Molecularly Thin BaTiO3 Nanosheets with Stable Ferroelectric Response

2D ferroelectrics provide great promise for future electronics owing to their unique properties at 2D limit. Although simple perovskite oxides (such as BaTiO3 and PbTiO3) are an important target for 2D ferroelectrics, controlled synthesis of 2D nanosheets still remains a challenge. Here, this issue is addressed by using a new chemical protocol; unilamellar Ti0.87O2 nanosheets can serve as templates, guiding the solvothermal synthesis of 2D BaTiO3 nanosheets with a molecularly thin thickness (< 3 nm). Raman spectroscopy, second‐harmonic generation, and transmission electron microscopy characterizations suggest the formation of tetragonal ferroelectric BaTiO3 nanosheets. Based on piezoresponse force microscopy, BaTiO3 nanosheets exhibit a clear ferroelectric response with a high Curie temperature (≈100 °C) even down to 1.8 nm thickness. This study provides new opportunities to investigate perovskite ferroelectrics with the critical thickness.


Introduction
Recent advancements in 2D nanosheets have brought about significant revival of ferroelectric science and technology; many candidates for 2D ferroelectricity have been theoretically predicted and experimentally verified in recent years. [1][2][3][4][5][6][7] Compared with bulk ferroelectrics, 2D fer roelectric nanosheets have many unique properties, such as a wide bandgap, [3,8] a negative piezoelectric constant [9] and high mechanical flexibility. [4] Owing to these distinct properties along with their mole cularly thin thickness, 2D ferroelectric nanosheets present the tantalizing pros pect of scaling electronic devices down to a truly atomic dimension, surpassing struc tures attainable by topdown approaches. Furthermore, layerbylayer engineering of 2D nanosheets provides a promising route for designing func tional heterostructures in which 2D ferroelectric nanosheets can be integrated with various functional nanosheets. [8,10] These 2D heterostructures have broad applications in memory, logic, sensing, optical, and energy harvesting devices.
Perovskite oxides are an important target for 2D ferroelec trics. As a prototype ferroelectric, simple perovskite oxides (such as BaTiO 3 and PbTiO 3 ) have long been central to fer roelectric research in bulk and thinfilm systems. However, due to the lack of appropriate layered compounds, simple perovskite oxides have not been considered as a target for 2D ferroelectrics; 2D nanosheets of nonlayered perovskites are inaccessible by typical exfoliation routes. For the synthesis of 2D nonlayered perovskite nanosheets, there have been many reports on wetchemical approaches; hydrothermal or solvothermal processes have been demonstrated as versatile methods to synthesize BaTiO 3 nanoplates or nanosheets. [11][12][13][14] In these approaches, the use of appropriate precursor tem plates often plays a key role in controlling the morphology of BaTiO 3 . Layered titanates such as H 1.07 Ti 1.73 O 4 ·H 2 O, [11,13] H 2 Ti 4 O 9 ·nH 2 O [14] and their exfoliated nanosheets [12] have been extensively used for precursor templates; owing to their unique lamellar structures and excellent ionexchange capa bilities, 2D nanoplates of layered titanates can be readily con verted into BaTiO 3 nanoplates or nanosheets via topochem ical reactions. Despite intensive studies on hydrothermal or solvothermal processes, achieving molecularly thin BaTiO 3 nanosheets is very difficult. The BaTiO 3 nanosheets synthe sized thus far have been rather thick (> 10 nm), which are not 2D ferroelectrics provide great promise for future electronics owing to their unique properties at 2D limit. Although simple perovskite oxides (such as BaTiO 3 and PbTiO 3 ) are an important target for 2D ferroelectrics, controlled synthesis of 2D nanosheets still remains a challenge. Here, this issue is addressed by using a new chemical protocol; unilamellar Ti 0.87 O 2 nanosheets can serve as templates, guiding the solvothermal synthesis of 2D BaTiO 3 nanosheets with a molecularly thin thickness (< 3 nm). Raman spectroscopy, second-harmonic generation, and transmission electron microscopy characterizations suggest the formation of tetragonal ferroelectric BaTiO 3 nanosheets. Based on piezoresponse force microscopy, BaTiO 3 nanosheets exhibit a clear ferroelectric response with a high Curie temperature (≈100 °C) even down to 1.8 nm thickness. This study provides new opportunities to investigate perovskite ferroelectrics with the critical thickness.
suitable for electrical characterization of 2D nanosheets with a truly atomic dimension.
