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

Подождите немного. Документ загружается.

Handbook of dielectric, piezoelectric and ferroelectric materials700

TEM samples were fabricated by standard mechanical and ion milling

procedures. Plan view and two types of cross-section TEM samples were

prepared from the PZT/SRO/STO heterostructures. The cross-sections were

cut in such a way that the electron beam was incident to the TEM sample

either from the [010] or the [110] direction. Conventional TEM was performed

in a Philips CM20T electron microscope at 200keV primary energy of the

electrons, and high resolution (HR)TEM in a Jeol 4010 high-resolution electron

microscope at 400keV. Some of the HRTEM images were Fourier-filtered to

reveal the nature of the observed defects. Figure 23.2(a) shows a cross-

section TEM micrograph of a PZT/SRO/STO heterostructure, seen from the

[010] direction. The tetragonal film was c-axis oriented, as shown by the

electron diffraction pattern (Fig. 23.2e), but contained narrow 90°

a-domains bounded by {101} twin planes (90° a–c boundaries) (see Fig.

23.2a, where the tilted white arrows are inserted). Typically two types of

vertical TDs were discovered in the PZT film. There exist pairs of partial

dislocations (partial dislocation dipoles) that run across the entire layer (Alpay

et al., 1999; Misirlioglu et al., 2006a) (indicated by the black arrows in Fig.

23.2a) and half-loops (Matthews, 1975; Fitzgerald, 1991; Suzuki et al., 1999;

Sun et al. 2004) (indicated by the white double-arrow in the figure) that

come from the top surface and end in the bulk of the PZT film. The TD

dipoles that cross the whole PZT layer go through the a-domains as well,

which indicates that the TD dipoles are already formed at the growth

temperature (Foster et al., 1996b; Alpay et al., 1999; Misirlioglu et al.,

2006a). The twin domains start being formed during the cooling of the

heterostructure through the cubic-to-tetragonal structural phase transition

(Foster et al, 1996a,b). The temperature at which the PZT layers are usually

0 1.00 2.00

µm

2.00

1.00

0 1.00 2.00

µm

2.00

1.00

23.1

Atomic force microscopy (AFM) images taken on 2µm × 2µm

areas on the surface of (a) a 70 nm thick step-flow grown SRO film

and (b) a 250nm thick defective PZT 20/80 film.

WPNL2204

Single crystalline PZT films 701

PZT

SrRuO

SrTiO

150 nm

100 nm

b50 nm

50 nm

002

STO

002

PZT

100

PZT/STO

23.2

Cross-section (a, b, seen from [010]) and plan view (c, d, seen

from [001]) TEM micrographs of PZT/SRO/STO (001)

heterostructrures. In (b) a detailed view is shown, and the circle

points out that the half-dislocation loop pins a 90°

-domain. The

plan view TEM image in (c) reveals the 90° domains, and the dark

spots are the terminations of threading dislocations on the PZT top

surface, which can be better seen in (d). In (e) a contrast-inverted

cross-sectional electron diffraction pattern (beam direction [010]),

correctly rotated with respect to (a), is given. Three reflections are

indexed, on the bottom right 45°-oriented streaks due to the habit

planes of the 90°

–

-boundaries are seen (from Vrejoiu

et al.

, 2006a,

http://www.tandf.co.uk/journals).

WPNL2204

Handbook of dielectric, piezoelectric and ferroelectric materials702

grown, T

, significantly exceeds the phase transition temperature for the

bulk PZT 20/80 (T

≈ 450°C). It can be seen that sometimes the half-loop

dislocations may pin a 90° domain (Fig. 23.2b).

