Polarization and Dielectric Properties of BiFeO3-BaTiO3 Superlattice-Structured Ferroelectric Films

Superlattice-structured epitaxial thin films composed of Mn(5%)-doped BiFeO3 and BaTiO3 with a total thickness of 600 perovskite (ABO3) unit cells were grown on single-crystal SrTiO3 substrates by pulsed laser deposition, and their polarization and dielectric properties were investigated. When the layers of Mn-BiFeO3 and BaTiO3 have over 25 ABO3 unit cells (N), the superlattice can be regarded as a simple series connection of their individual capacitors. The superlattices with an N of 5 or less behave as a unified ferroelectric, where the BaTiO3 and Mn-BiFeO3 layers are structurally and electronically coupled. Density functional theory calculations can explain the behavior of spontaneous polarization for the superlattices in this thin regime. We propose that a superlattice formation comprising two types of perovskite layers with different crystal symmetries opens a path to novel ferroelectrics that cannot be obtained in a solid solution system.


Introduction
Chemical tuning of the dielectric, ferroelectric, and piezoelectric properties of perovskite oxides (ABO 3 ) is traditionally based on the formation of solid solutions. Lead zirconate titanate, Pb(Zr, Ti)O 3 , is representative, composed of ferroelectric PbTiO 3 in tetragonal symmetry, and antiferroelectric PbZrO 3 in rhombohedral symmetry [1,2]. In this system, the dielectric and piezoelectric properties are maximized near the compositiondriven phase boundary [2], called the morphotropic phase boundary (MPB) [3], between the tetragonal and rhombohedral structures. The similar materials strategy has provided an extremely high piezoelectric response [4,5] in solid solutions such as Pb(Mg, Nb)O 3 -PbTiO 3 and Pb(Zn, Nb)O 3 -PbTiO 3 , where an electric field (E) is considered to induce a rotation of spontaneous polarization (P s ) [6].
Recently, bismuth ferrite (BiFeO 3 ) [7,8] has attracted considerable attention because of its multiferroic nature [9,10], i.e., the simultaneous presence of ferroelectric P s and an incommensurate spin cycloid structure, even at room temperature. Bulk BiFeO 3 has a rhombohedral structure in space group R3c and possesses a large P s along the pseudo-cubic [111] c direction [11,12]. Moreover, BiFeO 3 exhibits an extremely high Curie temperature (T C ) of 830°C [11,12], which can provide piezoelectric devices operating at high temperatures. In analogy to Pb(Zr, Ti)O 3 , considerable efforts have been made to investigate the solid solutions of rhombohedral BiFeO 3 and other perovskites in tetragonal symmetry. The BiFeO 3 -BaTiO 3 system [13][14][15][16][17] has been widely studied mainly in ceramic form because the MPB is expected to appear between rhombohedral R3c and tetragonal P4mm. Detailed structural analysis reveals that an increase in the BaTiO 3 content causes a structural change from rhombohedral R3c to a pseudo-cubic structure [17], where the phase boundary is boundary is ambiguous at around 33% BaTiO3 content. Moreover, it has been reported that BiFeO3-BaTiO3 solid solutions do not have a ferroelectric nature in the BaTiO3 content range of 40-50% [18].
Another approach exploiting the interplay of two types of perovskite oxides is to build a superlattice by thin-film growth technology [19,20]. Epitaxially grown superlattices composed of BiFeO3 and BaTiO3 have been reported to show a high magnetoelectric coupling coefficient compared with pristine films or bulk ceramics of BiFeO3, where the interface plays a crucial role [21,22]. At present, ferroelectric and dielectric properties of BiFeO3-BaTiO3 superlattices have been reported in a few reports [23][24][25], and thereby the fundamental questions remain unanswered concerning how the layers of BiFeO3 and Ba-TiO3 are structurally and ferroelectrically coupled, and how the coupling of the two layers is activated.
