Ramanathan Sh. (Ed.) Thin Film Metal-Oxides: Fundamentals and Applications in Electronics and Energy

Подождите немного. Документ загружается.

1 In Situ Synchrotron Characterization of Complex Oxide Heterostructures 17

Fig. 1.6 (a) Calculated intensities along the 00L for SrTiO

.001/ surfaces with SrO- and TiO

terminations (solid and dashed, respectively). Results from both 15.0 and 16.2 keV X-ray energies

are shown. (b) The intensity ratio, I(16.2 keV)/I(15.0 keV), shown for SrO- and TiO

-terminations

Information regarding the nature of the terminating plane is contained within the

CTRs. However, the distinction between AO- and BO

-terminated surfaces can be

minute unless the incident photon energy is tuned near an absorption edge [114]. As

showninFig.1.6a, the calculated intensity along the 00L is very similar for both

the SrO- and TiO

-terminations at 15.0 keV. When the X-ray energy is adjusted to

16.2 keV (just above the Sr K-edge), sharp differences are observed near the 001 and

003 peaks. This difference can be further accentuated by taking the ratio between

the scattered intensity near and away from an absorption edge (Fig. 1.6b).

Other perovskite (001) substrates can be prepared with BO

-termination in-

cluding LaAlO

[115], .LaAlO

0:3

-.Sr

AlTaO

0:7

[115], and DyScO

[116],

although complete BO

-termination is difﬁcult to achieve. Substrates such as

NdGaO

can be prepared with AO-termination [115].

18 T.T. Fister and D.D. Fong

Fig. 1.7 Depiction of the surface morphology during step-ﬂow, layer-by-layer, and 3D growth,

and the corresponding behavior of CTR intensity when growth is initiated

1.3.2 Growth Oscillations

Epitaxial growth can be characterized by the following growth modes: step-ﬂow,

layer-by-layer, or three-dimensional [117]. During homoepitaxial growth, the cor-

responding anti-Bragg intensity for each of the growth modes is shown on the right

of Fig. 1.7. During step-ﬂow growth, the deposited adatoms have sufﬁcient time

and mobility to attach to step-edges on a vicinal surface. Thus, existing surface

steps simply propagate across the surface, leaving the CTR intensity unchanged.

When the adatoms have some mobility but insufﬁcient time to reach the step-edges,

two-dimensional islands nucleate, grow, and coalesce on the crystal terraces, result-

ing in layer-by-layer growth and an oscillating CTR intensity. Three-dimensional

growth occurs when the depositing atoms remain essentially where they ﬁrst land,

roughening the surface and producing a sharp drop in CTR intensity.

The roughness oscillations observed during layer-by-layer growth have a thick-

ness periodicity determined by the island height: this is typically a single unit cell

for perovskite (001) systems. These oscillations are similar to those observed by

reﬂection high-energy electron diffraction (RHEED). A careful study of these oscil-

lations during PLD or MBE permits determination of the terminating plane since

perovskite ﬁlms grown on perovskite substrates are inclined to continue the al-

ternating AO-BO

sequence [118]. A prominent distinction between oxide MBE

and other growth techniques is the ability to independently control A- and B-site

layer deposition because of the use of elemental sources. While this level of control

1 In Situ Synchrotron Characterization of Complex Oxide Heterostructures 19

0.0

2.0

4.0

Thickness (ML)

0246810121416

4 6 8 10121416

Thickness (ML)

(Å)

Fig. 1.8 (a) 00

intensity oscillations during growth of La

0:7

0:3

MnO

on SrTiO

.001/ at 600

and P

D 0:3 Torr . (b) Simulated total, continuous, and discrete RMS roughnesses (Reprinted

with permission from [120])

permits growth of layered structures like the Ruddlesden–Popper phases, in situ

monitoring of these oscillations is mandatory for growth of MBE ﬁlms with the

desired cation stoichiometry [119].

