Zuo-Guang. Ye Advanced Dielectric Piezoelectric and Ferroelectric Materials: Synthesis, Characterisation and Applications

Подождите немного. Документ загружается.

670

22.1 Introduction

Thanks to their high polarisability, ferroelectric materials are very sensitive

to charged defects such as vacancies or substituted ions. In this respect,

oxide perovskites of general formula ABO

display very specific features:

• their ferroelectric transition temperature can be tuned continuously across

room temperature;

• the density of oxygen vacancies that they can host may be tuned on

annealing under well-defined reducing/oxidising atmospheres;

• a large number of heterovalent substitutions can be performed so as to

tune the local charge.

As expected, interfaces play a major role in localising charged defects in

ferroelectrics. The surface and the electrode/ferroelectric boundary are the

first macroscopic locations where charge accumulation can influence dielectric

properties. Mesoscopic space charges can be found in single crystals and

ceramics. In the former case, a well-modulated space charge can be produced

and released by laser irradiation; in the latter, solid-state chemistry is able to

fix charge gradients at the grain boundaries. As a consequence, the overall

dielectric permittivity and conductivity of electroceramics can be tuned. In

some cases, the tuning of dielectric properties through interfaces has met

real world applications. Sometimes, the use of a single ferroelectric compound

is not enough to comply with industry needs. In such instances, composites

made of a ferroelectric material mixed with a simple dielectric are used.

Solid-state chemistry and soft chemistry have been used in order to monitor

the architecture of these composites and thus the density of interfaces even

on the nanoscopic scale.

In Section 22.2, we will recall the basics of charged defects chemistry in

perovkites taking the BaTiO

-type compounds as an example and moving to

the CaCu

double perovskite which has been the focus of a great deal

of research effort in recent years. In Section 22.3, we will address the ionic

Interface control in 3D ferroelectric

nanocomposites

C ELISSALDE and M MAGLIONE,

University of Bordeaux 1, France

WPNL2204

Interface control in 3D ferroelectric nanocomposites 671

concentration gradient, whether controlled or not, in ceramics and composites,

including a ferroelectric phase. We will then describe the most recent routes

that have been used to control the microstructure in micro- and nanocomposites

including a ferroelectric phase.

22.2 Interface defects and dielectric properties of

bulk ferroelectric materials

Charged defect chemistry in ferroelectric perovskites is a long-standing topic

which has stimulated a lot of research with special emphasis on BaTiO

based compounds (Chan et al., 1981; Vollman and Waser, 1994; Smyth

2000). General agreement has been reached when the density of such defects

is small but the heavily doped case is still a matter of debate which is mostly

about the compensation mechanism of charged defects (Morrison et al.,

2001; Smyth, 2002). In this section, we will review the various contributions

of charged defects to the overall dielectric permittivity of perovskites.

22.2.1 Charged defects at electrode/ferroelectric

interfaces

Like any dielectric material, ferroelectrics can be considered as broadband

gap semiconductors at least in their paraelectric state where no polarisation

or ferroelectric domain contribution are to be included. When semiconductor

models apply, the electrode/ferroelectric interface can be considered as a

metal/semiconductor junction (Sze, 1969). Such semiconductor models were

used in the 1960s particularly in the case of Nb-doped SrTiO

which shows

appealing conductivity behaviour (Wemple et al., 1969). Very early on,

numerical models were used to solve the tricky problem of band bending and

charge localisation at the metal/ferroelectric interface (Bardet, 1979). More

recently, such analytical and computer simulations have been greatly renewed

in the context of ferroelectric thin films (Ishibashi, 1990; Stolichnov and

Tagantsev, 1998; Baudry, 1999; Dawber and Scott, 2000). The main result of

these investigations is that the density of charged defects such as oxygen

vacancies can greatly increase the penetration length of the space charge in

the ferroelectric material. With a defect density of 0.1%, a space charge

could be extended to more than 25% in depth of submicrometre films (Baudry,

1999). When going to the ferroelectric state, the depolarising electric field

and the polarisation decay length (Kretschmer and Binder, 1979) have to be

added in the equations which lead to very unusual behaviour. Recently, ab

initio computations have clearly shown that not only does the ferroelectric

material undergo a continuous decrease of the local electric field but a space

charge is also induced in the metal electrode (Junquera and Ghosez, 2003).

