Room Temperature Ferrimagnetism and Ferroelectricity in Strained, Thin Films of BiFe0.5Mn0.5O3

Highly strained films of BiFe0.5Mn0.5O3 (BFMO) grown at very low rates by pulsed laser deposition were demonstrated to exhibit both ferrimagnetism and ferroelectricity at room temperature and above. Magnetisation measurements demonstrated ferrimagnetism (TC ∼ 600K), with a room temperature saturation moment (MS) of up to 90 emu/cc (∼ 0.58 μB/f.u) on high quality (001) SrTiO3. X-ray magnetic circular dichroism showed that the ferrimagnetism arose from antiferromagnetically coupled Fe3+ and Mn3+. While scanning transmission electron microscope studies showed there was no long range ordering of Fe and Mn, the magnetic properties were found to be strongly dependent on the strain state in the films. The magnetism is explained to arise from one of three possible mechanisms with Bi polarization playing a key role. A signature of room temperature ferroelectricity in the films was measured by piezoresponse force microscopy and was confirmed using angular dark field scanning transmission electron microscopy. The demonstration of strain induced, high temperature multiferroism is a promising development for future spintronic and memory applications at room temperature and above.


Introduction
In multiferroic materials, magnetism and ferroelectricity coexist, and if they are coupled they give rise to magnetoelectricity (ME) (or ferroelectromagnetism). Since most ferromagnets are conductive (with partially fi lled d -bands) and ferroelectrics are insulating (with fi lled bands), ferromagnetic ferroelectrics are scarce and currently non-existent at room temperature. [ 1 ] Multiferroic materials must not have either time reversal or spatial inversion symmetry which means there is only a small class of materials with both symmetries, typically those with frustrated magnetism, resulting from competing spin interactions. Room temperature multiferroics are currently the subject of intense scientifi c research because of various novel technological applications including magnetoelectric random access memory (MERAM), tunable multifunctional spintronic devices including four-state memory devices, and spin fi lters. [2][3][4] However, to date there have been no clear demonstrations of both room temperature (RT) ferroelectricity (FE) and strong ferro/ferrimagnetism (FM/FIM).
Among single-phase multiferroic materials, BiMnO 3 (BMO) and BiFeO 3 (BFO) have been extensively investigated. BMO is a rare multiferroic material which is FM (saturation moment, M S , of 3.6 µ B /f.u). While the FE transition temperature is relatively high at 400 K, the FM transition temperature ( T C ) is far below RT ( T C ∼ 105 K), and the polarization is small (0.1 µC/cm 2 at ∼90K). [ 5 ] BMO shows a weak negative magnetodielectric effect (MDE) near T C and exhibits a large gap between the FE and FM transition temperature. [ 6 ] To achieve practical ME applications of BMO, La-substitution on the A -site has been studied. [ 7 ] However, despite the enhancement of the ME effect, the T C remains low and the piezoresponse amplitude ( d 33 ) of 2 pm/V is lower than BFO fi lms (60 pm/V). [ 7,8 ] BFO has an incommensurate G -type antiferromagnetic (AFM) structure ( T N ∼ 650 K) and ferroelectricity ( T C ∼ 1103 K) caused by 6 s 2 lone pair distortions of Bi 3+ at the A -site. [ 9 ] In spite of the high ordering transition temperature of BFO, there are drawbacks for practical applications, most notably a weak magnetic moment due to the cycloidally modulated G-type AFM structure with a large period of 62 nm. [ 10 ] With the aim of enhancing the magnetic moment in BFO while maintaining FE, various approaches have been followed including attempting to induce a local spin confi guration in a B -site ordered structure, or creating a complex chiral or canted spin structure via substitution of different 3d transition metal cations (transition metal, TM = Cr, Mn, Co, Ni, and Cu) at the B -sites. [ 11,12 ] The highest moment values achieved were for TM = Co, where 6 emu/cc was obtained at 10 K. [ 12 ] As far as Mn doping of the B -site goes, even though the theoretical predictions of Pálová et al. suggested that it may be possible to achieve co-existent FM and FE, in both bulk and thin fi lms over a wide doping range only low moments have been observed. [10][11][12][13][14][15][16][17][18][19][20][21][22][23] However, in these earlier studies relatively thick fi lms (>100 nm) were studied which were not highly strained. [19][20][21][22] The hypothesis behind this work is that it may be possible to achieve a magnetic FE system in BiFe 0.5 Mn 0.5 O 3 (BFMO) at RT by strain engineering. Recent reports in oxide systems have demonstrated that the magnetic state can be engineered by epitaxial strain. [24][25][26] Earlier, we showed that 30 nm fi lms of BFMO grown on (001) SrTiO 3 (STO) have a highly strained, coherent structure with a 2.5% higher c parameter than in high pressure bulk material and a high magnetic transition temperature (∼600 K) and saturation magnetic moment (90 emu/cc, 0.58 µ B /f.u). [ 27 ] However, details of the strain effect, the elucidation of precise magnetic structure and the measurement of ferroelectric properties were not determined previously.
