By controlling the growth of complex oxide films with atomic precision, emergent phenomena and fascinating properties have been discovered, and even been manipulated. With oxide molecular beam epitaxy (OMBE) we grow high-quality SrTiO3(110) films by evaporating Sr and Ti metals with separate controls of the open/close timing of the shutters. The incident electron beam angle of the reflective high energy electron diffraction (RHEED) is adjusted to make the (01) beam sensitive to surface chemical concentration. By monitoring such an intensity, we tune the shutter timing to synchronize the evaporation amount of Sr and Ti in real-time. The intensity is further used as a feedback control signal for automatic growth optimization to fully compensate the possible fluctuation of the source flux rates upon extended growth. A 22 nm-thick film is obtained with the precision of metal cation stoichiometry better than 0.5%.

For both fundamental research of emergent behaviors in correlated electron systems and material applications with designed functionalities, it is important to control the stoichiometry of oxide films precisely due to its critical role in determining the film properties.1 A low energetic, high purity technique, oxide molecular beam epitaxy (OMBE) has been employed for the epitaxial growth of complex oxide films with controllability in atomic scale.2 Unlike the adsorption and incorporation processes during GaAs film growth in which only one low vapor pressure element Ga is involved and the stoichiometry can be obtained within the self-regulation mechanism,3,4 the OMBE growth of complex oxide films involves multiple low vapor pressure metals and therefore requires precise flux control of each individual evaporating beam.5–8 The in situ quartz crystal microbalance (QCM) or atomic absorption spectroscopy (AAS) technique offers a few percent precision for the calibration of the flux rates of metal sources.4 However, the optimization of OMBE growth with QCM or AAS has to be based on the assumption that adhesion coefficient of the evaporating species keeps the same on the growing surface and the quartz crystal regardless their temperature difference. It is not always true in the real OMBE growth, so the stoichiometry precision of the film is deteriorated.

Furthermore, it is quite challenging to maintain stable flux rates from different metal sources since they are evaporating at high temperature in oxygen atmosphere. The oxidation of the source metals often results in the uncontrollably fluctuating flux rates,9–12 especially during the long growth for relatively thick films. A real-time, in situ monitoring technique is indispensable for the OMBE growth of high-quality films with precise stoichiometry.13,14 Reflective high energy electron diffraction (RHEED) has been demonstrated as a powerful tool to optimize the OMBE growth. Schlom et al. analyzed the fine structure of RHEED intensity oscillations during the homoepitaxial growth of SrTiO3(001) film and achieved the cation stoichiometry within ±1%.7 Even better stoichiometry control (±0.3%) has been obtained for SrTiO3 on Si(001) by monitoring the evolution of surface reconstructions induced by excess Sr or Ti with RHEED.15 

Titanate with perovskite structures have attracted extensive research interests due to their myriad of remarkable properties.16–19 Great efforts were made to tailor the functionalities by constructing low-dimensional titanate structures,20–23 which provided a strong incentive to develop the precise control methods for the OMBE growth, especially for the SrTiO3-based heterostructures along (001) direction. On the other hand, as a polar surface with anisotropy, SrTiO3(110) offers even more tunability of physical properties.24,25 There have been reports on SrTiO3(110) that facilitated the formation of off c-axis-oriented YBa2Cu3O7−δ Josephson junctions,26 controlled the magnetic anisotropy in manganite films,27 stabilized the charge-ordered phase in (Nd1−xPrx)0.5Sr0.5MnO3 films,28 and offered orientation-ordered percolating network in manganite films.29 However, comparing to SrTiO3(001), there is few studies so far on the growth of SrTiO3(110) films.30,31

We have reported that the reconstruction of SrTiO3(110) surface evolves upon the cation concentration ratio [Ti]/[Sr] reversibly.32 And this evolution is observable at high temperature (HT) by RHEED in real time during the film growth, so the flux rates of Sr and Ti sources can be synchronized by monitoring the change of the fractional diffractions that reflect the long-range order of the growing surface.33 The atomic precision was achieved by extending the calibration time, since any deviation of cation concentration in relative to the ideal bulk stoichiometry accumulates on the growing surface as revealed by RHEED and high-resolution transmission electron microscopy (HRTEM) characterizations, consistent with the density functional theory calculations.34,35 In the current work, we develop a technique for the homoepitaxial growth of SrTiO3(110) film with atomically precise control. Instead of tuning the source temperatures as in the previous work, we adjust the open/close timing of the shutter of Sr evaporation source that changes the relative ratio of metal cation deposition rate quickly and reliably. By selecting the RHEED incident beam angle carefully, we monitor the (01) diffraction intensity that is sensitive to the surface chemical concentration. This intensity is used as the real-time feedback signal for the optimization of shutter timing, and high quality SrTiO3(110) films are obtained with the precision of cation stoichiometry better than 0.5%. More importantly, such an automated OMBE growth technique allows the growth of film with any thickness without being affected by the possible fluctuation of the source flux rate.

