitor is dropped across the two

thin dielectric layers, the working voltage is low, typically about 10 V.

Capacitance values per unit electrode area of approximately 10 mF m

are

readily achieved.

Internal barrier layers

The thinnest reoxidized layers, which result in very large eﬀective permittivities,

have properties similar in some respects to those of varistors as described in

Section 4.3. They contain Schottky barriers in the semiconducting surfaces of the

grains which result in properties similar to those of two back-to-back diodes (see

Fig. 2.22). Their working voltages are therefore limited to the range within which

the current is low. In order to withstand higher voltages it is necessary to have a

ceramic structure that comprises a number of such barrier layers in series

between the electrodes. There must also be an intergranular component that

allows the diﬀusion of oxygen and dopant ions to the crystallite surfaces during

oxidation.

The grain size in these units averages about 25 mm. Crystallites smaller than

about 10 mm have a large fraction of their volume taken up by Schottky barriers

and associated space charges that increase their resistivity, and if the crystals are

larger than 50 mm there is the possibility that only a few crystals will separate the

electrodes in some places, resulting in a greater likelihood of breakdown.

Additions of small amounts (about 1%) of silica and alumina provide an

intergranular layer that allows ionic movement and access to oxygen at high

temperatures. Dysprosium or other donor ions are added to assist in the

reduction process.

Discs or other shapes are ﬁrst ﬁred in air to remove organic matter and then

sintered in air to obtain the required level of crystal growth. A reducing

atmosphere of carbon monoxide or hydrogen is then introduced. This is found to

HIGH-PERMITTIVITY CERAMICS 327

inhibit crystal growth and so cannot usually be combined with the sintering stage.

After cooling, a boric oxide frit containing acceptor ions such as Cu, Mn, Bi or Tl is

painted on the surfaces of the pieces which are then reheated in air to 1300–

1400 8C. The acceptor ions diﬀuse along the grain boundaries and modify the

surface properties of the crystallites in much the same way as the acceptors that

protect dielectrics for base-metal-electroded multilayer capacitors. However, their

precise behaviour has not been established since it is extremely diﬃcult to

determine the structure of the thin intergranular layers and their interfaces with the

crystallites.

Model calculations

The structure of internal-barrier-layer units can be represented by the simpliﬁed

model shown in Fig. 5.52. The overall capacitance can be calculated as follows.

The capacitance C

of an individual element, assuming t

 t

, is given by

ð5:49Þ

and that of a series connected column by

=ðno. of elements in columnÞ¼

t=t

ð5:50Þ

328 DIELECTRICS AND INSULATORS

Fig. 5.52 Schematic diagram of a section through an internal-barrier-layer capacitor.

The capacitance C per unit area is

C ¼

ð5:51Þ

since there are 1/t

columns per unit area. It follows that the eﬀective relative

permittivity e

of the composite dielectric is

ð5:52Þ

Assuming t

¼ 50 mm, t

¼ 0.2 mm and e

¼ 200, we obtain e

¼ 50 000. Values

as high as this can be achieved in practice with tan d values of typically 0.03.

Units based on SrTiO

are more stable with respect to ﬁeld and temperature

than those based on BaTiO

. Their capacitance is only reduced by 5% under

maximum d.c. ﬁeld and their variation with temperature can be kept within

+20% over a 720 to +85 8C range; their eﬀective permittivity is 10 000–20 000.

BaTiO

units have eﬀective permittivities of up to 50 000.

5.7.5 Ferroelectric memories

No text covering electroceramics can omit some discussion of ‘ferroelectric

memories’ even though the engineering of the ceramics into the memory elements

of a computer is so highly specialized that it is arguably more properly dealt with

in the context of semiconductor technologies. Certainly the topic demonstrates

very clearly the interdisciplinary nature of electroceramics. The ‘materials’

aspects – especially the production of very well deﬁned precursors for fabricating

thin ferroelectric ﬁlms, characterizing them and improving understanding of

their basic physics are pursuits well within the ceramist’s ﬁeld of interest. There

are evident common interests between those working to maximize the volumetric

eﬃciency of a multilayer capacitor by reducing dielectric layer thickness and

those engaged on fabricating and optimizing the properties of thin ferroelectric

ﬁlms for memories.

The topics very brieﬂy touched upon here are dealt with in greater depth in the

monograph by J.F. Scott [25] which is recommended to those wishing or needing

to pursue the subject; an apt starting point is to quote from the introduction to

the monograph.