In this work, around 60 growth runs were undertaken to elucidate the optimum conditions for achieving high quality epitaxy and high magnetic moment (a summary of the range of growth conditions are shown in Scheme 1 . In addition, we tried to grow BFMO fi lms on various substrates, (001) LaAlO 3 , (001) yttria-stabilized zirconia, and (111) STO with different in-plane lattice mismatch (3 -9%). For all these substrates, the BFMO perovskite phase was not well stabilized and many impurity phases formed. The Bi-Fe-Mn-O is a very complex system because of the possible mixed Mn and Fe valences and because there are a number of different possible phases. [ 28 ] Hence, we found that the BFMO perovskite phase only forms if there is a close lattice match with the substrate (i.e., (001) STO), and if the fi lms are very thin and highly strained. Meticulous growth control is required to achieve clean, highly epitaxial, strained BFMO ( Table 1 ).

Results and Discussion
As shown in Scheme 1 and Table 1 , it was necessary to use 'very slow' growth rates to achieve near-phase pure BFMO with strong magnetic moments. Faster grown fi lms contained several impurity phases and weak magnetic signals. As observed by transmission electron microscopy (TEM) and AFM, all the 'very slow' growth fi lms contained only one second phase,  Since the growth temperature must be low to avoid excessive Bi loss during growth, the growth rate also needs to be low to give suffi cient kinetics to achieve the required epitaxy (it is clear from Figure 2 below that very high epitaxial quality and coherent strain are required to give strong magnetism). Of course, slow growth will also mean some Bi loss, and so it is a matter of very fi nely balancing BFMO phase stabilization plus epitaxial quality, versus Bi loss. The likely reason that Fe 3 O 4 + MnFe 2 O 4 + Bi 2 O 3 phases were found in the fast grown fi lms and not in the slower grown fi lms is that under fast growth conditions, the thermodynamically preferred perovskite BFMO phase (stabilized through epitaxy with the perovskite substrate) is kinetically not preferred: here the growth is too fast to allow the complex composition to crystallize suffi ciently. Hence, the simpler binary and ternary component oxides are kinetically stabilized.
It was absolutely necessary to be certain that the magnetism did not originate from Bi 1-δ (Mn 1-x Fe x ) 2 O 4 . Hence, we measured the properties of many 'slow growth' fi lms where the phase was observed to be present at a level of up to 20% (as determined from AFM and TEM). We found that there was no 'slow growth' fi lm which contained only BFMO+ Bi 1-δ (Mn 1-x Fe x ) 2 O 4 and which had a moment above 10 emu/cc. For example, Film5 (Table 1 ) contained at least 20% of Bi 1-δ (Mn 1-x Fe x ) 2 O 4 (as seen in TEM and AFM), but only a diamagnetic signal from the STO substrate was obtained. Hence, it was certain that Bi 1-δ (Mn 1-x Fe x ) 2 O 4 was non-magnetic. Therefore, the magnetic signals in the 'very slow growth' fi lms (where only Bi 1-δ (Mn 1-x Fe x ) 2 O 4 was observed in addition to the BFMO fi lm) were intrinsic to highly strained BFMO.