The experiments were carried out in an ultra-high vacuum scanning tunneling microscopy (STM) system equipped with an OMBE chamber. The base pressure was better than 1×10−10 mbar. The Nb (0.7 wt%) doped SrTiO3 (110) single crystalline substrates were treated by Ar+ sputtering (1 keV/8 μA) followed by 1-hour annealing at 800°C in oxygen at 5.5×10−6 mbar. Atomically well defined surfaces were obtained with different reconstructions depending on the Ar+ sputtering time.32 In the current work the sputtering time was fixed at 8 minutes to get the substrate surface in (4×1) and (5×1) reconstructions coexisting with each other.

The SrTiO3(110) films were grown by evaporating Sr and Ti metals by a low-temperature and a high-temperature effusion cells, respectively, under an oxygen pressure of 5.5×10−6 mbar with the substrate temperature held at 700°C. By preparing the Si(001)-Sr(2×1) reconstructed interface, the flux rate of Sr source was calibrated as 1.23 ML/min [1 ML=4.64×1014 atoms/cm2 in relative to the bulk-truncated SrTiO3(110) surface] at 316°C. The temperature of Ti source was adjusted to 1560°C to obtain the same flux rate as Sr by monitoring the change of fractional RHEED patterns along [001] direction during the homoepitaxial growth of the (110) film by coevaporating Sr and Ti in oxygen.32,33 Then we intentionally decreased the temperature of the Ti source to 1530°C (giving the flux rate of ∼0.6 ML/min) and opened/closed the Sr shutter repeatedly. By optimizing the Sr shutter timing while keeping the Ti shutter opened all the time [see Fig. 1(a)], we synchronized the deposition rates from both metal sources precisely. The RHEED incident beam was aligned along the [001] direction of SrTiO3 substrate. The diffraction patterns were recorded by a charge-coupled device (CCD) camera and their intensities were analyzed and utilized as the feedback signals to trigger the action of Sr shutter. The films were characterized by the in situ STM, as well as the ex situ HRTEM (FEI Tecnai F20) and energy dispersive spectroscopy (EDS) with the sample prepared by mechanical polishing and ion milling.

Since the reconstruction on SrTiO3(110) surface evolves upon the surface cation cencentration ratio [Sr]/[Ti], the flux rates of Sr and Ti sources can be synchronized by the fine tuning of their temperatures.32,33 However, the source temperature stabilizes slowly when varied, so the optimization procedure requires a long time to achieve the satisfactory precision. To avoid such a common problem in the coevaporation growth of compound films, we further notice that the evolution of the reconstruction is totally reversible following the change of chemical potential of Sr or Ti. When there is a mismatch between the flux rates of Sr and Ti, the relative ratio of reconstruction changes accordingly while the whole surface remains atomically smooth. This means that the Sr and Ti flux rates do not have to be equivalent all the time to achieve the high-quality OMBE growth. We can even deposit Sr and Ti alternatively by controlling the open/close status of the shutters to realize the growth of high-quality single crystalline film. The shutter timing changes the deposition amount of Sr or Ti onto the surface quickly and reliably, making it a sensitive parameter to optimize the growth. Moreover, any deviation of [Sr]/[Ti] from 1 accumulates on the growing surface33–35 and the difference of surface [Sr]/[Ti] ratio between reversibly changing phases provides a rather broad window for optimization [e.g., the (6×8) phase can be changed to (4×1) by evaporating 0.85 ML Sr metal]. Therefore it is easy to keep the stable chemical environment in which the SrTi3 is thermodynamically dominant during growth.