The difficulties in synthesis of BaTiO 3 nanosheets motivated us to investigate new chemical protocols for designed synthesis of molecularly thin BaTiO 3 nanosheets. Here, we report the synthesis of molecularly thin BaTiO 3 nanosheets using a 2D templateassisted approach. Unilamellar Ti 0.87 O 2 nanosheets are composed of two TiO 6 octahedra, the key units of Tibased ferroelectrics, which make the nanosheet an ideal template with a critical thickness. Ti 0.87 O 2 nanosheets are used as a hard template, and the solvothermal process involving a Ti 0.87 O 2 nanosheet suspension with Ba(OH) 2 and ethanol can induce the transformation of titania to BaTiO 3 nanosheets with a molecularly thin thickness (<3 nm). We discuss various charac teristics of BaTiO 3 nanosheets based on detailed investigations of their structural and ferroelectric properties.

Template Synthesis
The solvothermal reaction of Ti 0.87 O 2 nanosheets with Ba(OH) 2 ·8H 2 O was used to synthesize BaTiO 3 nanosheets. We used largesize, unilamellar Ti 0.87 O 2 nanosheets (with a thick ness of ≈1 nm and lateral size of 5-10 µm) (Figure 1a, Figure S1, Supporting Information), which are suitable for 2D templates. A colloidal suspension of Ti 0.87 O 2 nanosheets was synthesized by delaminating a layered titanate (H 1.07 Ti 1.73 O 4 ·H 2 O) with a tetrabutylammonium hydroxide solution (TBAOH). [15] The solvothermal reaction was carried out under rather mild con ditions (50-100 °C), different from high reaction temperatures (100-240 °C) used in previous studies. [11][12][13][14] Considering the boiling point of water (100 °C) and ethanol (78.3 °C) and the azeotrope point (78.2 °C), some treatments should be called as "reflex"; however, for simplicity, all processes were described as "solvothermal". Figure 1b shows powder Xray diffraction (XRD) pat terns of samples synthesized at different temperatures for 10 h. The XRD patterns of the collected precipitates revealed the formation of the BaTiO 3 phase. The phase composi tion of the samples varied with the reaction temperature. At 50 °C, the samples contained BaTiO 3 and BaCO 3 , with small amounts of unreacted Ti 0.87 O 2 phase. As the reaction temper ature increased, the peak intensities of BaTiO 3 increased. For the samples synthesized at 60 and 80 °C, the XRD patterns can be identified as BaTiO 3 with good crystallinity, and small amounts of the BaCO 3 impurity were detected. At 100 °C, the peak intensities of BaCO 3 increased, while the peak intensities of the BaTiO 3 phase were reduced. We note that Ti 0.87 O 2 peaks were hardly noticeable in the XRD patterns of the samples syn thesized at above 60 °C; no significant amount of unreacted Ti 0.87 O 2 phase remained ( Figure S2, Supporting Information). These results indicate that the Ti 0.87 O 2 nanosheets were effec tively converted into the BaTiO 3 phase for the solvothermal treatments above 60 °C.
The morphological features were characterized by scanning electron microscopy (SEM) (Figure 1c). For the lowtemperature treatments at 50 and 60 °C, sheetlike products were obtained.
