Composition‐Dependent Ferroelectricity of LuFeO3 Orthoferrite Thin Films

This work characterizes the structural, magnetic, and ferroelectric properties of epitaxial LuFeO3 orthoferrite thin films with different Lu/Fe ratios. LuFeO3 thin films are grown by pulsed laser deposition on SrTiO3 substrates with Lu/Fe ratio ranging from 0.6 to 1.5. LuFeO3 is antiferromagnetic with a weak canted moment perpendicular to the film plane. Piezoresponse force microscopy imaging and switching spectroscopy reveal room temperature ferroelectricity in Lu‐rich and Fe‐rich films, whereas the stoichiometric film shows little polarization. Ferroelectricity in Lu‐rich films is present for a range of deposition conditions and crystallographic orientations. Positive‐up‐negative‐down ferroelectric measurements on a Lu‐rich film yield ≈13 µC cm−2 of switchable polarization, although the film also shows electrical leakage. The ferroelectric response is attributed to antisite defects analogous to that of Y‐rich YFeO3, yielding multiferroicity via defect engineering in a rare earth orthoferrite.


Introduction
Oxides of rare earths (R) and iron can form a range of crystal structures depending on their R to Fe ratio, including RFe 2 O 4 , R 3 Fe 5 O 12 (rare earth iron garnet), R 2 Fe 3 O 7 , orthorhombic RFeO 3 , and hexagonal RFeO 3 . These materials exhibit a diverse variety of useful properties such as magnetism, [1] ferroelectricity, [2] multiferroicity, [3] magnetoresistance, [4] magnetooptical activity, [5] and catalytic activity. [6] The electrical, magnetic, optical, and transport properties of these and other complex predicted polarization. [19] It is therefore of interest to investigate the effect of antisite defects on the ferroic properties of LFO.
Previous studies of the stoichiometric orthoferrite LFO report that remnant polarization (P r ) ranges from several nC cm −2 (polycrystalline LFO) [22][23][24] up to ≈15 µC cm −2 (thin film LFO) [25] at room temperature. The ferroelectricity in polycrystalline LFO was first explained by exchange striction at magnetic domain walls, [22] or by charge disproportionation from increased covalency, [24] and the higher polarization in thin films was ascribed to the tetragonally strained structure with an out-of-plane to in-plane lattice parameter ratio of 1.045. [25] Polycrystalline Y 1−x Lu x FeO 3 also showed unsaturated polarization versus electric field (P-E) loops, and the hysteresis was attributed to symmetry breaking from the mixing of Lu and Y ions. [26] A metastable hexagonal LFO phase was found to be ferroelectric due to its noncentrosymmetric crystal structure, with T C = 1020 K and P r up to ≈10 µC cm −2 at room temperature. [27][28][29] Both orthorhombic and hexagonal LFO phases have been grown in the form of thin films using pulsed laser deposition (PLD), [25,27,30] radio frequency magnetron sputtering, [31,32] or molecular beam epitaxy. [28] When the two phases are grown together, the film can show multiferroicity with a magnetization of ≈0.24 µ B per f.u. from the orthorhombic phase and a polarization of ≈5 µC cm −2 from the hexagonal phase. [29] Orthorhombic LFO is a good candidate for defect engineered multiferroic material because the magnetic order persists above room temperature (T N = 623 K), in contrast to hexagonal LFO with T N = 155 K. [28] However, so far, orthorhombic LFO has not received as much attention as hexagonal LFO, and the stoichiometry range this material can tolerate as a single-phase film has not been explored.
Here, we grow epitaxial orthorhombic LFO thin films with Lu/Fe ratio varying from 0.6 to 1.5, and characterize their structural, magnetic, and ferroelectric properties at room temperature. LFO films with different orientations grown on SrTiO 3 (STO) form a single-phase perovskite despite their non-ideal cation stoichiometry, which in bulk would favor the formation of secondary phases. The observation of ferroelectric hysteresis by different methods (piezoresponse force microscopy (PFM), P-E loops, and positive-up-negative-down (PUND) measurements) all support the hypothesis of antisite-defect-mediated ferroelectricity in orthoferrites with nonstoichiometric R:Fe ratio.