Figure 1 shows how we optimize the Sr shutter timing while keeping Ti shutter always opened for the growth of SrTiO3(110) film. The change of Sr concentration on the growing surface in relative to the monophased (4×1) surface, Δ[Sr], fluctuates in accordance with the Sr shutter status, as shown by the blue line in Fig. 1(a). Since the flux rate of Sr (1.23 ML/min) is higher than Ti (∼0.6 ML/min), Δ[Sr] increases when both shutters are opened and decreases when Sr shutter is closed. The change of Δ[Sr] is sensitively reflected by the varying intensity of the RHEED (01) pattern (IR) as a real-time monitor (the detail mechanism will be discussed in the following), i.e., IR drops down immediately when Sr shutter is opened (Δ[Sr] rises up) and vice versa. The change of the surface reconstruction is seen in the STM images taken on the surfaces with growth stopped at t0 and t1.36 The initial substrate surface contains about 67% (5×1) and 33% (4×1) in terms of area [Fig. 1(b)]. After ∼35 s growth with both shutters opened, the surface turns to about 93% (6×1) and 7% (5×1) [Fig. 1(c)]. Comparing to the reconstruction evolution reported previously,32 the surface Sr concentration increases by ∼0.4 ML, quantitatively consistent with the estimation made by the flux rates (1.23 ML/min for Sr and ∼0.6 ML/min for Ti).

FIG. 1.

(a) The status of shutter (open/close), the change of Sr concentration on the growing surface in relative to the monophased (4×1) surface (Δ[Sr]), and the intensity of RHEED (01) pattern during the homoepitaxial growth of SrTiO3(110) film. (b)-(e) STM images (unoccupies-states) taken on the surfaces with growth stopped at t0, t1, t2, and t3 as labeled in (a).

FIG. 1.

(a) The status of shutter (open/close), the change of Sr concentration on the growing surface in relative to the monophased (4×1) surface (Δ[Sr]), and the intensity of RHEED (01) pattern during the homoepitaxial growth of SrTiO3(110) film. (b)-(e) STM images (unoccupies-states) taken on the surfaces with growth stopped at t0, t1, t2, and t3 as labeled in (a).

Close modal

The values of IR at the points where shutter status is inversed in each cycle also sensitively represents the change of Δ[Sr]. Here we set 72 seconds as a cycle of shutter action. From t0 to t2, the open/close time of Sr shutter is 35.5/36.5 sec, which results in the slow increase of [Sr] after complete cycles of shutter action. At each point of a multiple of 72 sec when Sr shutter is being opened, IR shows a maximum with decreasing value upon the cycle number. The corresponding increase of Δ[Sr] is directly visible in the STM images [Figs. 1(b) and 1(d)]. The 7-cycle growth (t0 to t2) changes the surface into about 43% (6×1) and 57% (5×1), indicating that Sr has been deposited ∼0.22 ML more than Ti through this period.33 Similarly if we set the open/close time of Sr shutter as 32.5/39.5 sec (between t2 and t3), the maxima of IR in each shutter cycle increases gradually, indicating the decrease of Sr concentration on the growing surface. It is also directly shown in the STM image [Fig. 1(e) at t3] where (5×1) and (4×1) reconstructions coexist with the same area ratio as the initial substrate surface [Fig. 1(b) at t0]. After 7 cycles from t2, the Sr excess of ∼0.22 ML is compensated at t3. Determining the Sr deposition dose linearly, the Sr shutter timing can be easily optimized as 34 s and 38 s for opening and closing status in a cycle, respectively. After t3, IR shows stable fluctuation with unchanged maxima and minima, indicating that the total deposition doses of Sr and Ti in each shutter cycle are the same.

The time duration of each shutter cycle does not critically affect the growth quality and can be selected arbitrarily, as long as two criteria are met: 1, it is not too long to make the surface reconstruction going beyond the reversible range of the phase transformation; and 2, the open and close time ratio in each cycle is optimized for fixed sources temperatures. Technically the current method of SrTiO3(110) film growth is similar to that of the (001) film proposed by Schlom et al.7 The difference is that, in the “shuttered”growth of (001) film, the SrO and TiO2 atomic layers were separately grown as the Sr and Ti shutters opened alternatively, while in the growth of (110) film, the separate deposition of Sr or Ti changes the atomic bonding configuration on the surface that is different from the bulk lattice. The observed IR value depends on the change of reconstructions or their area ratio, but not on the surface roughness induced by the atomic-layer steps as characterized by RHEED oscillations in the common layer-by-layer MBE growth.