During heteroepitaxial growth, the electron density difference between the ﬁlm

and substrate leads to additional (Kiessig) oscillations visible for both step-ﬂow and

layer-by-layer growth. At an anti-Bragg position, the period for Kiessig oscillations

is two-unit-cells. Employing the previously described roughness formulation, Dale

et al. [120] were able to model the growth oscillations observed during PLD of

1x

MnO

(LSMO) on SrTiO

.001/ (Fig. 1.8a). Here, the continuous rough-

ness obeyed a power–law relationship,



C .˛t

; (1.25)

whereas the discrete roughness depended on the coverage  :



D c

.t/.1  .t//; (1.26)

as shown in Fig. 1.8b.

1.3.3 Case Studies of Oxide Growth

There are now several oxide PLD chambers installed on synchrotron beamlines

around the world including ones at the APS [121], CHESS [44], ESRF [122], and

SLS [123]. An oxide MOCVD system is also installed at the APS [124]. In this

section, we discuss results from work with a few of these systems.

20 T.T. Fister and D.D. Fong

1.3.3.1 PLD of SrTiO

on SrTiO

.001/

Oxide growth by PLD has several beneﬁcial features [125, 126]. Foremost is its

ability to reproduce the stoichiometry of the target after optimization of deposi-

tion parameters [5, 127]. For oxides, adequate oxygen stoichiometry is ensured by

growth in relatively high P

(e.g., 10 mTorr), although some groups favor growth

in low P

or with other oxidants to prevent surface roughening and/or to stabilize

perovskites with low oxidation states [128]. The huge instantaneous deposition rate

per shot (0:01 ML in 5s or 2,000ML per second [123]) leads to a high nucle-

ation density, and the arrival of 25 eV particles at the surface promotes enhanced

surface diffusion. Therefore, two-dimensional growth is strongly favored by PLD,

and a ﬂat growth surface can be maintained even after the deposition of hundreds

of monolayers. The laser repetition rate is typically 10 Hz for continuous deposi-

tion, but surface kinetics are often studied with an “interrupted” timing pattern with

variable dwell times between shots.

Tischler et al. [129] performed time-resolved measurements (10 s time resolu-

tion) of the 00

intensity during homoepitaxial growth of SrTiO

.001/. The growth

of two MLs with a 50 s dwell time is shown in Fig. 1.9a.

The layer-by-layer deposition could be adequately modeled by assuming the ex-

istence of only two incomplete monolayers throughout growth [131]:

I.t/ D I

Œ1  2

.t/ C 2

nC1

.t/

; (1.27)

where 

.t/ and 

nC1

.t/ are the fractional coverages of the bottom and top layers,

respectively. The resulting 

.t/ and 

nC1

.t/ are shown as stepped lines in Fig. 1.9a.

At the ﬁrst oscillation maximum, it is observed that 

0:85 while 

nC1

0:1.

When 

ﬁnally reaches 1, completing the bottom layer, the coverage of the top

layer is already at 

nC1

0:5, i.e., at the initial stages of island coalescence. Mag-

niﬁed versions of the circled regions in Fig. 1.9aareshowninFig.1.9b. After the

nearly instantaneous initial deposition from the shot, and during the 50 s dwell, a

small amount of material .0:02 ML/ is added to the underlying layer from the 2D

islands on top. This small amount of interlayer transport is shown as a function of

bottom layer coverage in Fig. 1.9c. Since each laser shot corresponds to the growth

of 0.1 ML, only 20% of this undergoes thermal interlayer transport; this falls to 5%

for a 0.2 s dwell time. Therefore, during continuous PLD growth .10 Hz/,mostof

the interlayer transport takes place far from thermal equilibrium. As 

nears 1, the

coverage changes more slowly with time because adatoms have greater difﬁculty

ﬁnding the remaining holes.

Using a CCD placed near the 00

, Fleet et al. [130] observed both specular and

diffuse intensity as a function of time during SrTiO

growth (Fig. 1.9d). During de-

position of the ﬁrst half monolayer, they found satellites with an in-plane correlation

length of 20nm. The diffuse intensity peaks at the completion of 0.5 ML (see inset).