In this paper, we report the crystal structure, polarization, and dielectric properties of superlattice-structured epitaxial thin films composed of BaTiO3 and BiFeO3 on singlecrystal SrTiO3 substrates prepared by pulsed laser deposition (PLD) (Figure 1). Here, we adopted Mn(5%)-doped BiFeO3 instead of BiFeO3 to avoid a considerable influence of oxygen vacancies on the polarization and leakage current properties [23,[26][27][28], because a trapping capability of oxygen vacancies by Mn 3+ at the Fe 3+ site, i.e., a strong attractive interaction between Mn 3+ and oxygen vacancy, inhibits the formation of an oxygen vacancy-rich layer at the interfaces. The total number of ABO3 unit cells were fixed at 600, and that of the BaTiO3 and Mn-BiFeO3 layers (N) varied from 300 down to 1 (Figure 1b), while the average composition of the entire superlattices remained unchanged, i.e., 50% Mn-BiFeO3-50% BaTiO3. We found that the samples for an N greater than 25 can be regarded as a simple series connection of their individual capacitors, while those for an N of 5 or less behave as a unified 'ferroelectric', where the BaTiO3 and Mn-BiFeO3 layers are structurally and electronically coupled.

Experimental
Thin films of BaTiO 3 [29] and Mn(5%)-doped BiFeO 3 [28,30], a (Ba 0.7 Sr 0.3 )TiO 3 buffer layer [29], and (Ba 0.1 Sr 0.9 )RuO 3 electrodes [31] were fabricated on (100) SrTiO 3 singlecrystal substrates (5 × 5 × 1 mm 3 ) by PLD (KrF excimer laser, λ = 248 nm) using ceramic targets. The details of the deposition conditions are summarized in Supplementary Tables S1 and S2. Figure 1 displays the schematic of the superlattice composed of Mn-BiFeO 3 and BaTiO 3 . The total number of ABO 3 unit cells was fixed at 600 (Figure 1a). The number (N) of those in each layer forming the superlattice varied: N = 300, 50, 25, 10, 5, 3, and 1. Figure 1b depicts the structure of N = 3 as an example, where the superlattice is constructed by an alternate stacking of the thin layers of BaTiO 3 and Mn-BiFeO 3 with three ABO 3 unit cells (N = 3). For all the samples, the Mn-BiFeO 3 layer was deposited on the bottom electrode because of its better in-plane lattice matching with it. As a result, the layer just beneath the top electrode was the BaTiO 3 layer. During the deposition, the following condition was adopted: a substrate temperature T sub of 640 • C, an oxygen pressure (Po 2 ) of 2.6 Pa, and a laser repetition rate of 1 Hz for BaTiO 3 and 7 Hz for Mn-BiFeO 3 . The diameter of the top electrode was 0.1 mm. The polarization electric field (P-E) hysteresis properties were measured at 25 • C (3 kHz); the direction from the bottom to the top electrode was defined as positive for E and P.
High-resolution X-ray diffraction (XRD) reciprocal space maps (RSMs) were observed by using a Cu-Kα 1 source. The data of the intensity profile I i (q x , q z ) of reflection i in the reciprocal space (q x , q z ) were used for the detailed analysis of the lattice parameters of the in-plane (a) and the out-of-plane (c) directions, where the parameters a and c denote those of the pseudo-cubic ABO 3 unit cell. Throughout this paper, we adopted the pseudo-cubic notation unless otherwise stated.

