Epitaxial TiOx Surface in Ferroelectric BaTiO
Transcript Of Epitaxial TiOx Surface in Ferroelectric BaTiO
Epitaxial TiOx Surface in Ferroelectric BaTiO3:
Native Structure and Dynamic Patterning at the
Atomic Scale
Maya Barzilay,1,2 Tian Qiu,3 Andrew M. Rappe3 and Yachin Ivry1,2,*
1Department of Materials Science and Engineering, Technion – Israel Institute of Technology, Haifa 3200003, Israel 2Solid State Institute, Technion – Israel Institute of Technology, Haifa 3200003, Israel 3Department of Chemistry, University of Pennsylvania, Philadelphia, Pennsylvania 191046323, U.S.A. *Correspondence to: [email protected]
Abstract Surfaces and interfaces of ferroelectric oxides exhibit enhanced functionality, and therefore serve as a platform for novel nano and quantum technologies. Experimental and theoretical challenges associated with examining the subtle electro-chemo-mechanical balance at metal-oxide surfaces have hindered the understanding and control of their structure and behavior. Here, we combine advanced electronmicroscopy and first-principles thermodynamics methods to reveal the atomic-scale chemical and crystallographic structure of the surface of the seminal ferroelectric BaTiO3. We show that the surface is composed of a native < 2-nm thick TiOx rock-salt layer in epitaxial registry with the BaTiO3. Using electron-beam irradiation, we successfully patterned artificially TiOx sites with sub-nanometer resolution, by inducing Ba escape. Therefore, our work offers electro-chemo-mechanical insights into ferroelectric surface behavior in addition to a method for scalable high-resolution beam-induced chemical lithography for selectively driving surface phase transitions, and thereby functionalizing metal-oxide surfaces.
1
1. Introduction:
The ability of oxides to endure extreme mechanical chemical and thermal conditions has been intriguing for researchers from a broad range of disciplines, e.g., earth science, nuclear engineering, space technology and dental care.[1–3] Subtle electro-chemo-mechanical balance at complex-oxide surfaces and interfaces allows the formation of phases with structural and functional characteristics that differ from the bulk.[4–6] The unique electric and magnetic properties of such phases arise from their divergence from stoichiometry, leading to variations in oxidation states of the participating ions. Hence, functional-oxide surfaces and interfaces constitute a rich platform for novel phases that are attractive for high-performance miniaturized electronic devices[7,8] as well as for chemical catalyses.[9]
Because ferroelectric oxides comprise regions with varying crystallographic and electric polarization orientations, there has been a growing interest in their outer surface and domain wall functionality, which is expressed as enhanced conductivity,[10–12] magnetism[4] and even superconductivity.[13] The chemical origin of such functional behavior is typically attributed to either oxygen vacancy dynamics[14– 17] or cation segregation,[18] while other studies look at the effects of intrinsic symmetry breaking.[4] Despite the accumulated knowledge on domain walls, the structure and behavior of ferroelectric surfaces, which are responsible for domain stabilization and are attractive for e.g. nano lithography[19– 23] and catalysis,[9,24–27] has remained elusive. Specifically, the longstanding challenge in understanding how the surface mediates between the absence of electric and mechanical fields in the vacuum and the polarization, and strain in the bulk is not merely experimental or theoretical, but even conceptual.[28] Hence, computational methods that were developed to explain the electro-chemical[29] and electromechanical[30] atomic-scale interactions in ferroelectrics have been adopted to describe experimental observations of the surface behavior. For example, Tsurumi et al.[31,32] combined dielectric measurements and DFT calculations to demonstrate that nanoparticles of the seminal non-toxic ferroelectric, BaTiO3 organize in a core-shell structure that helps release strain. These authors suggested that the unit cells at the shell (a few nanometers thick) assume a cubic structure that helps mediate between the tetragonal symmetry of the core and the vacuum. Likewise, a collaborative experimental (STM) – computational (DFT thermodynamics) study demonstrated the stability of various titaniumoxide terminations during surface reconstruction in BaTiO3.[33] By contrast, although transition electron
2
microscopy (TEM) provides us with significant input regarding domain-wall structure and functionality at the atomic scale,[34–40] such TEM characterization of the surface is lacking. Consequently, experimental or theoretical data regarding the surface formation and dynamics or even the surface structure, is absent.