This means that even without charged defects in the ferroelectric material,

WPNL2204

Handbook of dielectric, piezoelectric and ferroelectric materials672

space charge must be taken into account. The effect of charge depletion on

the ferroelectric properties of thin films has also been investigated (Bratkovsky

and Levanyuk, 2000].

22.2.2 Space charge relaxation in bulk ferroelectric

perovskites

Experimental probing of space charge at metal/ferroelectric interfaces is not

an easy task since the broadband gap and the low electronic density of states

precludes the use of standard dc current techniques which are efficient in

usual semiconductors. Triggered by the reliability issues of ferroelectric

films, improved experimental methods and models have been used to explain

the dc current behaviours of such devices which can change their polarisation

switching (Waser and Klee, 1992; Dawber and Scott, 2000). Even if all the

issues have not yet been settled, general agreement is being reached as to the

band bending and charged defects at metal–ferroelectric interfaces. One of

the key ingredient is the density of oxygen vacancies which can reach levels

as high as 10

–3

at the very top atomic layers. A long time ago, dielectric

spectroscopy was able to show that such charged interfaces can lead to large

dielectric relaxation, in many ferroelectric perovskites whatever their chemical

composition, morphology and ferroelectric properties. A monodispersive

Debye-type relaxation was recorded in the temperature range 500–100K and

frequency range 1kHz–10MHz, leading to an activation energy of about

1eV (Maglione and Belkaoumi, 1992; Bidault et al., 1994). The close

connection of this activation energy with the one of the static conductivity

showed that macroscopic space charges at the metal/ferroelectric interfaces

are contributing to the dielectric permittivity (Fig. 22.1). Such relaxation has

been evidenced in numerous bulk ferroelectrics, most of them including

heterovalent substitutional cations like BaFe

1/2

(Raevski et al., 2003;

Abdelkafi et al.), as well in thin films.

22.2.3 Charged boundaries in ceramics and

single crystals

Away from the surface and from the electrode/ferroelectric interfaces, charged

defects can be induced in the bulk of ceramics and single crystals. Before

addressing the grain boundary localisation of such charged defects, one has

to recall that the basic equations for bulk ceramics rely on the mass action

law and the charge neutrality conditions (Chan et al., 1981). The charge state

of host lattice ions and dopants may be tuned under annealing and oxidation/

reduction at several oxygen partial pressures. As a consequence the density

of free holes and electrons can be systematically varied. The resulting Kröger–

Vink diagrams in pure and doped BaTiO

have been probed through isothermal

WPNL2204

Interface control in 3D ferroelectric nanocomposites 673

resistance experiments (Chan et al., 1981). This average bulk model for

impurities is refined in the case of ceramics. As was clearly evidenced by

Waser and coworkers, even the most perfect grain boundary between two

grains in a ceramic carries a positive charge (Waser and Klee, 1992; Vollman

and Waser, 1994). As a consequence, negatively depletion layer called space

charge is created in the grains close to the grain boundary. Such a grain

boundary space charge has been probed through time domain relaxation as

well as impedance spectroscopy and its width estimated in Ni-doped SrTiO

through a Schottky model (Vollman and Waser, 1994). One of the main

assumptions in such a model is that the grain boundaries are perfect, which

means that there is a crystallographic continuity between two adjacent

grains. Even if this is close to reality in the most dense ceramics, imperfect

grain boundaries are increasing as the grain size decreases to the nanometre

scale.

There is one case where micrometre size space charges are well controlled

in ferroelectric materials, which is the storage of laser interference patterns

in bulk single crystals: the so-called photorefractive activity of ferroelectrics.

The spatial periodic modulation of the light intensity localises the free charges

in the selected part of the single crystals, leading to strong space charges.