Plan view TEM micrographs are shown in Fig. 23.2(c) and (d). 90° domains

(partly etched away during the TEM sample thinning process) are visible in

Fig. 23.2(c), as well as black spots that are spread over the entire viewed

area. These spots can be better seen in the higher magnified image, given in

Fig. 23.2(d). They turned out to be dislocation dipoles with a typical separation

distance of 10–50 nm. To further elucidate the nature of these threading

dislocation dipoles, (110)-Fourier filtered plan view HRTEM images were

taken. Figure 23.3 shows such an image. It reveals that the TD dipole consists

of two partial edge dislocations of equal in-plane Burgers’ vector component,

= (a/2)[110] (Matthews, 1975; Fitzgerald, 1991; Wunderlich et al., 2000)

(0.28 nm), but of opposite sign, separated by a distance of about 15nm. The

vector (a/2)[110] represents half of a lattice vector, in view of the primitive

cubic unit cell of PZT. As indicated by cross-sectional HRTEM of a single

half-loop TD (Fig. 23.4a,b), half-loop TDs have an out-of-plane Burgers’

vector component b

⊥

= (c/2)[001] and thus represent partial dislocations, too.

Although neither the half-loops nor the dipoles have been entirely characterized

in terms of their Burgers’ and line vectors, we may state that partial in-plane

Burgers’ vector components b

= (a/2)[110] and partial out-of-plane components

⊥

= (c/2)[001] have been detected.

Misirlioglu et al. (2006a) studied the microstructure of epitaxial PZT

20/80 films of ~300nm thickness grown directly on (001) SrTiO

by pulsed

laser deposition. Their TEM observations revealed that the films were

PZT (110) planes

5 nm

23.3

(110)-Fourier filtered plan-view HRTEM of a single TD dipole,

with a separation distance of about 15nm (from Vrejoiu

et al.

, 2006a,

http://www.tandf.co.uk/journals).

WPNL2204

Single crystalline PZT films 703

predominantly c-oriented with a

- and a

-oriented domains (Li et al., 2005)

lying on {101} planes. The substrate/film interface contained arrays of edge-

type misfit dislocations and there was a high density of TDs (>> 10

–2

Dai et al. (1996) showed that interface dislocations in BaTiO

/LaAlO

were

directly linked up with the 90° domain boundaries in BaTiO

and thus the

domain boundaries were likely to be pinned at the dislocation site. Lu et al.

(2003a) studied the defect structure of a 350nm-thick epitaxial Ba

0.3

0.7

TiO

film grown on (001) LaAlO

and revealed that the dominant defects in the

BSTO film were edge-type TDs with Burgers’ vectors b = <100> and <110>.

Diffraction contrast TEM and HRTEM images of the PZT/SRO/STO

heterostructures on which Vrejoiu et al. (2006a) reported showed that the

partial dislocation dipoles include a stacking fault (SF) in between (Matthews,

1975; Suzuki et al., 1999; Wunderlich et al., 2000; Lu et al., 2003a,b), as

seen in Fig. 23.5(a) and (b). The black arrow indicates the location of one of

the numerous SFs. Figure 23.5(c) is a HRTEM image seen from the

[110]

direction, so that the (001) PZT planes and the SF are visible edge-on. This

image was taken on a very thin sample area (weak phase object) at Scherzer

defocus (–40nm), which allows one to assume that cations are visualized as

dark spots (‘intuitively interpretable image’, cf. van Dyck, 1997). In view of

these imaging conditions, the very dark elongated contrast spots, which

occur along the stacking fault, can most probably be assigned to pairs of lead

cations. Thus the stacking faults are most probably lead-rich. An unambiguous

allocation of the elongated contrast spots to lead ions would, however, require

some detailed contrast simulation work; thus the possibility of a titanium-

rich stacking fault, as described by Lu et al. (2003b), cannot be entirely

10 nm

Seen from [010]

23.4

HRTEM of a single threading dislocation half-loop, with a

separation distance of about 20nm, seen from the [010] direction in

(a), and a detailed view in (b). In (a) the white line is added as a

guide to the position of the dislocation. The black boxes and arrows

in (b) designate the vertical position of the unit cells on the left and

right of the stacking fault (from Vrejoiu

et al.