2. Results and Discussion:
Combining advanced and conventional electron-microscopy techniques for structural, chemical and oxidation-state analyses with computational methods, we looked at the native chemical, ionic and crystallographic structure of BaTiO3 in ca. 50-nm particles. Moreover, by increasing the dose exposure we used the electron beam to excite the surface, allowing us to image its real-time dynamics. Such particles are on the one hand large enough to be considered bulk-like,[31] while on the other hand they are thin enough to allow atomic-resolution TEM imaging. We demonstrate that the native surface of the tetragonal BaTiO3 crystal is composed of a non-stoichiometric (i.e. high defect concentration) nearly cubic titanium oxide phase (TiOx, x≈1). We also show that by increasing the dose exposure of the material to the electron beam, the electron beam can be used for chemical patterning with sub-nanometer controllability. That is, using the electron beam as a source for exerting localized electric field and heat, we induced Ba escape contactless and expanded the native TiOx phase by will with nearly atomic resolution on the cost of the bulk BaTiO3. Our imaging methods allowed us not only to realize the Ba escape mechanism, but also to quantify the process.
To reveal the structure near the native surface, we imaged with high-resolution transition electron microscopy (HRTEM) the particles from both {100} and {110} zone axes (ZAs) as seen in Figures 1ac and 1d-f, respectively. We also performed high-angle annular dark-field imaging (HAADF- STEM) from these ZAs (Figures 1c, f), in which the brightness of each atomic column is proportional to the atomic weight squared (𝑍2), allowing us to deduce the chemical structure near the surface. The electron micrographs clearly show a difference between the bulk tetragonal BaTiO3 unit cells and a ca. 1-nm thick terrace-like epitaxial structure of different chemical and crystallographic footprints. The HAADF images (Figures 1c, g) indicate that although columns of both Ti and Ba appear clearly in the bulk region (oxygen atoms are too light to be detected in these images), the surface region contains only Ti columns
3
and no Ba. A careful look at the surface in the HRTEM images, where the oxygen columns can be detected reveals neighboring columns with different contrast only from {110} ZA (Fig. 1e-h) and not from the {100} ZA (Fig. 1a-d). Combining the information from the HAADF-STEM and HRTEM methods suggests that the columns are composed of alternating Ti-O columns that can be observed separately only from {110} ZA. Likewise, the inter-atomic distance is similar for both ZAs (eliminating the possibility of rutile or anatase structures). Therefore, we can conclude confidently that the native surface of the BaTiO3 is a rock-salt-like titanium oxide. This conclusion is in agreement with our measurement of the Ti-Ti inter-atomic distance at the interface (2.09±0.005 Å, see Fig. 1b, f), which is substantially smaller than in BaTiO3 (3.99-4.04 Å[41] ) and is in agreement with the 4.18 Å lattice parameter of TiO[42] (we should note that this value is much smaller than e.g. the 5.54 Å lattice parameter of BaO).
Figure 1| Epitaxial TiO surface layer in BaTiO3 crystals. (a) High-resolution TEM of the BaTiO3 surface from a [010] zone axis. (b) A closer look at the area highlighted in (a) shows the rock-salt structure of the TiO surface as well as the epitaxial growth of the TiO surface on the perovskite BaTiO3 crystal. (c) HAADF image of a different grain from a similar zone axis. (d) Schematic illustration of the [100] non-stoichiometric TiOx surface – BaTiO3 bulk native structure. The complementary (e) large-scale and (f) closer-look high-resolution TEM images as well as (g) HAADF image taken from the [110] zone axis shows the BaTiO3 crystal and rock-salt TiO from a perpendicular orientation. (h) Schematic illustration of the [110] nonstoichiometric TiOx surface – BaTiO3 bulk native structure. The border between the bulk BaTiO3 and the TiO surface is highlighted (orange lines in a, c, e and g), while the location of Ba, Ti and O atoms is designated in (b) and (f).
4
To support our chemical analysis, we used a novel microscopy system that allows us to perform atomic-resolution electron-diffraction x-ray spectroscopy (EDX) mapping[43] simultaneously with HAADF imaging at the same region. Figure 2 shows such cross-identification of the Ba, Ti and oxygen columns. Likewise, to further support our structural analysis, we used integrated differential phase contrast (iDPC)[44] imaging, which is a method that has recently been developed for mapping the location of individual atomic columns with very high accuracy (also simultaneously with the HAADF and EDX imaging). In this method, the contrast is proportional to the atomic weight (∝ Z), allowing for high detectability of heavy and light atoms alike (see the Experimental Section for more details about the iDPC and EDX imaging). Figure 2h shows the iDPC signal from a small area near the surface (same area as in Fig. 1c), verifying the exact location of the rock-salt-like TiOx atomic columns (as well as of the perovskite BaTiO3 structure).
Figure 2| Cross-chemical and structural analysis of the BaTiO3 crystal and TiO surface. (a) HAADF image of the BaTiO3 crystal and its surface. (b) Atomic-scale chemical mapping (EDX) showing higher Ti concentration and lower Ba concentration at the surface (here, the signal is not of an instantaneous beam detection. Rather, the signal is integrating over a large time, hence deteriorating the imaging stability and the spatial resolution). (c) High-accuracy location mapping (iDPC) of the same area demonstrates the oxygen-titanium rock-salt structure of the TiO surface. Simultaneous high-resolution chemical analysis of the (d) barium, (e) titanium and (f) oxygen EDX signals as well as the combined chemical analysis (g) and high-precision location mapping (iDPC) are given as a reference. The areas in (a) from which the surface and bulk images were taken are highlighted, while the HAADF image of the same surface area is given in Figure 1c.
