Epitaxy from a Periodic Y–O Monolayer: Growth of Single-Crystal Hexagonal YAlO3 Perovskite

The role of an atomic-layer thick periodic Y–O array in inducing the epitaxial growth of single-crystal hexagonal YAlO3 perovskite (H-YAP) films was studied using high-angle annular dark-field and annular bright-field scanning transmission electron microscopy in conjunction with a spherical aberration-corrected probe and in situ reflection high-energy electron diffraction. We observed the Y–O array at the interface of amorphous atomic layer deposition (ALD) sub-nano-laminated (snl) Al2O3/Y2O3 multilayers and GaAs(111)A, with the first film deposition being three cycles of ALD-Y2O3. This thin array was a seed layer for growing the H-YAP from the ALD snl multilayers with 900 °C rapid thermal annealing (RTA). The annealed film only contained H-YAP with an excellent crystallinity and an atomically sharp interface with the substrate. The initial Y–O array became the bottom layer of H-YAP, bonding with Ga, the top layer of GaAs. Using a similar ALD snl multilayer, but with the first film deposition of three ALD-Al2O3 cycles, there was no observation of a periodic atomic array at the interface. RTA of the sample to 900 °C resulted in a non-uniform film, mixing amorphous regions and island-like H-YAP domains. The results indicate that the epitaxial H-YAP was induced from the atomic-layer thick periodic Y–O array, rather than from GaAs(111)A.


Materials and Methods
The samples were prepared in a growth/analysis ultra-high vacuum (UHV) multi-chamber system (Designed and constructed by M. Hong, J. Kwo, and the group members, Taiwan, with chambers/parts from various countries.) [34,35]. All of the chambers were connected through UHV modules under~10 −10 torr to ensure intactness of the pristine surfaces and interfaces during the sample transfers. Epitaxial GaAs layers were grown on GaAs(111)A in the solid-source GaAs-based MBE chamber. The epi-wafers were transferred under UHV to the ALD reactor for the deposition of the snl multilayers, which included 24 periods of ALD-Al 2 O 3 (three cycles)/-Y 2 O 3 (three cycles) [30]. Y(EtCp) 3 /deionized H 2 O and TMA/deionized H 2 O were used as co-reactants for constituents of Y 2 O 3 and Al 2 O 3 , respectively. Y(EtCp) 3 denotes tris(ethylcyclopentadienyl) yttrium and TMA denotes trimethylaluminum. Two samples of different stacking orders were prepared. Figure 1 shows the schematics of sample A and B, where A has the ALD-Y 2 O 3 (three cycles) as the first layer and B has the ALD-Al 2 O 3 (three cycles) as the first layer. We used in situ RHEED to monitor the sample growth. For comparison, we studied the growth of pure ALD-Y 2 O 3 and -Al 2 O 3 films from the initial stage to a nm thickness. After the ALD, all of the samples were taken out from the UHV system and annealed to 900 °C in a helium atmosphere for 30-60 s using RTA. We characterized the crystalline structure of the samples using synchrotron radiation X-ray diffraction (SR-XRD) (Huber, Rimsting, Germany) at the National Synchrotron Radiation Research Center (NSRRC). We studied the detailed atomic packing by highresolution high-angle annular dark-field (HAADF) and annular bright-field (ABF) spherical aberration (Cs)-STEM located at National Taiwan University (NTU) and the Industrial Technology Research Institute (ITRI). The STEM experiments were performed on an aberration-corrected (0.9 Å probe size) JEOL 2100F scanning transmission electron microscope (JEOL, Tokyo, Japan), operated at an accelerating voltage of 200 kV at NTU, and spherical aberration (Cs) corrected STEM (JEOL, JEM-ARM200F, Tokyo, Japan), operated at an accelerating voltage of 200 kV, at the Industrial Technology Research Institute (ITRI). We prepared the STEM samples using mechanical polishing and a focused ion beam (FIB) at the Taiwan Semiconductor Research Institute (TSRI).