In the X-ray diffraction (XRD) spectra of Figure 1 a, we show highly epitaxial growth with sharp (00 l ) peaks of BFMO to the left of the (00l) STO peaks. The ω -rocking curve of Figure 1 b shows that the more highly crystalline fi lms have the higher moments. Looking more closely at the best fi lm (Film2, 30 nm thick) by X-ray reciprocal space mapping (RSMs) near the (103) STO peak and (103) BFMO peak (Figure 1 c), we see epitaxial straining of the BFMO to the STO (indicated by the vertical line). The in-plane lattice parameter of the STO was measured to be 3.907 ± 0.002 Å. The out-of-plane and in-plane lattice parameters of Film2 were c = 4.015 ± 0.003 Å and a = 3.900 ± 0.019 Å. To further probe the details of the strained fi lm structure, bright fi eld TEM images were taken along the [100] zone axis of the Nb:STO substrate of Film2 (see Supplementary Note and Figure S1 online). In short, the fi lm was found to be clean, stoichiometric in Fe and Mn (i.e., 1:1 ratio from energy dispersive X-ray spectroscopy in TEM within the error of 2-5%), highly epitaxial and coherently strained throughout the thickness. It is not expected that there will be coherent strain in fi lms of thickness much greater than 30 nm. Figure 2 demonstrates the strain effect on the magnetic properties of BFMO. As shown in Figure 2 a, the fi lms have maximum magnetic moments of 50-60 emu/cc (∼0.4 µ B /f.u) at 5 kOe at 300 K, which is just less than half of the theoretical 1 µ B /f.u. value for fully ordered double perovskite BFMO, indicating partial or full disorder of Fe and Mn. [ 24,29 ] We discuss this point more later on.
The need for excellent epitaxial growth and highly strained fi lms is highlighted by comparing XRD and magnetic measurements. Figure 2 b shows the relationship between the c-axis parameter and full width half maximum (FWHM) of the ω -rocking curves, and the saturation magnetic moment ( M S ) at 5 kOe: the thinner fi lms are more highly strained, have higher c/a values, have better epitaxial quality, and have higher M S values. As shown from Table 1 it is clear that very high quality substrates (determined from the substrate FWHM values of ω -rocking curve of (002) peak (<0.02°) and a single sharp (103) peak (<0.03°)) are necessary to give strong magnetism, but this is not a necessary and suffi cient condition, because if the fi lms are grown fast or thick (>70 nm), the substrate epitaxial quality is not transferred to the fi lm.
According to the trend of Figure 2 b, fi lms above ∼120 nm should have a negligible magnetic moment. Hence it is understandable why previously studied thicker fi lms from the literature had very small or zero moments, e.g., a 180 nm thick fi lm with a c-axis of 3.952 Å had an M S value of 4 emu/cc at RT and 1.5 T. [ 20 ] The magnetic moments for the most strained fi lms of our study (<70 nm) are around an order of magnitude higher than literature values of thicker (>100 nm) BFMO fi lms. [ 11,12,[19][20][21][22] On the other hand, the values are around 30% lower than the best values (90 emu/cc) we obtained for ∼30 nm fi lms on higher quality pure STO substrates (FWHM values of ω -rocking curve of (002) peak of 0.019° and (103) peak of 0.03°) where there is better epitaxy than on Nb:STO (FWHM values of ω -rocking curve of (002) peak of 0.023° and (103) peak of 0.10°), but it was necessary to use the Nb:STO conducting substrates in this work to measure FE properties. [ 27 ] The magnetic transition temperatures ( T C ) are much higher than bulk BMO (∼105 K). Film2 shows a sharp T C above 600 K and the less strained No long range chemical ordering of Mn and Fe could be evidenced in these studies, and any such ordering, if present, must be very short range, in regions much smaller than the thickness of a typical scanning transmission electron microscope (STEM) sample (i.e., <<20 nm). A cation disordered perovskite structure with the possible metal oxygen bonds along the in-plane and out-of-plane directions is shown in Figure 2 d. Studies of the ordering in the fi lms did, however, demonstrate that there is distinct structural ordering in the form of a symmetry reduction, at least in parts of the upper layers of these fi lms (See supplementary Note and Figure S2 online). This ordering consists of a displacive transformation resulting in the formation of the √2 a p × 4 a p × 2√2 a p ( a P is the parameter of the cubic perovskite subcell) orthorhombic Pnma phase previously reported by Belik. [ 30 ] All the evidence suggests that this Pnma phase is formed in such a way that domains with the [010] axis of the orthorhombic phase lie in plane, and this is expected, as this domain confi guration would reduce misfi t strain to the Nb:STO substrate.