5
We should note that the universality of the existence of a native TiOx layer at the surface of BaTiO3 was confirmed by a large-area panoramic view (Supporting Information, Figure 3), while the appearance of a TiOx surface was observed in all the different BaTiO3 crystals we examined (> 50 different crystals), including the five different particles that are presented in this paper. These crystals arrived from different sources and were prepared by different methods (see Experimental Section). Moreover, revisiting existing literature (e.g. Figure 5 from Zhu et al.[45]), we believe that the TiOx surface has been observed previously, but the chemical and crystallographic structure have not yet been identified probably due to the lack of the novel imaging methods that have been used in the current study. Finally, the thickness of the native TiOx surface layer is ca. 1 nm suggests that there are only one to three monolayers, and hence this observation is in agreement with the previous observations of titanium-oxide termination in BaTiO3,[33] which typically cannot determine the exact structure of the inner layers.
Figure 3| Long-range presence of the native epitaxial TiOx surface of BaTiO3 crystals. High-precision atomic-scale
mapping (iDPC) of the native TiOx surface over a long range. The panoramic image is composed of four iDPC images.
The deviation from TiO stoichiometry in a rock-salt structure (TiOx with 1.3 > x > 0.7) leads to high defect concentration.[46] Using the 3D nature of TEM imaging, we identified three types of defects at the very thin native surface: stacking fault, twinning and vacancies. These defects appear clearly in Figure 4. Moreover, using electron energy-loss spectroscopy (EELS) signals, we were able to identify the chemical shift[47] of the Ti peaks between the ions at the bulk and the ions at the surface. This shift corresponds to the variations in the Ti oxidation state, while no shift was observed in either the Ba or O peaks. Figure 5 shows that the position of the Ti cation peaks recorded from the bulk BaTiO3 corresponds to Ti4+. However, the signal recorded from the surface demonstrates that the peak shifts towards lower energies, indicating on the existence of Ti2+, as expected from the TEM imaging.[47–49]
6
Figure 4| Defects in the native TiOx surface. Highlighted (red arrows) crystallographic defects in the TiO surface at the areas presented (a) in Figure 1a and (b) in Figure 1d.
Figure 5| Chemical shift between the bulk and surface Ti binding energies in BaTiO3. A comparison between the EELS signals from the bulk and from the surface (HAADF image of the areas from which the EELS signals were taken is given as a reference). Although no change was observed in the picks related to the Ba or to the O ions, a closer look (insert) at the L2 and L3 energies of the Ti ions show a chemical shift that corresponds to the reduction in oxidation state from Ti4+ in the perovskite BaTiO3 bulk towards the Ti2+ in rock-salt structure TiO surface.[47–49]
The experimental identification of a TiO epitaxial surface for native BaTiO3 indicates that TiO is a stable phase on the BaTiO3 surface. Because the stability of heterogeneous thin films depends strongly on the interfacial interactions and cannot be directly predicted from their bulk stability, we perform ab initio DFT calculations to determine the stability of different surfaces on BaTiO3 and to derive the stability map of this interface. To cover a large span of possible surface phases, we include all phases in
7
the Ba-Ti-O2 ternary system found in Materials Project[50] that can be epitaxially placed on BaTiO3 (001) surface. The energy of each phase Pi is expressed in Equation (1).
𝐸P𝑖 = 𝐸BTO+P𝑖 − 𝐸BTO
(1)
Where EBTO+Pi is the total energy of BaTiO3 with Pi phases on the surface, and EBTO is the energy of the bare BaTiO3 crystal. A phase is considered stable if none of the decomposition paths is spontaneous, i.e., when for any other three phases Pj, Pk, and Pl that satisfy Equation (2).