Figure 2 shows the STM images taken on the surfaces with growth stopped at t4 and t5, respectively, when IR are exactly the same [see Fig. 1(a)]. There are steps with the height of a single unit cell on the surface corresponding to the layer-by-layer growth of the SrTiO atomic plane. And the area ratio of the higher terraces indicates the amount of the growing layer, i.e., ∼0.1 ML at t4 and ∼0.6 ML at t5. While the step roughness changes dramatically, the high-resolution STM images show no detectable change of the area ratio of (5×1) and (4×1) reconstructions, i.e., the surface stoichiometry almost keeps unchanged.36 On the other hand, IR changes much more significantly with Δ[Sr] as shown in Fig. 1(a). In brief, the intensity of the integral (01) RHEED pattern is sensitive to the surface chemical concentration (actually the atomic configurations), instead of the atomic layer roughness.

FIG. 2.

(a) and (b) STM images (unoccupies-states) taken on the surfaces with growth stopped at t4 and t5, respectively, as labeled in Fig. 1(a). The insets are the corresponding zoom-in images.

FIG. 2.

(a) and (b) STM images (unoccupies-states) taken on the surfaces with growth stopped at t4 and t5, respectively, as labeled in Fig. 1(a). The insets are the corresponding zoom-in images.

Close modal

The key of the RHEED monitoring in the current work is to adjust the incident angle of the electron beam (θi in relative to the surface). Although with some certain θi values, the dependence of IR on the growth time follows the change of the atomic layer roughness, we can shift θi within the grazing incidence range (1.5° ∼5°) to make IR sensitive to the surface reconstruction, i.e., the metal cation concentration. We analyze such a behavior by calculating the RHEED (01) beam intensity with a kinematic model in the single scattering limit.37,38 As presented in detail in the Supplement,39 the dependence of IR on θi is considered in two cases. One is where the growing surface contains randomly distributed terraces in two adjacent layers, both in the same microscopic structure. The minimum of IR is reached when the area proportion of the upper terrace is 50% (

${\rm I}^{{\rm 0.5ML}}_{R}$
IR0.5 ML ⁠), while the maximum is reached when the proportion is 0 (
${\rm I}^{{\rm 0ML}}_{R}$
IR0 ML
). The sensitivity of IR to the atomic layer roughness is characterized by
$\Delta I^{{\rm rough}}_{R}$
ΔIR rough
=
${\rm I}^{{\rm 0ML}}_{R}$
IR0 ML
-
${\rm I}^{{\rm 0.5ML}}_{R}$
IR0.5 ML
. The other case is where the surface is atomically flat without steps. The sensitivity of IR to the chemical concentration can be characterized, e.g., by the difference between monophased (4×1) and (5×1) reconstructions,
$\Delta I^{{\rm conc}}_{R}$
ΔIR conc
=
${\rm I}^{{\rm 4\times 1}}_{R}$
IR4×1
-
${\rm I}^{{\rm 5\times 1}}_{R}$
IR5×1
. As shown in Fig. 3,
$\Delta I^{{\rm conc}}_{R}$
ΔIR conc
is significantly larger than
$\Delta I^{{\rm rough}}_{R}$
ΔIR rough
when θi ≈1.8o, i.e., IR can be much more sensitive to the change of surface chemical concentration than to the atomic layer roughness.

FIG. 3.

The calculated

$\Delta I^{{\rm conc}}_{R}$
ΔIR conc (red) and
$\Delta I^{{\rm rough}}_{R}$
ΔIR rough
(blue) with different θis.

FIG. 3.

The calculated

$\Delta I^{{\rm conc}}_{R}$
ΔIR conc (red) and
$\Delta I^{{\rm rough}}_{R}$
ΔIR rough
(blue) with different θis.

Close modal

In principle, the surface cation concentration can be monitored by fractional RHEED patterns in real-time since it is represented by the evolution of surface reconstructions. But technically, it takes a relatively long time to accumulate the change of surface long-range order that can be clearly reflected by RHEED fratcional patterns, and therefore to improve the precision of cation stoichiometry.33 Instead, IR sensitively changes upon the surface concentration (we estimate that IR reflects the change of surface [Sr]/[Ti] by a few percents), providing an ideal real-time signal for the optimization of source flux rates. Once the surface [Sr]/[Ti] ratio is kept within the range that the reconstruction evolves reversibly, we provide the stable chemical environment in which the SrTiO3 phase is thermodynamically dominant and obtain the high-quality, stoichiometric films with atomic precision. More importantly, the (01) integral diffraction pattern is bright and sharp, so it is much easier to be detected in real-time than the fuzzy fractional patterns, especially when there are mixed phases on the growing surface.32 