Results and Discussion
LFO thin films with Lu/Fe ratios of 0.6, 0.8, 1.0, 1.2, and 1.5 were grown on (001)-oriented STO (insulating) and Nb-doped STO (NSTO, conductive) substrates with a = 3.905 Å (Figure 1a). There were no secondary phases observed in the wide angle range X-ray diffraction (XRD) scans (15°-80 o , Figure S1, Supporting Information) for any of the films, despite the fact that according to the Lu-Fe-O ternary phase diagram, [33] LFO does not accommodate significant excess Lu or Fe, instead forms additional phases such as LuFe 2 O 4 , Lu 3 Fe 5 O 12 , Lu 2 Fe 3 O 7 , hexagonal LuFeO 3 , or binary oxides when Lu/Fe differs from 1.0. Due to the ionic radius difference between Lu 3+ (86 pm) and Fe 3+ (60 pm), as the film becomes Lu-rich, the XRD peak position 2θ of (002) p (p denotes the pseudocubic notation) moves to a lower angle and the out-of-plane lattice parameter c p increases from 3.789 Å (Lu/Fe = 0.6), 3.793 Å (Lu/Fe = 0.8), 3.800 Å (Lu/Fe = 1.0), 3.820 Å (Lu/Fe = 1.2), to 3.846 Å (Lu/Fe = 1.5). The monotonic trend in the peak position differs from that observed in YFO, when the 2θ angle increased then decreased as the film composition moved from Y-rich to Fe-rich. [19] XRD of Lu/Fe = 1.2 films grown at different oxygen partial pressure (P O2 ) is shown in Figure 1b. The 2θ angle shifts to a higher value as the pressure increases, consistent with a reduction in the oxygen vacancy (V O ) concentration with increasing oxygen pressure. Lu/Fe = 1.2 films were also grown on (111) and (110)-oriented NSTO substrates ( Figure S1, Supporting Information), and similar to the (100)-oriented substrates, only reflections arising from epitaxial perovskite LFO are observed. In Figure 1c, Lu/Fe = 1.2 films were grown on different substrates, LaAlO 3 (LAO) with a p = 3.788 Å and (LaAlO 3 ) 0.3 -(SrAl 0.5 Ta 0.5 O 3 ) 0.7 (LSAT) with a p = 3.868 Å. Film 2θ positions show minimal change with respect to the substrate. This is different from the behavior of YFO where the peak position shifted for different substrates. [19] The reciprocal space mapping (RSM) data of the (103) p reflection, shown in Figure 1d for Lu/Fe = 1.2 and Figure 1e for Lu/Fe = 1.0, indicates cube-on-cube epitaxy of the LFO on STO substrates. Although the crystal quality of Fe-rich films was not good enough to produce a peak in RSM (Figure 1f), the peak positions of the other films indicate in-plane incoherency. Inplane pseudocubic lattice parameters extracted from RSM are 4.017 and 3.987 Å for Lu/Fe = 1.2 and Lu/Fe = 1.0, respectively, compared with out-of-plane lattice pseudocubic lattice parameters 3.820 and 3.800 Å. Unit cell volumes are greater by ≈10% compared to bulk LFO, which is consistent with a greater oxygen deficiency in thin films versus bulk. In contrast, the YFO film unit cell volume was only 1% greater than that of bulk YFO. [19] Atomic force microscope (AFM) images of the Lu/Fe = 1.2 film, Figure 1g, shows island-like roughening common to PLD-grown films; the root mean square roughness is 0.460 nm for a ≈30 nm thick film. Figure 2 is the magnetic hysteresis loop of the Lu/Fe = 1.2 film measured at room temperature, with the magnetic field applied in both out-of-plane (OP) and in-plane (IP) directions. The outof-plane moment saturates at around 1 T with a saturation magnetization M s ≈0.8 kA m −1 , which is about an order of magnitude lower than that of bulk. [14] However, the in-plane moment does not saturate up to 2 T. This suggests that the orthorhombic c axis in Pbnm notation is perpendicular to the film because the weak ferromagnetism from DMI is along the c axis. Based on the magnetic properties and structural analysis, we conclude that the epitaxy of LFO films on perovskite substrates can be represented as Figure 1h. The orthorhombic a and b axes lie in-plane, 45 o rotated with respect to the pseudocubic unit cell, and the orthorhombic c axis lies out-of-plane with lattice parameter = 2c p .