DFT Calculations
Density functional theory (DFT) calculations were conducted using the generalized gradient approximation [32] with a plane wave basis set. We used the projector-augmented wave method [33] as implemented in the Vienna ab initio simulation package (VASP) [34]. We employed the Perdew-Burke-Ernzerhof gradient-corrected exchange correlation functional revised for solids (PBEsol) [35] and a plane wave cut-off energy of 520 eV. A Γ centered k-point mesh was used, and the details are provided later. Within the simplified generalized gradient approximation (GGA)+U approach [36], we added on-site Coulomb interaction parameters of U−J of 6 eV to Fe-3d throughout the calculations. As the spin configuration in BiFeO 3 can be approximated as the G-type antiferromagnet [37], we set the spin arrangement in which the adjacent Fe ions have an antiparallel spin configuration as much as possible. The experimental results for Mn-doped BiFeO 3 films reveal that the crystal symmetry and the spontaneous polarization (P s ) are not influenced by the doping of Mn up to 10%, and therefore we considered BiFeO 3 instead of Mn-doped BiFeO 3 for simplicity.
For building a superlattice cell, we took the following lattice constraint. Based on the experimental results of XRD for an N of 5 or less, the superlattice cell had a tetragonal structure with the lattice parameters of in-plane a DFT and out-of-plane c DFT in space group P4mm; its a DFT was fixed at the experimental a of 0.3985 nm, i.e., a DFT = a (experiment). The parameter c DFT is given by the following equation, c DFT = Nc BiFeO 3 + Nc BaTiO 3 , where c BiFeO 3 denotes the parameter c of the BiFeO 3 unit cell, and c BaTiO 3 that of the BaTiO 3 unit cell. The c BiFeO 3 and c BaTiO 3 were determined from the lattice volumes (V) derived from the geometrical optimizations of the BaTiO 3 cell (5 × 5 × 5 k-point) and the BiFeO 3 cell in P4mm symmetry. Considering the antiparallel spin configuration, we performed the optimization calculation of the BiFeO 3 cell with 2c BiFeO 3 (5 × 5 × 3 k-point) and regarded the half cell with c BiFeO 3 as the BiFeO 3 unit cell. For imposing the antiparallel spin configuration for N = 1, the long lattice with 2c super was taken as the superlattice cell, as depicted in Figure 2a. The structural optimizations were performed under a fixed a DFT and c DFT with 5 × 5 × 3 k-point mesh for all the supercells. From the structural parameters of the optimized cell, we obtained the atomic displacements (∆z) from the corresponding positions in the hypothetical non-polar paraelectric lattice. We also calculated the Born effective charges (Z*) [38] in the superlattice cells by density-functional perturbation theory. We estimated P s , as expressed by the following equation: where m i denotes the site multiplicity of the constituent atom i, and ∆z i ·Z * i is its dipole moment. The summation in Equation (1) is taken over the superlattice cell with the cell volume (V).
where denotes the site multiplicity of the constituent atom i, and ∆ • * is its dipole moment. The summation in Equation (1) is taken over the superlattice cell with the cell volume (V).

Crystal Structure
Supplementary Figure S1 shows the θ-2θ XRD patterns around the 002 reflection. In addition to the peaks of the SrTiO3 substrate at 46.5°, the (Ba0.1Sr0.9)RuO3 electrodes at 46.4 °, and the (Ba0.7Sr0.3)TiO3 buffer at 44.7°, the sample with N = 300 exhibits peaks individual to the layers of BaTiO3 and Mn-BiFeO3 because their layers are sufficiently thick for providing their corresponding reflections. With decreasing N, the integrated intensities of these peaks are weakened and eventually vanish for an N less than 5.
Supplementary Figure S2 shows the wide-area XRD-RSMs for N = 300 and 5. For N = 300 ( Figure S2b), the apparent reflections of 3/2 3/2 1/2 and 1/2 1/2 3/2 of Mn-BiFeO3 in monoclinic symmetry appear, whereas those were not observed for an N of 5 ( Figure S2d) or less. Figure 3 shows the integrated intensity of the 1/2 1/2 3/2 reflection as a function of N. With decreasing N, the intensity is weakened and then zero for N = 1-5. These results indicate that the monoclinic distortion, similar to the bulk (rhombohedral), is maintained in the Mn-BiFeO3 layer for the superlattice with N ≥ 10, while that is lost with N ≤ 5. The details of the structural analysis are described in Supplementary Note 2.