Clearly, the morphology was changed at 60 °C; SEM detected regular sheet morphologies, indicating templatelike reaction from Ti 0.87 O 2 nanosheets into BaTiO 3 nanosheets. As the reac tion temperature increased, the morphology of the samples varied from sheetlike to spherical particles; the solvothermal reaction above 80 °C resulted in a collapse of the sheetlike morphology. By considering the XRD and SEM results, the spherical particles could be identified as BaTiO 3 nanoparticles produced by the dissolutionprecipitation reaction.
To investigate the optimal conditions, solvothermal treat ments were carried out at 50-80 °C for 5-25 h (Figure 1d). Figure S3a (Supporting Information) shows powder XRD pat terns of samples synthesized at 60 °C for 5-25 h. According to the XRD data, BaTiO 3 nanosheets could be obtained at 60 °C even for 5 h. A prolonged reaction time resulted in improved crystallinity, but no significant change was observed for 25 h. As is evident from SEM image of 25 h case ( Figure S3b, Sup porting Information), nanoparticles and small nanosheets www.advelectronicmat.de appeared and were arranged on the surface of largesized nanosheets, indicating partial decomposition at the prolonged reaction time. Note that this solvothermal process was sensitive to both reaction time and temperatures. For mild conditions (at 50 °C for 5-10 h), precipitates contained unreacted Ti 0.87 O 2 phase ( Figure S4, Supporting Information). These results suggest that the optimal condition for synthesizing BaTiO 3 nanosheets can be estimated to be 60 °C for 10 h (reflux condi tion) (Figure 1d).

Structural Characterizations
The formation of BaTiO 3 nanosheets was confirmed by direct observations using atomic force microscopy (AFM) and transmission electron microscopy (TEM). For AFM characteri zations, we prepared colloidal suspensions by suspending the collected nanosheets in ethanol. We used the supernatant of the colloidal suspensions after centrifuge treatment and depos ited BaTiO 3 nanosheets on Si and SrTiO 3 :Nb substrates by the spincoating method (Figure 2a, Figure S5, Supporting Infor mation). AFM detected regular sheet morphologies of BaTiO 3 nanosheets without unexfoliated patches and impurities, as is also confirmed by piezoresponse force microscopy (PFM) measurements (discussed later). This indicates that the cen trifuge treatment effectively removes the unexfoliated patches and impurities. From the statical evaluation of the thickness variation (Figure 2b), the typical thicknesses of the obtained nanosheets were ranging from 1.4 to 1.8 nm, corresponding to the crystallographic thicknesses of two or three TiO 6 units. The difference between the experimental height and crystallo graphic thickness may stem from adsorbed water molecules. The lateral size was approximately 1-2 µm.
TEM and selectedarea electron diffraction (SAED) char acterizations ( Figure 2c) revealed the high crystallinity of 2D nanosheet. The SAED pattern exhibited clear diffraction spots as expected for a square lattice inherent from the tetragonal BaTiO 3 . The estimated lattice constant for the a axis (0.399 nm) was consistent with the typical value for tetragonal BaTiO 3 . A highangle annular darkfield scanning TEM (HAADFSTEM) image ( Figure 2d) clearly resolved the atomic structure of BaTiO 3 nanosheet; the cation columns (Ba and Ti) could be distinguished, indicating an ordered structure of tetragonal BaTiO 3 without cation vacancies.
