Oxygen-driven anisotropic transport in ultra-thin manganite films

Transition metal oxides have a range of unique properties due to coupling of charge, spin, orbital and lattice degrees of freedom and nearly degenerate multiple ground states. These

properties make them interesting for applications and for fundamental investigations. Here we report a new phase with abnormal transport anisotropy in La0.7Sr0.3MnO3 ultra-thin films under

large tensile strain. This anisotropy is absent in films under smaller tensile strain or compressive strain. Furthermore, thickness and magnetic-field-dependent experiments suggest that the

tensile-strain-induced two-dimensional character is crucial for the observed phenomena. X-ray absorption spectroscopy results indicate that this anisotropy is likely driven by O 2p orbital,

which hybridizes with Mn 3d. Ab initio calculations confirm this result. Our results may help to understand the anisotropic transport behaviour observed in other systems.

The emergence of exotic electronic phases from intricate coupling of charge, spin, orbital and lattice degrees of freedom has been a central theme of research in modern condensed matter

physics1,2,3. Among all the systems studied, complex oxides that contain 3d orbital transition metals are of particular importance for both fundamental study and applications. It is well

known that the physical properties of transition metal oxides depend strongly on the occupancy of Mn 3d orbitals4,5,6,7,8,9,10,11. However, the effect of O 2p orbital has attracted much less

attention so far.

Bulk La0.7Sr0.3MnO3 (LSMO) is a large-bandwidth material with ferromagnetic (FM) metallic ground state and its physical properties can be described successfully by the double-exchange

mechanism12. By depositing LSMO films on different substrates, epitaxial strain ranging from compressive to tensile can be induced (Fig. 1a). Here, we report pronounced in-plane transport

anisotropy in highly strained ultra-thin LSMO films at low temperature. X-ray absorption spectroscopy (XAS) results indicate that this anisotropy is likely driven by O 2p orbital and its

hybridization with Mn 3d, in connection with the strong correlations in the highly tensile-strained LSMO film. This is yet another example of exotic electronic phase induced by strong

charge–spin–orbital–lattice coupling in transition metal oxides.

(a) Lattice mismatch between LSMO and the substrates used in this study: −2.1% for LAO, −0.3% for NGO, +0.8% for STO and +1.9% for DSO. (b) XRD L scans around the (002)pc peaks for 8 nm LSMO

films on various substrates. The stars (*) and arrows (↓;) indicate the substrates and LSMO (002)pc peaks, respectively. Clear Kiessig fringes indicate the high quality of the films. (c)

Topography image of the 8 nm LSMO film on DSO substrate. Scan range is 3 μm × 3 μm. (d) Cross-sectional profile, averaged over the red rectangular box in Fig. 1c, showing 1 unit cell (~3.9 

Å) high steps.

We have deposited high-quality epitaxial LSMO thin films using pulsed laser deposition on (001)pc-oriented LaAlO3 (LAO), (110)o-oriented NdGaO3 (NGO), (001)c-oriented SrTiO3 (STO) and

(110)o-oriented DyScO3 (DSO) substrates (the subscripts ‘o’, ‘c’ and ‘pc’ represent orthorhombic, cubic and pseudocubic structures, respectively), all of which have pseudocubic lattice

constants close to that of LSMO (see Methods for details). Figure 1b shows the XRD patterns of 8 nm LSMO films on these substrates. The arrows indicate the (002)pc peak of LSMO, which shifts

from the left of substrate peak to the right, indicating that the strain changes from compressive to tensile. Figure 1c shows the topography image of the 8 nm LSMO film on DSO. The unit

cell level steps clearly suggest step flow growth mode of the films (Fig. 1d). The films are fully strained to the substrate, which is confirmed by synchrotron reciprocal space mappings (see

Supplementary Fig. S1 and Supplementary Note 1 for details). The results reveal that the 8 nm LSMO on DSO has c=3.826 Å and a tetragonal distortion of c/a=0.968. According to the calculated

temperature–strain phase diagram5, fully strained LSMO film on LAO and DSO have C-type and A-type antiferromagnetic (AFM) ground states at low temperature, respectively, whereas those grown

on NGO and STO have the FM ground state. Indeed, this is confirmed by the temperature dependence of in-plane resistivity as shown in Fig. 2a, which is consistent with previous reports6,7,8.