Crystal Structure
Supplementary Figure S1 shows the θ-2θ XRD patterns around the 002 reflection. In addition to the peaks of the SrTiO 3 substrate at 46.5 • , the (Ba 0.1 Sr 0.9 )RuO 3 electrodes at 46.4 • , and the (Ba 0.7 Sr 0.3 )TiO 3 buffer at 44.7 • , the sample with N = 300 exhibits peaks individual to the layers of BaTiO 3 and Mn-BiFeO 3 because their layers are sufficiently thick for providing their corresponding reflections. With decreasing N, the integrated intensities of these peaks are weakened and eventually vanish for an N less than 5.
Supplementary Figure S2 shows the wide-area XRD-RSMs for N = 300 and 5. For N = 300 ( Figure S2b), the apparent reflections of 3/2 3/2 1/2 and 1/2 1/2 3/2 of Mn-BiFeO 3 in monoclinic symmetry appear, whereas those were not observed for an N of 5 ( Figure S2d) or less. Figure 3 shows the integrated intensity of the 1/2 1/2 3/2 reflection as a function of N. With decreasing N, the intensity is weakened and then zero for N = 1-5. These results indicate that the monoclinic distortion, similar to the bulk (rhombohedral), is maintained in the Mn-BiFeO 3 layer for the superlattice with N ≥ 10, while that is lost with N ≤ 5. The details of the structural analysis are described in Supplementary Note 2. Figure 4 shows the high-resolution XRD-RSMs around the 103 reflections. For all the samples, the peak positions (q x , q z ) exhibit the following features: the (Ba 0.7 Sr 0.3 )TiO 3 buffer and the (Ba 0.1 Sr 0.9 )RuO 3 electrodes have an apparently small q x compared with the SrTiO 3 substrate, demonstrating that the parameter a of the (Ba 0.7 Sr 0.3 )TiO 3 buffer is sufficiently expanded to the bulk value, and also that the (Ba 0.1 Sr 0.9 )RuO 3 bottom electrode is coherently grown on the buffer. The detailed structural analysis for N = 300 (Figure 4a along with the 113 reflection; see Supplementary Note 2) indicates that the Mn-BiFeO 3 layer has a rhombohedral-like monoclinic M A structure. The splitting into two peaks of the 103 reflection of the Mn-BiFeO 3 layer stems from the ferroelastic domain variants. With further decreasing N, the splitting of the Mn-BiFeO 3 layer is smaller, and then the reflection can be regarded as a single peak for N = 25 and 10. At the same time, the q z of the Mn-BFO layer with N = 50, 25, and 10 becomes larger than that of N = 300, suggesting a structural change from the M A to monoclinic M B phases owing to an in-plane tensile strain (see Supplementary Note 2). The experimental results, i.e., the single peak of the 103 reflection, the q z shift, and the apparent 1/2 1/2 3/2 reflection (Figure 3), indicate that the Mn-BiFeO 3 layer for N = 25 and 10 has a pseudo-tetragonal structure, with a small monoclinic (M B ) distortion [39]. We note that for an N less than 5, the reflections from the Mn-BiFeO 3 and BaTiO 3 layers cannot be distinguished. These results enable us to consider that the superlattice has a unified tetragonal cell with a c/a of 1.01-1.02 as an average structure.  Figure 4 shows the high-resolution XRD-RSMs around the 103 reflections. For all the samples, the peak positions (qx, qz) exhibit the following features: the (Ba0.7Sr0.3)TiO3 buffer and the (Ba0.1Sr0.9)RuO3 electrodes have an apparently small qx compared with the SrTiO3 substrate, demonstrating that the parameter a of the (Ba0.7Sr0.3)TiO3 buffer is sufficiently expanded to the bulk value, and also that the (Ba0.1Sr0.9)RuO3 bottom electrode is coherently grown on the buffer. The detailed structural analysis for N = 300 (Figure 4a along with the 113 reflection; see Supplementary Note 2) indicates that the Mn-BiFeO3 layer has a