In order to characterize the ferroelectric properties, we performed switching spectroscopy via PFM (SS-PFM) and used the PFM tip to write polarization patterns on LFO thin films. Figure 3a,b,c shows the SS-PFM amplitude and phase hysteresis of LFO films with Lu/Fe of 1.2, 1.0, and 0.8, along with the remnant polarization after writing a "box-in-box" pattern, i.e., polarizing a 1 µm square region at −8 V tip bias then a smaller square region at +8 V. The Lu-rich film shows ferroelectricity which can be interpreted using the antisite defect mechanism. [19] SS-PFM hysteresis was observed in several Lurich films grown with different orientations or at different P O2 (Figure 3d,e). Moreover, films grown with top and bottom conductive SrRuO 3 (SRO) layers also exhibit ferroelectric switching ( Figure S2a-c, Supporting Information). With the SRO top layer, the tip is in contact with a conductive layer which rules out surface charging artifacts of PFM.
All the Lu-rich, stoichiometric, and Fe-rich LFO films showed hysteretic behavior in PFM, although the amplitude of SS-PFM and the phase contrast produced by domain writing is much smaller for Lu/Fe = 1.0. This behavior differs from that of YFO films where the ferroelectric response was limited to Y-rich compositions. The ferroelectricity in Y-rich YFO was explained by antisite defects Y Fe on the octahedral sites, which lower the symmetry to a noncentrosymmetric R3c structure. [19] If the same mechanism exists in LFO, it can account for the ferroelectric Adv. Electron. Mater. 2023, 9,2300059  response in the Lu-rich composition. For the film with Lu/Fe = 1.0, we hypothesize that the (weaker) ferroelectricity may also be caused by antisite defects, even though the composition would suggest that few such defects need be present. The XRD peak intensities provide support for the presence of antisite defects (i.e., both Lu Fe and Fe Lu ) in the film with Lu/Fe = 1.0. Using a pseudocubic ABO 3 perovskite unit cell, the structure factors F hkl of the (001) and (002) reflections Adv. Electron. Mater. 2023, 9, 2300059   Table 1, which were calculated using the equation under the assumption of Fe-rich and O-rich conditions and with in-plane pseudocubic lattice parameter of 3.905 Å (crystal structures are shown in Figure S4 in the Supporting Information). While the formation energies of cation or anion vacancies range from 4.74 to 6.97 eV, the formation of antisite defects is energetically more favorable, suggesting abundant distribution of these defects in thin films. This highlights the feasibility of antisite defect formation in LFO, which may explain the presence of ferroelectricity in LFO of various compositions including Lu/Fe = 1.0 via the mechanism presented for Y-rich YFO. [19] It is worth noting that a negative formation energy of the Fe Lu defect is associated with intrinsic instability of bulk orthorhombic phase LFO. In agreement with ref. [33], the thermodynamic phase diagram ( Figure S5, Supporting Information) constructed based on the calculated enthalpies of Fe 2 O 3 , Lu 2 O 3 , and LFO oxides indicates that the orthorhombic LFO structure is not stable, consistent with the depiction as a point in the ternary phase diagram, and can be stabilized by epitaxial growth. In order to qualitatively compare the magnitude of polarization among different orthoferrites, SS-PFM was performed using the same tip on Lu-rich LFO, Y-rich YFO, and BFO, and the results are shown in Figure 3f. The PFM amplitude is the highest in BFO, followed by LFO, and YFO. In a ferroelectric material with 4mm symmetry, the piezoelectric coefficient d 33 and polarization P 3 are related by d 33 = 2ε 0 ε 33 Q 3333 P 3 where ε 0 is the permittivity of vacuum, ε 33 is the relative permittivity, and Q 3333 the electrostrictive coefficient, [34] therefore we expect the PFM amplitude to be proportional to the polarization. We also note that perovskite ferroelectrics tend to show higher PFM amplitude when they are thicker. [35][36][37] Assuming LFO and YFO have similar ε 33 and Q 3333 and considering that the LFO sample is thinner than YFO, the measurements of PFM amplitude, which is proportional to d 33 , implies that LFO has a higher polarization than YFO (as predicted [19] ), but lower than BFO.
