Introduction

Interfaces of transition metal oxides are promising to offer fascinating physical properties because of the strong correlations between charge, spin, orbital and lattice degrees of freedom1. In oxide heterostructures, novel electrical and magnetic ground states may emerge as a result of charge transfer, electronic and orbital reconstruction, which has attracted increasing attention2,3,4,5. Particularly, multiferroic materials simultaneously exhibit at least two types of ferroic or antiferroic orders6,7,8,9,10,11 and the coupling between ferroelectric and magnetic orders facilitates the interconversion of electric and magnetic energies which may find novel applications in memory and logic devices12,13,14,15. Recently, exchange bias (EB), which was originally discovered in bilayers of ferromagnetic (FM) and antiferromagnetic (AFM) materials as a result of the magnetic coupling and pinning effects at interface16,17,18,19,20, was observed in multiferroic BiFeO3 (BFO) heterostructures13,21,22,23,24,25,26,27. Such multiferroic heterostructures with EB represents a milestone on the path towards next generation magnetoelectric devices27. However, although the EB effect, which is characterized by a shift in the magnetization loops away from the zero field axis, was discovered half a century ago and has found important technological applications in data storage and magnetoresistive sensors, its microscopic origin has not yet been fully elucidated. For instance, the issues regarding the coupling configurations at the interface28,29 and the correlations between exchange bias field (HE), coercivity (HC) and interfacial exchange coupling30,31,32 are under a lot of debate. In the past decades, EB effect in magnetic heterostructures has remained as an active research area and there have been reports on EB effect in a myriad of oxide bilayers and supperlattices33,34,35,36,37,38. Incorporating multiferroic oxides into such oxide heterostructures adds a new dimension to the EB research and offers exciting opportunities.

In this work, we focus on the magnetic coupling between a multiferroic TbMnO3 (TMO) with noncollinear spin order and a prototypical collinear ferromagnet La0.7Sr0.3MnO3 (LSMO). To the best of our knowledge, this is the first study on the magnetic interaction at a collinear/noncollinear spin heterointerface. TMO is a perovskite-structured multiferroic with a frustrated spiral spin order, exhibiting intricate magnetoelectric and magnetocapacitance effects39,40,41,42,43,44,45,46. The Mn sublattice in TMO single crystals shows a sinusoidal spin order along the b axis at the Néel temperature TN ~ 41 K (Fig. 1(a)) and the spin order becomes spiral (Fig. 1(b)) in the bc plane at the ferroelectric transition temperature Tlock ~ 28 K. The broken inversion symmetry due to the noncollinear magnetic structure leads to the formation of a ferroelectric polarization along the c direction39,40,41,42,43. Upon decreasing temperature, the Tb3+ ions show a long-range spin ordering at TTb ~ 7 K (schematically shown in Fig. 1(b))39,43,46. We should note here that there have been only a few reports on the magnetic order in TMO at T < 7 K39,43,46 and the multiple high-order reflection peaks detected in a recent neutron diffraction study suggest that the order of the Tb spins and their coupling with the Mn spins can be much more complex than the simplified schematic in Fig. 1(b). In particular, neutron scattering and x-ray resonance scattering studies have confirmed the existence of a strong coupling between Tb and Mn ions via the spin-polarized 5d conduction band44,45,46 and their states are hybridized. In such noncollinear multiferroics, the complex spin orders and the exchange interactions are important for not only the multiferroic ground states but also potentially provide novel electronic paths towards controlling the magnetic degree of freedom in heterostructures47,48,49.

Figure 1
figure 1

Schematics of possible spin configurations of the LSMO/TMO interface at the temperature just below Néel temperature TN (a) and below the Tb spin ordering temperature TTb (b).

Note that these schematics only serve the purpose of illustration and do not represent the exact spin orders at interface.

Full size image

So far, most works on TMO have focused on the properties of bulks and thin films39,50,51 and the properties of TMO-based heterostructures were seldom studied. Recently, it was demonstrated that rectifying junctions can be formed by growing TMO thin films on conducting Nb-doped SrTiO3 substrates52,53, but the study was limited to transport properties. In general, there is a lack of effort on incorporating multiferroics with noncollinear spin orders like TMO into functional thin film heterostructures. It has been proposed that the magnetic frustration and the noncollinear spin structure in the AFM layer contribute to the EB effect at the AFM/FM interface32,54. Multiferroic TMO exhibits unique and complex spin orders, thus offering new perspectives to exploit the exchange coupling in multiferroic heterostructures. In the LSMO/TMO bilayer illustrated in Fig. 1(a), both Mn and Tb spin sublattices in TMO may couple, either directly or indirectly, with the Mn spin lattice in LSMO. In such a case, we expect to observe not only the onset of EB below the Néel temperature of TMO (Fig. 1(a)), but also additional features in the interfacial magnetic coupling at the ordering temperature of Tb spins (Fig. 1(b)). Such an exchange coupling at the oxide interface highlights the intricate interactions between the multiple spin sublattices, enriching the magnetic properties of multiferroic heterostructures.