When the input laser beams are in the visible range, the resulting spatial

modulation of space charges takes place on the micrometre size. The effect

of pre-existing charged defects such as iron on the ability of single crystals

to store large space charges has been studied in great detail (Schwartz et al.,

1992). For example, it was only when the density of residual Fe defects

could be brought down to a few ppm that the contribution of intentionally

introduced defects could be investigated (Rytz et al., 1990). The contribution

1/τ(s

–1

)

0.9 1 1.1 1.2 1.3 1.4 1.5 1.6

(K)

–1

–2

–3

–4

σ (Ω

–1

)

22.1

Relaxation time (left scale) and dc conductivity (right scale) in

BaTiO

single crystals on an Arrhenius scale showing the similarity

between the thermal activation of both processes.

WPNL2204

Handbook of dielectric, piezoelectric and ferroelectric materials674

from oxygen vacancy-related centres could be evidenced through electron

spin resonance (ESR) (Vanhorst et al., 1996).

In the two cases reviewed in this section, the space charged localised at

boundaries does not affect the low-frequency dielectric permittivity. We will

show in the next section that grain boundaries are able to artificially increase

this permittivity up to giant effective values.

22.2.4 Grain boundary enhancement of the effective

dielectric permittivity: the case of CaCu

When the density of multivalent cations such as copper is high, as in the

multiple perovskite CaCu

, the probability of increased localisation of

charged defects increases. In ceramics and single crystals of this composition,

a huge dielectric permittivity (up to more than 10

) was recorded with a

sharp decrease in the MHz frequency range (Subramanian et al., 2000). The

thermal activation of this relaxation as well as its link with the inner grain

conductivity showed that the giant permittivity originated in barrier layer at

grain boundaries (Morrison et al., 2001). In this way, the giant dielectric

properties of CaCu

could be ascribed to space charges analogous to

the macroscopic one already mentioned but on a more local scale. In any

instance, the giant permittivity can hardly be attributed to a lattice polarisability

which could result from soft-phonon modes or from ionic dipoles. In this

respect, the recent description of CaCu

in terms of a relaxor model

(Ke et al., 2006) is questionable. On the other hand, this example shows that

the control of charged defects at interfaces is of key relevance to the effective

macroscopic dielectric properties. This opens the route towards the use of

grain boundary engineering as a tool to control the dielectric properties.

Some examples of such engineering will be described in the following sections.

22.3 Interdiffusion in bulk ceramics and composites

As mentioned above, tuning the macroscopic properties of polycrystalline

ferroelectric materials is desirable for a targeted application. From the early

investigations up to now, such improvements largely rely on the bulk chemistry

as well as on the standard shaping of ferroelectric ceramics. Recently, alternative

routes have been proposed as to overcome some intrinsic drawbacks of

ferroelectrics. The first one, which will be described below, is the use of

ferroelectrics for telecommunications in the GHz frequency range. To reduce

the dielectric losses, composites have been proposed that include ferroelectric

and dielectric components. The improved control of these composites on the

nanometre scale is the most recent achievement in this field. The second one

is the use of ferroelectric materials in multilayer capacitors which require

very low-temperature sensitivity, a requirement that is not fulfilled in

WPNL2204

Interface control in 3D ferroelectric nanocomposites 675

ferroelectrics. Composition gradients in individual ferroelectric grains are a

way to decrease their temperature sensitivity. Again, the control of compositions

on the nanoscale is the ultimate way to provide stable and reliable ferroelectric

ceramics.

22.3.1 Interdiffusion in BST/MgO and BST/MgTiO

ceramic composites

Ferroelectric materials are the topic of a great deal of research and development

in the high-frequency range (between 100MHz and 10GHz) due to the

increasing interest in applications in telecommunication and radar fields.

Low dielectric losses are required in order to allow for their integration as

low dimension capacitors in electronic devices and even microwave devices.

Close to their transition temperature, ferroelectrics usually display non-linear

properties under a dc electric field: the permittivity can be easily changed by

applying a dc bias. Such property is used for agile devices such as tunable

capacitors, tunable resonators and phase shifter in radars.

The required performances in the GHz frequency range for integration in

telecommunication electronic devices are: (i) a moderate permittivity (100–

1000), (ii) low insertion losses (tan δ <<1%), (iii) a good thermal stability

and (iv) a high tunability of the dielectric permittivity (>10%). Ferroelectric

materials such as BaTiO

(BT) show high permittivity but strong temperature

sensitivity at temperatures close to phase transition

()

max

′

. Moreover, it

has been shown that the maximal tunability is reached at

′

max

and that

losses increase in the GHz range. The challenge is to find a balance between

all these mismatched specifications.

