Introduction

Multiferroic materials with coexisting and coupled ferroelectric and magnetic orders, have drawn intensive interests due to their vast application potential and fundamental significance since the first discovery of multiferroicity and giant magnetoelectric (ME) coupling in BiFeO3 (BFO) and TbMnO3 (TMO) in 20031,2,3. Substantial experimental and theoretical efforts have found a variety of novel multiferroic phases such as the spiral-spin-order (SSO) in (Tb,Dy)MnO34, E-type antiferromagnetic (AFM) phase in HoMnO35,6, spin-dependent p-d hybridization induced ferroelectricity in Ba2CoGe2O77, di-atomic Ising spin chains in Ca3CoMnO68, possible stripe multiferroic phase in quarter-doped manganites9, etc. Although fascinating physics, such as the concurrent time and spatial inverse symmetry breaking, has been revealed, the small ferroelectric (FE) polarizations (P) (~102 μC/m2), the AFM nature and the low Curie temperatures for FE transitions (TC) (~20 K) keep these materials still far from technological applications.

Recently, quite a few works aiming to explore additional multiferroic materials with excellent physical properties, reported giant improper ferroelectricity (P = 2900 μC/m2 and TC = 90 K) in the so-called “ferroaxial” magnet CaMn7O1210,11 and gigantic magnetic field induced FE polarization (ΔP = 17000 μC/m2 and TC = 64 K) in ferrimagnetic oxide CaBaCo4O712. Some earlier studies revealed above room temperature multiferroic states and low magnetic field induced ME effect in hexaferrites, although their FE polarizations are still small13. These exciting results point to a promising future of multiferroics for the fabrication of multifunctional devices near room temperature. Meanwhile, recently the way to explore novel multiferroicity by finely tuning the multiple exchange interactions in these classic multiferroic materials has been confirmed to be efficient too. Giant tunability of the FE polarization (ΔP = 5000 μC/m2) through modifying the Gd-Mn symmetric exchange striction by magnetic field was reported in GdMn2O514. Interesting strain engineered ferromagnetic (FM) tendency was widely probed in multiferroic manganite (TbMnO3, YMnO3 and LuMnO3) thin films15,16,17,18,19. In particular, while the corresponding bulk crystal are well known as a robust antiferromagnet, a clear FM hysteresis loop and a large ferromagnetism (~1.0 μB) were captured in the orthorhombic LuMnO3 thin film by polarized neutron reflectometry19. Additionally, remarkable FE polarization (~12 μC/cm2) was theoretically predicted in highly compressively strained HoMnO3 thin films20, which is already comparable with that of traditional ferroelectric oxides such as BaTiO321.

For the strain effect in thin films, one of the most direct consequences is the modified microstructure due to lattice misfit between the films and substrates. This misfit would modulate the physical properties massively. One typical example is the multiferroic YMnO3/SrTiO3 (STO) thin films, in which the ground state can be changed from E-AFM to SSO by strain corresponding to the reduced othorhombicity of the films22. Another example is BiFeO3 which shows similar SSO phase in the bulk crystals, while this non-collinear spin state can be easily replaced with a collinear spin order by means of strain engineering (no matter compressive or tensile) in thin films23. In addition to the modified microstructure, the electron-lattice coupling due to the Jahn-Teller (JT) distortion could also be appreciable to physical properties. In HoMnO3 thin films, it was predicted that the strain tuned electron-lattice coupling could induce a progressive electronic contribution to the FE polarization20, although it still needs to be examined experimentally. Here one interesting feature is worth mentioning that the nano-scale twin-like domains are formed naturally to reduce the strain energy in multiferroic RMnO3/SrTiO3 thin films17,18. This multi-domain structure, in addition to the electron-lattice coupling tuning, was proposed to be responsible for the ferromagnetic tendency in RMnO3 thin films17,18, while its possible impact on the magnetically driven ferroelectricity is so far not clear.