Aided by scanning probe microscopy, they determined that the correlation length

was associated with the network of holes described above.

1 In Situ Synchrotron Characterization of Complex Oxide Heterostructures 21

Fig. 1.9 (a) 00

intensity for PLD of SrTiO

.001/ homoepitaxy at 650

C for a 50 s dwell time.

(b) Magniﬁed plot of layer coverages for the circled laser shots in (a). (c) Fraction of material

transferred by the slow interlayer transport step at 0.2 s dwell time (diamonds) and 50 s dwell time

(circles). The solid lines are Gaussian ﬁts. Parts (a), (b), and (c) are reprinted with permission

around 00

during SrTiO

homoepitaxy at T D 697

C;P

D 10

5

Torr and a pulse frequency

of 0.1 Hz. Inset: Diffuse intensity during deposition of the ﬁrst half monolayer. Each curve is the

integrated intensity following a pulse. Part (d) is reprinted with permission from [130] (Copyright

2006 by the American Physical Society)

1.3.3.2 PLD of La

1x

MnO

on SrTiO

.001/

Studies on the heteroepitaxial growth of LSMO .x

D 0:34/ on SrTiO

.001/

were carried out by Willmott et al. [132]. At 750

C and in a background of

1:5  10

4

Torr oxygen partial pressure, they performed detailed studies on the

growth of a single monolayer of LSMO with variable dwell time to examine the

relaxation behavior between pulses. Growth was carried out with a synchronized

O gas pulse for improved oxygen stoichiometry. Using energetic arguments, they

deduced that during the ﬁrst half of ML growth . < 0:5/, the impinging 25 eV

particles break up existing 2D islands into daughter islands, thereby inhibiting coars-

ening and increasing the density of nuclei until the average distance between them

is similar to their own size (at   0:5). During the second half of ML growth, is-

land coarsening occurs with enhanced surface diffusion. In this way, PLD promotes

surface smoothing throughout growth. A diagram illustrating the processes they de-

scribe is shown in Fig. 1.10.

22 T.T. Fister and D.D. Fong

Fig. 1.10 Schematic of the processes inﬂuencing 2D ﬁlm growth by PLD. At low coverages, the

impinging particles (white) nucleate small, relatively densely packed islands (black). Impinging

particles can cause them to split (gray). The density of the small islands increases until   0:5,

where the islands coalesce. For >0:5, the impinging particles diffuse to step edges and are

accelerated by the energy of impinging particles (Reprinted with permission from [132]. Copyright

2006 by the American Physical Society)

Although PLD does favor the layer-by-layer growth mode, this depends on T

and P

as shown in Fig. 1.11 [120].

It is observed that higher P

can lead to improved smoothness at high temper-

atures and step-ﬂow growth (e.g., at T D 950

CandP

D 300  10

3

Torr). At

600

C, however, this same pressure gives rise to increased surface roughness and

eventual three-dimensional growth. Lower pressures .P

D 10

3

Torr/ demon-

strate a similar trend, tending toward step ﬂow and three-dimensional growth at

higher and lower temperatures, respectively. For the 950

CandP

D 10

3

Torr

data set, growth was interrupted after 95, 123, 158, and 188 ML for annealing. These

annealing stages increased the 00

intensity (decreasing the roughness), but addi-

tional growth resulted in quick resumption of the prior roughness behavior. The

black lines are models of continuous surface roughness using (1.25), with ˇ D 0:5.

When depositing a perovskite alloy (i.e., the A- or B-sites are dopant substituted),

another consideration is the distribution of the dopant throughout the ﬁlm. Using the

COBRA phase retrieval method, Herger et al. [51] converted ten inequivalent and

ﬁve equivalent CTRs into the A- and B-site proﬁles shown in Fig. 1.12 for six LSMO

ﬁlms of different thickness grown on SrTiO

.001/.