, 2006a, http://

www.tandf.co.uk/journals).

WPNL2204

Handbook of dielectric, piezoelectric and ferroelectric materials704

excluded. In any case, the stacking fault should be metal-rich in nature. In

Fig. 23.6, a schematic model is given which refers to the possibility of a

lead-rich stacking fault. This model involves an in-plane displacement vector

component R

= (a/2)[110] (schematic top) and an out-of-plane displacement

vector component R

⊥

= (c/2)[001] and is thus in accordance with the

experimental findings for the respective Burgers’ vector components. The

changed linking of the oxygen-octahedra at the stacking fault is equivalent

to an oxygen deficiency (see Fig. 23.6, schematic seen from

[110]

direction).

In our case, the dislocation loops may form by aggregation of point defects

(i.e. oxygen vacancies) (Matthews, 1975; Fitzgerald, 1991; Suzuki et al.,

2000). Because the growth of the thin film often occurs under non-equilibrium

and high supersaturation conditions, as it is certainly the case for typical

PLD conditions (Chrisey and Hubler, 1994; Bäuerle, 2000; Willmott and

Huber, 2000), it can be anticipated that an excess of vacancies or gas atoms

could be trapped in a growing film. These point defects could aggregate to

form dislocation loops that would subsequently grow and coalesce to form

25 nm

200 nm

Seen from [110]

(001) PZT

Seen from [010]

Seen from

[11 0]

23.5

Threading dislocation dipoles with a stacking fault (SF),

indicated by the black arrow. (a) Diffraction contrast cross-section

TEM image (seen from [110] direction) and (b) cross-section HRTEM

micrograph (seen from [110] direction). A HRTEM micrograph (seen

from

[110]

direction), where the (001) PZT planes and the stacking

fault are seen edge-on, is given in (c) (from Vrejoiu

et al.

, 2006a,

http://www.tandf.co.uk/journals).

WPNL2204

Single crystalline PZT films 705

dislocations threading the thin film. Therefore, it is likely that non-stoichiometry

of the growing film could lead to dislocations being produced by the mechanism

under discussion (Matthews, 1975).

The PZT target we employed for PLD, although of Pb

1.1

(Zr

0.2

0.8

nominal composition, may have already been oxygen-deficient, as it is

suggested by its dark grey colour, and this may have contributed to the

slightly non-stoichiometric deposits. The target was deliberately purchased

with 10% PbO excess, as a precaution for the possible loss of PbO during the

deposition, because PbO is highly volatile at the usually elevated growth

temperature. Moreover, the correct oxygen content of the growing film was

23.6

A possible schematic model of the stacking fault seen from

[110]

direction: along the SF edge, oxygen ions are missing and

metallic Pb atoms are paired. The small top schematic is seen from

the [001] direction (from Vrejoiu

et al.

, 2006a, http://www.tandf.co.uk/

journals).

–Pb ‘left’ lattice

–Pb ‘right’ lattice

[110]

[001]

Seen from [001]

Seen from [110]

WPNL2204

Handbook of dielectric, piezoelectric and ferroelectric materials706

expected to be ensured by the background oxygen atmosphere, as the

depositions were performed in 150–200 mtorr O

. As mentioned in Section

23.1, PLD is known to be a deposition technique that, under favorable

conditions, allows for stoichiometric transfer of the material from the bulk

target to the growing thin film (Chrisey and Hubler, 1994; Bäuerle, 2000;

Willmott and Huber, 2000). However, we would expect that small deviations

from the nominal composition are likely to be exhibited by these PZT films.

Electron-energy loss spectroscopy (EELS) and energy dispersive X-ray (EDX)

analyses performed along the entire cross-section of the PZT layers ascertained

that the Ti/Zr ratio stays within 4 ± 1.

