The properties of manganite/ruthenate superlattices are reviewed with a specific focus on the manganite/ruthenate interface. La0.7Sr0.3MnO3/SrRuO3 and Pr0.7Ca0.3MnO3/SrRuO3 superlattices grow with a high crystalline perfection as illustrated in the figure to the right: at the interface the individual cation species can be clearly identified, interdiffusion is marginal. The superlattices show magnetization processes with an intricate interplay between magnetocrystalline anisotropy, size of the layer magnetization, spin confinement and interfacial antiferromagnetic interlayer coupling. There is further an unprecedented Curie temperature stabilization at room temperature values of the La0.7Sr0.3MnO3 layers in the superlattices down to layer thicknesses of one unit cell. The magnetotransport properties, especially the Hall effect, indicate the existence of a quasi-two-dimensional hole gas at the La0.7Sr0.3MnO3/SrRuO3 interface; this is further supported by an analysis of cation displacements as determined from scanning transmission electron microscopy. The manganite/ruthenate interface might be considered as a model system for the study of interfacial reconstruction and charge transfer in a highly correlated ferromagnetic system.
HAADF-STEM micrograph of two La0.7Sr0.3MnO3/SrRuO3 interfaces in a superlattice with four unit cell thick La0.7Sr0.3MnO3 and eight unit cell thick SrRuO3 layers. The spheres indicate the cations as determined from the Z-contrast.
In oxide materials many physical properties such as ferroelectricity, multiferroicity and high temperature superconductivity are found that are absent in metals. In principle, oxides might be combined in heterostructures and nanostructures to generate new functionalities at the interfaces. In practice, however, the growth of high quality samples often fails for various reasons. Interdiffusion might lead to strong intermixing and the formation of solid solutions; slight changes of the interfacial strain might induce specific growth defects; polar discontinuities might lead to interface defects in the alternating growth of polar and non-polar layers 1, 2; the oxygen pressure during deposition might be different for the various constituents making some material classes, e.g. the perovskites and the ferrites, incompatible to a large extent. On the other hand, model systems indeed exist in which new functionalities appear in heterostructures. As examples one might mention SrTiO3 films grown on DyScO3 in which ferroelectricity is generated by strain 3, SrTiO3/LaAlO3 interfaces at which superconductivity and ferromagnetism are created by interfacial electronic reconstruction from two insulating materials 4–8 and YBa2Cu3O7/La0.7Ca0.3MnO3 superlattices showing a magnetic proximity effect 9. In this review two new model systems, the La0.7Sr0.3MnO3/SrRuO3 and Pr0.7Ca0.3MnO3/SrRuO3 interfaces, are introduced. These systems grow with excellent crystalline perfection and – as will be seen in the course of this review – show rich physics ranging from structural transitions over intricate magnetic properties to the existence of a quasi-two-dimensional hole gas at the interfaces.
All samples discussed in this review were fabricated by pulsed laser deposition (PLD). Our own samples were made at a temperature of 650 °C and in an oxygen partial pressure of 0.14 mbar; these samples were deposited on vicinal SrTiO3 (STO) (001) single crystal substrates, the latter having a low miscut angle of about 0.1°. Prior to deposition the substrates were etched in buffered HF and annealed at 1000 °C for 2 hours in air. This treatment assured substrate surfaces with atomically flat terraces of a width (depending on the miscut angle) between 100 and 500 nm separated by unit-cell high steps. The starting layer was a manganite layer for all superlattices. Wherever samples from other research groups are discussed which were fabricated at significantly different deposition conditions, these conditions will be mentioned.
Bulk SrRuO3 (SRO) crystallizes in an orthorhombic structure (space group Pbnm, lattice parameters a =0.55670 nm, b =0.55304 nm, c =0.78446 nm at 300 K 10); bulk Pr0.7Ca0.3MnO3 (PCMO) also has orthorhombic symmetry (Pbnm, a =0.5426 nm, b =0.5478 nm, c =0.7679 nm 11); bulk La0.7Sr0.3MnO3 (LSMO) has rhombohedral symmetry (Rc, a =0.3876 nm, α = 90.46° 12). The symmetry of thin films of these materials in general depends on the strain between substrate and film. SrRuO3 films grown on SrTiO3 (001) substrates have orthorhombic symmetry with the c-axis parallel to the substrate and the  direction along the substrate normal; c-axis ordering occurs along terrace edges such that SrRuO3 films with one crystallographic domain can be realized on vicinal substrates 13, 14. Actually there is a small distortion of the orthorhombic cell with an angle of 89.8° between the a - and b-axes lowering the crystalline symmetry to monoclinic 13. Pr0.7Ca0.3MnO3 films grow on SrTiO3(001) substrates in the same orientation as SrRuO3 films. La0.7Sr0.3MnO3 films on SrTiO3 have cubic symmetry with a small tetragonal distortion along the substrate normal.
The superlattice samples discussed in this review are designated by the layer thickness in unit cells. The latter depends on the symmetry of the material and the growth direction. Keeping in mind, however, that the rhombohedral, orthorhombic and monoclinic distortions are small, the materials can be characterized by the respective pseudocubic cells. These have a lattice parameter of about 0.39 nm which is used to determine the layer thickness of the individual layers; a designation such as n /m LSMO/SRO stands for a superlattice with La0.7Sr0.3MnO3 and SrRuO3 layers with a thickness of n and m pseudocubic unit cells, respectively. Layer repetition of our own samples was 15.