P-E loop and PUND results of Lu-rich LFO with a Pt top electrode are shown in Figure 4a,b. In Figure 4a, a clear ferroelectric saturation feature is observed at high positive bias and switching current peaks are observed for both bias directions; however, the P-E loop at negative bias shows a more resistive-like behavior. Current leakage is also confirmed by the increase in the current at higher bias, similar to that observed in Hf 0.5 Zr 0.5 O 2 . [38] The P-E loop shape indicates that the LFO has a lower resistivity compared to other ferroelectrics (i.e., PbZr x Ti 1−x O 3 , PZT) and a diode-like behavior from the asymmetric electrode structure. The resistivity of LFO was 1.5 × 10 8 Ω cm (Figure 4c), measured by impedance spectroscopy to minimize the contribution of contact resistance. This value agrees with the resistivity calculated from the current flow at 1 V bias. The LFO resistivity is two or more orders of magnitude lower than that of PZT. [39] We attempted to increase the resistivity by using Ti 4+ as a dopant under the assumption that LFO is a p-type conductor. [40] Although Ti was successfully incorporated in the LFO without producing any secondary phases, the resistivity decreased to ≈10 6 Ω cm, implying the leakage was due to a different mechanism.
Adv. Electron. Mater. 2023, 9,2300059   Asymmetric P-E loops (ferroelectric hysteresis for one direction of bias and resistive for the other) have been reported in BFO and PZT. V O accumulation at the interface in BFO [41][42][43][44] caused such loops, and different diode characteristics at the interface may have contributed to asymmetry as well. We attempted to reduce the leakage and its directionality by using SRO as both top and bottom electrodes, but the asymmetry and leakage were still present as shown by the rounded hysteresis loop shape ( Figure S2d, Supporting Information). This may be a result of inequivalent growth condition of the SRO, in which the bottom layer is grown directly on the substrate, whereas the top layer is grown on the strain-relaxed LFO film. These results, along with studies on BFO or PZT, [27,39] highlight the importance of process parameters in ferroelectric film growth. During P-E measurements, some electrodes exhibited capacitive behavior at first, but applying a pulse similar to or higher than the coercive field could activate the film into showing ferroelectric behavior. A similar phenomenon was observed in PZT by removing pinned domains, [45] although the mechanism for LFO may not be the same, considering that LFO is an improper ferroelectric whereas PZT (or BFO) are proper ferroelectrics. The primary order parameter for proper ferroelectric transitions is polarization, while in improper ferroelectrics, polarization is derived from other order parameters (e.g., structural distortion arising from Lu Fe ). [13,27] PUND measurements were performed to exclude parasitic contributions from the P-E hysteresis and to extract the switchable polarization of LFO. The difference between the remnant polarization from the switching pulse (P r * ) and the nonswitching pulse (P r^) is equivalent to the net switchable polarization (Q SW , or 2P r ). Q SW and −Q SW derived from Figure 4b are 10.8 and −15.8 µC cm −2 , respectively, and the difference presumably rises from the asymmetric electrode configuration. The average value of 13.3 µC cm −2 is higher than that of Q SW = 7.2 µC cm −2 for YFO, which is in line with the hypothesis that Lu-rich LFO would have a higher polarization than Y-rich YFO. [19] Nevertheless, high ±P r^ in the PUND measurement indicates that the leakage, which is concurrent with ferroelectric switching, is significant when the field approaches and overcomes the coercivity, and it is also consistent with the leaky behavior in the P-E loop in Figure 4a. Reducing the leakage of LFO films is necessary to improve the performance of LFO as a room temperature multiferroic material.

Conclusion
This study presents the structural, magnetic, and ferroelectric behavior of epitaxial LFO thin films grown using PLD, particularly revealing ferroelectricity depending on Lu/Fe stoichiometry which is attributed to an antisite-defect-mediated mechanism. The out-of-plane magnetic hysteresis of LFO films shows small canted ferromagnetism and thus corresponds to the orthorhombic c axis in Pbnm notation. Using SS-PFM, particularly focusing on the Lu-rich composition with Lu/Fe = 1.2, we demonstrate room temperature ferroelectric behavior over a range of growth conditions. Moreover, we attribute the presence of ferroelectricity across all compositions to the abundance of antisite defects in LFO, which have a low formation energy according to DFT. The remnant polarization obtained from the PUND measurement is 13.3 µC cm −2 for Lu-rich LFO, higher than that of Y-rich YFO. However, the ferroelectric performance is limited by the low resistivity of LFO. This work realizes room temperature multiferroicity in films of an orthoferrite, LFO, and supports the model of ferroelectricity derived from antisite defect engineering in rare earth orthoferrites.

