The development of multiferroic materials aims to realize direct electric-field controlled switching of magnetization at room temperature.1,2 However, identifying stable and single-phase multiferroic materials to achieve this objective has been challenging, because electrical polarization and magnetization must be strongly coupled together in such systems.1–6 Recently, a new class of materials, called hybrid improper ferroelectrics (HIFs), was theoretically proposed as a potential room-temperature multiferroic system.2 The principle of the HIF mechanism is to induce ferroelectricity, ferromagnetism, and magnetoelectricity simultaneously with the same set of lattice instabilities.2,7 In particular, oxygen octahedron rotations and tilts, which are ubiquitous in perovskite materials and couple strongly to magnetism, combine with a layered crystal structure to induce polarization.8,9 Hybrid improper ferroelectricity at room temperature was first experimentally confirmed in the n=2 Ruddlesden-Popper material Ca3Ti2O710 and was subsequently observed in Sr3Sn2O7 and other layered perovskites.11–13
Ca3Mn2O7 (CMO) is another HIF candidate, which was the first theoretically proposed system for achieving room temperature multiferroicity.2 Ca3Mn2O7 is an n=2 member of the Ruddlesden Popper layered perovskite family with a general formula (ABO3)n(AO) or An+1BnO3n+1. The CMO crystal structure consists of double CaMnO3 perovskite blocks stacked along the [001] direction, with an extra CaO rocksalt sheet inserted after each double perovskite block. At room temperature, CMO crystallizes in the polar A21am space group. The condensation of an out-of-phase (a-a-c0 in Glazer notation14) octahedral tilting distortion and an in-phase (a0a0c+) octahedral rotation distortion establish the A21am symmetry and induce a polarization.2 Figure 1(a) shows schematic views of the A21am crystal structure viewed along the a ([100] ortho) and b ([010] ortho) projections, with the polarization (P) pointing along [ 00]. The polarization arises primarily from a two-against-one displacement of the Ca ions along the a-axis in each perovskite block (Figure 1a, right). In addition, the a-a-c0 octahedral tilting distortion involves alternating left/right displacements along the b-axis of the Ca ions in each layer (Figure 1a, left).
Several studies have revealed the complex domain structure and phase transition sequence of CMO. Early transmission electron microscopy (TEM) studies showed twin variants of the A21am phase with an (001) interface.15 In addition, Gao et al. observed irregular orthorhombic twins with curved boundaries and further demonstrated the rearrangement of the oxygen octahedral tilts and rotations during a complex phase transition from I4/mmm, to an intermediate nonpolar phase Acaa (a0a0c-), and finally to the polar A21am (a-a-c+) phase.7 This transition leads to a degeneracy of polarization orientations along either the a-axis ([100]orth) or b-axis ([010]orth) of the CMO crystal, and a large number of 90° ferroelectric domain walls perpendicular to the c-axis.7 Liu et al. experimentally demonstrated ferroelectric switching by measuring the ferroelectric hysteresis loop, but this could only be accomplished at a very low temperature (T < 28K) due to the relatively high electrical conductivity of CMO.16 On the other hand, the complex domain morphology of CMO leads to many novel physical properties.16,17 Most notably, negative thermal expansion (NTE) behavior was observed in both CMO and Ca3−xSrxMn2O7 (CSMO) crystals and was closely related to the coexistence of competing polar A21am and nonpolar Acaa phases over a large temperature range.18,19 It was further demonstrated that NTE can be tuned by controlling the phase competition between the polar A21am and nonpolar Acaa phase with different levels of Sr doping.19
Although several studies have focused on understanding the macroscale properties of CMO, little has been done to uncover the structure of the polar-nonpolar phase coexistence at the atomic scale.
In this work, for the first time, we uncover polar nanoregions (PNRs) in a nonpolar matrix of a layered perovskite at the atomic scale using scanning/transmission electron microscopy (S/TEM) imaging and quantify its structure with picometer precision. We further explore the underlying physics and chemistry of the phase competition transition dynamics as a function of temperature using in-situ high-resolution TEM and monochromated electron energy loss spectroscopy (EELS) techniques. We observe that the formation of Mn antisites drives the phase competition, which is a mechanism that may stabilize similar polar/nonpolar phase competition in other layered perovskite crystals and beyond.