All samples were routinely characterized by atomic force microscopy (AFM) measurements and X-ray diffractometry. Further, on selected samples, high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM), electron energy loss spectroscopy (EELS) and energy dispersive X-ray (EDX) mappings were done in a TITAN 80-300 FEI microscope (300 keV energy of the primary electrons) with a spherical aberration corrected (c = 0) probe forming system. High-resolution transmission electron microscopy (HRTEM) investigations were performed in a Jeol 4010 (400 keV energy of the primary electrons). Magnetization measurements were performed in a Quantum Design SQUID magnetometer. Magnetoresistance was measured in a He-flow cryostat equipped with an 8 T superconducting solenoid and a rotatable stage with an angular resolution better than 0.01 degree.
Topography of LSMO/SRO superlattices
The topography of the top surface of the samples was imaged by atomic force microscopy. Figure 1 shows an AFM micrograph taken on the surface of a 3/3 LSMO/SRO superlattice and of a 2/2/2/2 LSMO/STO/SRO/STO superlattice, showing that the layers grew in step-flow growth or layer-by-layer on top of the vicinal STO (001) substrate, following the step-and-terrace morphology of the substrate.
Interfacial structure of LSMO/SRO superlattices
The microstructure of the superlattices was investigated by high resolution transmission electron microscopy and high angle annular dark field scanning transmission electron microscopy (HAADF-STEM). Details about the study of the structural quality of interfaces and intermixing in LSMO/SRO superlattices by means of HAADF-STEM are given by Hillebrand et al. . Figure 2 shows HAADF-STEM images, electron energy loss spectroscopy (EELS) maps and Mn and Ru elemental profiles for the 3/3 LSMO/SRO superlattice. The investigations proved the structural integrity of the two types of layers and that sharp interfaces formed between LSMO and SRO, with less intermixing than in the case of the superlattices with thicker layers . LSMO/SRO superlattices with layers nominally as thick as 2 unit cells were also found to exhibit sharp interfaces and good overall microstructure . In order to check the stability of the LSMO/SRO superlattices, annealing experiments were performed by heating the samples ex situ in a furnace in air atmosphere, at temperatures up to 1200 °C for 2 hours. For annealing temperatures of up to about 1000 °C no degradation of the sample microstructure and no extension of the interfacial intermixing, compared with the as-grown samples, were observed by doing STEM and EDX analyses of the annealed samples. Figure 3 shows a STEM micrograph of the sample annealed at 1000 °C, for which no severe intermixing of the LSMO and SRO layers could be observed. Annealing at 1200 °C resulted in a complete intermixing of the LSMO and SRO, forming a solid-solution, but preserving, however, an epitaxial crystalline structure.
Structural transition in PCMO/SRO superlattices
Ziese et al. 17 investigated Pr0.7Ca0.3MnO3/ SrRuO3 superlattices, in order to probe the magnetic interlayer coupling of another manganite compound with SrRuO3. Figure 4 shows a low magnification conventional cross section TEM micrograph of a 4/10 PCMO/SRO superlattice, which proves that the superlattice has very good layer thickness homogeneity and is free of extended structural defects over large areas. The inset in Fig. 4 is a high magnification HAADF-STEM image of the same sample, demonstrating the sharpness of the PCMO/SRO interfaces with intermixing over maximum one unit cell. In the growth direction, the interfaces are asymmetric, the SRO and PCMO layers being terminated by BO2-planes (i.e. RuO2 or MnO2) 17. Structural modifications of the SRO layers of PCMO/SRO superlattices were observed by high resolution TEM. The structure of the ≃4 nm thick SRO layers was tetragonal for superlattices with thicker PCMO layers (i.e. more than 8 unit cells thick PCMO) and orthorhombic for superlattices with thin PCMO layers (4 unit cells thick, such as the one in Fig. 4) 17. The structure of the PCMO layers (4–10 unit cells thick) stayed orthorhombic for all investigated samples.
Other manganite based superlattices
It was pointed out in various studies that structural defects negatively affect the magnetic properties of manganite based superlattices. Fitting Kourkoutis et al. 18 reported HAADF-STEM and EELS investigations of LSMO/SrTiO3 superlattices with 5 unit cells thick individual layers and correlated the microstructure with the overall magnetization behavior of the LSMO layers in the superlattices 18, see Section 4.5. These LSMO/STO superlattices were fabricated by PLD, keeping all parameters constant, but varying the size of the laser spot on the target, thereby changing the laser energy density (laser fluence) on the target. The best microstructure, with reduced Mn/Ti intermixing and no extended structural defects, was obtained for the superlattice grown in a low oxygen pressure of 10–6 Torr O2, with the largest spot size (10.2 × 10.2 cm2), i.e. for the lowest laser fluence.
Bruno et al. 19 also fabricated and studied LSMO/ SrTiO3 superlattices; to investigate possible changes in doping due to interfacial charge transfer, the authors examined the interface chemistry by combined aberration-corrected electron microscopy and electron-energy-loss spectroscopy. The analysis of their STEM and EELS data led to the following conclusions: the Ti oxidation state evolves from Ti3.7+ when the STO is 2 unit cells thick to stoichiometric Ti4+ when the STO is 8 unit cells thick. While Ti atoms in the bulk of the layer were in a 4+ oxidation state when the bilayer repetition (and the STO layer thickness) increased, the interfacial Ti in the thin STO layers exhibited a reduced oxidation state. The Mn oxidation state remained constant and around 3.2+, less than the 3.3+ nominal stoichiometry. The reduced Ti oxidation state in ultrathin STO layers is consistent with an electronic reconstruction at the manganite-titanate interface. The extra La0.7Sr0.3O layer at the interface would provide a nominal extra (1/3)e– per interfacial TiO2 plane reducing the Ti oxidation state. La–Sr intermixing could also cause the reduced oxidation state of Ti. However, this could be excluded, because a negligible La signal was detected in the STO layer and, furthermore, electron doping of STO caused by La enrichment would imply a hole enrichment of the LSMO near the interface which was not observed. The reduced Mn oxidation state indicated a possibly small density (2%) of O vacancy doping.