Experimental Section
Thin Film Growth: Epitaxial LFO films were grown by PLD using a 248 nm wavelength KrF excimer laser with a fluence of ≈2 J cm −2 from oxide targets prepared with two different Lu/Fe ratios, 0.6 (Lu 3 Fe 5 O 12 , the stoichiometry of garnet) and 1.4. The Lu/Fe stoichiometry of the targets and film end members were measured by wavelength dispersive spectroscopy (WDS) using a JEOL-JXA-8200 Superprobe. The Lu/Fe = 0.6 target resulted in a film with the same Lu/Fe ratio, however, the Lu/Fe = 1.4 target resulted in a film enriched in rare earth (Lu/Fe = 1.5). A similar result was also seen in the growth of Y-rich YFO from a stoichiometric YFO target. [19] The Lu/Fe ratio of the film was varied by adjusting the shot ratio between the two targets based on their growth rates to achieve the desired stoichiometry. The substrate heater temperature setpoint was 900 °C (actual substrate temperature ≈650 °C) and P O2 was 10-150 mTorr during growth. For comparison, YFO was grown at the same deposition conditions as LFO from an oxide target with Y/Fe = 1 resulting in a film with Y/Fe = ≈1.2, and BFO was grown in a different chamber using a Bi/Fe = 1.2 oxide target with substrate heater setpoint of 700 °C (actual substrate temperature ≈550 °C). For some samples, conductive layers of SRO were grown in situ, above and below LFO, at the same P O2 and temperature as LFO.
Structural, Magnetic, and Ferroelectric Characterization: Structural characterization by XRD was done using a Rigaku Smartlab high resolution diffractometer with a Cu Kα 1 (wavelength 1.5406 Å) X-ray source. RSM was performed with a Bruker D8 Discover diffractometer, also using a Cu Kα 1 source. Magnetic hysteresis was measured in a Quantum Design MPMS3 SQUID magnetometer at 300 K. AFM imaging was performed using a Bruker Dimension Icon XR scanned probe microscope, and PFM was performed using an Asylum Research Cypher VRS AFM in dual AC resonance tracking (DART) mode. Both probe microscopies were done using Pt-coated Si probes from MikroMasch (HQ:NSC18/Pt). SS-PFM data was processed with a simple harmonic oscillator model using built-in software. A Radiant Technologies Precision Premier II tester was used for P-E hysteresis and PUND pulse measurements. Prior to electrical measurements, Pt electrodes with a diameter of 200 µm were deposited by sputtering using a shadow mask, and for samples with SRO layers, the top SRO layer was patterned by photolithography (Heidelberg MLA 150) and ion milled, while the underlying SRO layer below the LFO remained continuous.
Computational Methods: DFT studies were done using the Vienna Ab initio Simulation Package (VASP). [46,47] Generalized gradient approximation (GGA) parameterized by Perdew-Burke-Ernzerhof (PBE) [48] functional was used for the exchange-correlation term. The plane-wave cutoff energy was 500 eV. The rotationally averaged Hubbard correction with an effective U eff = 4 eV for Fe 3d electrons was used. Lu 5p 6 5d 1 6s 2 , Fe 3p 6 3d 7 4s 1 , and O 2s 2 2p 4 electrons were treated as valence. Enthalpies of oxide formation and defect formation energies were adjusted, taking into account O 2 overbinding and the use of Hubbard U correction scheme according to refs. [49] and [50], respectively. In the case of LFO, a 2 × 2 × 1 supercell with 8 f.u. was used with a 4 × 4 × 4 Monkhorst-Pack k-point grid. For defect formation energy calculations, the structures were epitaxially strained to the lattice parameter of STO (a = 3.905 Å), but relaxed in the out-of-plane direction. The formation enthalpy of each compound was calculated by 55 eV per f.u., and the choice of parameters with Fe Hubbard U correction [50] showed a good agreement with the experimental formation enthalpies of binary oxides Fe 2 O 3 and Lu 2 O 3 . [51] Supporting Information Supporting Information is available from the Wiley Online Library or from the author.