To probe the origin of the magnetic properties, we investigated the valence states by X-ray magnetic circular dichroism (XMCD) and X-ray absorption spectroscopy (XAS, See Supporting Information Figure S3 online) for Film1. EELS was also carried out to investigate the chemical composition and valence states of Fe and Mn. XMCD measurements were performed at Mn and Fe L 2,3 -edges on Film1. XMCD provides, through the use of sum rules, access to the spin and orbital moments of the atomic level as well as element sensitivity. [ 31 ] Moreover, from the response of the XMCD signal as a function of the magnetic fi eld, we could determine the spin directions. 2g e 1g ), and the Fe and Mn cations are ordered, then there will be antiferromagnetic coupling, with FIM magnetization of 1 µ B /f.u. [ 18,32 ] In our fi lms, there is no evidence of long range Fe 3+ and Mn 3+ ordering, i.e., we have the following possible bonds Fe 3+ -O-Fe 3+ , Mn 3+ -O-Mn 3+ or Fe 3+ -O-Mn 3+ (Figure 2 d) which all favor AFM interactions. Since only the Fe 3+ -O-Mn 3+ bond gives rise to an overall moment, the reduced moments measured of ≤90 emu/cc (0.58 µ B /f.u.) from the ordered 1 µ B /f.u. value are understood. [ 24,29 ] Both XAS spectra of with a predominance of 3+. These mixed valences are consistent with the presence of some non-magnetic Bi 1-δ (Mn 1-x Fex) 2 O 4 outgrowths at the fi lm surface. The moment estimated by XMCD (0.13 µ B /f.u.) was lower than the superconducting quantum interference device (SQUID) measurement (90 emu/cc, 0.58 µ B /f.u.), likely because the XMCD penetration depth is only 5 -10 nm, and so the signal is infl uenced by surface degradation effects and non-magnetic Bi 1-δ (Mn 1-x Fe x ) 2 O 4 outgrowths. [ 33 ] Polarized neutron refl ectivity (PNR) was undertaken after the SQUID measurements. PNR allows the extraction of the structural and magnetic depth profi le in nanoscale systems. [ 34 ] The refl ectivity is measured as a function of the spin eigenstate of the neutron being either parallel R ( ↑ ) or antiparallel R ( ↓ ) to a quantization axis defi ned by the applied magnetic fi eld which is in the plane of the sample. The refl ectivity, along with the model fi t, is shown in Figure 3 c. The substrate/fi lm interface is relatively sharp (located at Z = 0 nm). From the scattering length density (SLD) it is clear that the magnetic induction of the BFMO is homogeneous throughout the fi lm thickness, which is an important requirement for technological applications.  (Figure 4 a). In the BFMO fi lm above the interface region, the maps reveal that the distribution of Fe 3+ and Mn 3+ is uniform (Figure 4 f-h), i.e., Fe 3+ and Mn 3+ were observed well into the bulk of the fi lm. At the interface region there is some Ti diffusion from the substrate into the fi lm (to around 2 unit cells) which presumably leads to the Mn 4+ near the interface. The volume of this region is too small to explain the magnetic moments observed. Now we discuss the possible origin of the high T C FIM phase in disordered, highly strained, crystalline BFMO. First we note that transition metal perovskites show a great sensitivity of properties to bond angles and bond lengths. [18][19][20][21][24][25][26]35 ] As already mentioned, one of three strain mediated mechanisms can explain the behavior observed.
The fi rst is that strain can affect the in-built molecular electric fi eld which enhances the strength of superexchange interaction to minimize the magnetoelectric energy. This is more effective in Bi compounds because of the Bi 6 s 2 loan pair than La-based ones, e.g. ordered FIM LaFe 0.5 Mn 0.5 O 3 (LFMO). [ 24 ] , [ 37 ] Indeed, despite having a disordered structure, strained thin fi lms of BiCo 0.5 Mn 0.5 O 3 (BCMO) have a very high FM T C of 800 K, which is 700 K higher than disordered polycrystalline BCMO ( T C = 95 K). [ 24 ] The angles is well known in BMO. [ 36 ] Also, recently in ultra thin (<10 unit cell) La 0.67 Sr 0.33 MnO 3 (LSMO) fi lms, a signifi cant enhancement of magnetic transition temperature up to 560 K was found and was explained by strain induced rotation of the BO 6 octahedra. [ 25 ] The third is the coupling between electron spin and lattice via strain which can tune multiferroic order parameters. Recently, Lee et al. reported that substrate strain can control multiple order parameters to achieve FM ( T C of ∼4.2 K)/FE ( T C of ∼250 K) in EuTiO 3 (ETO) thin fi lms under tensile strain, whereas bulk EuTiO 3 is AFM and paraelectric (PE). [ 26 ] We note however that even though strongly enhanced magnetic properties have been achieved by strain engineering, none of the above mentioned compounds, ETO, BCMO, or LFMO show both FM/FIM and FE near to RT.