rhombohedral-like monoclinic MA structure. The splitting into two peaks of the 103 reflection of the Mn-BiFeO3 layer stems from the ferroelastic domain variants. With further decreasing N, the splitting of the Mn-BiFeO3 layer is smaller, and then the reflection can be regarded as a single peak for N = 25 and 10. At the same time, the qz of the Mn-BFO layer with N = 50, 25, and 10 becomes larger than that of N = 300, suggesting a structural change from the MA to monoclinic MB phases owing to an in-plane tensile strain (see Supplementary Note 2). The experimental results, i.e., the single peak of the 103 reflection, the qz shift, and the apparent 1/2 1/2 3/2 reflection (Figure 3), indicate that the Mn-BiFeO3 layer for N = 25 and 10 has a pseudo-tetragonal structure, with a small monoclinic (MB) distortion [39]. We note that for an N less than 5, the reflections from the Mn-BiFeO3 and BaTiO3 layers cannot be distinguished. These results enable us to consider that the superlattice has a unified tetragonal cell with a c/a of 1.01-1.02 as an average structure.   Figure 5 shows the P-E loops (E//[001] c at 3 kHz), and Figure 6a,b display the resultant remanent polarization (P r ) and the maximum polarization (P max ) at the highest positive E as a function of N, respectively. It is interesting to note that the superlattice samples exhibit an apparent ferroelectric polarization with an apparent P r , which is completely different from the solid solutions in the same composition (50% BaTiO 3 content) featuring a non-ferroelectric nature [18]. The N = 300 sample has a P r of 22 µC cm −2 . The P-E loop exhibits an imprint, i.e., a shift in the negative E direction. This behavior is assumed to stem from a flexoelectric effect [29,40,41], where a strain gradient in the out-of-plane direction in the ferroelectric layer stabilizes the upward polarization compared with the downward one. Compared with the buffered electrode with a = 0.3986 nm, the BaTiO 3 layer has the same a, whereas the Mn-BiFeO 3 layer possesses a slightly small a = 0.3965 nm. This result indicates that a strain gradient driving the flexoelectric effect is present in the Mn-BiFeO 3 layer adjacent to the boundary with the bottom electrode.   Figure 5 shows the P-E loops (E//[001]c at 3 kHz), and Figure 6a,b display th ant remanent polarization (Pr) and the maximum polarization (Pmax) at the highest E as a function of N, respectively. It is interesting to note that the superlattice exhibit an apparent ferroelectric polarization with an apparent Pr, which is com different from the solid solutions in the same composition (50% BaTiO3 content) f a non-ferroelectric nature [18]. The N = 300 sample has a Pr of 22 μC cm −2 . The exhibits an imprint, i.e., a shift in the negative E direction. This behavior is ass stem from a flexoelectric effect [29,40,41], where a strain gradient in the out-of-p rection in the ferroelectric layer stabilizes the upward polarization compared downward one. Compared with the buffered electrode with a = 0.3986 nm, the layer has the same a, whereas the Mn-BiFeO3 layer possesses a slightly small a nm. This result indicates that a strain gradient driving the flexoelectric effect is p From the data shown in Figure 6a-c, we think that the polarization and dielectric behavior can be divided into three regions: I. the simple series connection of the capacitors (N ≥ 25, see Figure 7a), II. the transition region (10 ≤ N < 25), and III. the unified ferroelectric regime (N < 10, see Figure 7b). In region I, with decreasing N, the hysteresis is slanted, and the resultant P r and P max are monotonically reduced (Figure 