In an extended OMBE growth, the flux rates of different metal sources fluctuate and therefore the film quality might be deteriorated. This is also the case observed in the current work. Although the Sr shutter timing has been optimized and fixed as 34 sec (opening) and 38 sec (closing), the according maxima of IR do not keep constant during a long growth, as shown in Fig. 4(a). The unstable envelop of IR indicates the fluctuation of [Sr] on the growing surface, which has been verified by STM that reveals the varied area ratio of different reconstructions. We further take the advantage of the easily monitored IR by using it as an automatic feedback control signal to obtain the metal cation stoichiometry of the SrTiO3(110) film. Since IR increases/decreases when the Sr shutter is closed/opened, two thresholds are set in the control software. When the monitored IR reaches the upper threshold, Sr shutter is triggered to open; and when the lower threshold is reached, Sr shutter is triggered to close. In such a way, the oscillation of IR is kept stable, indicating that the metal cation stoichiometry has been maintained. The upper panel of Fig. 4(b) shows the time-dependent IR during the growth with automatic feedback control of the Sr shutter.

FIG. 4.

(a) IR monitored during the growth with the Sr shutter timing fixed at 34 sec (opening) and 38 sec (closing). (b) IR during the growth with automatic feedback control of the Sr shutter (upper panel) and the recorded corresponding opening time of Sr shutter (lower panel).

FIG. 4.

(a) IR monitored during the growth with the Sr shutter timing fixed at 34 sec (opening) and 38 sec (closing). (b) IR during the growth with automatic feedback control of the Sr shutter (upper panel) and the recorded corresponding opening time of Sr shutter (lower panel).

Close modal

On the other hand, the recorded shutter timing that determines the variation of IR deviates by a few percent randomly, as shown in the lower panel of Fig. 4(b). By fixing the corresponding fluctuation of the source flux rates by the feedback control in real-time, high-quality homoepitaxy SrTiO3(110) film is obtained. The grown surface is atomically smooth without detectable change of reconstruction, as shown in Figs. 5(a) and 5(b). The perfect lattice structure of the film is revealed by the cross-sectional HRTEM [Fig. 5(c)], without columnar features resulted from Ruddlesden-Popper planar faults or disorders associating with Sr deficiency.40,41 The interface between the film and substrate is indistinguishable in the HRTEM image of this homoepitaxial system. It can only be estimated by counting the atomic layers from the surface with the growth rate of Ti since Ti shutter is always opened in the current work. With EDS equipped on HRTEM, the difference of the cation concentration ratio [Sr]/[Ti] in the film and in the single crystalline substrate is smaller than 0.5%, which is the precision limit of the analysis method. Again, it is worth of noting that the 0.5% precision of the cation stoichiometry is not directly obtained by RHEED sensitivity, but from the dominant thermodynamic stability of the SrTiO3 phase in the relatively stable chemical environment during growth.

FIG. 5.

(a) and (b) STM images (unoccupied-states) of the substrate and the 22 nm-thick film surfaces, respecively. The insets are the corresponding zoom-in images. (c) The cross-sectional HRTEM image of the homoepitaxial SrTiO3(110) thin film.

FIG. 5.

(a) and (b) STM images (unoccupied-states) of the substrate and the 22 nm-thick film surfaces, respecively. The insets are the corresponding zoom-in images. (c) The cross-sectional HRTEM image of the homoepitaxial SrTiO3(110) thin film.

Close modal

In summary, we develop a precise control method for OMBE growth of complex oxide films. High-quality homoepitaxial SrTiO3(110) films are obtained by evaporating Sr and Ti metals with separate controls of the open/close timing of the shutters. The incident electron beam angle of RHEED is adjusted to make the (01) beam intensity sensitive to the surface atomic configuration that is determined by the metal cation concentration. This intensity is monitored and used as the feedback control signal to automatically optimize the shutter timing in real-time. Thus the possible fluctuation of the source flux rates upon extended growth can be fully compensated, resulting in the stable chemical environment in which the SrTiO3 phase is thermodynamically dominant and ensuring the precision of metal cation stoichiometry better than 0.5%. This method can be applied to the OMBE growth of SrTiO3(110) films with any thickness, from a single atomic layer to, in principle, infinite.