Results and Discussion
The high-resolution HAADF-STEM image (Figure 2a) of sample A (Figure 1a) in the asdeposited condition shows an amorphous ALD snl layer on crystalline GaAs(111)A. No diffraction peaks except for the GaAs(111) and (222) reflections were found in the SR-XRD specular scan, shown in Figure 2d, which confirmed the lack of a long-range order of the deposited film along the surface normal. The Ga-As dumbbell pairs in the GaAs(111)A epilayer are clearly resolved in the STEM image and the corresponding crystalline orientation is labeled in the figure. Note that the atomic stacking is always terminated with Ga in GaAs(111)A substrate and with the epitaxial growth. We studied the surface electronic structure of the epi-GaAs(111)A using in situ synchrotron radiation photoelectron spectroscopy (SRPES), confirming the Ga termination [36]. After the ALD, all of the samples were taken out from the UHV system and annealed to 900 • C in a helium atmosphere for 30-60 s using RTA. We characterized the crystalline structure of the samples using synchrotron radiation X-ray diffraction (SR-XRD) (Huber, Rimsting, Germany) at the National Synchrotron Radiation Research Center (NSRRC). We studied the detailed atomic packing by high-resolution high-angle annular dark-field (HAADF) and annular bright-field (ABF) spherical aberration (Cs)-STEM located at National Taiwan University (NTU) and the Industrial Technology Research Institute (ITRI). The STEM experiments were performed on an aberration-corrected (0.9 Å probe size) JEOL 2100F scanning transmission electron microscope (JEOL, Tokyo, Japan), operated at an accelerating voltage of 200 kV at NTU, and spherical aberration (Cs) corrected STEM (JEOL, JEM-ARM200F, Tokyo, Japan), operated at an accelerating voltage of 200 kV, at the Industrial Technology Research Institute (ITRI). We prepared the STEM samples using mechanical polishing and a focused ion beam (FIB) at the Taiwan Semiconductor Research Institute (TSRI).

Results and Discussion
The high-resolution HAADF-STEM image (Figure 2a) of sample A (Figure 1a) in the as-deposited condition shows an amorphous ALD snl layer on crystalline GaAs(111)A. No diffraction peaks except for the GaAs(111) and (222) reflections were found in the SR-XRD specular scan, shown in Figure 2d, which confirmed the lack of a long-range order of the deposited film along the surface normal. The Ga-As dumbbell pairs in the GaAs(111)A epilayer are clearly resolved in the STEM image and the corresponding crystalline orientation is labeled in the figure. Note that the atomic stacking is always terminated with Ga in GaAs(111)A substrate and with the epitaxial growth. We studied the surface electronic structure of the epi-GaAs(111)A using in situ synchrotron radiation photoelectron spectroscopy (SRPES), confirming the Ga termination [36]. It is noteworthy that a periodic array of a single atomic layer or two is visible at the heterointerface above the topmost layer of GaAs dumbbells. The periodic array is very likely a Y-O atomic template from the first three ALD cycles using Y-precursor Y(EtCp)3 and water. The Y-O bonded with Ga, forming Y-O-Ga. The in-plane symmetry of GaAs(111)A, a thin Y-O adatom array, and thicker Y2O3 was studied using in situ RHEED, and will be discussed later.
In comparison, the high-resolution HAADF-STEM image of sample B in the as-deposited condition ( Figure 2b) shows an amorphous ALD snl layer on the crystalline GaAs(111)A similar to that in sample A; however, no periodic array of adatoms on top of GaAs(111)A was observed, different from what was observed in sample A. Note that sample B in Figure 1b has 24 periods of ALD-Y2O3 and -Al2O3 in an snl structure, similar to sample A, but with the initial layer being three cycles of ALD-Al2O3. The Ga-As dumbbells in the GaAs(111)A epilayer are clearly resolved and remain intact from the bulk to the top surface of the substrate in sample A in both unfiltered and filtered images ( Figure 2a). In sample B, the Ga-As dumbbells in GaAs(111)A are also clearly resolved; however, the very top dumbbells are blurred in the unfiltered image. The filtered image revealed the existence of As, which was bonded with some atoms, whose contrast is not as clear as that of Ga in sample A. See and compare the unfiltered and filtered images in Figure 2a,b. The TMA and H2O of the three ALD cycles in sample B could interact with the top Ga, resulting in a less ordered interface structure.