6a,b). We note that the relative dielectric permittivity (ε r ) remains constant at~120. This constant ε r can be understood in terms of a simple series connection of the capacitors of the Mn-BiFeO 3 and the BaTiO 3 layers. Considering an ε r of 399 for the Mn-BiFeO 3 capacitor, and that of 93 for the BaTiO 3 one (those were measured individually for their respective capacitors), we obtain ε r~1 50 (=2ε r (BaTiO 3 )·ε r( Mn-BiFeO 3) /[ε r (BaTiO 3 )+ε r (Mn-BiFeO 3 )]). This is qualitatively in good agreement with the experiment (ε r~1 20). In region III, with decreasing N, the P r is reduced, while the ε r is higher.  From the data shown in Figure 6a-c, we think that the polarization and dielectri behavior can be divided into three regions: I. the simple series connection of the capacitor (N ≥ 25, see Figure 7a), II. the transition region (10 ≤ N < 25), and III. the unified ferroelec tric regime (N < 10, see Figure 7b). In region I, with decreasing N, the hysteresis is slanted and the resultant Pr and Pmax are monotonically reduced (Figure 6a,b). We note that th relative dielectric permittivity (εr) remains constant at ~120. This constant εr can be unde  From the data shown in Figure 6a-c, we think that the polarization and di behavior can be divided into three regions: I. the simple series connection of the cap (N ≥ 25, see Figure 7a), II. the transition region (10 ≤ N < 25), and III. the unified fer tric regime (N < 10, see Figure 7b). In region I, with decreasing N, the hysteresis is s and the resultant Pr and Pmax are monotonically reduced (Figure 6a,b). We note t BaTiO3 layers. Considering an εr of 399 for the Mn-BiFeO3 capacitor, and that of 93 for the BaTiO3 one (those were measured individually for their respective capacitors), we obtain εr~150 (=2εr(BaTiO 3 )·εr ( Mn-BiFeO3 ) /[εr(BaTiO 3 )+εr(Mn-BiFeO3)]). This is qualitatively in good agreement with the experiment (εr~120). In region III, with decreasing N, the Pr is reduced, while the εr is higher.  Figure 7 shows the schematics of the superlattice structures along with the Ps com ponent along the out-of-plane direction (Ps//[001]c). In region I (N ≥ 25), the presence of the 1/2 1/2 3/2 reflection from the Mn-BiFeO3 layer ( Figure 3) and the polarization and dielec tric properties ( Figure 6) indicate that the superlattice can be regarded as the simple series connection of the capacitors of BaTiO3 and Mn-BiFeO3. In the BaTiO3 layer, the Ps vector is present along [001]c; our DFT calculations reveal that the Ps strength is 28.5 μC cm −2 which is close to the bulk value [42]. In contrast, the Mn-BiFeO3 layer has a Ps nearly along [111]c, and the value is reported to be 90-100 μC cm −2 [37]. As the polarization components along [001]c in these layers are markedly different, the interface effect plays an importan role. It is assumed that the interface region of several to several tens of unit cells in width needs to accommodate the difference in the direction and strength of the Ps vector across it, as in ferroelastic domain walls [43][44][45][46][47][48][49][50]. As a result, a depolarization field (Edep.) is buil up in the interface region, where the Edep. is present in a direction that prevents the change in the polarization component. Given that the Ps vectors are switched by an E application the Pr is expected to be ~40 μC cm −2 . The Pr of 25 μC cm −2 for N = 300 is smaller than this expected value, which is caused by a domain clamping by the Edep. In region I, the Pr is reduced when the N is smaller, which is because the volume fraction of the clamped do mains is raised by a denser interface with the Edep.