Finally, we turn to the FE properties of the fi lms. Firstly the leakage current density vs electric fi eld characteristics ( J -E ) were measured. As shown in Figure 5 a, an asymmetric J -E loop was observed with a sharp increase in the positive E -fi eld region from 150 kV/cm when a positive voltage was applied on the Pt electrode. This originates from the difference in the electron injection from the top (Pt) and bottom (Nb:STO) electrodes which have different work functions (Pt: 5.3 eV and Nb:STO: 4.3 eV). [ 38 ] The level of leakage current lies between pure BFO and BMO thin fi lms [ 13,38,39 ] and agrees with typical resistivity measurements of BFMO compared to BFO and BMO, namely 10 5 versus 10 9 (best reported for BFO, but possibly atypical) versus 10 2 Ωcm for BMO. [ 26,40,41 ] The piezoresponse force microscopy (PFM) amplitude and phase as a function of bias voltage (Figure 5 b and c), while sweeping the DC voltage from +8 to −8 to +8V, show characteristic hysteretic behavior indicative of ferroelectricity at RT. Figure 5 d and e show PFM images. The poling sequence was the following: a) application of a voltage of +10 V over a large area of 6 × 6 µm 2 square, b) application of a voltage of −20 V over small square of size 2 × 2 µm 2 , c) reading the phase right after poling shown in Figure 5 d, d) reading again after 24 h (Figure 5 e). The piezoresponse phase can be reversibly switched by 180° (Figure 5 d), and the phase contrast still remains after 24 h (Figure 5 e). The piezoresponse amplitude of 45 pm/V is marginally lower than that of a pure BFO thin fi lm with 50 -60 pm/V [ 8 ] but higher than previously reported for 0.15% Mn-substituted BFO fi lms (30 pm/V) [ 11 ] and La-substituted BMO fi lms (2 pm/V). [ 7 ] To directly decipher the polarization behavior of BFMO on a subunit-cell scale, we performed atom position quantifi cation from aberration-corrected ADF STEM images of Film1 ( Figure 6 a). This technique allows the mapping the relative cation displacements of the B -site atoms in the ABO 3 perovskite structure with picometer precision. [ 42 ] A schematic of the quantifi cation method to measure the relative displacements is given as inset in Figure 6 a. Figure 6 b and c show out-of-plane ( dz , normal to the interface) and in-plane ( dy , parallel to the interface) displacement maps of B -site cations in the structure, respectively; the profi les averaged over the vertical rows are plotted in Figure 6 d. From these results, we can see that both  www.afm-journal.de www.MaterialsViews.com the interface and the bulk of the fi lm have characteristic polarization. At the interface, the out-of-plane ( dz ) displacements of B -site cations are sharply increased; thereafter, there is smaller displacement with some scatter as we move into fi lm. In the bulk of the fi lm, substantial negative (∼ -0.2 Å) dy displacements are detected. Displacements show substantial inhomogeneity as evidenced by error bars in Figure 6 d. If we employ the approximation of a simple linear relationship between displacement and polarization, then high strain levels ( Figure 1 c and Figure S1c) and displacement of the B -site cations (Figure 6 d) can explain the piezoresponse amplitude of BFMO along the out-of-plane direction.