The in situ RHEED patterns upon the initial growth of ALD Y2O3 on the GaAs(111)A revealed ordered and crystalline Y2O3 from 3 to 10 cycles, similar to our earlier work reported in Ref. [23]. The bottom panels of Figure 2c show sharp, streaky reconstructed RHEED patterns of the clean epi-GaAs(111)A-(2 × 2) with Kikuchi arcs. With 3-cycle ALD-Y2O3 deposition (middle panels), faint broad It is noteworthy that a periodic array of a single atomic layer or two is visible at the hetero-interface above the topmost layer of GaAs dumbbells. The periodic array is very likely a Y-O atomic template from the first three ALD cycles using Y-precursor Y(EtCp) 3 and water. The Y-O bonded with Ga, forming Y-O-Ga. The in-plane symmetry of GaAs(111)A, a thin Y-O adatom array, and thicker Y 2 O 3 was studied using in situ RHEED, and will be discussed later.
In comparison, the high-resolution HAADF-STEM image of sample B in the as-deposited condition ( Figure 2b) shows an amorphous ALD snl layer on the crystalline GaAs(111)A similar to that in sample A; however, no periodic array of adatoms on top of GaAs(111)A was observed, different from what was observed in sample A. Note that sample B in Figure 1b has 24 periods of ALD-Y 2 O 3 and -Al 2 O 3 in an snl structure, similar to sample A, but with the initial layer being three cycles of ALD-Al 2 O 3 . The Ga-As dumbbells in the GaAs(111)A epilayer are clearly resolved and remain intact from the bulk to the top surface of the substrate in sample A in both unfiltered and filtered images (Figure 2a). In sample B, the Ga-As dumbbells in GaAs(111)A are also clearly resolved; however, the very top dumbbells are blurred in the unfiltered image. The filtered image revealed the existence of As, which was bonded with some atoms, whose contrast is not as clear as that of Ga in sample A. See and compare the unfiltered and filtered images in Figure 2a,b. The TMA and H 2 O of the three ALD cycles in sample B could interact with the top Ga, resulting in a less ordered interface structure.
The in situ RHEED patterns upon the initial growth of ALD Y 2 O 3 on the GaAs(111)A revealed ordered and crystalline Y 2 O 3 from 3 to 10 cycles, similar to our earlier work reported in Ref. [23].
The bottom panels of Figure 2c show sharp, streaky reconstructed RHEED patterns of the clean epi-GaAs(111)A-(2 × 2) with Kikuchi arcs. With 3-cycle ALD-Y 2 O 3 deposition (middle panels), faint broad streaks, superimposed with the underlying GaAs pattern, were observed. After 10 cycles (top panels), the broad streaks became dominant, which manifested the crystalline nature of the thicker Y 2 O 3 layer. The RHEED k-spacing in the thin ALD oxide is slightly larger than that of epi-GaAs, marked by dashed lines. The trend continues with 10-cycle ALD-Y 2 O 3 , where no Kikuchi arcs were observed.