The structural features of BaTiO 3 nanosheets were char acterized by Raman spectroscopy. Figure 2e shows a Raman spectrum of BaTiO 3 nanosheets synthesized at 60 °C for 10 h. For comparison, the reference spectrum of a BaTiO 3 powder sample with the size of 500 nm is shown. In our study, we have applied the microRaman spectroscopy to evaluate the composition of the nanosheets. The laser beam (with a diam eter about 1 µm) was focused on the sheetlike products with a 100× objective lens, giving a local analysis of the nanosheets while eliminating possible interferences from byproducts. In the BaTiO 3 nanosheets, the main features at 182, 254, 304, and 515 cm −1 were assigned to the A 1 (1TO), A 1 (2TO), B 1 (TO), and A 1 (3TO)+E(TO) modes, respectively. [16,17] In the nanosheets, all the modes became broader with reduced intensities. These changes can be explained by reduced dimensionality from 3D to 2D. Such 2D effect is most significant in the outofplane A 1 (TO) modes. In general, Raman activity is governed by the symmetry of a material (selection rules), which is different between the 3D and 2D forms. In the 2D nanosheets, the absence of long range interaction along caxis causes the reduced intensity of the outofplane modes. We note the appearance of the B 1 (TO) mode at 304 cm −1 , characteristic of the tetragonal ferroelec tric BaTiO 3 . Since the 304 cm −1 mode is inactive in the cubic paraelectric phase, the observed features suggest the possible formation of the tetragonal ferroelectric BaTiO 3 . The second harmonic generation (SHG) measurements also revealed the broken inversion symmetry, supporting the tetragonal ferro electric phase ( Figure S6, Supporting Information).
The surface chemistry was investigated using Xray photoelectron spectroscopy (XPS) ( Figure S7, Supporting The inset of (c) shows the SAED pattern. e) Raman spectrum of BaTiO 3 nanosheets synthesized at 60 °C for 10 h. The reference spectrum of a BaTiO 3 powder sample with the size of 500 nm (Sakai Chemical Industry, Co. Ltd.) is shown for comparison. All the measurements were carried out at room temperature.

www.advelectronicmat.de
Information). In the Ba3d 5/2 and Ti 2p spectra ( Figures S7a,b, Supporting Information), the main components were character istic of Ba 2+ and Ti 4+ states in the perovskite structure. The main peak of the Ba3d 5/2 corelevel at 778.1 eV was assigned to the bulkcoordinated Ba. The Ti 2p corelevel spectra ( Figure S7b, Supporting Information) revealed two main components at 457.6 and 463.6 eV due to Ti 4+ as in the perovskite structure. [18] In the Ba 3d, Ti 2p and O 1s core levels (Figures S7a−c, Sup porting Information), each spectrum also had a high binding energy component, which is thought to be due to water adsorp tion. [19] There are two competing sites for OH − adsorption: oxygen vacancies and surface layer. In this context, the Ti 2p and O 1s corelevel spectra did not detect any trace of Ti 3+ and oxygen vacancies. The dominant site for OH − is the chemisorp tion at the nanosheet surface. Raman spectroscopy is also sen sitive to the chemistry of BaTiO 3 since introduction of vacan cies into BaTiO 3 leads to significant changes of Raman spectra. We note the absence of vacancyinduced modes at 184 and 470 cm −1 , which are often observed in BaTiO 3 with Ti and oxygen vacancies. [20] These results indicate that the composi tion of the final products is closed to the stoichiometric BaTiO 3 .
Key to controlled synthesis of 2D BaTiO 3 nanosheets is the use of solvothermal reactions involving colloidal unilamellar nanosheets under a rather mild condition (60 °C for 10 h) (Figure 1d). In previous studies, hydrothermal or solvothermal reactions using layered titanium oxides and their exfoli ated nanosheets were investigated for controlled synthesis of BaTiO 3 ; these strategies relied on high reaction temperatures (100-240 °C) for long reaction times (12-24 h). [11][12][13][14] Such rather harsh conditions possibly cause collapse of the sheetlike mor phology via dissolution-recrystallization. In fact, the products were thick nanoplate or rodlike structures. In contrast, the 2D template used in this study was a unilamellar nanosheet; Ti 0.87 O 2 nanosheets possess fully exposed surfaces, which exhibit a high reactivity.