This study explores Sr-doped CMO because the presence of the Sr cation promotes the coexistence of the A21am and Acaa phases at room temperature. The addition of 3% Sr is expected to decrease the ferroelectric transition temperature of Ca2.9Sr0.1Mn2O7 to below room temperature, and thus at room temperature, we expect a small fraction of the A21am polar phase region to be embedded in a nonpolar matrix with Acaa symmetry, as indicated by previous X-ray measurements.19
Figures 1b and c show the annular dark-field (ADF) and annular bright-field (ABF) STEM micrographs of the CSMO crystal with the nonpolar Acaa space group, respectively. The ADF-STEM shows the contrast from heavier atomic species, Ca, Sr, and Mn. The presence or absence of Ca/Sr displacements can be used to distinguish between the A21am phase where displacements are expected (Figure 1a) and the Acaa phase, where such displacements are not allowed by symmetry. The Ca/Sr atoms in the image do not exhibit displacement with regard to the unit cell center, which is expected since the matrix should belong to the nonpolar Acaa space group. On the other hand, we observe many nanometer-sized features along the (001) interface distributed randomly all across the sample on the low magnification TEM micrograph taken from [100]orth zone axis (Figure 1d). To determine the crystal structure, we obtain the selected area electron diffraction (SAED) patterns from the same sample (Figure 1e). A row of weak spots at (0, Ɩ, Ɩ) (with Ɩ = 2n+1) connected with near-continuous streaks is observed in the SAED patterns, suggesting the formation of a superstructure along the (001) interfacial planes, similar to the features in the electron diffraction study reported in an undoped Ca3Mn2O7 crystal with a polar A21am space group.15 High-resolution transmission electron microscopy (HREM) images taken in this region confirm that the weak spots and the streaks in the SAED pattern arise from the linear features within the crystalline matrix (Figure S1).
To directly observe the linear features at the atomic scale, we perform aberration-corrected high-resolution scanning transmission electron microscopy (AC-STEM). Numerous linear features (bright lines) along the (001) interface are observed in the ADF-STEM image in Figure 2a. The high-resolution ADF-STEM image of the linear features reveals two distinct types of double bilayer polar nanoregions (a-type and b-type db-PNRs), as shown in Figures 2b and 2c, respectively. Both types of db-PNRs consist of two adjacent double perovskite blocks (a double bilayer). Importantly, this is the first observation of db-PNRs in a layered perovskite system. In the a-type db-PNRs (Fig. 2b), we observe alternating left/right displacements of the Ca/Sr atoms in the rocksalt sheet between two double perovskite blocks. In the b-type db-PNRs (Fig. 2c), the Ca/Sr atoms all displace in the same direction. The a-type and b-type db-PNRs show displacement patterns similar to those of the CMO A21am polar ground state structure viewed from the [010] and [100] zone axis, respectively (Figure 1a). To quantify the structural distortions, we use an atom position refinement algorithm to accurately assess the atom positions and measure the center Ca/Sr atomic displacements in the double perovskite blocks.20,21 The displacement measurement is presented in Figure 2b and 2c as colored vector maps superimposed on the ADF-STEM images of the db-PNRs. The left panel of Figure 2b shows the a-type db-PNRs with a relative displacement of the Ca/Sr atoms of close to 40 pm, and the left panel of Figure 2c shows the b-type db-PNRs with an alternating left/right displacement of the Ca/Sr atoms of approximately 20 pm. Additionally, as illustrated in the vector map, the center Ca/Sr atoms exhibit much larger displacements than the matrix of the crystal (close to 0 pm as indicated by purple color, also see Figure S2). The strain mapping from our geometric phase analysis (GPA) overlaid on the ADF-STEM image (Fig. 2b and 2c, right panel) shows a large strain of around 5-10% along the [001] direction. This local strain is likely the cause of the diffraction contrast in the lower magnification ADF-STEM image in Figure 2a.
To investigate the possibility of oxygen vacancies22,23 as well as to quantify the magnitude of the octahedral tilts, we make use of ABF-STEM. The ABF-STEM mode has been widely used for imaging light elements in a wide range of materials and is especially useful in visualizing oxygen octahedral distortions in oxide materials.24–26 In this work, our ABF-STEM images uncover the oxygen atom columns in CSMO, and show an intact crystal structure at and away from the db-PNRs regions, as shown in Figure 2d. There is no clear indication of oxygen vacancy ordering as it would modulate the intensity of the oxygen atom columns locally. Additionally, the ABF-STEM images clearly illustrate a negligible tilt of the oxygen octahedra in the matrix as expected in the nonpolar Acaa phase, whereas significant octahedral tilting is observed at the vicinity of the db-PNRs. We measure the oxygen octahedra tilting angle to be 158 degrees on average with a standard deviation of 4.7 degrees locally in the vicinity of the db-PNRs, as opposed to approximately 180 degrees in the matrix.