P𝑖 → 𝑛𝑗P𝑗 + 𝑛𝑘P𝑘 + 𝑛𝑙P𝑙
(2)
where nj, nk, and nl are stoichiometric numbers, the following inequality holds in Equation (3):
𝐸P𝑖 ≤ 𝑛𝑗𝐸P𝑗 + 𝑛𝑘𝐸P𝑘 + 𝑛𝑙𝐸P𝑙
(3)
where energies are calculated from Equation (1). Similarly, two phases Pi and Pj could coexist if, for any other two phases Pk and Pl , as given in Equation (4)
P𝑖 + 𝑛𝑗P𝑗 → 𝑛𝑘P𝑘 + 𝑛𝑙P𝑙
(4)
Pi and Pj have lower energy, as shown in Equation (5)
𝐸P𝑖 + 𝑛𝑗𝐸P𝑗 ≤ 𝑛𝑘𝐸P𝑘 + 𝑛𝑙𝐸P𝑙
(5)
Figure 6 shows that based on this method (more computational details can be found in Supporting Information), we not only show the stability of the native TiOx layer, but we also predict that a thicker layer may also be stable, while the stoichiometry (x) varies with the film thickness. This effect can be seen in the difference between the ternary diagrams of two atomic monolayers (Figure 6a) and four atomic monolayers (Figure 6c) of TiOx on the BaTiO3 surface. We think that introducing external excitations in the experimental conditions discussed here (i.e. TEM) a removal of oxygen from the system can be induced, driving it down from the point of BaTiO3 into a ternary region on the phase diagram. This ternary region evolves from BaTiO3-BaO-Ti4O5 to BaTiO3-BaO-TiO when increasing
8
thickness of overlayers, suggesting that Ti4O5 is the stable composition for the thinner TiOx layer (Figure 6Figure 6Figure 6a-b), while TiO is also stable for thicker layers (Figure 6c-d).
These results allow us to draw two main conclusions regarding the observed surface overlayers. First, TiO incurs less lattice mismatch penalty compared with Ti4O5 because the relative stability of TiO increases for thicker TiO layers. This conclusion is confirmed by comparing the lattice mismatch from their bulk states, where Ti4O5 has 5.0% lattice mismatch and TiO has 3.5% lattice mismatch. Secondly, Ti4O5 has a stronger interaction with the BaTiO3 (001) lattice plane. A consequence is that the TiOx layer closest to the BaTiO3 (e.g. the native TiOx) is Ti4O5, while subsequent layers of TiO may be induced. These results are in close agreement with our experimental measurements, while our prediction of the TiO-Ti4O5 interplay may also explain the observations of high defect concentration (Figure 4). Finally, these calculations allow us to predict that introducing external excitations to the BaTiO3-TiOx system may favor controlled growth of the pseudo-cubic TiOx (Ti4O5-TiO) near the native TiOx surface. We should note that Ti is smaller than Ba and hence presumably we expect that the growth of the surface will include barium oxide formation rather than titanium oxide formation. However, Ba ions doe not have many intermediate oxidation states and tend to be stable only as Ba2+. Hence, as soon as the surface becomes suboxide barium oxide would become unstable. On the other hand, Ti has a broader range of oxidation-state stability and can even be stable as Ti3+, e.g. in Ti2O3. This analysis is in agreement also with previous studies regarding BaTiO3 surface termination that show that under reducing conditions (i.e., when oxygen atoms are driven away), the Ba also goes away, leaving a Ti-rich level.[33]
Figure 6| Calculated phase-stability map of the BaTiO3 surface. (a) Phase diagram of phases with thickness of two layers on BaTiO3 surface. (b) Top view of Ti4O5 on BaTiO3 surface. (c) Phase diagram of phases with thickness of four layers on BaTiO3 surface, noticing that the ternary region BaTiO3-BaO-TiO replaces the ternary region BaTiO3-BaO-Ti4O5 observed in the two-layer case. (d) Top view of TiO on BaTiO3 surface.
9
Following the prediction that exciting the system can grow the TiOx phase, we used the electronbeam irradiation as a contactless source for such excitations at a length scale that is determined by the beam size. Because we knew that TEM imaging does not affect the surface, the attempts to excite the surface included increase of the electron-beam doses with respect to the doses that are used for TEM imaging. Figure 7a-d shows the growth of the TiOx surface on the expense of BaTiO3 as a function of electron-beam irradiation time over a large area. Calculating carefully (see Experimental Section and Fig. SI2) the irradiation of the high-resolution TEM electron beam, the dose was 0.003 nA nm-2. The growth of the TiOx surface as a function of dose exposure is clearly observed at the images. We should note that for very long exposure times, we can confidently assume that the TiOx grew significantly also at the direction parallel to the beam and not only perpendicular to the surface, so that a distinction between the bulk BaTiO3 and the TiOx surface is not so clear any more (see Figure 7d). Thanks to the accurate EDX measurements we were able to quantify the rate of Ba escape (Figure 7f-g), i.e. the dynamic BaTiO3-to-TiOx transition under a constant electron-beam irradiation. Here, the dose was much higher that in the TEM imaging: 80 nA nm-2 at 200 keV (see Experimental Section and Fig. SI2 for further details regarding the dose calculation), so that the Ba escape process was also much faster. Given the above theoretical analysis regarding the Ba escape, we can deduce that the electron-beam irradiation lowers the oxidation state of the metal atoms, so that to become more stable, there must be less oxygen atoms around them. This state of oxygen deficiency would favorite Ba escape over Ti escape, even though the Ba atoms are larger.