Note that three to five cycles of ALD-Y 2 O 3 as the initial deposition are employed to give a complete coverage, namely 1 monolayer, on GaAs surface, which may not be attained with two cycles of ALD-Y 2 O 3 [37]. The more cycles of ALD-Y 2 O 3 or ALD-Al 2 O 3 layer may cause less mixing uniformity. Therefore, three cycles of ALD-Y 2 O 3 and ALD-Al 2 O 3 are suitable for the purpose of homogenous chemical composition and attaining the monolayer limit at the initial growth stage. The k-spacing in the RHEED patterns increases from GaAs (111) (Figure 2b). The as-deposited condition of sample A showed a periodic adatom array, whereas that of sample B revealed no visible periodic adatom array. Moreover, the surface Ga-As dumbbells interacted with the 3-cycle of ALD-Al 2 O 3 differently from those with the 3-cycle of ALD-Y 2 O 3 in the as-deposited sample A, whose top GaAs dumbbells remained intact upon 3-cycle ALD-Y 2 O 3 deposition. The HAADF-STEM images show amorphous ALD snl multilayered thin films for both samples A and B in the as-deposited condition.
Previously, we studied the initial chemical bonding of the first half-and one-cycle ALD-Y 2 O 3 on GaAs(001)-4 × 6 on an atomic scale using in situ SRPES [39]. Y(EtCp) 3 precursors reside on the faulted As atoms and undergo a charge transfer to the bonded As atoms. The next ALD half-cycle of H 2 O molecules removes the bonded As atoms, and the oxygen atoms bond with the Ga atoms underneath, forming Y-O-Ga bonding at the interface. Y-O-Ga prevented the interdiffusion between Y 2 O 3 and GaAs upon RTA to high temperatures, regardless of ALD-or MBE-Y 2 O 3 , as evidenced from the attainment of low D it and a low electrical leakage current in the Y 2 O 3 /GaAs(001)-4 × 6 [40,41]. Note that the top surface atoms of GaAs(111)A are Ga [36,42], while those of (001)-4 × 6 are As [38,43]. Therefore, forming Y-O-Ga bonding is easier for ALD-Y 2 O 3 on GaAs(111)A than that on GaAs(001). Figure 3a. The hetero-interface was atomically ordered and morphologically smooth and sharp. The corresponding SR-XRD normal scans are shown in Figure 4a.  In comparison, the structure of sample B after 900 °C RTA was not uniform. Island-like H-YAP crystalline grains formed on parts of the substrate, while the rest were amorphous, as shown in a highresolution transmission electron microscopy (HRTEM) image (Figure 3b). The interface of the annealed sample B is not as smooth as that of the annealed sample A, particularly in the amorphous region. The HAADF-STEM image of a crystalline H-YAP region in the annealed sample B is shown in Figure 3c, which reveals an atomic-scale microstructure of a similar quality to that of sample A. Nonetheless, the H-YAP region exhibits an island-like morphology, similar to what is exhibited in Figure 3b.

RTA to 900 • C transformed the amorphous snl ALD multilayer in sample A to single-crystal single-domain H-YAP, as shown in an HAADF-STEM image in
The XRD radial scans along the surface normal of samples A and B after RTA to 900 °C for 30 and 60 s are illustrated in Figure 4a,b. The abscissa denotes the scattering vector, q, whose magnitude is 4π sin(2θ/2)/λ, where 2θ and λ are the scattering angle and X-ray wavelength, respectively. Apart from the intense GaAs (111)    In comparison, the structure of sample B after 900 °C RTA was not uniform. Island-like H-YAP crystalline grains formed on parts of the substrate, while the rest were amorphous, as shown in a highresolution transmission electron microscopy (HRTEM) image (Figure 3b). The interface of the annealed sample B is not as smooth as that of the annealed sample A, particularly in the amorphous region. The HAADF-STEM image of a crystalline H-YAP region in the annealed sample B is shown in Figure 3c, which reveals an atomic-scale microstructure of a similar quality to that of sample A. Nonetheless, the H-YAP region exhibits an island-like morphology, similar to what is exhibited in Figure 3b.