Since ferroelectricity was discovered in 1921 it has been obvious to many scientists

and engineers that the two stable polarization states þP and P could be used to

encode the 1 and 0 of the Boolean algebra that forms the basis of memory and

logic circuitry in all modern computers. Yet until very recently this has been

unsuccessful. In fact, although ferroelectric materials are used in a wide variety of

commercial devices, it has until now always been the case that some other

property of the material – especially pyroelectricity or piezoelectricity – is the

characteristic actually employed. Ironically, no devices using ferroelectrics have

actually required ferroelectricity to work.

HIGH-PERMITTIVITY CERAMICS 329

For non-volatile random access memories (NV-RAMs; in which the stored

information is retained even if power to the chip is interrupted), ferroelectrics

serve not just as capacitors (as in the case of the DRAM, described below), but as

the memory element itself. Their principal advantages in this application are low-

voltage (1.0 V) operation, small size (about 20% of a conventional EEPROM*

cell – and cost is proportional to size once high-yield production is achieved),

radiation hardness (not just for military applications but also for satellite

communications systems) and very high speed (60 ns access time in commercial

devices, sub-nanosecond in laboratory tests on single cells).

Additional advantages oﬀered by a NV-RAM over a conventional EEPROM

are fast write-access (530 ns) and high number of write cycles (410

The principle on which a computer memory is based is illustrated in Fig. 5.53.

The binary-coded information is stored in an array of cells addressed by

voltage pulses applied to the ‘word’ and ‘bit’ lines. The response of a particular

cell, measured by the associated electronic circuitry, determines its state.

Figure 5.53(a) indicates the situation for a ‘passive’ FeRAM (ferroelectric

random access memory). If the capacitor is in the þP state (say a ‘1’) and the

application of two V/2 voltages is suﬃcient to switch the polarization to the P

state, there is a resulting ‘large’ current pulse. However were the cell in the P

state (a ‘0’) then the current pulse would be very much smaller. If the state of the

cell is such that the polarization is switched then a follow-on pulse must be

applied to return it to its original state.

In practice there are complications because the ferroelectric elements do not

have a suﬃciently well deﬁned switching voltage. This can lead to unintentional

switching of elements other than the one addressed and, consequently, to an

unacceptable error rate. The problem can be overcome by coupling each memory

element with a transistor which provides the sharply deﬁned switching voltage.

In its simplest form each node of such an ‘active’ matrix would then take the

form shown in Fig. 5.53(b). In fact, for a variety of reasons associated with the

changes which occur in the physical properties of ferroelectrics with time, the

transistor-memory cell circuitry at each node is more complex than outlined

above, but the general principles are the same.

In the case of dynamic random access memories (DRAMs) the cell is a thin

ﬁlm capacitor. The state of the cell (‘0’ or ‘1’) is read by the current pulse

following an addressing voltage pulse. Because the charge stored in the capacitor

leaks away in time (t ¼ RC) then, for the information to be retained it must be

periodically refreshed, the rate of leakage determining the necessary refresh

interval.

In the conventional DRAM the capacitor dielectric is SiO

 5) or a

combination of SiO

and Si

(‘ONO’ ﬁlm) and is formed by controlled

oxidation or nitridation of the silicon crystal. It is necessary for the memory

capacitor to have a capacitance value of approximately 30 fF (i.e. 3010

15

330 DIELECTRICS AND INSULATORS

* EEPROM, electrically erasable programmable read-only memory – a ‘FLASH’ memory.

and this places a lower limit on its size. A lower limit on the thickness of the

dielectric is necessary to retain adequate resistance to electrical breakdown

arising from ‘tunnelling’ currents. Capacitor size, and therefore component

density, can be reduced by increasing the permittivity of the dielectric. A

signiﬁcant reduction in capacitor size could be achieved by replacing the silica

with Ta

 25) and this step seems likely in the foreseeable future. If a

‘ferroelectric’ capacitor (e

of order 10

to 10

) could be used then a bit density

increase of about ten-fold might be achieved. This is the stimulus for

considerable research and development eﬀort, especially on SrTiO

4 200)

and Ba

ð1xÞ

TiO

e.g. Ba

0.7

0.3

TiO

4 300), the composition being chosen

so that the material is in the paraelectric phase at room temperature.

For FeRAMs materials attracting interest are PbZr

ð1xÞ

(with x  0:6)

and the Aurivillius layer-structured perovskite, SrBi

(SBT).

HIGH-PERMITTIVITY CERAMICS 331

Fig. 5.53 Schematic of FeRAM matrix: (a) a matrix of memory cells (b) each memory cell

coupled with a switching transistor to more closely deﬁne the threshold switching voltage.

(a)

(b)