In the present work, we report remarkably enhanced spontaneous electric polarization along the a-axis, Pa ~ 4900 μC/m2, high ferroelectric phase transition temperature TC = 75 K and weak ferromagnetism emerging at T ~ 105 K in modestly strained GdMnO3 (GMO) thin films grown on (001) STO substrates, while these thin films accommodate nano-scale twin-like domains. These novel multiferroic properties probably can be explained in the framework of strain stabilized ab-plane SSO phase and twin-like domain structure. There are three motivations for choosing GMO. First, the orthorhombic RMnO3 with small rare-earth element R is a typical multiferroic material and the multiferroic state shows plentiful variations upon tuning the inherent orthorhombic distortion24. For example, the SSO in TMO and DyMnO3 (DMO) can be transformed to an E-type AFM state by simply replacing the Tb/Dy with smaller Ho24. On the other side, if the Tb/Dy is substituted by larger rare earth Gd, the situation becomes complicated25,26. GMO locates rightly at the phase boundary between the A-type AFM and ab-SSO, as illustrated in Fig. 1 and a small spontaneous Pa (~40 μC/m2) arises along the a axis, which could be significantly tuned by disorder or defects25. Importantly, the Pa can be enhanced up to 100 times by strengthening the ab-SSO phase with a magnetic field H > 6 T (Pa ~ 4000 μC/m2 at H > 6 T and Ea = 8 MV/m), which is the largest value reported so far among the SSO multiferroic manganites24. Considering the high sensitivity of multiferroicity to physical excitations in GMO, it would be possible to stabilize the remarkable Pa by some practically accessible methods such as the strain in thin films instead of high magnetic field. Second, GMO is on the border of A-type AFM phase, which allows an easy access to weak ferromagnetism due to spin-canting induced by the highly distorted crystalline structure24,27. The canted A-AFM with a net ferromagnetic moment Mc is schematically drawn in the top left of Fig. 1. Therefore, it would be interesting to explore the possible coexistence of large spontaneous polarization and ferromagnetism engineered by strain in GMO thin films. Third, according to the earlier studies17,18, we note that the nano-scale domain structure is possible in multiferroic RMnO3/SrTiO3 thin films. We expect similar nano-domains appearing in GMO/STO thin films. Moreover, possible reconstruction of the magnetic ordering due to size effect would be anticipated, which could induce distinct multiferroicity in thin films considering the close relationship between ferroelectric and magnetic orders.

Figure 1
figure 1

Schematic drawing of the canted A-type antiferromagnetic phase (left) and ab-plane spiral spin ordering phase (right).

Sketch of the magnetic phase diagram which shows the phase evolution with bandwidth W.

Full size image

Results

Fig. 2(a) displays the high resolution X-ray diffraction (HRXRD) pattern of a typical GMO/STO thin film with thickness t = 110 nm. The film shows pure phase and single c-axis orientation. The out-of-plane lattice parameter is determined to be ~7.479 Å which is slightly larger than the c ~ 7.431 Å for GMO bulk single crystal28, proving a small in-plane compressive strain. To check the epitaxy of the film growth, we further did Φ-scan which is shown in Fig. 2(b). Four GMO (111) peaks with an equal separation of 90° can be seen clearly, which is accompanied by the expected four-fold STO (111) reflections but with a 45° shift. This feature verifies epitaxial growth of the GMO films with respect to the STO substrates. Moreover, the 45° deviation of the GMO (111) peaks as for the STO (111) reflections suggests that the orthorhombic unit cell of GMO is grown on a 3.905√2 Å × 3.905√2 Å square STO lattice, which can accommodate the in-plane film lattice better. Since the orthorhombic GMO shows strong structural anisotropy with lattice parameters a = 5.318 Å, b = 5.866 Å and c = 7.431 Å28, huge lattice misfit (−3.7% along a axis and 6.2% along b axis) would still be expected intuitively for the epitaxial growth of GMO on the top of STO, which is drastically different from the modest compressive strain as revealed by the XRD data. It is then implied that additional microstructural features appear in order to release these huge in-plane strains, consistent with the quite small variation of the out-of-plane lattice constant c.