Results

Figure 2(a) shows the oscillation of the in-site RHEED specular intensity recorded during the growth of bilayer. A layer-by-layer growth mode can be observed for the eight unit cells of LSMO and the first several unit cells of TMO. In addition, as shown in Figure 2(b), the atomic force microscopy (AFM) image measured on the LSMO layer shows a clear step-and-terrace surface with the height of steps being around one unit cell, which is consist with the layer-by-layer growth mode of LSMO. As shown in the XRD θ–2θ scan (Fig. 2(c)), only reflections corresponding to the substrates and the TMO (001) planes were observed, indicating that the films were c-axis oriented. Because of the ultrathin nature of the LSMO layer, we used synchrotron in the measurements. The XRD data of the reference LSMO single layer shows clear fringes, indicating a high crystalline quality and the lattice constant of LSMO can be calculated as 3.811±0.05 Å. The compression in the out-of-plane direction is expected since the lattice constant of LSMO (rhombohedral perovskite bulk structure, pseudocubic a = 3.873 Å) is smaller than that of STO (cubic, a = 3.905 Å). Figure 2(d) further shows the reciprocal space mapping (RSM) data collected around the STO (103) for the TMO reference film. The RSM data clearly reveals that the in-plane lattice parameter of TMO is identical to that of STO, indicating that the films are coherently grown on the substrate and slightly compressively strained55,56.

Figure 2
figure 2

(a) Intensity oscillation of the in-site RHEED recorded during the growth of LSMO/TMO bilayer. (b) AFM image taken after the growth of eight unit cell LSMO layer. (c) XRD θ–2θ data for the LSMO/TMO bilayer and the corresponding reference single layers. Cu-Kα radiation was used to measure the bilayer and the TMO single layer, whereas synchrotron was used to measure the ultrathin LSMO single layer. (d) RSM data around the (103) Bragg reflection of STO measured on the TMO reference single layer. Dashed line is guide to eyes.

Full size image

Figure 3(a) shows the zero field cooling (ZFC) and field cooling (FC) data of magnetization versus temperature measured on the reference single layer TMO sample. The magnetization shows an upturn at low temperature as a result of the ordering of Mn spins. In the plot of inverse susceptibility () vs. temperature, a deviation from the paramagnetic linear behavior occurs at TN ~ 41 K. On decreasing temperature, the cusp feature at ~10 K can be related to the long range Tb spin ordering (TTb)39,43,46.

Figure 3
figure 3

(a) Low temperature ZFC and FC curves measured on the reference TMO single layer. The magnetic field applied during the measurement is 200 Oe. Also shown is the inverse susceptibility () of the FC curve. Dashed lines are guides to the eyes. (b) Temperature dependence of the remanent magnetization measured under the ZFC and FC conditions. The data were normalized according to the value of the FC magnetization at 5 K. Also shown is the difference () between the ZFC and FC data. The weak anomalous feature at ~26 K (*) presumably is related to the onset of the spiral spin structure.

Full size image

Furthermore, we measured the thermal remanent magnetization after ZFC (RM-ZFC) and FC (RM-FC), which often gives valuable information regarding the irreversible magnetization in disordered systems57. During the RM-ZFC (RM-FC) measurements, sample was first cooled down under zero field (6 Tesla) from 300 K, then a 6 Tesla field is applied and removed, which was followed by magnetization measurements under zero field on increasing temperature. As shown in Fig. 3(b), in addition to the magnetic transitions mentioned above, there is a weak anomalous feature at ~26 K in the difference of the ZFC and FC data, which is close to the Mn spiral spin ordering temperature observed in bulk TMO (Tlock) at ~28 K, but it is much less pronounced than the bulk result39.