The XRD radial scans along the surface normal of samples A and B after RTA to 900 °C for 30 and 60 s are illustrated in Figure 4a,b. The abscissa denotes the scattering vector, q, whose magnitude is 4π sin(2θ/2)/λ, where 2θ and λ are the scattering angle and X-ray wavelength, respectively. Apart from the intense GaAs (111)   In comparison, the structure of sample B after 900 • C RTA was not uniform. Island-like H-YAP crystalline grains formed on parts of the substrate, while the rest were amorphous, as shown in a high-resolution transmission electron microscopy (HRTEM) image (Figure 3b). The interface of the annealed sample B is not as smooth as that of the annealed sample A, particularly in the amorphous region. The HAADF-STEM image of a crystalline H-YAP region in the annealed sample B is shown in Figure 3c, which reveals an atomic-scale microstructure of a similar quality to that of sample A. Nonetheless, the H-YAP region exhibits an island-like morphology, similar to what is exhibited in Figure 3b.
The XRD radial scans along the surface normal of samples A and B after RTA to 900 • C for 30 and 60 s are illustrated in Figure 4a,b. The abscissa denotes the scattering vector, q, whose magnitude is 4π× sin(2θ/2)/λ, where 2θ and λ are the scattering angle and X-ray wavelength, respectively. Apart from the intense GaAs(111) and (222) Bragg peaks, four additional diffraction peaks were observed, where the peak locations agree well with the expected positions of the H-YAP(0002), (0004), (0006), and (0008) reflections, verifying the c-axis-oriented H-YAP. Moreover, the extensive persistence of Pendellösung fringes near the H-YAP reflections in sample A provides additional evidence of its sharper interfaces and better crystallinity compared with those in sample B. These XRD observations are consistent with the STEM results.
The high-resolution HAADF-and ABF-STEM images (Figure 5a,b) provide valuable information about the bonding at the interface and inside the H-YAP. The images are identical for the whole film, which is H-YAP in the annealed sample A and the H-YAP domains in sample B. We overlaid the atomic models of H-YAP and GaAs on both figures. The good match between the models and the imaged atomic columns indicates that the constituent atoms were located at the expected positions, having the expected bonding length and angle. We observed the Ga-As pairs and the Y atoms periodically located in the hexagonal YAP lattices. The observed interplanar distances between neighboring Y atomic planes along both lateral [0110] and normal [0001] directions are well matched with those in the atomic models, which were drawn using the VESTA software [44]. In the HAADF image, the intensity scattered by an atom scales with the atomic number Z as Z 1.7 [45]. Therefore, the HAADF-STEM image contrast is dominated by heavier Y, Ga, and As atoms; the light Al and O atoms, which are situated between Y atoms, are barely observable in Figure 5a.
Nanomaterials 2020, 10, x 7 of 10 the whole film, which is H-YAP in the annealed sample A and the H-YAP domains in sample B. We overlaid the atomic models of H-YAP and GaAs on both figures. The good match between the models and the imaged atomic columns indicates that the constituent atoms were located at the expected positions, having the expected bonding length and angle. We observed the Ga-As pairs and the Y atoms periodically located in the hexagonal YAP lattices. The observed interplanar distances between neighboring Y atomic planes along both lateral [011 0] and normal [0001] directions are well matched with those in the atomic models, which were drawn using the VESTA software [44]. In the HAADF image, the intensity scattered by an atom scales with the atomic number Z as Z 1.7 [45]. Therefore, the HAADF-STEM image contrast is dominated by heavier Y, Ga, and As atoms; the light Al and O atoms, which are situated between Y atoms, are barely observable in Figure 5a. In the ABF-STEM images, with black atom contrast, the heavier atoms correspond to dots with a larger size and darker colors. Although ABF-STEM images exhibit smaller image contrast than the HAADF-STEM ones, both light and heavy atomic columns are visible simultaneously [46]. The ABF-STEM image thus shows not only the Ga-As pairs of the epi-GaAs and the Y of H-YAP films, but also, critically, the Al and O sites. The enlarged ABF-STEM image of an interface region of annealed sample A is shown in Figure 5b, in which the largest and darkest dots are the heaviest Y atoms. Between the Y layers stacked along the H-YAP c-axis, the gray and smaller dots reveal the locations of the Al atom and match those in the atomic models, as depicted in Figure 5a,b.