This work was supported by “973” Program of China (2012CB921700), NSFC Project 11027406 and Specific Funding of Discipline and Graduate Education Project of Beijing Municipal Commission of Education.

1.
M.
Imada
,
A.
Fujimori
, and
Y.
Tokura
,
Rev. Mod. Phys.
70
,
1039
(
1998
).
2.
A.
Tsukazaki
,
S.
Akasaka
,
K.
Nakahara
,
Y.
Ohno
,
H.
Ohno
,
D.
Maryenko
,
A.
Ohtomo
, and
M.
Kawasaki
,
Nature Mater.
9
,
889
(
2010
).
3.
J. R.
Arthur
,
J. Appl. Phys.
39
,
4032
(
1968
).
4.
D. G.
Schlom
and
J. S.
Harris
 Jr.
,
MBE Growth of High Tc Superconductors. in: Molecular Beam Epitaxy
, edited by
R. F. C.
Farrow
(
Noyes
,
Park Ridge
,
1995
), pp.
505
622
.
5.
J. N.
Eckstein
and
I.
Bozovic
,
Annu. Rev. Mater. Sci.
25
,
679
(
1995
).
6.
D. J.
Rogers
,
P.
Bove
, and
F. H.
Teherani
,
Supercond. Sci. Technol.
12
,
R75
(
1999
).
7.
J. H.
Haeni
,
C. D.
Theis
, and
D. G.
Schlom
,
J. Electroceram.
4
,
385
(
2000
).
8.
D. G.
Schlom
,
J. H.
Haeni
,
J.
Lettieri
,
C. D.
Theis
,
W.
Tian
,
J. C.
Jiang
, and
X. Q.
Pan
,
Sci. Eng. B
87
,
282
(
2001
).
9.
J. N.
Eckstein
,
I.
Bozovic
, and
G. F.
Virshup
,
MRS Bull.
19
,
44
(
1994
).
10.
E. S.
Hellman
and
E. H.
Hartford
,
J. Vac. Sci. Technol. B.
12
,
1178
(
1994
).
11.
C. D.
Theisand
and
D. G.
Schlom
,
J. Vac. Sci. Technol. A.
14
,
2677
(
1996
).
12.
Y. S.
Kim
,
Namrata
Bansal
,
Carlos
Chaparro
,
Heiko
Gross
, and
Seongshik
Oh
,
J. Vac. Sci. Technol. A.
28
(
2
),
271
(
2010
).
13.
M.
Naito
,
H.
Yamamoto
, and
H.
Sato
,
Physica C.
305
,
233
(
1998
).
14.
S. A.
Chambers
,
Surf. Sci. Rep.
39
,
105
(
2000
).
15.
M. P.
Warusawithana
,
C.
Cen
,
C. R.
Sleasman
,
J. C.
Woicik
,
Y.
Li
,
L. F.
Kourkoutis
,
J. A.
Klug
,
H.
Li
,
P.
Ryan
,
L.
Wang
,
M.
Bedzyk
,
D. A.
Muller
,
L.
Chen
,
J.
Levy
, and
D. G.
Schlom
,
Science
324
,
367
(
2009
).
16.
J. F.
Schooley
,
W. R.
Hosler
, and
M. L.
Cohen
,
Phys. Rev. Lett.
12
,
474
(
1964
).
17.
K. A.
Müller
and
H.
Burkard
,
Phys. Rev. B
19
,
3593
(
1979
).
18.
Y.
Tokura
,
Y.
Taguchi
,
Y.
Okada
,
Y.
Fujishima
,
T.
Arima
,
K.
Kumagai
, and
Y.
Iye
,
Phys. Rev. Lett.
70
,
2126
(
1993
).
19.
K.
Ueno
,
S.
Nakamura
,
H.
Shimotani
,
A.
Ohtomo
,
N.
Kimura
,
T.
Nojima
,
H.
Aoki
,
Y.
Iwasa
, and
M.
Kawasaki
,
Nature Mater.
7
,
855
(
2008
).
20.
A.
Ohtomo
,
D. A.
Muller
,
J. L.
Grazul
, and
H. Y.
Hwang
,
Nature
419
,
378
(
2002
).
21.
J. H.
Haeni
,
P.
Irvin
,
W.
Chang
,
R.
Uecker
,
P.
Reiche
,
Y. L.
Li
,
S.
Choudhury
,
W.
Tian
,
M. E.
Hawley
,
B.