Suzuki et al. (2000) reported on the effect of non-stoichiometry on the

microstructure of epitaxial BaTiO

thin films grown by PLD. They varied

the cation non-stoichiomety over a wide range (up to 50% excess of either Ti

or Ba) and observed the modification of the microstructure. Based on the

observation that the defect structures were characteristic to the titanium-

excess films and were not seen in stoichiometric films (Suzuki et al., 1999),

the authors drew the conclusion that the dislocation cores could act as a sink

for insoluble excess cations and for relevant point defects such as barium

and oxygen vacancies. In titanium-excess films (BaTi

1.2

3.4

), planar defects

were found and described as being terminated by a pair of partial dislocations

with projected Burgers’ vectors of type a/2 [110]. These planar defects with

the (001)-projected shear vector component of a/2 [110] formed the so-

called crystallographic shear (CS) structure described by the (100) stacking

irregularity of either double TiO

-layers or double BaO-layers, closely related

to non-stoichiometry. Lu et al. (2003b) argued that the formation of stacking

faults with a double layer of edge-sharing TiO

-octahedra in epitaxial

0.3

0.7

TiO

films was probably related to a small amount of Ti excess

during the PLD growth of the film as well.

In the case of Aurivillius-phase ferroelectrics, such as SrBi

and

SrBi

, bismuth volatility makes the growth of phase-pure films difficult

(Zurbuchen et al., 2003). Loss of Bi during the slow cooling of a SrBi

film resulted in a 5% Bi-deficient film and the generation of a high density

of out-of-phase boundaries (OPBs). These OPBs, by symmetry, were

demonstrated to be ferroelectrically inactive. Although the OPBs did not

represent a considerable volume of material, the film had severely reduced

ferroelectric properties compared with stoichiometric films, even though it

appeared to be of very good quality by XRD. This could explain why many

reported films of apparent high quality by XRD have such poor ferroelectric

properties.

Another observation related to the formation of TDs starting from the

PZT/SRO interface shows that the TDs may originate on particulates, as it

can be seen in Fig. 23.7. This can be a source for heterogeneous nucleation

of dislocation dipoles (Fitzgerald, 1991). The particle ‘P’ (10–15nm in diameter)

WPNL2204

Single crystalline PZT films 707

appeared to be amorphous (better visible in the top left inset of Fig. 23.7)

and may be either an impurity, a particulate ejected from the PLD target or

formed on the SRO surface at the initial stage of the PZT layer growth. It is

assumed that the presence of such a particle or impurity results in stress

concentrations which can extend into the epilayer.

23.3.2 Piezo-force microscopy investigations on defective

epitaxial PZT 20/80 films

Piezo-force microscopy (PFM) investigations, performed on that PZT/SRO/

STO heterostructure whose TEM is given in Fig. 23.5, revealed some intriguing

aspects regarding the switching of 180° ferroelectric domains (Vrejoiu et al.,

2006a). Figure 23.8a shows the surface topography of a 2µm × 2µm area of

the as-grown PZT film. The two-dimensional grid exhibited by the surface

confirmed that 90° twin a–c domains (Ganpule et al., 2000; Le Rhun et al.,

2006) were formed in the PZT layer. The piezoresponse image of the as-

grown PZT film, associated with the topography image is given in Fig.

23.8(b). The piezoresponse image confirms the existence of 90° domains,

with the dark lines (forming a grid) representing the a-domains (i.e. domains

with the polar c-axis lying in the plane of the film) (Ganpule et al., 2000;

10 nm

PZT

SRO

20 nm

Seen from [010]

23.7

Cross-section HRTEM micrograph revealing that a threading

dislocation dipole may nucleate on an amorphous particle ‘P’ at the

PZT/SRO interface that can be better seen in the inset (upper left

corner) (from Vrejoiu

et al.

, 2006a, http://www.tandf.co.uk/journals).