The interface quality of LSMO/SrTiO3 superlattices with nominally 4 unit cell thick layers was investigated by Gray et al. 20. Interface compositional mixing or roughness over 0.6 nm was found, as well as a significant change in the soft X-ray optical coefficients of LSMO near the interface.
The microstructure of superlattices composed of non-magnetic metallic La0.5Sr0.5TiO3 and ferromagnetic LSMO were reported by Yang et al. 21 and Gu et al. 22. The motivation for these works was to study the role played by the possible charge transfer between the B-site cations across the interfaces in the magnetic and transport properties of the superlattices. HAADF-STEM and EELS investigations of the LSMO/La0.5Sr0.5TiO3 superlattices showed that it had sharp interfaces, was free of structural defects and that the interfaces were affected by an interdiffused region less than 2 unit cells thick. Gu et al. 22 studied LSMO/La0.5Sr0.5TiO3 superlattices grown on different substrates, such as SrTiO3 (001), GdScO3 (001) and LaAlO3 (001).
Recently Boschker et al. 23 proposed a route how to prevent the possible adverse effects of surface reconstructions driven by the polar discontinuity at the interface on the magnetic and transport properties of LSMO ultrathin films grown on SrTiO3 (001). Insertion of a single La0.33Sr0.67O atomic layer resulted in improved interface magnetization and electrical conductivity of ultrathin LSMO films. However, the Curie temperature of such interface-engineered 5 unit cells thick films still stayed below 150 K. Structurally, a high level of La diffusion into the STO substrate (up to 2 nm depth) was observed in the case of the non-interface-engineered LSMO films, by EELS-STEM investigations, and much less diffusion occurred for the interface-engineered samples.
Vrejoiu 24 grew superlattices of La0.7Sr0.3Mn0.95Ru0.05O3/SrTiO3 on SrTiO3(001) with 2 unit cells thick manganite layers separated by 3 unit cells thick SrTiO3 layers. A HAADF-STEM micrograph of such a superlattice is shown in Fig. 5 and demonstrates its good microstructural quality.
Antiferromagnetic coupling: experiments and theory
It was experimentally established by Ke et al. 25, 26 and Padhan et al. 27 that the magnetic interlayer coupling across the La0.7Sr0.3MnO3/SrRuO3 interface is antiferromagnetic, i.e. the magnetic moments of Mn3+/4+ and Ru4+ ions are antiparallel. This was either manifested by a positive exchange bias field, i.e. a right-shift of the magnetization loop 25, 26, see below, or a small anomaly in the magnetization curve vs. temperature at the Curie temperature of the SrRuO3 layers 27, see also early work on La0.7Ca0.3MnO3/SrRuO3 superlattices by Gong et al. 28. A clear antiferromagnetic interlayer coupling was observed by Ziese et al. 29 in magnetization measurements as a function of temperature. Since the average magnetic moment per manganese site is about 30 and therefore considerably larger than the magnetic moment per Ru site of 31, the antiferromagnetic interlayer coupling is most prominent in magnetization curves, if the LSMO–SRO layer thickness ratio is around 2.3. Figure 6 shows magnetization curves of a 4/12 LSMO/SRO superlattice with a layer thickness ratio of 3. The Curie temperature of the LSMO layers is close to room temperature, that of the SRO layers is about 140 K. Cooling the superlattice in an applied field leads to the formation of a ferromagnetic magnetization of the LSMO layers along the applied field and, below 140 K, of the SRO layers antiparallel to the applied field. Depending on the strength and direction of the applied field at low temperatures an almost compensated magnetization can be reached, especially in the remanent state and in small fields applied parallel to the superlattice. The existence of a nearly compensated ferrimagnetic layer magnetization state at a thickness ratio of 3 (instead of 2.3 as expected from the bulk magnetization values) indicates that some magnetic interface reconstruction occurs. SrTiO3 layers of one unit cell thickness inserted between the LSMO and SRO layers are sufficient to suppress the antiferromagnetic coupling 29.
Figure 7 shows hysteresis loops of this 4/12 LSMO/ SRO superlattice at 10 K and above the Curie temperature of the SRO layers at 150 K. At 10 K and in small fields applied parallel to the superlattice an almost compensated magnetization is found. When the parallel field is increased the SRO layer magnetization is gradually rotated towards the magnetic field direction, thus forming Bloch walls at the interfaces. In contrast, when the magnetic field is applied perpendicular to the superlattice, both Mn and Ru magnetic moments are gradually rotated towards the perpendicular direction. Above the Curie temperature of the SRO layers the superlattice behaves as a homogeneous ferromagnet.