10
Native Structure and Dynamic Patterning at the
Atomic Scale
Maya Barzilay,1,2 Tian Qiu,3 Andrew M. Rappe3 and Yachin Ivry1,2,*
1Department of Materials Science and Engineering, Technion – Israel Institute of Technology, Haifa 3200003, Israel 2Solid State Institute, Technion – Israel Institute of Technology, Haifa 3200003, Israel 3Department of Chemistry, University of Pennsylvania, Philadelphia, Pennsylvania 191046323, U.S.A. *Correspondence to: [email protected]
Abstract Surfaces and interfaces of ferroelectric oxides exhibit enhanced functionality, and therefore serve as a platform for novel nano and quantum technologies. Experimental and theoretical challenges associated with examining the subtle electro-chemo-mechanical balance at metal-oxide surfaces have hindered the understanding and control of their structure and behavior. Here, we combine advanced electronmicroscopy and first-principles thermodynamics methods to reveal the atomic-scale chemical and crystallographic structure of the surface of the seminal ferroelectric BaTiO3. We show that the surface is composed of a native < 2-nm thick TiOx rock-salt layer in epitaxial registry with the BaTiO3. Using electron-beam irradiation, we successfully patterned artificially TiOx sites with sub-nanometer resolution, by inducing Ba escape. Therefore, our work offers electro-chemo-mechanical insights into ferroelectric surface behavior in addition to a method for scalable high-resolution beam-induced chemical lithography for selectively driving surface phase transitions, and thereby functionalizing metal-oxide surfaces.
1
1. Introduction:
The ability of oxides to endure extreme mechanical chemical and thermal conditions has been intriguing for researchers from a broad range of disciplines, e.g., earth science, nuclear engineering, space technology and dental care.[1–3] Subtle electro-chemo-mechanical balance at complex-oxide surfaces and interfaces allows the formation of phases with structural and functional characteristics that differ from the bulk.[4–6] The unique electric and magnetic properties of such phases arise from their divergence from stoichiometry, leading to variations in oxidation states of the participating ions. Hence, functional-oxide surfaces and interfaces constitute a rich platform for novel phases that are attractive for high-performance miniaturized electronic devices[7,8] as well as for chemical catalyses.[9]
Because ferroelectric oxides comprise regions with varying crystallographic and electric polarization orientations, there has been a growing interest in their outer surface and domain wall functionality, which is expressed as enhanced conductivity,[10–12] magnetism[4] and even superconductivity.[13] The chemical origin of such functional behavior is typically attributed to either oxygen vacancy dynamics[14– 17] or cation segregation,[18] while other studies look at the effects of intrinsic symmetry breaking.[4] Despite the accumulated knowledge on domain walls, the structure and behavior of ferroelectric surfaces, which are responsible for domain stabilization and are attractive for e.g. nano lithography[19– 23] and catalysis,[9,24–27] has remained elusive. Specifically, the longstanding challenge in understanding how the surface mediates between the absence of electric and mechanical fields in the vacuum and the polarization, and strain in the bulk is not merely experimental or theoretical, but even conceptual.[28] Hence, computational methods that were developed to explain the electro-chemical[29] and electromechanical[30] atomic-scale interactions in ferroelectrics have been adopted to describe experimental observations of the surface behavior. For example, Tsurumi et al.[31,32] combined dielectric measurements and DFT calculations to demonstrate that nanoparticles of the seminal non-toxic ferroelectric, BaTiO3 organize in a core-shell structure that helps release strain. These authors suggested that the unit cells at the shell (a few nanometers thick) assume a cubic structure that helps mediate between the tetragonal symmetry of the core and the vacuum. Likewise, a collaborative experimental (STM) – computational (DFT thermodynamics) study demonstrated the stability of various titaniumoxide terminations during surface reconstruction in BaTiO3.[33] By contrast, although transition electron
2
microscopy (TEM) provides us with significant input regarding domain-wall structure and functionality at the atomic scale,[34–40] such TEM characterization of the surface is lacking. Consequently, experimental or theoretical data regarding the surface formation and dynamics or even the surface structure, is absent.
2. Results and Discussion:
Combining advanced and conventional electron-microscopy techniques for structural, chemical and oxidation-state analyses with computational methods, we looked at the native chemical, ionic and crystallographic structure of BaTiO3 in ca. 50-nm particles. Moreover, by increasing the dose exposure we used the electron beam to excite the surface, allowing us to image its real-time dynamics. Such particles are on the one hand large enough to be considered bulk-like,[31] while on the other hand they are thin enough to allow atomic-resolution TEM imaging. We demonstrate that the native surface of the tetragonal BaTiO3 crystal is composed of a non-stoichiometric (i.e. high defect concentration) nearly cubic titanium oxide phase (TiOx, x≈1). We also show that by increasing the dose exposure of the material to the electron beam, the electron beam can be used for chemical patterning with sub-nanometer controllability. That is, using the electron beam as a source for exerting localized electric field and heat, we induced Ba escape contactless and expanded the native TiOx phase by will with nearly atomic resolution on the cost of the bulk BaTiO3. Our imaging methods allowed us not only to realize the Ba escape mechanism, but also to quantify the process.