Based on these results, we consider the formation mecha nism of BaTiO 3 nanosheets during a lowtemperature solvo thermal reaction (reflux condition) (Figure 3). The conversion of Ti 0.87 O 2 nanosheets to BaTiO 3 is considered to be competi tive reactions between (i) templatelike reaction ( Figure 3a) and (ii) dissolution-recrystallization reaction (Figure 3b). In reac tion (i), Ba 2+ ions can absorb onto the negatively charged sur face of Ti 0.87 O 2 nanosheets, facilitating the rearrangements of TiO 6 units from edgesharing to cornersharing form. Then, the Ba 2+ ions are intercalated into cornersharing TiO 6 octahedra, inducing the templatelike reaction to BaTiO 3 nanosheets. In the Ti 0.87 O 2 nanosheet, the unit cell along the thickness direc tion comprises two edgesharing TiO 6 octahedra. This structure may regulate the critical thickness obtained by 2D template approach, which is consistent with the smallest thickness www.advelectronicmat.de (2 units) in this study. In reaction (ii), the solvothermal reac tion occurred in an alkaline condition with Ba(OH) 2 and TBAOH. In such a case, TBA + ions are capable of forming stable TBA + −O 2− complexes, [21] facilitating the splitting of the Ti 0.87 O 2 nanosheet into monomer TiO 6 units. Then, the TiO 6 units are connected along the a, b and caxis directions, and the perovskite lattice is formed by the dissolution-recrystal lization reaction. We note that the lateral size (1-2 µm) of the BaTiO 3 nanosheets was reduced compared to that (5-10 µm) of starting Ti 0.87 O 2 nanosheets. This indicates that a partial dissolution of Ti 0.87 O 2 nanosheets is involved during the solvo thermal reaction. In this context, we also checked the products synthesized at slightly higher temperature (70 °C for 10 h), and the soobtained nanosheets were rather thick, with thicknesses varying from 3 to 5 nm ( Figure S8, Supporting Information). For further harsh conditions at high temperatures (>80 °C), dissolution-recrystallization reactions dominated, promoting the growth of rod-like or spherical nanoparticles. These results indicate that the lowtemperature solvothermal reaction (reflux condition) using Ti 0.87 O 2 nanosheets is favorable for controlled synthesis of molecularly thin BaTiO 3 nanosheets.

Ferroelectric Properties
For characterization of the ferroelectricity, we performed PFM measurements on individual BaTiO 3 nanosheets with different thicknesses (Figure 4). BaTiO 3 nanosheet with a thickness of 3.0 nm exhibited a typical butterfly amplitude loop in the amplitude−voltage curve. From the phase−voltage curve, 180° phase flipping was observed at the coercive voltages, suggesting a good ferroelectric switching nature of BaTiO 3 nanosheets. In addition, the amplitude and the phase curves possessed iden tical coercive voltages (±2 V), indicating the ferroelectric nature of the 3.0 nmthick BaTiO 3 nanosheet. The effective piezo electric coefficient (d) is defined as d = 2Qε 0 εP, where Q is the electrostriction coefficient, ε 0 ε is the dielectric constant and P is the polarization. [22] These PFM responses, characteristic of ferroelectrics, could still be detected even down to 1.8 nm. The piezoelectric coefficients (d 33 ) for the 1.8 and 3.0 nm nanosheets were 23 and 33 pm V −1 , respectively, which are much higher than those of nanoparticles or nanocubues (1-2 pm V −1 ) [23,24] and even comparable to those of 100 nmthick thin films (30 pm V −1 ) [25] (Figure 4g). For further reduced thickness, 1.4 nm thick nanosheets exhibited degradation of the PFM amplitude, with no detectable hysteresis in the phase−voltage curve. These results indicate that stable ferroelectric characteristics are main tained at the thickness of 1.8 nm.