Figure 2
figure 2

(a) Typical θ–2θ scan of GdMnO3/SrTiO3 thin films. The inset shows the epitaxial growth model, where the green squares denote the STO lattice and the red and blue rectangles denote the twin-like film lattice. The arrows indicate the [100] direction of the film or the substrate. (b) Φ-scan of GdMnO3 (111) and SrTiO3 (111) reflections. (c) Sketch of the device structure. (d) Surface morphology of the film.

Full size image

In order to capture further details of the film microstructures, plane-view transmission electron microscopy (TEM) measurements were conducted and some typical images are highlighted in Fig. 3. As shown in Fig. 3(a), the film has a nano-scale multi-domain structure and the mean size of the domains is around 30 nm and relatively uniform. This is consistent with the domain structure observed by atomic force microscopy measurement shown in Fig. 2(d). A closer plane-view TEM image reveals two variants in the film, which are perpendicular to each other as shown in Fig. 3(b). The corresponding electron diffraction pattern shown in Fig. 3(c) confirms this twin-like domain structure. From the high-resolution TEM image shown in Fig. 3(d), some weak fringes can be seen at/near the domain boundaries, suggesting to be the Morie fringes during the overlapping of domains at the boundaries and the rotation with respect to the variants 1 and 2. The observed twin-like domain structure can further reduce the anisotropic strain energy during the epitaxial growth, giving rise to the small compressive strain of the film. Similar twin-like domain structure was reported in other multiferroic RMnO3 thin films15,16,17,18,29,30, which is probably a common feature for the epitaxial growth of such films once the mismatching strain is anisotropic.

Figure 3
figure 3

(a) Plane view transmission electron microscopy image of a 110 nm GdMnO3/SrTiO3 thin film. (b) An enlarged plane view TEM image showing clear twin-like domain structure labeled with ‘variant 1’ and ‘variant 2’. Small region (labeled by 3) shows ~45° rotation with respect to the two variants. The inset shows electron diffraction pattern of variant 1. (c) Electron diffraction pattern including both of variant 1 and 2. (d) High resolution TEM image of the same sample. The areas showing different variants are enclosed by blue and red lines, respectively.

Full size image

The in-plane magnetization (M) as a function of T and H is presented in Fig. 4. Although the GMO bulk crystals show successive magnetic phase transitions24,25,26, the M(T) curves of the GMO thin film show smooth variation over the whole measured T range, marking modified magnetic properties for the films. Although no evident phase transition can be seen, the M(T) curves of field cooling (FC) and zero field cooled (ZFC) cases present noticeable divergence at low T, which is accompanied by a striking increase in M with decreasing T. This feature leads us to argue a possible net magnetization in the GMO thin film. The irreversible temperature is determined to be Tirre ~ 39 K by plotting ΔM = MFC − MZFC as a function of T. The clear magnetic hysteresis loop with coercive field of Hc = 1.2 kOe measured at T = 5 K confirms the existence of low-T FM phase, as shown in Fig. 4(b). Interestingly, a closer check captures a second anomalous at Ta ~ 105 K on the plot of ΔM-T, which signals the onset of long-range magnetic correlation. To probe further, a M(H) curve was measured at 100 K and a striking FM hysteresis with Hc = 0.1 kOe is observed, verifying an exotic FM orders.

Figure 4
figure 4

(a) Temperature dependence of magnetization (magnetic contribution from substrate has been subtracted.) under FC and ZFC cases. The open circles represent the ΔM plotted as a function of T. Two possible phase transitions, Tirre and Ta, are indicated by the arrows, respectively. (b) M–H curves measured at T = 5 K (the main panel) and T = 100 K (the inset).