To reveal the structure near the native surface, we imaged with high-resolution transition electron microscopy (HRTEM) the particles from both {100} and {110} zone axes (ZAs) as seen in Figures 1ac and 1d-f, respectively. We also performed high-angle annular dark-field imaging (HAADF- STEM) from these ZAs (Figures 1c, f), in which the brightness of each atomic column is proportional to the atomic weight squared (𝑍2), allowing us to deduce the chemical structure near the surface. The electron micrographs clearly show a difference between the bulk tetragonal BaTiO3 unit cells and a ca. 1-nm thick terrace-like epitaxial structure of different chemical and crystallographic footprints. The HAADF images (Figures 1c, g) indicate that although columns of both Ti and Ba appear clearly in the bulk region (oxygen atoms are too light to be detected in these images), the surface region contains only Ti columns
3
and no Ba. A careful look at the surface in the HRTEM images, where the oxygen columns can be detected reveals neighboring columns with different contrast only from {110} ZA (Fig. 1e-h) and not from the {100} ZA (Fig. 1a-d). Combining the information from the HAADF-STEM and HRTEM methods suggests that the columns are composed of alternating Ti-O columns that can be observed separately only from {110} ZA. Likewise, the inter-atomic distance is similar for both ZAs (eliminating the possibility of rutile or anatase structures). Therefore, we can conclude confidently that the native surface of the BaTiO3 is a rock-salt-like titanium oxide. This conclusion is in agreement with our measurement of the Ti-Ti inter-atomic distance at the interface (2.09±0.005 Å, see Fig. 1b, f), which is substantially smaller than in BaTiO3 (3.99-4.04 Å[41] ) and is in agreement with the 4.18 Å lattice parameter of TiO[42] (we should note that this value is much smaller than e.g. the 5.54 Å lattice parameter of BaO).
Figure 1| Epitaxial TiO surface layer in BaTiO3 crystals. (a) High-resolution TEM of the BaTiO3 surface from a [010] zone axis. (b) A closer look at the area highlighted in (a) shows the rock-salt structure of the TiO surface as well as the epitaxial growth of the TiO surface on the perovskite BaTiO3 crystal. (c) HAADF image of a different grain from a similar zone axis. (d) Schematic illustration of the [100] non-stoichiometric TiOx surface – BaTiO3 bulk native structure. The complementary (e) large-scale and (f) closer-look high-resolution TEM images as well as (g) HAADF image taken from the [110] zone axis shows the BaTiO3 crystal and rock-salt TiO from a perpendicular orientation. (h) Schematic illustration of the [110] nonstoichiometric TiOx surface – BaTiO3 bulk native structure. The border between the bulk BaTiO3 and the TiO surface is highlighted (orange lines in a, c, e and g), while the location of Ba, Ti and O atoms is designated in (b) and (f).
4
To support our chemical analysis, we used a novel microscopy system that allows us to perform atomic-resolution electron-diffraction x-ray spectroscopy (EDX) mapping[43] simultaneously with HAADF imaging at the same region. Figure 2 shows such cross-identification of the Ba, Ti and oxygen columns. Likewise, to further support our structural analysis, we used integrated differential phase contrast (iDPC)[44] imaging, which is a method that has recently been developed for mapping the location of individual atomic columns with very high accuracy (also simultaneously with the HAADF and EDX imaging). In this method, the contrast is proportional to the atomic weight (∝ Z), allowing for high detectability of heavy and light atoms alike (see the Experimental Section for more details about the iDPC and EDX imaging). Figure 2h shows the iDPC signal from a small area near the surface (same area as in Fig. 1c), verifying the exact location of the rock-salt-like TiOx atomic columns (as well as of the perovskite BaTiO3 structure).
Figure 2| Cross-chemical and structural analysis of the BaTiO3 crystal and TiO surface. (a) HAADF image of the BaTiO3 crystal and its surface. (b) Atomic-scale chemical mapping (EDX) showing higher Ti concentration and lower Ba concentration at the surface (here, the signal is not of an instantaneous beam detection. Rather, the signal is integrating over a large time, hence deteriorating the imaging stability and the spatial resolution). (c) High-accuracy location mapping (iDPC) of the same area demonstrates the oxygen-titanium rock-salt structure of the TiO surface. Simultaneous high-resolution chemical analysis of the (d) barium, (e) titanium and (f) oxygen EDX signals as well as the combined chemical analysis (g) and high-precision location mapping (iDPC) are given as a reference. The areas in (a) from which the surface and bulk images were taken are highlighted, while the HAADF image of the same surface area is given in Figure 1c.