The nominal Sr concentration, x

D 0:35, was set using a translatable PLD

target rod comprised of LaMnO

on one end and SrMnO

on the other [133]. For

the three thicker ﬁlms, x

is observed to change from 0.8 to 0.4 as z increases from

0:5 to 1.5. Above z D 1:5; x

 0:3 up to the topmost layer where x

increases

again because of Sr surface segregation. For the three thinnest ﬁlms, the LSMO

is La-deﬁcient with x

never exceeding 0.5. This may result from a Sr-rich layer

1 In Situ Synchrotron Characterization of Complex Oxide Heterostructures 23

Anti-Bragg Intensity

Film Thickness (monolayers)

Fig. 1.11 00

intensity vs. La

0:7

0:3

MnO

ﬁlm thickness in P

D 10

3

and 0.3 Torr, for 950

and 600

C. The black lines are models of the evolution of continuous surface roughness with

ˇ D 0:5 power-law thickness dependence (Reprinted with permission from [120])

“ﬂoating” to the surface during growth. Sr surface segregation was also observed in

LaVO

=SrTiO

superlattices grown by PLD, producing vanadate layers with diffuse

lower and abrupt upper interfaces [134].

The LSMO ﬁlms grown by Herger et al. [51] were observed to have MnO

termination, as expected based on the growth oscillations. More interesting is the

appearance of Mn in what is nominally the top TiO

plane of the SrTiO

.Theau-

thors believe this may be additional Ti at the interface in actuality; the similar atomic

numbers between Mn and Ti make their distinction difﬁcult. As will be discussed in

Sect. 4, a TiO

double-layer has been observed on bare SrTiO

.001/ under condi-

tions similar to that for the growth of LSMO.

1.3.3.3 MOCVD of PbZr

1x

on SrTiO

.001/

Unlike PLD, growth by MOCVD relies on reactions between the incoming or-

ganic precursors and the hot sample rather than direct transfer of material from

target to substrate. The high pressures and temperatures involved make other in situ

24 T.T. Fister and D.D. Fong

0.0

0.5

1.0

0.0

0.5

1.0

0.0

0.5

1.0

0.0

0.5

1.0

0.0

0.5

1.0

−202468

z [unit cell]

Occupancy of layer

0.0

0.5

1.0

1 ML

2 ML

3 ML

4 ML

6 ML

9 ML

Fig. 1.12 Cation occupancies for LSMO ﬁlms of various thickness grown on SrTiO

.001/

1 In Situ Synchrotron Characterization of Complex Oxide Heterostructures 25

probes difﬁcult to apply, and the oxide MOCVD system at the APS is unique in its

ability to monitor oxide growth by MOCVD. Studies with this system have focused

on PbZr

1x

(PZT), a prototypical ferroelectric oxide ideal for X-ray stud-

ies because of its large ferroelectric displacements and large scattering signal. For

deposition of PZT, Wang et al. [135] employed tetraethyl lead (TEL), titanium tert-

butoxide (TTB), and zirconium tert-butoxide (ZTB) cation precursors. Nitrogen was

used as the carrier gas, oxygen was the oxidant .P

 2 Torr/, and the total pres-

sure in the MOCVD chamber was ﬁxed at 10 Torr. While growth of the alloy PZT

requires compatible B-site precursors like TTB and ZTB [136], titanium isopropox-

ide (TIP) is the favored Ti precursor for PbTiO

for the reasons discussed below.

Homoepitaxial growth oscillations of PbTiO

on a thick (relaxed) PbTiO

ﬁlm

are shown in Fig. 1.13, with results from the TIP precursor in Fig. 1.13a and results

from the TTB precursor in Fig. 1.13b.