Figure 4 shows the ZFC and FC curves and the M-H loops measured on the LSMO/TMO bilayer and reference samples. The bifurcation between the ZFC and FC curves which was already observed in the TMO single layer (Fig. 3(a)) becomes stronger in the bilayer (Fig. 4(a)), which likely indicates the magnetic frustration at the LSMO/TMO interface58. It was previously proposed that the strain effect can induce a weak ferromagnetic signal at low temperatures55,59, which was observed in the TMO reference single layer. This weak ferromagnetism of TMO may contribute to the overall magnetism of the LSMO/TMO bilayer. Regarding to the LSMO component, as shown in Fig. 4(a), the Curie temperature of the reference LSMO single layer is ~220 K, which is much lower than the bulk value ~ 369 K. The strong suppression of magnetic properties is likely the result of stain which is known to induce distortion of MnO6 octahedra in the Jahn-Teller systems60,61. In the bilayer, the Curie temperature of LSMO is further suppressed to ~165 K, which indicates that the magnetism of the 8 unit cell LSMO is significantly affected by the presence of the TMO overlayer. We note that the possible intermixing at the LSMO/TMO interface must be limited to one or two unit cells because the Curie temperature of five unit cell LSMO was reported to drastically decease to ~100 K62. However, the exact origin of the Curie temperature reduction in the bilayer and the detailed structure of the interface clearly warrant further studies.

Figure 4
figure 4

(a) ZFC and FC data measured on the LSMO/TMO bilayer and reference single layer samples. The applied magnetic field during measurements is 200 Oe. (b) and (c) show the corresponding magnetic hysteresis loops measured at 50 K and 10 K, respectively. The measurements follow a FC procedure from 300 K under a magnetic field of +4 Tesla. (d) Hysteresis loops measured at 5 K on the LSMO/TMO bilayer after the +4/−4 Tesla FC procedure.

Full size image

The definite evidence of magnetic coupling at the LSMO/TMO interface is the enhancement of coercivity in the bilayer. As shown in Fig. 4(b), at 50 K, HC of the bilayer is enhanced to 178 Oe, which is much larger than the value of the reference LSMO single layer (44 Oe). Further enhancement was observed at lower temperatures: at 10 K HC of the bilayer reaches as high as 960 Oe (Fig. 4(c)). Furthermore, we observed the EB effect in the bilayer as evidenced by the shift of the hysteresis loop towards the opposite direction of the cooling field (Fig. 4(d)); HE at 5 K under a cooling field of 4 Tesla is 42 Oe. The observed exchange coupling and bias unambiguously suggest the existence of interfacial magnetic coupling between LSMO and TMO.

The temperature dependence of HC and HE shown in Fig. 5 sheds more light on the magnetic coupling at LSMO/TMO interface. As the temperature goes down, the EB emerges at the Néel temperature TN of TMO (Fig. 5(a)), which is expected because the AFM order in the TMO layer needs to develop and pin the FM domains of the LSMO layer. However, the magnetic order in TMO is quite complex: besides the A-type and G-type modulated structures, a neutron diffraction study revealed the coexistence of C-type and F-type orderings below TN46. On further decreasing temperature, HE shows an anomalous plateau around TTb and then it increases by almost four fold to a value of 128 Oe. This non-monotonous temperature dependence of HE reflects the complex magnetic interactions between various spin sublattices. It was proposed that the competing exchange interactions, i.e., magnetic couplings between nearest () and next nearest () Mn spin sublattices, along with the clamping magnetic interaction between Tb and Mn () spin sublattices, collectively determine the Mn–O–Mn bound angles and modulate the strength of the exchange interaction between Mn ions in TMO45. As a result of the strong coupling between Tb and Mn spin sublattices of TMO44,45,46, the formation of long-range Tb3+ spin ordering could lead to a significant canting of the neighboring Mn spin order. Consequently, the spin frustration within the AFM TMO layer affects the coupling strength across the interface between the Mn sublattices of TMO and LSMO, resulting in the nontrivial temperature-dependent variation of HE.

Figure 5
figure 5

Temperature dependence of the exchange bias HE (a) and the coercivity Hc (b) measured on the LSMO/TMO bilayer.

For comparison, the HC data of a reference LSMO single layer are also shown.

Full size image

However, it is noteworthy that the schematic in Fig. 1(b) apparently over-simplifies the spin configuration at the LSMO/TMO interface. There remain open questions regarding the magnetic ordering of Tb spins in TMO at low temperatures as well as the exchange interactions between the Tb and Mn sublattices43,44. Furthermore, although in the schematics of Fig. 1(a) and (b) the MnO2 layer at the LSMO/TMO interface adopts the spin order of TMO, it is apparently shared by both layers and spin-flop coupling may develop at the interface31. The detailed elucidation may require neutron and synchrotron experiments. As far as we know, there has been no theory so far on elucidating the magnetic order/disorder at such an interface between a spiral multiferroic and a ferromagnet. The anomalous temperature dependence of EB at TTb invites future theoretical efforts to shed light on the ground-state spin configuration and exchange coupling at such strongly frustrated interfaces.