5
We should note that the universality of the existence of a native TiOx layer at the surface of BaTiO3 was confirmed by a large-area panoramic view (Supporting Information, Figure 3), while the appearance of a TiOx surface was observed in all the different BaTiO3 crystals we examined (> 50 different crystals), including the five different particles that are presented in this paper. These crystals arrived from different sources and were prepared by different methods (see Experimental Section). Moreover, revisiting existing literature (e.g. Figure 5 from Zhu et al.[45]), we believe that the TiOx surface has been observed previously, but the chemical and crystallographic structure have not yet been identified probably due to the lack of the novel imaging methods that have been used in the current study. Finally, the thickness of the native TiOx surface layer is ca. 1 nm suggests that there are only one to three monolayers, and hence this observation is in agreement with the previous observations of titanium-oxide termination in BaTiO3,[33] which typically cannot determine the exact structure of the inner layers.
Figure 3| Long-range presence of the native epitaxial TiOx surface of BaTiO3 crystals. High-precision atomic-scale
mapping (iDPC) of the native TiOx surface over a long range. The panoramic image is composed of four iDPC images.
The deviation from TiO stoichiometry in a rock-salt structure (TiOx with 1.3 > x > 0.7) leads to high defect concentration.[46] Using the 3D nature of TEM imaging, we identified three types of defects at the very thin native surface: stacking fault, twinning and vacancies. These defects appear clearly in Figure 4. Moreover, using electron energy-loss spectroscopy (EELS) signals, we were able to identify the chemical shift[47] of the Ti peaks between the ions at the bulk and the ions at the surface. This shift corresponds to the variations in the Ti oxidation state, while no shift was observed in either the Ba or O peaks. Figure 5 shows that the position of the Ti cation peaks recorded from the bulk BaTiO3 corresponds to Ti4+. However, the signal recorded from the surface demonstrates that the peak shifts towards lower energies, indicating on the existence of Ti2+, as expected from the TEM imaging.[47–49]
6
Figure 4| Defects in the native TiOx surface. Highlighted (red arrows) crystallographic defects in the TiO surface at the areas presented (a) in Figure 1a and (b) in Figure 1d.
Figure 5| Chemical shift between the bulk and surface Ti binding energies in BaTiO3. A comparison between the EELS signals from the bulk and from the surface (HAADF image of the areas from which the EELS signals were taken is given as a reference). Although no change was observed in the picks related to the Ba or to the O ions, a closer look (insert) at the L2 and L3 energies of the Ti ions show a chemical shift that corresponds to the reduction in oxidation state from Ti4+ in the perovskite BaTiO3 bulk towards the Ti2+ in rock-salt structure TiO surface.[47–49]
The experimental identification of a TiO epitaxial surface for native BaTiO3 indicates that TiO is a stable phase on the BaTiO3 surface. Because the stability of heterogeneous thin films depends strongly on the interfacial interactions and cannot be directly predicted from their bulk stability, we perform ab initio DFT calculations to determine the stability of different surfaces on BaTiO3 and to derive the stability map of this interface. To cover a large span of possible surface phases, we include all phases in
7
the Ba-Ti-O2 ternary system found in Materials Project[50] that can be epitaxially placed on BaTiO3 (001) surface. The energy of each phase Pi is expressed in Equation (1).
𝐸P𝑖 = 𝐸BTO+P𝑖 − 𝐸BTO
(1)
Where EBTO+Pi is the total energy of BaTiO3 with Pi phases on the surface, and EBTO is the energy of the bare BaTiO3 crystal. A phase is considered stable if none of the decomposition paths is spontaneous, i.e., when for any other three phases Pj, Pk, and Pl that satisfy Equation (2).