In magnetic studies of the Pr0.7Ca0.3MnO3/SrRuO3 system also an antiferromagnetic interlayer coupling between SRO and the weak ferromagnet PCMO was found. In this case the situation is more complex, since, depending on the thickness of the PCMO layers, the SRO layers had either tetragonal or orthorhombic symmetry 17. It was found that the antiferromagnetic interlayer coupling was stronger for orthorhombic SRO layers 32. Figure 8 shows magnetization curves of a 4/11 PCMO/SRO superlattice for magnetic fields applied both parallel and perpendicular to the layers. In this case the Curie temperature of the SRO layers of about 143 K 33 is higher than that of the PCMO layers of about 110 K. Therefore, the SRO layers order their magnetization parallel to the applied field and below about 100 K, in not too strong applied fields, a decrease of the total magnetization was observed due to the antiparallel orientation of the PCMO magnetization. A direct comparison of the magnetic field dependence of the magnetization in Figs. 6 and 8 already shows that the antiferromagnetic coupling is broken up more easily by the application of magnetic fields in the PCMO/SRO superlattices.
Phenomenologically, the antiferromagnetic coupling between Mn3+/4+ and Ru4+ can be understood as arising from the first Goodenough-Kanamori rule: superexchange between two half-filled d-orbitals mediated by oxygen in a 180° bond is strongly antiferromagnetic 34. Detailed density functional theory calculations 29, 35 yield essentially the same result, see Fig. 9. The Mn ions induce a spin polarization of the same orientation on the π bonds, but of opposite orientation on the σ bonds of neighbouring oxygen atoms, whereas the Ru ions induce the same spin polarization on adjacent oxygen atoms. Therefore the spin orientation of Mn and Ru ions, as mediated by the oxygen ions at the interface, is antiparallel 29, 35. The calculation of the total magnetic moment of a 4/12 LSMO/SRO superlattice showed that this does not have a compensated magnetic moment in contrast to the experimental result shown in Fig. 6, although the suppression of the Ru magnetic moment at the interfaces was taken into account 29. This might be explained by a certain amount of Mn/Ru intermixing at the interface as also found experimentally 29. On the other hand He et al. 36 suggested that cooperative tilts of the oxygen octahedra might occur at the LSMO/SRO interface that modify the interfacial magnetic moments and might explain the compensated moment of the supercell. This issue has not yet been settled with probably both effects contributing to the size of the magnetic moment.
(110)-oriented SrRuO3 films grown on SrTiO3 (001) have a somewhat intricate magnetocrystalline anisotropy 37, 38. The in-plane  axis is magnetically hard; a second magnetic hard axis is found in the (001) plane directed under about 30° with respect to the in-plane axis; an easy axis is located in the (001) plane tilted away from the  axis under about 30° towards the axis. Correspondingly, in single orthorhombic SRO films the magnetization is mainly oriented along the substrate normal. The magnetocrystalline anisotropy constants are large of the order of 500 kJ/m3 at low temperatures 38. La0.7Sr0.3MnO3 films grown on SrTiO3 (001) substrates have a cubic magnetocrystalline anisotropy with a negative anisotropy constant K1≃ –1 KJ/m3 at low temperatures 39. Therefore the shape anisotropy of LSMO films dominates the magnetocrystalline anisotropy and the magnetization lies in the film plane.
The magnetic anisotropy of the LSMO/SRO and PCMO/SRO superlattices is rather complex. As discussed in Section 3 in PCMO/SRO superlattices a structural transition of the SRO layers from orthorhombic to tetragonal occurs, when the PCMO layer thickness is varied. This structural transition is accompanied by a strong modification of the magnetocrystalline anisotropy 17: orthorhombic SRO layers have the same magnetocrystalline symmetry as single SRO films, whereas the out-of-plane direction (c-axis) of tetragonal SRO layers is magnetically hard. Accordingly, in PCMO/SRO superlattices with tetragonal SRO layers the SRO layer magnetization lies in the superlattice plane 17. This structural transition further leads to a sign change of the anomalous Hall constant from a negative value at low temperatures in orthorhombic to a positive value in tetragonal SRO layers, clearly showing that the structural transition affects the electronic structure 33. The Hall effect sign change was also found in orthorhombic and tetragonal single SRO films 40.
The situation in LSMO/SRO superlattices is less clear. The magnetization data shown in Fig. 6 show that the out-of-plane direction is the overall magnetically hard direction, since the in-plane magnetization in the remanent state and in small applied fields is considerably larger than the out-of-plane magnetization. For thicker SRO layers the SRO layer magnetization is clearly oriented perpendicular to the layers, see 29. However, the magnetization direction in superlattices with thinner SRO layers need not be homogeneous. Indeed, spin polarized neutron reflectometry measurements on 5/12 and 5/20 LSMO/SRO superlattices show that the SRO magnetic moments close to the interface and the LSMO magnetic moments lie in the superlattice plane, whereas the SRO magnetic moments in the cores of the SRO layers are rotated toward the out-of-plane direction 41. A similar conclusion was reached by Solignac et al. 42 on LSMO/SRO bilayers by magnetometry and polarized neutron reflectometry measurements. Therefore, it appears that the magnetic moment direction in LSMO/SRO superlattices is determined by the bulk magnetocrystalline anisotropy of the constituents and the strong antiferromagnetic interlayer coupling; a structural transition of the SRO layers does not seem to occur.