It is now well recognised that ferroelectric materials with transition

temperatures close to, but below, room temperatures are of key interest. That

is the reason why mixed perovskites such as Ba

0.6

0.4

TiO

(BST) attract so

much interest. BST displays a Curie temperature close to 280K, permittivity

values usually about 2000 at room temperature for micrometre grain ceramics,

high tunability (>10% for a 20 kV/cm field at room temperature) and dielectric

losses higher than 1% up to high frequency (Sengupta and Sengupta, 1999).

To reduce the temperature sensitivity, decrease the permittivity and lower

dielectric losses, ceramic/ceramic composites were processed. Composite

materials such as BST combined with low permittivity and low-loss non-

ferroelectric oxides (MgO, MgTiO

) are excellent candidates for

telecommunications devices (Fig. 22.2) (Weil et al., 1998). Pioneer works in

the field of BST composites with tunable properties are attributed to Sengupta

et al. (1995a,b).

BST–alumina (Liang et al., 2003), zirconia and magnesia ceramic

composites have also been studied. Interesting performances in terms of

tunability and losses were reported. However, the properties improvement is

WPNL2204

Handbook of dielectric, piezoelectric and ferroelectric materials676

limited due to interdiffusion between the ferroelectric and non-ferroelectric

phases (Sengupta et al., 1996; Nagai et al., 2000; Chang and Sengupta,

2002). As a result, the cubic–tetragonal phase transition peak shifted to a

lower temperature, leading to a decrease in both permittivity and tunability

of the composite at room temperature. Such an interdiffusion was also reported

for the BST/MgO and BST/MgTiO

ceramic composites by Nenez et al.

(2002) In this latter case, a secondary phase BaMg

was detected. In

addition, ceramics with nanometre grain size were compared with micrometre

grain size ones. No improvement of the dielectric performance was obtained.

The reduction of the grain size allowed the reduction of the permittivity

values but to the detriment of both tunability and losses (Nenez, 2001).

These results underline the influence of sintering temperature, grain boundary

control and microstructure on the dielectric performance of these composites.

Improvements have been proposed in the literature. As an example, several

authors have reported the preparation of BST/MgO in the forms of thick and

thin films as an interesting alternative to increase the tunability (Sengupta et

al., 2000; Joshi and Cole, 2000). Tunability enhancement was also obtained

by mixing micrometre-size BST with nano-size MgO (Yu et al., 2003). In

order to decrease the synthesis temperature and thus to limit interdiffusion,

sol–gel processes have been also used (Jain et al., 2002; Synowczynski et

Dielectric loss tangent, tan δ

0.0180

0.0160

0.0140

0.0120

0.0100

0.0080

0.0060

Weight percent of non-ferroelectric oxide: 60

300 K

10 GHz

′

tan δ

0.48 0.50 0.52 0.54 0.56 0.58 0.60

Real permittivity,

′

125

120

115

110

105

100

22.2

Measured complex permittivity data for a BST/non ferroelectric

oxide composite as a function of Ba/Sr ratio for BST phase (Weil

et al

., 1998).

WPNL2204

Interface control in 3D ferroelectric nanocomposites 677

al., 2002). However, the downshift of the Curie temperature resulting from

the interdiffusion in such composite materials is still hard to avoid.

Improved control of interfaces in ceramic composites

In the field of nanosciences, it is well established that surfaces and interfaces

play a key role in affecting the properties of materials. Full control of interfaces

is needed to monitor the interdiffusion occuring during the solid-state synthesis

of composites. This can be achieved via a multidisciplinary approach involving

nanoscale wet chemistry, solid-state chemistry and materials science. Our

strategy was to use nanoparticle technology to process ferroelectric/dielectric

composites with controlled interfaces between the two components.

Nanoscience applied to ferroelectrics thus opens great opportunities for the

design of new functional materials for microelectronics and

telecommunications. Another way to decrease the chemical reactivity between

the ferroelectric and the non-ferroelectric phases is to control the microstructure

using an advanced sintering process.