P𝑖 → 𝑛𝑗P𝑗 + 𝑛𝑘P𝑘 + 𝑛𝑙P𝑙
(2)
where nj, nk, and nl are stoichiometric numbers, the following inequality holds in Equation (3):
𝐸P𝑖 ≤ 𝑛𝑗𝐸P𝑗 + 𝑛𝑘𝐸P𝑘 + 𝑛𝑙𝐸P𝑙
(3)
where energies are calculated from Equation (1). Similarly, two phases Pi and Pj could coexist if, for any other two phases Pk and Pl , as given in Equation (4)
P𝑖 + 𝑛𝑗P𝑗 → 𝑛𝑘P𝑘 + 𝑛𝑙P𝑙
(4)
Pi and Pj have lower energy, as shown in Equation (5)
𝐸P𝑖 + 𝑛𝑗𝐸P𝑗 ≤ 𝑛𝑘𝐸P𝑘 + 𝑛𝑙𝐸P𝑙
(5)
Figure 6 shows that based on this method (more computational details can be found in Supporting Information), we not only show the stability of the native TiOx layer, but we also predict that a thicker layer may also be stable, while the stoichiometry (x) varies with the film thickness. This effect can be seen in the difference between the ternary diagrams of two atomic monolayers (Figure 6a) and four atomic monolayers (Figure 6c) of TiOx on the BaTiO3 surface. We think that introducing external excitations in the experimental conditions discussed here (i.e. TEM) a removal of oxygen from the system can be induced, driving it down from the point of BaTiO3 into a ternary region on the phase diagram. This ternary region evolves from BaTiO3-BaO-Ti4O5 to BaTiO3-BaO-TiO when increasing
8
thickness of overlayers, suggesting that Ti4O5 is the stable composition for the thinner TiOx layer (Figure 6Figure 6Figure 6a-b), while TiO is also stable for thicker layers (Figure 6c-d).
These results allow us to draw two main conclusions regarding the observed surface overlayers. First, TiO incurs less lattice mismatch penalty compared with Ti4O5 because the relative stability of TiO increases for thicker TiO layers. This conclusion is confirmed by comparing the lattice mismatch from their bulk states, where Ti4O5 has 5.0% lattice mismatch and TiO has 3.5% lattice mismatch. Secondly, Ti4O5 has a stronger interaction with the BaTiO3 (001) lattice plane. A consequence is that the TiOx layer closest to the BaTiO3 (e.g. the native TiOx) is Ti4O5, while subsequent layers of TiO may be induced. These results are in close agreement with our experimental measurements, while our prediction of the TiO-Ti4O5 interplay may also explain the observations of high defect concentration (Figure 4). Finally, these calculations allow us to predict that introducing external excitations to the BaTiO3-TiOx system may favor controlled growth of the pseudo-cubic TiOx (Ti4O5-TiO) near the native TiOx surface. We should note that Ti is smaller than Ba and hence presumably we expect that the growth of the surface will include barium oxide formation rather than titanium oxide formation. However, Ba ions doe not have many intermediate oxidation states and tend to be stable only as Ba2+. Hence, as soon as the surface becomes suboxide barium oxide would become unstable. On the other hand, Ti has a broader range of oxidation-state stability and can even be stable as Ti3+, e.g. in Ti2O3. This analysis is in agreement also with previous studies regarding BaTiO3 surface termination that show that under reducing conditions (i.e., when oxygen atoms are driven away), the Ba also goes away, leaving a Ti-rich level.[33]
Figure 6| Calculated phase-stability map of the BaTiO3 surface. (a) Phase diagram of phases with thickness of two layers on BaTiO3 surface. (b) Top view of Ti4O5 on BaTiO3 surface. (c) Phase diagram of phases with thickness of four layers on BaTiO3 surface, noticing that the ternary region BaTiO3-BaO-TiO replaces the ternary region BaTiO3-BaO-Ti4O5 observed in the two-layer case. (d) Top view of TiO on BaTiO3 surface.
9
Following the prediction that exciting the system can grow the TiOx phase, we used the electronbeam irradiation as a contactless source for such excitations at a length scale that is determined by the beam size. Because we knew that TEM imaging does not affect the surface, the attempts to excite the surface included increase of the electron-beam doses with respect to the doses that are used for TEM imaging. Figure 7a-d shows the growth of the TiOx surface on the expense of BaTiO3 as a function of electron-beam irradiation time over a large area. Calculating carefully (see Experimental Section and Fig. SI2) the irradiation of the high-resolution TEM electron beam, the dose was 0.003 nA nm-2. The growth of the TiOx surface as a function of dose exposure is clearly observed at the images. We should note that for very long exposure times, we can confidently assume that the TiOx grew significantly also at the direction parallel to the beam and not only perpendicular to the surface, so that a distinction between the bulk BaTiO3 and the TiOx surface is not so clear any more (see Figure 7d). Thanks to the accurate EDX measurements we were able to quantify the rate of Ba escape (Figure 7f-g), i.e. the dynamic BaTiO3-to-TiOx transition under a constant electron-beam irradiation. Here, the dose was much higher that in the TEM imaging: 80 nA nm-2 at 200 keV (see Experimental Section and Fig. SI2 for further details regarding the dose calculation), so that the Ba escape process was also much faster. Given the above theoretical analysis regarding the Ba escape, we can deduce that the electron-beam irradiation lowers the oxidation state of the metal atoms, so that to become more stable, there must be less oxygen atoms around them. This state of oxygen deficiency would favorite Ba escape over Ti escape, even though the Ba atoms are larger.
10