To check the stability of ferroelectricity, we performed tem peraturedependent PFM measurements on a 1.8 nmthick BaTiO 3 nanosheet (Figure 5a). For temperatures up to 100 °C, a ferroelectric hysteresis was observed, which started to deform above 100 °C. At 110 °C, the PFM response transformed into a paraelectric behavior without hysteresis, which is related to the ferroelectric transition. BaTiO 3 nanosheet possesses a high Curie temperature (T C ) value (≈100 °C), in marked contrast to BaTiO 3 nanostructures, such as nanoparticles, [27] nanocubes, [24] nanowires, [28] and thin films. [29] In BaTiO 3 nanoparticles, for example, the T C decreased with a decrease in the size; BaTiO 3 nanoparticles with size of less than 100 nm exhibited a dif fuse phase transition at 50-100 °C. We note that the T C value of the BaTiO 3 nanosheet is much higher than those of nano structures [24,27,28] and almost comparable to those of strain engi neered thin films (T C >100 °C). [30] Even in the molecularly thin form (1.8 nm thickness), BaTiO 3 nanosheets exhibited a stable ferroelectric response up to 100 °C.
Insulating property is also a key factor for the stability of fer roelectricity; previous experiments on ultrathin BaTiO 3 films often suffered from large leakage current. We investigated the insulating property of a 1.8 nmthick BaTiO 3 nanosheet with fer roelectric response ( Figure S9, Supporting Information). BaTiO 3 The measurements were carried out at room temperature. g) Thickness dependence of the piezoelectric coefficient (d 33 ) for BaTiO 3 nanosheets and various nanostructures. Data are included for nanoparticles, [23] nanocubes, [24] nanowires [26] and thin films. [25] www.advelectronicmat.de nanosheet showed a highly insulating nature with a low cur rent (<10 pA), which was in contrast to the highly conducting behavior of the SrRuO 3 substrate. From the I-V curve, we esti mated the bandgap energy. The observed insulating gap of 3.8 V roughly corresponds to the bandgap energy, which is much larger than the bulk value (≈3.2 eV). We also note that BaTiO 3 nanosheet showed strong dielectric endurance with a high breakdown voltage. For a positive voltage, breakdown occurred at +5.4 V (27 MV cm −1 ). Due to 2D confinement effects, 2D nanosheets cause both an enlarged band gap and improved breakdown strength, which are favorable for stable ferroelec tricity even in the molecularly thin form. To our best knowledge, the 1.8 nmthick BaTiO 3 nanosheet is the thinnest freestanding perovskite to experimentally confirm ferroelectricity so far.
In perovskite thin films, the polarization decreases or even disappears when the film thickness reaches several unit cells or a critical thickness. [33][34][35][36] Such behavior is often called the size effect, which is a longstanding problem in ferroelectricity. Despite significant progress in theoretical studies and nanoscale characterizations, the underlying mechanism behind the size effect and critical thickness is still debatable. The picture of the size effect has continued to evolve, from early studies suggesting intrinsic effects [34,35] to recent developments showing depolar izing field effects. [33] The polarization properties of perovskite thin films could also be influenced by many competing factors, such as intrinsic size effects, [34,35] deadlayer effects (local chem ical environments and defects/lowk layers at the interface), [37,38] substrate strains, [39] and electrical boundary conditions; [40,41] these combinations cause complicated behaviors in measured properties. In previous studies, the SrRuO 3 electrodes often suf fered from competing effects including the contact effects with an incomplete screening and processing issues such as strains and defects at the film/electrode interface. In particular, high temperature postannealing (>400 °C) caused large extrinsic effects arising from strains and intermixing of the film/elec trode. [42] A clear benefit of our approach is roomtemperature deposition without hightemperature annealing, which real izes an ideal nanocapacitor with a clean metallic SrRuO 3 elec trodes with low internal and contact resistances. Such a supe rior interface property was not specific to BaTiO 3 /SrRuO 3 , but also achieved in highk dielectric nanosheets (Ti 0.87 O 2 and Ca 2 Nb 3 O 10 ). [8] In these cases, the polarizing field can be applied efficiently to BaTiO 3 nanosheets. From density functional theory (DFT) calculations, Stengel et al. found that the relatively weak interaction between the films and electrodes enhanced the spontaneous polarization in BaTiO 3 films. [35] This situation is quite similar to our system. AFM and XPS characterizations ( Figures S5 and S7, Supporting Information) revealed the exist ence of adsorbed water layers on nanosheet surfaces, yielding the relatively weak interaction between the nanosheet and substrate. Since large piezoelectric coefficients are a signature for highly polarizable nature of BaTiO 3 nanosheets; the highly polarizable nature of BaTiO 3 nanosheets with the relatively weak interaction might compensate for the relatively inefficient electronic screening capabilities of SrRuO 3 , realizing the full potential of molecularly thin BaTiO 3 nanosheets.