Exchange bias strength and inverted hysteresis loops
The magnetization processes in the LSMO/ SRO superlattices are determined by various factors: (i) the size of the layer magnetization determines the Zeeman energy in an applied magnetic field and therefore influences the direction of the overall magnetization; (ii) the magnetocrystalline anisotropy strongly influences the direction of the magnetization; (iii) the antiferromagnetic Mn–Ru coupling establishes a strong antiparallel coupling at the interface; (iv) the individual layer thickness sets constraints on the width of domain walls of the Bloch type with a spiral direction perpendicular to the interfaces. In principle, the LSMO–SRO superlattices constitute hard-soft ferromagnetic systems such that the existence of exchange spring effects 43, 44 might be anticipated. These, however, do only occur in a minority of samples; in the samples investigated by us two types of magnetization reversal processes occur. Type 1 is the classical magnetization loop of a hard–soft system: starting in high applied magnetic fields with parallel layer magnetizations, the magnetization of the magnetically soft LSMO layers is reversed first on de creasing the field, whereas the magnetization of the magnetically hard SRO layers is reversed only in magnetic fields of opposite polarity. This type of loop is shown for a 6/12 LSMO/SRO superlattice in Fig. 10(a); it looks unconventional in this case, since the LSMO layer magnetization is larger than the SRO layer magnetization and the magnetization reverses sign in the wrong quadrant leading to an inversion of the central loop 45. Thermodynamically, this loop shape is allowed, since the work done along the complete hysteresis loop is still positive. Reversal processes of type 2 are illustrated in Fig. 10(b); these consist of a three-step reversal process: again starting in high applied magnetic fields with parallel layer magnetizations, the magnetization of the magnetically hard SRO layers reverses first, i.e. the Bloch walls expand into the SRO layers; at low fields a ferrimagnetic layer magnetization state is formed that is reversed by reversing the applied field; in increasing fields of opposite polarity the SRO magnetization is gradually rotated towards the magnetic field direction. Most of the samples investigated show magnetization reversal processes of type 2. Samples showing type 1 reversal processes only do so at low temperatures below a compensation temperature, whereas these show type 2 reversal processes above 45. Experimentally, we could not clearly establish which parameters determine the magnetization reversal process; it appears, however, that type 1 is favoured by larger LSMO layer thicknesses as well as reduced antiferromagnetic interlayer couplings, whereas type 2 is favoured by a reduced magnetocrystalline anisotropy of the SRO layers. The engineering of magnetization hysteresis loop shapes might be used in a magnetocaloric device 46.
The exchange-bias field can only be determined in superlattices with a type 1 reversal process; in this case the exchange-bias field HX is given by the horizontal shift of the LSMO magnetization loop. The inset to Fig. 10(b) shows HX of the 6/12 superlattice below the compensation temperature of 62 K and further the apparent coercive field HC The sign change of HC indicates the crossover from type 2 to type 1 magnetization reversal. The exchange-bias field HX is of the order of 1 T and considerably larger than in typical exchange coupled systems. An effect of field cooling on the value of HX was demonstrated 47 and is probably related to the freezing of Bloch walls by field cooling.
Exchange bias in asymmetric superlattices of the type [LSMO/SrTiO3/SRO]n and [LSMO/BaTiO3/SRO]n was studied by Ziese et al. 48. In case of SrTiO3 interlayers the structural perfection is so high that the magnetization reversal mechanism is of type 2 and an exchange-bias field cannot be measured. This does not mean that the exchange bias vanishes, but that it is so strong that the resulting magnetization processes do not allow for its measurement. In case of BaTiO3 interlayers structural defects in the form of unit cell high steps at the BaTiO3/SRO interface were observed 48 that reduce the exchange-bias strength and lead to a type 1 magnetization reversal mechanism in these superlattices. This results in the counterintuitive finding that a reduction in the exchange strength enables the measurement of the exchange-bias field. HX was about 70 mT at low temperatures in [LSMO/BaTiO3/SRO]n superlattices; this demonstrates the eminent effect of structural perfection on the antiferromagnetic interlayer coupling in these samples.
Curie temperature and layer thickness
It is a common observation that the Curie temperature of ferromagnetic films decreases with decreasing film thickness. This might be related to general physical concepts such as finite size scaling 49, to material-specific intrinsic properties such as electronic phase separation 50 or to growth characteristics and microstructure 51. The strong coupling between electron, spin, orbital and phonon degrees of freedom in the colossal magnetoresistance manganites leads to the formation of insulating antiferromagnetic states in thin layers of originally metallic ferromagnetic compounds 52, 53. On a phenomenological basis this can be either understood by orbital ordering and the weakening of the double exchange mechanism in a particular direction in a strained manganite lattice 53–55 or by interfacially driven electronic phase separation 56, 57. It appeared natural to look for the Curie temperature dependence on layer thickness in LSMO/SRO superlattices, since this might be influenced by both the antiferromagnetic interlayer exchange coupling as well as interfacial electronic reconstruction.
In a systematic study on LSMO/SRO and LSMO/STO/ SRO/STO superlattices a strong dependence of the Curie temperature of the LSMO layers on the presence of a direct LSMO/SRO interface was found 16. All these superlattices showed clear ferromagnetic transitions near room temperature, sizable fractions of the spin-only moment of LSMO of and the antiferromagnetic interlayer coupling between LSMO and SRO layers below the TC of the SRO layers. Thus in LSMO/SRO superlattices the ferromagnetic order in the LSMO layers is stabilized with TC values far above the TC of the SRO layers 16.