Discussion
In region III, the 1/2 1/2 3/2 reflection of the Mn-BiFeO3 layer is absent (Figure 3), and the polarization and dielectric properties ( Figure 6) cannot be explained by the series con nection of the capacitors of BaTiO3 and Mn-BiFeO3. It is reasonable to consider that the  Figure 7 shows the schematics of the superlattice structures along with the P s component along the out-of-plane direction (P s //[001] c ). In region I (N ≥ 25), the presence of the 1/2 1/2 3/2 reflection from the Mn-BiFeO 3 layer ( Figure 3) and the polarization and dielectric properties ( Figure 6) indicate that the superlattice can be regarded as the simple series connection of the capacitors of BaTiO 3 and Mn-BiFeO 3 . In the BaTiO 3 layer, the P s vector is present along [001] c ; our DFT calculations reveal that the P s strength is 28.5 µC cm −2 , which is close to the bulk value [42]. In contrast, the Mn-BiFeO 3 layer has a P s nearly along [111] c , and the value is reported to be 90-100 µC cm −2 [37]. As the polarization components along [001] c in these layers are markedly different, the interface effect plays an important role. It is assumed that the interface region of several to several tens of unit cells in width needs to accommodate the difference in the direction and strength of the P s vector across it, as in ferroelastic domain walls [43][44][45][46][47][48][49][50]. As a result, a depolarization field (E dep. ) is built up in the interface region, where the E dep. is present in a direction that prevents the change in the polarization component. Given that the P s vectors are switched by an E application, the P r is expected to be~40 µC cm −2 . The P r of 25 µC cm −2 for N = 300 is smaller than this expected value, which is caused by a domain clamping by the E dep . In region I, the P r is reduced when the N is smaller, which is because the volume fraction of the clamped domains is raised by a denser interface with the E dep .

Discussion
In region III, the 1/2 1/2 3/2 reflection of the Mn-BiFeO 3 layer is absent (Figure 3), and the polarization and dielectric properties ( Figure 6) cannot be explained by the series connection of the capacitors of BaTiO 3 and Mn-BiFeO 3 . It is reasonable to consider that the superlattice has a unified unit cell, where electronic orbitals of the BaTiO 3 and the Mn-BiFeO 3 layers are hybridized. In other words, these two layers are no longer distinguished, but the structural and electronic features are completely different from the solid solutions [18]. On the assumption that the superlattice has a unified unit cell (Figure 2), our DFT calculations show that the N = 1 cell has a P s of 27.3 µC cm −2 , which is close to the Nanomaterials 2021, 11, 1857 9 of 11 experimental P r (21.6 µC cm −2 ) of N = 1. Moreover, the enhancement in P r with increasing N (Figure 6a) can be qualitatively explained by the theoretical calculations ( Figure 6d): P s is 31.4 µC cm −2 for the N = 2 cell, and 43.7 µC cm −2 for the N = 4 cell.
Finally, we comment on an additional degree of freedom in superlattice design by adopting an unequal N in the BaTiO 3 and the Mn-BiFeO 3 layers, where material properties can be tuned by different N(BaTiO 3 ) and N(Mn-BiFeO 3 ). For example, we can expect that N(BaTiO 3 ) < N(Mn-BiFeO 3 ) delivers an enhanced P s in a unified cell in the superlattice. Moreover, superlattice design based on different unit cell numbers is anticipated to provide a means to control the strain effect at will.

Conclusions
We investigated the crystal structure and dielectric and polarization properties of superlattice-structured epitaxial thin films composed of Mn(5%)-doped BiFeO 3 and BaTiO 3 with a total thickness of 600 perovskite (ABO 3 ) unit cells. The number of ABO 3 unit cell (N) in the layers of Mn-BiFeO 3 and BaTiO 3 varied from 300 down to 1. It was revealed that the superlattices for an N greater than 25 can be regarded as a simple series connection of their individual capacitors. In the thin regime of an N of five or less, the superlattice behaves as a unified ferroelectric, where the BaTiO 3 and Mn-BiFeO 3 layers are structurally and electronically coupled. With decreasing N from five to one, the εr is markedly enhanced, whereas the P r is reduced. DFT calculations show that the P s is suppressed with decreasing N, which is in good agreement with the experimental P r . We conclude that superlattices formed by two types of perovskite layers with different crystal symmetries represent a path to novel ferroelectrics that cannot be obtained in a solid solution system.

Data Availability Statement:
The data that support the findings of this study are available upon reasonable request from the corresponding author.