WPNL2204

Handbook of dielectric, piezoelectric and ferroelectric materials708

Roelofs et al., 2002; Le Rhun et al., 2006). 180° domains of ‘up’ preferential

orientation of the polarization are present in the PZT film (Alpay et al.,

1999; Le Rhun et al., 2006; Vrejoiu et al., 2006b), their piezoresponse being

displayed as bright irregular spots with various lateral sizes between about

50 and 200nm. In order to investigate the switching behaviour of these 180°

domains, the following experiment was performed. The top left quadrant

(1µm × 1µm) of the 2µm × 2µm area was poled by –4V, and the bottom

right quadrant (1µm × 1µm) was poled oppositely by +4V. The piezoresponse

image of the overall area, acquired immediately after the poling procedure is

shown in Fig. 23.8(c). Complete switching to positive polarization occurred

in the positively biased area, whereas in the negatively poled area, some of

23.8

PFM investigations on a PbZr

0.2

0.8

/SrRuO

/SrTiO

(001)

heterostructure: (a) image of the PZT surface topography (2 µm × 2

µm) and (b) piezoresponse image acquired by scanning the area

shown in (a); (c) piezoresponse image acquired in the very same

area, immediately after poling the top left corner (1 µm × 1 µm) with

–4 V and the bottom right corner by +4 V; (d) piezoresponse image of

the very same area acquired by scanning 15 hours after the poling.

In (c) and (d) the dashed lines were added to guide the reader’s eye.

(From Vrejoiu

et al.

, 2006a, http://www.tandf.co.uk/journals).

400 nm

Topology Piezoresponse

WPNL2204

Single crystalline PZT films 709

the positive 180° domains did not completely switch to negative polarization.

The sample was left on the PFM stage for 15 hours and the same 2µm ×

2µm area was scanned again, the resulting image being displayed in Fig.

23.8(d). In the previously positively poled quadrant, no modification of the

polarization pattern was observed. However, in the negatively poled area

backswitching from the negative polarization to the positive polarization

state took place, as many of the initial bright spots reappeared at the very

same location.

According to up-to-date observations, TD half loops, TD dipoles and

backswitching 180° domains are all absent in layer-by-layer grown, defect-

free PbZr

0.2

0.8

films deposited on SrRuO

/SrTiO

(001) (Vrejoiu et al.,

2006b). Since they all show up simultaneously in the present films, it is

reasonable to assume that TD half loops and/or TD dipoles (with their lead-

rich stacking faults) may represent pinning centres for backswitching 180°

domains. The lateral spacing between 20nm and about 120 nm of the TD

dipoles and of the TD half loops at the surface of the films, and the lateral

dimension of the backswitching ferroelectric 180° domains (which is between

50 and about 200nm) are also in accordance with this assumption.

23.4 Single crystalline PbTiO

and PZT 20/80 films

free from extended structural defects

The fabrication of single crystalline ferroelectric thin films that are free from

extended structural defects requires the careful choice of a single crystalline

substrate and its crystallographic surface termination, a suitable growth

technique and – most important – appropriate growth parameters. It is necessary

to employ a single crystalline substrate with closely matched lattice constants,

because in mismatched epitaxial films which grow in a two-dimensional

mode (i.e. step flow or layer-by-layer growth), biaxial stress is generated. If

the thickness of the epitaxial film is below a certain critical value (Matthews,

1975) it may be expected that the misfit strain is not relaxed and the epilayer

is coherently grown on the substrate. When the layer-by-layer growth regime

is achieved, the indications are that epitaxy results in almost all circumstances

provided that the substrate surface is clean enough and well prepared. In this

regime the substrate has a very strong influence on the form of the film

produced, and the growing film has little option but to choose the best (i.e.

necessarily epitaxial) orientation in which to grow.

PLD is a suitable technique for synthesis of high-quality homo- and

heteroepitaxial films. This is due to the highly supersaturated ablation plasma

plume, its pulsed nature and the adjustable deposition rate attainable by

changing the laser energy density, laser repetition rate, background reactive

gas pressure and target-to-substrate distance.

Bearing in mind the aforementioned reasoning, Vrejoiu et al. (2006b)

WPNL2204