The oxygen atoms have the smallest size and are the most difficult to observe among all the elements in this work. The dark features extended from the nearby metal atoms in H-YAP can be identified as oxygen atoms, as indicated by the red dots in Figure 5b. Furthermore, the clouded region between the H-YAP film and GaAs substrate indicates the existence of O atoms at the interface In the ABF-STEM images, with black atom contrast, the heavier atoms correspond to dots with a larger size and darker colors. Although ABF-STEM images exhibit smaller image contrast than the HAADF-STEM ones, both light and heavy atomic columns are visible simultaneously [46]. The ABF-STEM image thus shows not only the Ga-As pairs of the epi-GaAs and the Y of H-YAP films, but also, critically, the Al and O sites. The enlarged ABF-STEM image of an interface region of annealed sample A is shown in Figure 5b, in which the largest and darkest dots are the heaviest Y atoms. Between the Y layers stacked along the H-YAP c-axis, the gray and smaller dots reveal the locations of the Al atom and match those in the atomic models, as depicted in Figure 5a,b.
The oxygen atoms have the smallest size and are the most difficult to observe among all the elements in this work. The dark features extended from the nearby metal atoms in H-YAP can be identified as oxygen atoms, as indicated by the red dots in Figure 5b. Furthermore, the clouded region between the H-YAP film and GaAs substrate indicates the existence of O atoms at the interface between Y and Ga atoms. We have elucidated the interfacial atomic arrangements, namely only O atoms being between Y and Ga atoms. The observations indicate that the interfacial bonding between H-YAP and GaAs(111)A is Y-O-Ga, which comes from the initial single atomic layer of periodic Y-O. The sharp interface also indicated that Y-O-Ga bonding is very strong and stable and sustains 900 • C RTA without detectable deterioration.

Conclusions
The substrate surface initiates and determines the epi-growth, which involves deposited films and the substrates underneath. It is rare to see epi-growth on a periodic array of adatoms, which is not part of the native substrate surface. In this work, we utilized atomic layer deposition (ALD), which is a self-limiting growth method with an atomic accuracy and conformal coverage, to tailor the configuration of the adatoms on the GaAs(111)A surface. The first film deposition of three cycles of ALD-Y 2 O 3 gave a periodic array of atoms with a thickness of a monolayer or two at the interface; this was observed using careful studies of HAADF-and ABF-STEM, in situ RHEED, and our previous work of the in situ synchrotron radiation photoemission. The Y-O monolayer, not GaAs, induced the epi-growth of single-crystal hexagonal yttrium aluminum perovskite (H-YAP) by annealing the amorphous sub-nano-laminated (snl) ALD-Al 2 O 3 /-Y 2 O 3 multilayers. The single-crystal H-YAP was uniform over the STEM-studied area in the GaAs wafer, and had an atomically sharp interface with the substrate. The persistence of the Pendellösung fringes of X-ray scattering radial scans along the surface normal near the H-YAP reflections over a large q range indicate sharp interfaces and excellent crystalline structures. In comparison, the amorphous snl ALD-Y 2 O 3 /-Al 2 O 3 multilayer on GaAs(111)A with the initial three ALD cycles of Al 2 O 3 resulted in an annealed film consisting of amorphous regions and island-like single-crystal H-YAP grains.
With the same substrate of GaAs(111)A, the initial film depositions of 3 ALD cycles of Y 2 O 3 or Al 2 O 3 produced films of entirely epitaxial single-crystal H-YAP with an atomically smooth interface or mixed amorphous/single-crystal regions, respectively. It is the periodic Y-O adatom array, rather than the GaAs substrate, which induced the epi-growth of single-crystal H-YAP. Single-crystal H-YAP was successfully grown on GaN, and is expected to be grown on Si using the method discussed here. Our results may lead to novel epitaxial growth by tailoring the substrate surfaces using a foreign atomic-layer thick periodic adatom array for producing desirable phases, opening up a new chapter in hetero-epitaxy.