Conclusion
In conclusion, we have demonstrated for the fi rst time, coexistent ferrimagnetism and ferroelectricity at RT in BiFe 0.5 Mn 0.5 O 3 (BFMO) by strain engineering. The most highly strained and crystalline fi lms had a ferrimagnetic transition temperature of ∼600 K which is 500 K higher than bulk BMO and a piezoresponse amplitude of 45 pm/V. XMCD proved the ferrimagnetism in the fi lms. Very slow growth rates and a narrow temperature window are critical for achieving good phase purity, very high crystallinity and coherent strain. There are three possible strain related mechanisms responsible for the high T C , with the Bi 6 s 2 lone pair effect playing an important role. The work has opened up a new window for designing multiferroic materials through epitaxial strain control.

Experimental Section
Film Fabrication: Films were grown on (001) SrTiO 3 (STO) and (001) Nb:STO substrates by pulsed laser deposition with a KrF laser. A mixed BFMO target was used for making BFMO. The BFMO target was prepared by thoroughly mixing stoichiometric amounts of high purity (99.99%) Bi 2 O 3 , Fe 2 O 3 , and MnO 2 , followed by sintering at 800 °C. For all the growth runs the fi lm heater (infra red) temperature was set to 820°C but the actual temperature on the surface of surface was ∼180 °C lower than this, namely 640 °C. The laser pulse rate was 2 -5 Hz. The oxygen pressure was fi xed at 100 mTorr.
Structural Analysis: The phase and the crystalline quality of the thin fi lms were investigated by θ -2θ and asymmetric X-ray diffraction (XRD) on a Bruker D8 theta/theta diffractometer with Cu-K α radiation and a www.afm-journal.de www.MaterialsViews.com graded mirror. Reciprocal space maps (RSMs) were used to investigate the strain between the fi lms and substrates. Transmission electron microscopy (TEM) was performed using a JEOL 4000EX microscope operating at 400 kV. Scanning transmission electron microscopy (STEM) was performed using a JEOL 4000EX microscope operating at 400 kV and two Nion UltraSTEMs, both operated at 100 kV and equipped with a Gatan Enfi na EELS spectrometer, and a JEOL ARM200F (cold FEG) operated at 200 kV. Magnetic Analysis: Magnetisation measurements ( M -T and M -H ) were performed using a Princeton vibrating sample magnetometer (VSM) and a superconducting quantum interference device (SQUID) magnetometer (Quantum Design, MPMS). The magnetic moment is converted from emu/cc to µ B /f.u. by using the volume of unit cell of BFMO measured by X-ray refl ectivity and TEM. To investigate whether the magnetic origin in the fi lms was FIM or FM, X-ray magnetic circular dichroism (XMCD) and X-ray absorption spectroscopy (XAS) were performed on beamline U4B at the National Synchrotron Light Source (NSLS), Brookhaven National Laboratory. The degree of circular polarization for the XMCD measurement was 70% with the incident photons aligned parallel to the magnetic fi eld. The XMCD spectra were obtained by using a fi xed circular polarization and switching the applied magnetic fi eld by pulsing it to 9.5 kOe and measuring in remanence. The spectra were measured by total electron yield mode (TEY) at 300 K. XAS measurements were taken at an angle of 45° between the E -fi eld of the X-rays and the out-of-plane direction of the sample. Polarized neutron refl ectivity (PNR) measurements were performed using the CRISP refl ectometer at ISIS [ 43 ] The measurements were carried out at RT in a fi eld of 3 kOe applied in the plane of the sample perpendicular to the scattering direction. The fi tting was done using the GenX refl ectivity package. [ 44 ] Piezoresponse Measurements: Leakage current density-voltage characteristics were measured using a Keithley 6487 picoammeter and a 2440 sourcemeter with a stabilization time of 10s. Nb:STO was used as the bottom electrode and for the top electrode; a 150 nm thick Pt pad of size 7.85 × 10 −5 cm 2 was deposited on the BFMO fi lm. To characterize the ferroelectric properties, an atomic force microscope (Agilent SPM 5500) with 3 (MAC-3) lock-in amplifi ers (LIAs) was used at RT in piezoresponse force microscopy (PFM) mode (out-of-plane signals). An Olympus Pt-coated tip (Asylum Research AC240TM) was used at the excitation frequency of 15 kHz with alternating current (AC) voltage V AC = 2.5 V, polarized by a direct current (DC) bias of V DC = ±10 to ±20 V. The piezoresponse signals were calibrated using a standard height ("Micro-Mesh") sample and the LiNbO 3 PFM factory standard sample ("Bruker") with d 33 = 7.5 pm/V.

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