To further understand the dynamics and stabilization of polar nanoregions in CSMO, we perform in-situ heating experiments inside the TEM column. Figure 3 shows the TEM micrographs of the morphology in the CSMO sample at different temperatures during heating. Initial heating of the sample starts from room temperature and goes to 200 °C, as shown in Figure 3a. As discussed previously, the densely populated, nanometer-sized features are the db-PNRs. During heating from 200 °C to 450 °C, the location and the density of the db-PNRs are relatively stable and do not show much change as shown in Figure 3a-c. Within this temperature range, the sample undergoes strain relaxation, and the diffraction contrast induced by strain is reduced. As the temperature is further raised to 550 °C, the db-PNRs start to disappear in the nonpolar phase (Figure 3d). By 600-650 °C, most of the db-PNRs undergo a phase transition into the nonpolar Acaa phase (Figure 3e-f). However, there are still a small number of db-PNRs remaining at 650 °C. This observation indicates a gradual transition between the polar and nonpolar phases in CSMO, but the mechanism that stabilizes db-PNRs at high temperatures is still unknown.
To understand the underlying physics of the db-PNR stabilization mechanism, we use monochromated EELS to determine the local chemical environment of the Mn ions in the db-PNRs. The L ionization edges of the transition metal elements (Mn in this case) reflect the electronic transition from 2p to 3d levels, with the L3 and L2 edges showing the transition from 2p3/2 to 3d3/2 3d5/2 levels and from 2p1/2 to 3d3/2 levels respectively.27–30 By performing monochromated EELS, the energy loss near edge structure (ELNES) of Mn L2,3 is resolved for fingerprinting the oxidation states across the db-PNRs in the crystal. Since the ELNES for the same Mn oxidation state in different compounds is very similar, it has been successfully demonstrated that mixed Mn oxidation states can be determined with very high resolution by fitting the experimental Mn ELNES spectra with the existing reference ELNES spectra.31–33 Figure 4a is a plot of a line-scan EELS data overlaid on the ADF-STEM image indicating the spatial location of the electron probe, and Figure 4b shows the averaged EELS Mn L edge spectra from the db-PNRs (red) and nonpolar phase(blue). We perform linear combination fitting on our line-scan EELS data set; fitting of the averaged spectra from the polar and nonpolar regions is shown in Figure 4b. We use the ELNES spectra from MnO (Mn2+) and SrMnO3 (Mn4+) collected by Garvie et al. as the reference spectra.32 The fitting results indicate that the ELNES of the Mn L-edge at the db-PNRs shows more similarity to the Mn2+ reference spectra. These fitting results are plotted and overlaid on the simultaneously acquired ADF-STEM image (Figure 4c) and indicate that the Mn2+ percentage is approximately 90% at the center of the db-PNRs.
To confirm this result, we also determine the integrated intensity ratio of the L3/L2 edges of the line-scan dataset acquired from another region near the db-PNRs and obtain a ratio of approximately 2.6 and 2 at polar and nonpolar regions respectively (Figure S3). The L3/L2 edge intensity ratio confirms that the oxidation state of Mn ions at the db-PNRs changes from the Mn4+ state to either Mn3+ or a mixture of Mn2+/ Mn4+ states.27 Additionally, we observe a much stronger Mn L edge intensity from the center rocksalt layer of the db-PNRs compared to the perovskite layer. Energy-dispersive X-ray spectroscopy (EDX) mapping of the same region confirms a higher Mn ion and lower Ca ion content in these rocksalt sheets (Figure S4 and S5). Our observations indicate that Mn antisite defects form at the rocksalt layer Ca site of the Ruddlesden-Popper structure. The multi-valent Mn ions change oxidation states when forming antisite defects on the Ca sites.
To validate our hypothesis that Mn antisite defects stabilize the db-PNRs in CSMO, we perform density functional theory (DFT)+U calculations on Ca3-xMn2+xO7 for a range of x. We do not include the Sr dopants in our calculations to keep our computational cell to a reasonable size. The main effect of this choice is that A21am is the ground state structure for all Mn dopant concentrations that we consider, however, the energy difference can still inform us about how the substitution of Mn for some Ca cations impacts the relative energy of the two phases. Figure 5(a) shows as a function of Mn-dopant concentration x for Ca3-xMn2+xO7, obtained from DFT+U structural relaxations. As the dopant concentration x increases, the energy difference grows, indicating that the A21am structure is further stabilized with respect to Acaa. This result supports the interpretation of the experimental data that regions with Mn dopants stabilize the polar structure with A21am-symmetry.