The Curie temperatures of the LSMO and SRO layers are shown in Fig. 11 as a function of the respective layer thicknesses. The data fall into two groups. The series of LSMO/SRO superlattices shows a stabilization of the ferromagnetic order in the LSMO layers down to at least a LSMO layer thickness of 2 unit cells; there is actually a trend for the Curie temperature to increase for decreasing layer thicknesses below 5 unit cells. However, there is also a considerable sample-to-sample variation as is evident from the data points at 4 unit cells. On the other hand, the SRO layer TC's in this LSMO/SRO series decrease gradually with decreasing SRO layer thickness as also found in SRO films 58 and SRO/STO superlattices 59. This is the conventional behavior of the Curie temperature in thin ferromagnetic films. In the series of LSMO/STO/SRO/ STO superlattices the Curie temperatures of both LSMO and SRO layers decrease with decreasing layer thickness rather sharply below a thickness of 4 unit cells. This is similar to the behaviour commonly observed in LSMO/ STO and SRO/STO superlattices. The central feature of Fig. 11 is the bifurcation of the LSMO TC curves below a layer thickness of 5 unit cells that clearly proves the stabilization of the ferromagnetic order in the LSMO layers of the LSMO/SRO superlattices.
The stabilization of the ferromagnetic order in the LSMO/SRO superlattices might be due to various factors such as a RKKY-like coupling of the LSMO layers across the conducting SRO layers, the magnetic coupling between Mn and Ru ions at the interface, charge transfer from SRO to LSMO layers or a reduction in the density of growth defects in the LSMO layers due to the modified epitaxy conditions. The first factor could be experimentally excluded 16. It is unlikely that the TC of the LSMO layers is stabilized by the antiferromagnetic Mn–O–Ru coupling, since the TC of the SRO layers is far too low for this. Although the antiferromagnetic coupling is not responsible for the Curie temperature stabilization in the LSMO layers, a direct interface between LSMO and SRO is essential, since the insertion of STO interlayers significantly reduces the magnetic moment and the Curie temperature. Therefore, a charge transfer mechanism 60 might play a vital role in the stabilization of the ferromagnetic order, see also 61. The results are further discussed in Section 5.
Other manganite based superlattices
There is extensive work on manganite/SrTiO3 superlattices 18–20, 23, 62–65. In general, with one exception 18, in these superlattices a strong decrease of the Curie temperature of the manganite layers was found when the manganite layer thickness was reduced below a thickness of a few unit cells. As an example and in spite of the control of the interfacial charge states, the results of Bruno et al. 19 for their LSMO/SrTiO3 superlattices follow the general trend: the superlattices with 4 unit cell thick LSMO layers showed a much reduced Mn XMCD signal even for 2 unit cells thick STO spacers, when compared with superlattices with 6 unit cell thick LSMO layers. Exceptionally high Curie temperatures in LSMO/SrTiO3 superlattices, although not reaching the values reported here for the LSMO/SRO superlattices, were found by Fitting Kourkoutis et al. 18 under specific deposition conditions, see Section 3.4. A Curie temperature approaching ≃250 K as well as metallic behavior were measured for the 5 unit cells thick LSMO layers of the best (oxygen pressure of 10–6 Torr O2, laser spot size 10.2 × 10.2 cm2) superlattice.
In superlattices with La-doped SrTiO3 layers the situation is quite similar. In the work of Yang et al. 21, the thinnest LSMO layers were 6 unit cells thick, had a Curie temperature of only 125 K and exhibited insulating behavior, whereas at a thickness of 12 unit cells the superlattices had a Curie temperature of 240 K and were metallic.
Moderate amounts of doping with Ru (about 5%) of La0.6Sr0.4MnO3 thin films grown on SrTiO3 (001) was reported to yield an enhancement of the Curie temperature and of the coercive field 66. This was corroborated by work by Vrejoiu 24 with the Curie temperature of a single 40 nm thick La0.6Sr0.3Mn0.95Ru0.05O3 film being 340 K and the coercive fields being somewhat enhanced. However, the Curie temperature of the ultrathin Ru-doped LSMO layers in corresponding superlattices with SrTiO324 was fairly low, around 50 K. This rules out doping effects as origin of the Curie temperature stabilization reported in Section 4.4.
Resistivity and magnetoresistance
LSMO/ SRO and PCMO/SRO superlattices are conducting 16, 33, 48. This might at first sight not seem to be surprising, since SRO is an itinerant ferromagnet; however, the behaviour of the resistivity of SRO thin films 58, 67, 68 and SRO layers in superlattices 59, 68, 69 in the thickness regime of a few monolayers is still a matter of debate. Whereas Chang et al. 58 reported metallic conductivity down to layer thicknesses of 2 unit cells, Xia et al. 67 and Bern et al. 68 observed a crossover to insulating behaviour below a critical thickness of about 4 unit cells. The differences in the resistivity of these thin SRO films on SrTiO3 (001) were attributed to the growth mechanism, i.e. step-flow vs. 3D island growth 58. In SRO superlattices with SrTiO359, 68 and LaAlO369 also a critical thickness for the transition to insulating behaviour was observed. This appears to depend on strain and is below 2 unit cells in SRO/SrTiO359 and about 4 unit cells in SRO/LaAlO369 superlattices. These data lead to the conclusion that SRO layers with a thickness of 4 unit cells under moderate strain are certainly metallic.