Full size image

The above magnetic data shows significant modulation of the magnetic properties in the GMO thin films as compared to the bulk counterpart, pointing to modified multiferroicity due to the intimate coupling between magnetic and ferroelectric orders in this material. Fig. 5(a) and (b) present the T-dependence of pyroelectric current (I) and corresponding integrated FE polarization, respectively. Surprisingly, a remarkable spontaneous polarization Pa ~ 4900 μC/m2 along the a-axis of the films is observed (the polarization can never be along the b-axis for both SSO and E-AFM phases in RMnO331). Nearly identical value of Pa (not shown here) was detected along the in-plane perpendicular direction consistent with the above proposed twin-like domain structure. The FE phase transition temperature is derived to be as high as T = TC ~ 75 K, which is almost among the highest values for discovered multiferroic materials10,11,12,13.

Figure 5
figure 5

(a) Pyroelectric current as a function of temperature measured for various heating rates. The circles denote the data measured along c axis. (b) Integrated polarization as a function of temperature under various magnetic fields. (c) Ferroelectric hysteresis loop measured using the PUND method at T = 5 K and 40 K.

Full size image

The magnetically driven FE state looks very robust against the magnetic field and the Pa just shows small variation even the applied field is as large as H = 9 T. For a comparison, we also measured the I-T curve along the c axis in another sample grown simultaneously, but with a conductive substrate Nb-doped STO (001), as shown in Fig. 5(a). Clearly, no pyroelectric current was found, suggesting the zero FE polarization along the c-axis of the film. To attest that the measured current along the a-axis do come from the pyroelectricity, we measured the I-T curves for three different heating rates (2–4 K/min) as shown in Fig. 5(a). No evident shift of the peak position of the I-T curves can be seen, suggesting the existence of the FE state. The minuscule move of the peak (less than 1 K) is probably due to the thick substrate (0.5 mm) which prevents quick thermal transport. Moreover, the spontaneous polarization can be reversed by reversing the poling electric field (Fig. 5(b)) and such Pa reversing process can also be realized by switching the electric field isothermally using a PUND method shown in Fig. 5(c). The values of Pa obtained from PUND method are close to that of pyroelectric measurements. These demonstrate once again the FE state in the GMO film.

Discussion

Here, we discuss the possible origins of the large spontaneous (Pa ~ 4900 μC/m2), high Curie temperature of the FE state and weak ferromagnetism up to 105 K in GMO/STO thin films. In essence, the observation of spontaneous Pa leads us to argue an ab-spiral spin ordering phase in the thin film similar to the case in GMO bulk crystal, motivated by the magnetic phase diagram of RMnO324,31. This is further supported by the absence of pyroelectric current along the c-axis (Fig. 5(a)). However, both the onset temperatures of the weak FM phase T ~ 105 K and the ferroelectric phase TC ~ 75 K of the film are much higher than that of the bulk, which suggests that the incommensurate AFM temperature (~40 K) and the SSO temperature (~8 K) of Mn moments known in bulk DMO have been uniformly moved to high temperature side in the thin films24,25,26.

First, we look at the high-T phase with weak FM component in the films. Since the films just show very small compressive strain as evidenced by the XRD data, a significant contribution from the lattice modulation by strain is not to be expected. Therefore, we propose that the nano-scale twin-like domain structure is essential for this. This twin-like domain structure is due to the strain relaxation during the epitaxial growth, which has been comprehensively investigated in Ref 18. There are two possible mechanisms responsible for the strain relaxation, which are misfit dislocations at the film-substrate interface and increased strain relaxation at the domain walls. In LuMnO3/YAlO3 thin films which show close lattice misfit (Δc ~ 0.8%) to the present case but without twin-like structure, misfit dislocations induced ferromagnetism was demonstrated, while the FM phase can only be observed below 50 K19. Therefore, more factors should be taken into account for the high-T weak ferromagnetism in the present films. As evidenced by the TEM measurements, the lattice at domain walls shows different arrangement over the domains, which would naturally break the perfect long rang AFM orders and thus generate net moment (FM orders) at the domain walls. Actually, surface (assembles the domain walls) ferromagnetism was widely observed in nano-scale perovskite manganites32,33,34. In particular, similar high temperature anomalous magnetic phase transition at T ~ 125 K was observed in TMO/STO thin films which show similar nano-scale domain structure17.