The core–shell concept applied to ferroelectric/non-ferroelectric

composites

The proposed concept is to achieve a network made of individual ferroelectric

particles coated with a dielectric shell (written in the following as core@shell)

with a full control control of both the size of the ferroelectric core (BT, BST)

and the thickness of the non-ferroelectric shell (SiO

). This concept has

already been applied to various inorganic materials from magnetic materials

(Fe

@SiO

) to structural oxides (Al

@SiO

) coated with silica. We

successfully adopted this approach to ferroelectric materials (Huber et al.,

2003; Mornet et al., 2005). Note that enhancement of the BT particle stability

in acidic media by silica coating had already been demonstrated by Shih et

al. (1995). Our method based on seeded growth process was developed to

coat ferroelectric particles (from 50nm to 1µm size) and has allowed the

control of the silica shell thickness from 2 to 80 nm with an accuracy of

1–2nm (Fig. 22.3).

In order to improve both the microstructure and the dielectric performance

of the final composites, we took advantage of the self-assembly properties of

size-sorted BST, BT@silica particles. Thanks to the silica shell, SEM

observations show neck formation at only 1000°C (Fig. 22.4). Reliable

dielectric characterisation was thus possible on as-calcined colloidal

composites. The ferroelectric transition temperature was kept in the final

composite, the dielectric permittivity reached expected values and the dielectric

losses were very low and stable in the temperature range of 100–400K (Fig.

22.5). This last result confirmed that the silica shell acts as a dielectric loss

WPNL2204

Handbook of dielectric, piezoelectric and ferroelectric materials678

barrier while keeping the intrinsic properties of the ferroelectric cores, i.e.

interdiffusion is avoided, which was difficult to achieve in standard ferroelectric/

dielectric composites. Sintering studies on nano-sized BaTiO

(30nm) coated

with silica showed complex densification behaviour as a function of SiO

content (Park and Han, 2005).

The core–shell approach was also investigated using a solvothermal synthesis

route. BaTiO

nanoparticles were coated with an amorphous Al

shell of

few nanometres using a chemical fluid deposition process in supercritical

state. After sintering of the core–shell nanoparticles, the final ceramic material

exhibited significant improvement in the lowering of dielectric losses while

keeping the Curie temperature unchanged (Fig. 22.6) (Aymonier et al., 2005).

More recently, a modified hydrothermal process was developed to

synthesized BST nanoscaled powder (17–20nm) and a core–shell structure

was formed by self-wrapping with MgO oxide. However, a radial composition

Measured silica thickness (nm)

100

0 20 40 60 80 100

Calculated silica thickness (nm)

22.3

Thickness evolution of the silica shell as determined by

transmission electron microscopy (TEM) as a function of the

calculated thickness.

22.4

SEM micrographs of BST@silica composites as obtained after

sintering at 1000°C for 3h of size-sorted BST–silica nanoparticles.

WPNL2204

Interface control in 3D ferroelectric nanocomposites 679

gradient from the shell to the core was revealed and no dielectric investigation

was performed (Tian et al., 2005).

Ferroelectric composites obtained by non-conventional sintering method

One of the problems in the fabrication of BST/MgO composites via the

conventional route is the high sintering temperature required to densify the

ceramics due to the refractory character of MgO (T

sint

> 1600°C). As a result

of high thermal treatment, grain growth and interdiffusion between the two

components are hard to avoid. To overcome these drawbacks, non-conventional

sintering processes which provide rapid heating, short sintering time and

low sintering temperature are particularly well suited.

The BST/MgO composites fabricated by microwave processing were

reported by Agrawal et al. (2004) (Fig. 22.7). The aim of this study was to

obtain a well-defined phase boundary and a fine grain size in the sintered

composites by lowering the chemical reactivity. Thanks to uniform connectivity

obtained between the BST grains, enhanced tunability was reached (maximum

′

100 k

1 MHz

0 100 200 300 400 500

(K)

tan δ

0.015

0.010

0.005

100 kHz

1 MHz

0 100 200 300 400 500

(K)

22.5

Dielectric permittivity and losses versus temperature measured

at 100kHz and 1MHz on sintered colloidal BST@silica (5nm)

composite.

WPNL2204