As shown in Fig. 5(b), the temperature dependence of HC shows a monotonous behavior, which is different from the trend of HE, suggesting that HE and HC perhaps have different origins. As reported previously, the enhancement of HC may occur without the onset of HE, which depends on not only the spin structure in the AFM layer (e.g., collinear or noncollinear) but also the detailed domain structures54. The frustration of interfacial spins and the spin-flop coupling between FM and AFM layers have been proposed to be responsible for a large uniaxial anisotropy and enhanced HC, whereas additional effects like interfacial defects are needed to explain the EB31,32. Moreover, because HC is not an intrinsic property of material, it is sensitively dependent on sample morphology63 and other factors.

The coercivity is supposed to have a power law dependence on the thickness of FM layer, i.e., , when the random interaction at the AFM/FM interface is assumed63. We measured bilayers with the thickness of LSMO layer varying from 8 to 13 unit cells while keeping the thickness of TMO layer fixed. Indeed, as shown in Fig. 6, both HC and HE decrease with increasing LSMO thickness. The enhancement of HC is quite significant within the range of LSMO layer thickness. In a phenomenological model, the exchange bias field is expressed as , where is the interfacial exchange energy density, which represents the strength of interfacial exchange coupling, and are the magnetization and thickness of the FM layer, respectively16,17. As show in Fig. 6, HE appears to follow the trend of 1/tFM, which is consistent with the interfacial nature of the EB effect.

Figure 6
figure 6

The dependence of HE and Hc on the thickness of LSMO, while the thickness of TMO is fixed at 40 nm.

Dashed lines are guides to eyes. Solid lines are theoretical fittings to .

Full size image

Discussion

As mentioned previously, EB has also been reported in heterostructures made of another multiferroic material BFO, which has attracted lots of attention4,21,27. In the previously reported bilayers composed of LSMO (2–10 nm) and BFO (600 nm), an EB shift as high as 225 Oe was observed27. A large HE suggests a high interfacial exchange energy density at the LSMO/BFO interface. On the other hand, the enhancement of Hc is more significant in our LSMO/TMO sample; the largest measured Hc is 1455 Oe, whereas it is about 750 Oe in the LSMO/BFO bilayer reported in reference 27. But we should note here that before utilizing the magnetoelectric properties of TMO in novel devices, the domain structures of TMO must be elucidated, which not only affects the polarization64,65, but also plays a critical role in the EB effect16,17,24,27. However, information on the domain structures of TMO is very limited56. Further studies focused on controlling the domain characteristics of TMO thin films through rational synthesis and local probing are needed to tailor the interface properties of TMO-based heterostructures.

In summary, the pronounced EB and the enhancement of coercivity in LSMO/TMO bilayer confirm the existence of a strong interfacial exchange coupling, which is related to the presence of uncompensated spins at the interface between AFM TMO and FM LSMO. The emerging coupling between the Mn and Tb spin sublattices in TMO frustrates the interfacial magnetic coupling in the collinear/noncollinear spin bilayers, leading to the anomalous plateau-like feature of HE at TTb. These data put constraints on future theoretical attempts of capturing the complex magnetic coupling in TMO and its heterostructures. Future experimental efforts using interface sensitive synchrotron and neutron probes are imperative to reveal the spin configurations and the domain structures in the vicinity of the LSMO/TMO interface. As the first study on the magnetic properties of such collinear/noncollinear magnetic heterostructures, our results also invite theoretical investigations on the exchange coupling at oxide interfaces involving noncollinear and frustrated spin structures. We can envisage that further elucidation of the coupling between multiple spin lattices will open new avenues towards manipulating the ferroic orders in oxide heterostructures.

Methods

We prepared thin film bilayers of LSMO/TMO on TiO2-terminated SrTiO3 (001) single crystal substrates at 750°C using pulsed laser deposition. The frequency of the excimer laser was 1 Hz and 2 Hz for LSMO and TMO, respectively. A laser fluence of ~1.5 J/cm2 was calibrated and used for the depositions. The growth of 8 unit cell LSMO took place under an oxygen pressure of 0.05 mbar and it was in situ monitored by reflection high energy electron diffraction. TMO layers with a thickness of 40 nm were grown at a relatively higher oxygen pressure (0.4 mbar) on top of the LSMO layer. The reference single layers were prepared under the identical conditions. The structure of the samples was characterized using high-resolution x-ray diffraction (XRD) with a diffractometer (Smartlab, Rigaku, Japan) using Cu-Kα radiation. For the ultrathin (eight unit cells) LSMO single layer, synchrotron light (wavelength: 1.5386 Å) was used in the XRD measurements to achieve sufficient intensity. We characterized the magnetic properties of the bilayers and reference single layers using a superconducting quantum interference device (SQUID, Quantum Design, US) magnetometer.