As seen in Fig. 1.13a, higher growth temperatures promote smaller amplitude

oscillations, signaling smoother growth; at reduced TIP ﬂows .0:06 mol=min/,

growth proceeds in the step-ﬂow mode [124]. The inset shows that growth rate is

roughly independent of temperature within a 100

window. The growth rate is also

independent of the TEL ﬂow, but scales linearly with the TIP ﬂow rate (not shown),

making it easy to control both growth rate and growth mode. Deposition with the

TTB precursor is considerably different (Fig. 1.13b). Here, the growth rate drops

by nearly a factor of three when the substrate temperature is decreased from 696

to 657

C, suggesting that incomplete cracking of the TTB precursor may occur at

lower temperatures.

0 10 20 30 40

0.2

0.4

0.6

0.8

1.0

1.2

1.4

Normalized CTR Intensity (offset by 0.1)

Normalized CTR Intensity (offset by 0.2)

Time after Growth Start (s)

0 10203040

Time after Growth Start (s)

736°C

716°C

696°C

677°C

657°C

650 700

750

120

140

160

180

200

T (°C)

G (nm/hr)

0.2

0.4

0.6

0.8

1.0

1.2

696 °C

657 °C

150

100

G (nm/hr)

700650

T (°C)

Fig. 1.13 Evolution of the 20

position before, during and after growth of 4 unit cells of PbTiO

various T using ﬁxed TIP and TEL precursor ﬂows. (a) Growths with ﬁxed TEL and TIP ﬂows of

0.25 and 0:26 mol=min. (b) Growths with ﬁxed TEL and TTB ﬂows of 0.25 and 0.95 mol/min.

Insets: growth rate as a function of temperature for ﬁxed precursor ﬂows [136]. Reproduced with

permission of the International Union of Crystallography

26 T.T. Fister and D.D. Fong

Fig. 1.14 Equilibrium phase diagram of the PbTiO

.001/ surface. Solid lines separate phase ﬁelds

corresponding to PbO condensation, c.22/,and.16/ reconstructions. Dotted lines are literature

values for the PbO condensation and PbTiO

decomposition boundaries [138] (Reprinted with

The ability to characterize ﬁlms in the MOCVD chamber permits high-

temperature studies as functions of P

and the PbO partial pressure. This can

be crucial for measurements of the ferroelectric transition temperature, which de-

pends strongly on the epitaxial strain state and can be as high as 725

CforPbTiO

on SrTiO

[137]. At this temperature, the PbO partial pressure must be kept between

4 10

6

and 2 10

4

Torr to prevent PbTiO

decomposition or PbO condensation,

as shown by the dotted lines in Fig. 1.14. By performing in situ studies within the

growth chamber, the necessary PbO pressure can be maintained by a steady ﬂow of

TEL [124].

Figure 1.15a depicts repeated in-plane .H / scans across the PZT 402 during

growth of PZT on SrTiO

.001/ at 736

C[135]. The nominal Zr composition, x

0:086, is based on the Zr vapor composition. Bulk PbZr

0:086

0:914

, however, has

a large compressive misﬁt with SrTiO

, and the ﬁlm begins to relax after about

175 s of growth (at 7 nm). The surface in-plane lattice parameter is shown on the

left axis of Fig. 1.15b, while the Zr composition at the surface, as measured by

an X-ray ﬂuorescence detector mounted above the sample, is shown on the right

axis. As seen, the coherently strained part of the ﬁlm has a low Zr mole fraction

of about 5%. After exceeding the critical thickness, the in-plane lattice parameter

relaxes toward its bulk value, and x

increases dramatically to nearly 10% at 25 nm.

Eventually, as the ﬁlm continues to relax, the surface composition matches that of

the Zr vapor mole fraction, although ﬁlms grown with larger Zr vapor compositions

were observed to incorporate more Zr than expected in the relaxed surface.

This behavior is caused by the larger Zr cations preferentially incorporating in

regions of larger lattice parameter, an effect known as “lattice pulling” [140]. It

is expected that similar behavior occurs in other alloy systems that have under-

gone strain relaxation during growth. The resulting compositional gradient may