Second, regarding the FE state with relatively high TC, the mechanism should be similar to that of the bulk crystals as discussed above. However, this ab-SSO phase is driven to high-T side since the long-range magnetic correlation has been developed in high temperature as evidenced by the magnetic data14,24,25,26,35. Another possibility is the size effect, owing to the finite size of the domains. It is known that in BiFeO3 nanoparticle, strong FM tendency could be induced as the particle size is reduced and comes close to the period length of the spiral spin structure36. In the present case, although the ab-SSO period length known in the bulk GMO (few unit cells)26 is shorter than the observed domain size (~30 nm) in the films, a possible modulation on the period of the SSO phase would still be expected considering its small energy scale (in meV). Therefore, the magnetically driven ferroelectricity would be tuned correspondingly because of the close correlation between SSO period and the related FE properties22.

In a nutshell, multiferroicity of GdMnO3 thin film grown on SrTiO3 (001) substrate has been investigated by performing extensive structural, magnetic and ferroelectric measurements. Large spontaneous polarization Pa ~ 4900 μC/m2 was detected and a possible ab-plane spiral spin ordering phase was proposed to be responsible for it. Interestingly, high ferroelectric phase with TC = 75 K and weak ferromagnetism at ~105 K were detected, which should be related to the nano-scale twin-like domain structure and strain-modified lattice distortion.

Methods

Sample preparation

Epitaxial GMO thin films with various thickness (t) ranging from 10 to 110 nm were grown on STO (001) substrates using pulsed laser deposition (PLD) method. A KrF excimer laser (248 nm wavelength) was used for the deposition with a repetition rate of 3 Hz and the laser energy density was set at ~1 J/cm2. A stoichiometric GMO target was used for the ablation. The depositions were made with substrate temperature of 800°C and an oxygen partial pressure of 10 Pa. After deposition, the samples were cooled down at 1 atm of oxygen pressure to achieve oxidation.

Structure characterization

High resolution X-ray diffraction analyses were carried out using a Rigaku Rotaflex Diffractometer. Surface morphology of the samples was checked using Bruker Multimode 8 atomic force microscope. The plane-view and cross-section Transmission Electron Microscopy were performed using JEM 2100F microscope.

Ferroelectric and magnetic measurements

In order to probe in-plane ferroelectricity, gold electrodes of an inter-digitated pattern shown in Fig. 2(c) were sputtered on the film surface through standard photolithography and lift-off procedures. The same device structure was used to probe the in-plane ferroelectricity of YMnO3 thin films early37. The ferroelectric polarization measurements were performed using standard pyroelectric current method using the Keithley 6514 A and 6517 electrometers connected to the Quantum Design Physical Properties Measurement System (PPMS). For each measurement, the sample was first cooled from 150 K to 2 K under a constant poling electric field (E). Then, the poling electric field was removed and the sample was short-circuited for sufficient time in order to release any charges accumulated during the poling process. The recorded background current noise amplitude was less than 0.1 pA. The pyroelectric current was collected during the heating process and various heating rates were used to ensure no contribution other than the pyroelectric current. The polarization was obtained by integrating the collected current with time. To double test the ferroelectricity, PUND method was employed to obtain the P-E loops38. Magnetization (M) as a function of temperature and magnetic field (H) was measured using a Magnetic Properties Measurement System (MPMS, Quantum Design). The cooling and measuring field was set at 0.1 T.