Since PCMO is insulating, the resistivity of PCMO/ SRO superlattices is determined by the SRO layers. LSMO bulk and thicker film samples show metallic conductivity 30, but with a resistivity higher than that of SRO films. Therefore one would expect that the resistivity of LSMO/SRO superlattices would also be dominated by the SRO layers. As shown in Fig. 12 this is indeed so: the resistivity of LSMO/SRO superlattices shows the same temperature dependence as that of a SRO single film and is of the same order of magnitude. There is a slope change of the resistivity at the Curie temperature, see arrow in Fig. 12. The bulk resistivity of a 38 unit cell thick LSMO film is shown for comparison in Fig. 12; it is larger than the resistivity of the superlattices and has a distinctly different temperature dependence. The slight enhancement of the superlattice resistivity compared to the thin film resistivity might be attributed to increased interfacial scattering.
It is well known 52, 53 that very thin LSMO films are insulating, since crystalline distortions favour superexchange over double exchange and lead to antiferromagnetic ground states of these thin films. One might therefore argue that the four unit cell thick LSMO layers 70 in the samples shown in Fig. 12 are insulating anyway and cannot contribute to the resistivity. This argument, however, is at odds with the observation in Section 4.4 that the LSMO layers in the LSMO/SRO superlattices are ferromagnetic with a TC close to room temperature down to layer thicknesses of one unit cell. Obviously, it is too simplistic to transfer results obtained on LSMO single films 52, 53, 70 onto LSMO layers embedded in superlattices. Here we have to look at interfacial reconstruction, charge transfer and hybridization effects in more detail.
The magnetoresistance (MR) of a SRO single film as well as LSMO/SRO superlattices 4/12 and 4/8 are shown in Fig. 13 as a function of magnetic field at 10 K. The magnetoresistance was measured in longitudinal (magnetic field parallel to current density) and transverse (magnetic field perpendicular to current density) configuration. In all samples the magnetoresistance has a pronounced direction dependence of the magnetization indicating that it is due to anisotropic magnetoresistance 71. In the SRO films shallow maxima/minima appear in the magnetoresistance curves at the coercive fields, as it is typical for MR loops. The LSMO/SRO superlattice 4/12 shows MR maxima/ minima at considerably higher field strengths; these can be related to the fields necessary to rotate the SRO layers towards the magnetic field direction, see the corresponding magnetization curves in Fig. 7. It is important to note, however, that the only features observed in the MR of this sample are due to magnetization processes in the SRO layers. This is different in the 4/8 LSMO/SRO superlattice, see Fig. 13(c), with further MR features appearing at lower magnetic fields. The magnetoresistance of this sample is shown in Fig. 14 at higher temperatures of 50, 100 and 150 K. Whereas at 50 and 100 K the MR maxima/minima caused by the magnetization rotation processes in the SRO layers are clearly seen, at all temperatures there are sharp MR features at lower magnetic fields that are due to the reversal of the ferrimagnetic layer structure. This observation might be interpreted in two ways: (i) The current passed through the superlattices is carried by LSMO and SRO layers in parallel, and the more so by the LSMO layers, the thinner the SRO layers. Therefore the magnetoresistance of the 4/8 superlattice is more strongly influenced by magnetization processes in the LSMO layers than that of the 4/12 superlattice. (ii) The electrical current flows in the SRO layers only. However, the magnetoresistance of the SRO layers is influenced by magnetization reversals in the LSMO layers through their coupling to the SRO layer magnetization. This is more severe in case of thinner SRO layers, since the coupling is stronger and since the unaffected core of the SRO layers becomes thinner. On the basis of the magnetoresistance data it is impossible to decide which interpretation is correct. Therefore we turn to the discussion of the Hall effect in the next section.
Hall effect and two-dimensional hole gas
The Hall effect of LSMO/SRO superlattices with various layer thicknesses was measured 16. For ferromagnetic samples the Hall resistivity is expected to follow with the ordinary and anomalous Hall constants; M denotes the magnetization component perpendicular to the layers 72. In principle, Hall effect measurements allow for the determination of the carrier density by identifying the high field slope of with the ordinary Hall constant and using to calculate an effective carrier density n; e denotes the electronic charge. Apart from charge carrier compensation effects affecting n, the experimental difficulty lies in a reliable determination of it is often found that the high field Hall resistivity slope, especially at higher temperatures, where the magnetization is not fully saturated, is affected by the anomalous Hall effect.
Figure 15 shows the Hall resistivity of LSMO/SRO superlattices 4/20 and 4/8 at temperatures 50, 150 and 250 K. In both cases an anomalous Hall contribution was observed at low magnetic fields and a linear field dependence in highest magnetic fields. There are, however, significant differences in the Hall effect of both samples that highlight the role of the LSMO layers in the transport processes in these samples. Overall, the Hall effect in superlattice 4/20 is rather similar to that of single orthorhombic SRO films 33, 73, 74. At low temperatures the anomalous Hall coefficient is negative, at higher temperatures it becomes positive. The high field slope at 50 K is negative indicating electron conduction as it is well established for SRO 14, 75. At temperatures of 150 and 250 K the high field slope is positive, since it is dominated by the anomalous Hall effect. The only difference in the data of superlattice 4/20 and single SRO films lies in the small, ferromagnetic-like anomalous Hall effect contribution observed at small fields at 250 K, i.e. at temperatures much higher than the TC of the SRO layers. Since this contribution is absent in single SRO films, it must be attributed to the presence of the ferromagnetic LSMO layers. The Hall effect of superlattice 4/8 in Fig. 15(b) looks distinctly different. The high field slope is positive at all temperatures indicating hole conduction; the anomalous Hall effect is negative at all temperatures in contrast to the behaviour of SRO films, but consistent with the Hall effect of manganite films 76. At first glance it appears that the Hall effect of LSMO/SRO superlattices with thinner SRO layers is dominated by the LSMO layers. However, assuming parallel conduction of LSMO and SRO layers that would mean that the LSMO layers in the superlattices had acquired a conductivity even higher than in the bulk 16.