Another factor that may explain the robust ferroelectricity is the charge screening by water absorption. Previous studies have shown that surface hydroxylation commonly occurs in oxides (including BaTiO 3 ), especially in the case of solutionbased synthetic routes, and that chemisorbed OH groups are very stable even at elevated temperatures (>400 °C) in an ultrahigh vacuum environment. [43] Recent DFT calculation studies [28,44] also revealed the important role of water absorption for ferro electricity in ultrathin BaTiO 3 films; water adsorption screens a significant amount of the polarization charge on the surface, reducing the depolarizing field relative to bare BaTiO 3 . From the AFM evaluation of the thickness, BaTiO 3 nanosheets studied here are covered with water molecules. Our study demonstrated that BaTiO 3 nanosheets with water absorption exhibit bulklike ferroelectricity even in the 1.8 nm thick (3 units), in agreement with previous DFT studies. This again suggests that water adsorption is quite effective for stabilizing the ferroelectricity.
For decades, the critical thickness for ferroelectricity has been extensively investigated via stateoftheart experimental tech niques, including PFM, TEM, Raman, and synchrotron XRD. These studies revealed that the polarization was significantly suppressed for thin films with <10 units (≈4 nm), and the fer roelectricity was still observed down to 4-7.5 units (1.6-3 nm) for BaTiO 3 , [29,45] 3-10 units (1.2-4 nm) for PbTiO 3 , [46,47] 1.5 units (≈0.6 nm) for PbZr 0.2 Ti 0.8 O 3 [48] and 7.5 units (≈3 nm) for BiFeO 3 . [49] Recent TEM studies also indicated possible absence of critical thickness and size effect in ultrathin perovskite Data are included for nanoparticles, [27] nanocubes, [24] nanowires, [28] and thin films. [29][30][31][32] www.advelectronicmat.de ferroelectric films. [49] Below 3 units (1.2 nm), the polarization was reduced but remained stable even for ultrathin PbZr 0.2 Ti 0.8 O 3 films (with ≈22 µC cm −2 for 2.5 units and ≈16 µC cm −2 for 1.5 units). However, the question of how thin perovskite layers can be retaining ferroelectricity has been challenging to address, because experimental studies require the synthesis of near perfect ultrathin perovskite layers and ways to profile the proper ties at the critical thickness. Our study provides a solution to these issues by using individual, freestanding BaTiO 3 nanosheets with the critical thickness (1.4-3.0 nm); the 1.8 nm thick (3 units) BaTiO 3 nanosheet is the thinnest freestanding perov skite to experimentally confirm ferroelectricity. The assembly of individual BaTiO 3 nanosheets on an atomically flat substrate enables characterization of intrinsic properties free from pro cessing damage and substrate effects (such as defects/lowk layers and strains). Our study provides new opportunities to investigate perovskite ferroelectrics with the critical thickness.