Figure 5(b-g) show the Mn-O-Mn bond angles for several Mn-dopant concentrations, obtained from our DFT+U structural relaxations. The structures in Figure 5(b-g) depict the lowest energy dopant configuration for each doping level. Other Mn-dopant configurations are shown in Figures S7-S8. The n=2 Ruddlesden-Popper A3B2O7 structure contains two symmetry-distinct A-sites: a larger A-site situated in the center of the perovskite bilayer, and a smaller A-site located in the rock salt layer. Our DFT calculations show that it is always energetically favorable to place the smaller Mn2+ dopants in the rocksalt A-sites, which agrees with prior experimental observations. Interestingly, when more than one dopant is present in the supercell (Figure 4c-f), we find that it is energetically favorable for the Mn2+ dopants to cluster in a single rocksalt layer. We also find that as the Mn-dopant concentration increases, the bond angles bend further from 180o. This effect can be rationalized by considering the tolerance factor: a decrease in the A-site cation ionic radius decreases the tolerance factor, which leads to increased octahedral rotation angles.34,35 The Shannon radii of Ca2+ and Mn2+ (in 8-fold coordination) are 1.12 and 0.96 respectively. Thus, as the Mn2+ dopant concentration increases, the tolerance factor decreases, and the octahedral tilting increases (more bending of the Mn-O-Mn bond). We further notice that the bond angles closest to the Mn2+ dopants change the most compared to their values in the undoped compound in Figure 5b, whereas the bond angles further away from the dopant change a smaller amount.
We also use our DFT+U calculations to probe the charge state of the Mn dopants substituted on the Ca sites. In Ca3Mn2O7, the Mn cations have a formal charge of 4+, whereas we expect a formal charge of 2+ if Mn substitutes onto a Ca site. Mn2+ can adopt high- or low-spin states, which would correspond to magnetic moments of 5 and 0.5 respectively. Table 1 reports the magnetic moments of the Mn-dopants obtained from our DFT+U calculations. We find that for all dopant concentrations, the moment of the Mn dopants is 4.3-4.4 , which is consistent with the Mn2+ charge state in the high-spin configuration. This agrees with the experimental observation of the existence of the Mn2+ oxidation state in the polar region. Overall, the DFT calculations exhibit excellent agreement with our experimental observations from atomic resolution STEM combined with EELS.
In conclusion, our study uncovers that the coexistence of polar A21am and nonpolar Acaa phases in CSMO leads to the formation of db-PNRs. PNRs are widely viewed as the embryo of the ferroelectric phase and are critical to the relaxor properties of ferroelectrics in ultrasonic applications due to their superior piezoelectric properties36, and the discovery of db-PNRs in a hybrid improper ferroelectric system is unprecedented. The interpretation of the origin of the polar nanoregions in the relaxor ferroelectric materials has been challenging because of the heterogeneity over different time and length scales.37–39 By employing atomic-resolution STEM imaging in combination with EELS and DFT calculations, this study demonstrates the polar A21am phase stabilization mechanism of db-PNRs in CSMO to be the formation of the Mn antisites on the Ca sites that increases the octahedral tilting amplitudes. This study utilizes an in-situ heating experiment to further explore the polar/nonpolar phase transition as a function of temperature and observed the presence of the db-PNRs at temperatures as high as 650 °C. Both previous studies19 and our in-situ heating experiment show that the polar A21am and nonpolar Acaa phases in CSMO exhibit a competition over a large temperature range, but the stabilization mechanism has not been clear until now. The stabilization mechanism of db-PNRs in CSMO due to Mn antisites is similar to Ti antisites producing PNRs and a switchable polarization in SrTiO340,41, as well as the recent report of Y antisites leading to room temperature ferroelectricity in yttrium orthoferrite YFeO3.42 Our work shows that antisite defects play an important role in stabilizing polar nanoregions in the family of layered perovskite crystals and beyond. This study provides a path toward engineering polar nanoregions and designing novel lead-free relaxor ferroelectrics in hybrid improper ferroelectric materials by tuning the stoichiometry during growth for sophisticated defect engineering.
Table 1. The magnetic moment of Mn-dopant atoms in Ca3-xMn2+xO7 for structures with symmetry A21am and Acaa calculated with DFT+U. For compositions with more than one dopant atom in the supercell, the reported magnetic moment is obtained by averaging the moments of all the dopant atoms in that cell. The magnetic moment of the B-site Mn atoms (that is, those within the oxygen octahedra) is for all structures, which is consistent with a Mn4+ charge state.
Material
|
x
|
Magnetic moment, A21am (μB)
|
Magnetic moment, Acaa (μB)
|
Ca2.75Mn2.25O7
|
0.25
|
4.40
|
4.45
|
Ca2.5Mn2.5O7
|
0.5
|
4.37
|
4.43
|
Ca2.25Mn2.75O7
|
0.75
|
4.37
|
4.42
|
Ca2Mn3O7
|
1.0
|
4.35
|
4.41
|
CaMn4O7
|
2.0
|
4.34
|
4.41
|