Craigo
,
A. K.
Tagantsev
,
X. Q.
Pan
,
S. K.
Steiffer
,
L. Q.
Chen
,
S. W.
Kirchoefer
,
J.
Levy
, and
D. G.
Schlom
,
Nature (London)
430
,
758
(
2004
).
22.
J.
Heber
,
Nature (London)
459
,
28
(
2009
).
23.
C. M.
Brooks
,
L.
Fitting Kourkoutis
,
T.
Heeg
,
J.
Schubert
,
D. A.
Muller
, and
D. G.
Schlom
,
Appl. Phys. Lett.
94
,
162905
(
2009
).
24.
I. C.
Infante
,
J. O.
Osso
,
F. S.
anchez
, and
J.
Fontcuberta
,
Appl. Phys. Lett.
92
,
012508
(
2008
).
25.
J. X.
Ma
,
X. F.
Liu
,
T.
Lin
,
G. Y.
Gao
,
J. P.
Zhang
,
W. B.
Wu
,
X. G.
Li
, and
J.
Shi
,
Phys. Rev. B
79
,
174424
(
2009
).
26.
H.
Akoh
,
C.
Camerlingo
, and
S.
Takada
,
Appl. Phys. Lett.
56
,
1487
(
1990
).
27.
Y.
Suzuki
,
H. Y.
Hwang
,
S.-W.
Cheong
, and
R. B.
van Dover
,
Appl. Phys. Lett.
71
,
140
(
1997
).
28.
Y.
Ogimoto
,
M.
Nakamura
,
N.
Takubo
,
H.
Tamaru
,
M.
Izumi
, and
K.
Miyano
,
Phys. Rev. B
71
,
060403
(
2005
).
29.
K. J.
Lai
,
M.
Nakamura
,
W.
Kundhikanjana
,
M.
Kawasaki
,
Y.
Tokura
,
M. A.
Kelly
, and
Zhi-Xun
Shen
,
Science
329
,
190
(
2010
).
30.
Y.
Mukunoki
,
N.
Nakagawa
,
T.
Susaki
, and
H. Y.
Hwang
,
Appl. Phys. Lett.
86
,
171908
(
2005
).
31.
J. H.
Hao
,
J.
Gao
,
Z.
Wang
, and
D. P.
Yu
,
Appl. Phys. Lett.
87
,
131908
(
2009
).
32.
Zhiming
Wang
,
Fang
Yang
,
Zhiqiang
Zhang
,
Yuanyuan
Tang
,
Jiagui
Feng
,
Kehui
Wu
,
Qinlin
Guo
, and
Jiandong
Guo
,
Phys. Rev. B.
83
,
155453
(
2011
).
33.
Zhiming
Wang
,
Jiagui
Feng
,
Yang
Yang
,
Yuan
Yao
,
Lin
Gu
,
Fang
Yang
,
Qinlin
Guo
, and
Jiandong
Guo
,
Appl. Phys. Lett.
100
,
051602
(
2012
).
34.
J. A.
Enterkin
,
A. K.
Subramanian
,
B. C.
Russell
,
M. R.
Castell
,
K. R.
Poeppelmeier
, and
L. D.
Marks
,
Nature Mater.
9
,
245
(
2010
).
35.
Fengmiao
Li
,
Zhiming
Wang
,
Sheng
Meng
,
Yongbao
Sun
,
Jinlong
Yang
,
Qinlin
Guo
, and
Jiandong
Guo
,
Phy. Rev. Lett.
107
,
036103
(
2011
).
36.
STM images are taken over different areas on the sample and the statistics indicate the uniformity of the entire surface.
37.
A.
Ichimiya
and
P. I.
Cohen
,
Reflection high-energy electron diffraction
(
Cambridge University Press
,
2004
).
38.
L. M.
Peng
,
Micron
30
,
625
648
(
1998
).
39.
See supplementary material at http://dx.doi.org/10.1063/1.4773555 for the calculations of the intensity of RHEED integral diffraction patterns.
40.
S. N.
Ruddlesden
and
P.
Popper
,
Acta Crystallogr.
10
,
538
(
1957
).
41.
S. N.
Ruddlesden
and
P.
Popper
,
Acta Crystallogr.
11
,
54
(
1958
).

Supplementary Material