Conclusion
We have demonstrated a facile approach for synthesizing molecularly thin BaTiO 3 nanosheets using 2D Ti 0.87 O 2 nanosheets. Key to controlled synthesis of 2D BaTiO 3 nanosheets is the use of the solvothermal reaction involving colloidal unilamellar nanosheets under rather mild conditions (60 °C for 10 h). Such a lowtemperature solvothermal reaction (reflux condition) with 2D templates is favorable for the trans formation into molecularly thin BaTiO 3 nanosheets. Charac terizations by AFM, Raman spectroscopy, SHG, XPS and TEM revealed the formation of tetragonal BaTiO 3 nanosheets with controlled thicknesses ranging from 1.4 to 3.0 nm. We utilized PFM of individual BaTiO 3 nanosheets with different thick nesses to examine the ferroelectric behavior down to the critical thickness. Our study demonstrated that BaTiO 3 nanosheets with water absorption exhibit bulklike ferroelectricity even in the 1.8 nm thick (3units), in agreement with previous DFT studies. Evaluation of individual, isolated BaTiO 3 nanosheets on an atomically flat substrate enables probing of intrinsic properties free from the process damage and substrate effects, providing a detailed glimpse of the ferroelectric properties of perovskites at the critical thickness.

Experimental Section
Synthesis of Ti 0.87 O 2 Nanosheets: A colloidal suspension of Ti 0.87 O 2 nanosheets was prepared by delaminating a layered titanate according to well-established soft-chemical procedures. [15] The starting layered titanate (K 0. 8 (3 mL), and ethanol was then added to the solution, keeping the total amount at 15 mL. The Ba/Ti mole ratio was set to 1.1. The mixture was then transferred into a Teflon-lined, sealed stainless steel vessel (with an inner volume of 25 mL), and solvothermal treatments were carried out at 50-100 °C for 5-25 h. We used rather mild condition (50-100 °C), different from high reaction temperatures (100-240 °C) used in previous studies. [11][12][13][14] After the solvothermal treatments, the as-synthesized white powders were washed with an acetic acid aqueous solution for 3 h to remove BaCO 3 that was possibly formed as a byproduct. The final samples were dried in an oven at 80 °C for 12 h for further characterization. For film fabrication, colloidal suspensions were prepared by suspending the collected nanosheets in ethanol. To remove unexfoliated patches and impurities, the colloidal suspension was centrifuged at 4000 rpm for 15 min, and the supernatant was collected. By spin coating the obtained supernatant, the nanosheets were deposited on substrates (such as Si, SrTiO 3 :Nb, SrRuO 3 and Pt).
Characterization: The phase of the final samples were investigated by powder XRD (RIGAKU SmartLab) with Cu Kα (λ = 0.154 nm) radiation. The size and morphology of the samples were observed using SEM (JEOL JSM-7500F). The thickness of the nanosheets was characterized by AFM (Hitachi E-Sweep or Asylum Research MFP-3D). For AFM characterizations, the nanosheets were deposited on a Si substrate by spin coating the obtained supernatant. The structures of the nanosheets were investigated by TEM and HAADF-STEM. TEM images were measured with a Hitachi H-9000 microscope operating at 200 kV. HAADF-STEM images were collected with a JEOL JEM-ARM200F microscope operating at 80 kV.
Raman spectra were acquired using a micro-Raman system (Horiba-Jobin Yvon T64000) with a 514.5 nm excitation laser. The samples were transferred onto the Pt/Si substrates to improve the signal-to-noise ratio. SHG measurements were carried out by a micro-SHG system (LVmicro-VIII/SHG) with a 1064 nm excitation laser. XPS data were collected by an XPS system (ULVAC-PHI VersaProbe III) with an Al Kα radiation (hν = 1486.6 eV). The crystal structure was illustrated with the VESTA software package. [50] The ferroelectric property of the nanosheets was characterized by PFM (Asylum Research Cypher S) in the DART (dual AC resonance tracking) mode. [51,52] The off-field PFM measurements were performed using a Pt/Ir-coated Si cantilever; PFM hysteresis loops were measured by recording the PFM amplitude and phase signals after turning off the individual DC pulse. In this way, a true PFM signal could be elucidated from molecularly thin nanosheets while greatly reducing the electrostatic effect. The insulating property was examined by conducting AFM (Hitachi E-Sweep) with a Pt/Ir-coated Si cantilever.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.