The Hall effect data of a single SRO film and several LSMO/SRO superlattices were further characterized by determining the carrier concentration as well as the saturation anomalous Hall effect 16. The number of carriers per unit cell is shown in Fig. 16(a). In agreement with literature values 75 SRO shows electron conduction. With decreasing thickness of the SRO layers the carrier concentration of the superlattices evolves from electron to hole conduction. This indicates the formation of a hole gas within the LSMO layers contributing to the overall carrier concentration. The temperature dependence of the Hall effect of orthorhombic SRO explains the data for the samples with thicker SRO films 73. However, the weak temperature dependence and comparatively large value of the anomalous Hall contribution of samples 3/3 and 2/2 is unexpected. This might be interpreted as the Hall effect contribution from the interfacial hole gas that emerges more and more clearly for thinner layers 16.
In recent work Borisevich et al. 77 analyzed the atomic displacements of Mn and Ru ions near a LSMO/SRO interface as obtained from STEM images. At the interface an electric polarization develops, although both SRO and LSMO are metallic. The electric polarization, electric field, potential and charge distribution reconstructed from the STEM images are shown in Fig. 17. Borisevich et al. 77 attributed the observed effects to oxygen vacancy segregation. However, in view of our Hall effect data that clearly indicate transport parallel to the interfaces, the observed phenomena might be related to real charge transfer effects. From Fig. 17 note that the potential variation occurs within about five unit cells to either side of the LSMO/SRO interface. This would mean that the LSMO layers in the superlattices presented here would be fully conducting.
In this review we have shown that manganite/ruthenate superlattices display a large variety of fascinating physical phenomena. In Pr0.7Ca0.3MnO3/SrRuO3 superlattices a structural transition of the SrRuO3 from orthorhombic to tetragonal symmetry was found, when the thickness of the Pr0.7Ca0.3MnO3 was varied. There are indications of a similar effect in La0.7Sr0.3MnO3/SrRuO3 superlattices, but detailed structural investigations are still needed for a confirmation. At present it is open, whether the structural transition is entirely driven by strain or whether an electronic mechanism, probably involving interfacial electronic, orbital or magnetic reconstructions, is involved. On the other hand, this might be related to the induced charge states observed at the La0.7Sr0.3MnO3/SrRuO3 interface. A priori one might expect that superlattices with ultrathin layers form a new artificial material, since electronic states hybridize across adjacent layers. The experimental evidence on the La0.7Sr0.3MnO3/SrRuO3 superlattices indicates otherwise: samples with one or two unit cell thick layers still show two separate ferromagnetic transitions close to the Curie temperatures of the constituent bulk materials. Therefore we view the properties of the superlattices in terms of the characteristics of the single layers adding magnetic interlayer couplings and charge transfer effects.
What remains? Evidence for a quasi-two-dimensional hole gas at the La0.7Sr0.3MnO3/SrRuO3 interface has come from extensive global resistivity and Hall effect measurements. Microscopic evidence for an interfacial potential and electric polarization has been obtained from the analysis of atomic displacements. Here it would be desirable to obtain further microscopic evidence on the charge states at the interface. It has been shown that ultrathin SrTiO3 interlayers suppress the formation of the hole gas; the effect of other interlayers, e.g. ferroelectric BaTiO3 or multiferroic BiFeO3 would be of high interest, since these materials would modify the interface potential. Ultimately, the studies should be extended to other constituent materials.
This work was supported by the German Science Foundation (DFG) within the Collaborative Research Center SFB 762 “Functionality of Oxide Interfaces”. We thank Dr. Eckhard Pippel for the excellent STEM work with the TITAN 80-300 FEI microscope at the MPI Halle and Prof. Dietrich Hesse for a critical reading of the manuscript.
Born 1964 in Uetersen, Schleswig-Holstein, Germany, Michael Ziese studied at the University of Hamburg where he graduated with a diploma in Physics in 1992. He obtained his Ph.D. from the University of Bayreuth in 1995 with a study on flux-line lattice dynamics in superconductors. After working as research associate in the Physics Departments at the Vrije Universiteit Amsterdam in 1995, the University of Leipzig in 1996 and the University of Sheffield from 1997 to 1999, he joined the Physics Department at the University of Leipzig. In 2005 he finished his habilitation and was appointed Privatdozent; since 2010 he is head of the undergraduate Physics Laboratory. Scientific interests are magnetic and transport properties of oxide materials, especially at interfaces.
Dr. Ionela Vrejoiu received her MSc degree from the Department of Physics, at Bucharest University, Romania, in year 2000 and her Ph.D. from the Department of Applied Physics at Johannes Kepler University in Linz, Austria, in 2004. During her Ph.D. she gained extensive experience in pulsed-laser deposition of thin films of various materials. After four years postdoctoral experience, from April 2009 she has been a Minerva Research Group leader at Max Planck Institute of Microstructure Physics in Halle, Germany, researching on nanoscale ferroelectric and multiferroic heterostructures. Now she continues the activity of her Minerva Group at the Max Planck Institute for Solid State Research in Stuttgart, Germany.