Under relatively small compressive (on STO) or tensile (on NGO) strain, the LSMO thin film shows metallic behaviour with slightly reduced Curie temperature. Under large compressive or

tensile strains (on LAO and DSO), it becomes insulating.

(a) Temperature dependence of resistivity of 8 nm LSMO thin films on different substrates under zero magnetic field. (b) Schematic illustration of the experimental set-up used to measure the

in-plane resistivity of the LSMO thin films along the two orthogonal directions (x and y). The samples are square with a side length of 5 mm. The Pt electrodes are 0.4 mm × 0.8 mm. (c)

Resistivity versus temperature curves for 9.6 nm LSMO on DSO along the two channels during cooling and heating processes. A hysteresis is observed in the intermediate region indicating

two-phase coexistence; while it is absent in the low- and high-temperature regions. (d) The in-plane resistivity anisotropy, (ρb–ρa)/ρa, of 8 nm LSMO thin films on different substrates. Note

that only the film on DSO under large tensile strain shows pronounced resistivity anisotropy. (e) The anisotropy versus temperature curves of three 8 nm LSMO films on DSO, confirming the

repeatability of the result. Note the switching of anisotropy sign from negative to positive upon cooling.

Interesting results are observed when the in-plane resistivity is measured along both of the orthogonal directions. The measurement set-up is shown schematically in Fig. 2b (the x, y and z

directions are defined along the pseudocubic a, b and c directions. See Supplementary Fig. S2 for details). For the film grown on DSO, abnormal in-plane transport anisotropy is observed

(Fig. 2c) and the resistivity along the longer a direction is lower, which has not been reported before. The resistivity anisotropy (ρb–ρa)/ρa is ~30% at 95 K for the 8 nm film and it does

not show saturation within our measurement range. More interestingly, there is a switch of easy axis (low resistivity axis) at ~130 K. This phenomenon is reproducible as shown in Fig. 2e for

three samples. On the contrary, films grown on STO, NGO and LAO show negligible in-plane resistivity anisotropy (Fig. 2d).

For the film on DSO, the hysteresis in the resistivity–temperature curves between ~130 K and 210 K obtained during cooling and heating suggest coexistence of two phases (Fig. 2c). It is

likely that the anisotropy within this temperature range is the result of anisotropic nucleation and growth of the low temperature phase following the percolation theory. Such behaviour has

been observed in other manganite systems, for example, Nd0.5Sr0.5MnO3 (ref. 13) and La0.325Pr0.3Ca0.375MnO3 (ref. 14). However, the new phenomenon here is the anisotropic resistivity at

temperatures below 130 K, which does not show hysteresis between heating and cooling and suggests new phase with intrinsic anisotropy.

To understand this phase, we have to start with the crystal structures of the substrates. STO has a cubic structure, thus no anisotropy is expected for the LSMO film grown on it. LAO has a

rhombohedral structure, and the two in-plane lattice are equivalent in the pseudocubic unit cell, thus no anisotropy is expected either. NGO, on the other hand, has an orthorhombic structure

with lattice constants of 5.428 Å (along [100]o), 5.498 Å (along [010]o), and 7.709 Å (along [001]o)15. The NGO (110)o surface has a nearly square mesh with pseudocubic lattices of a =3.855

 Å along [001]o and b =3.863 Å along [1–10]o direction, which leads to ~0.2% strain anisotropy in the LSMO films along the two orthogonal in-plane directions. We do not observe any

anisotropy in the transport, indicating that the small strain difference alone is not enough to induce detectable changes in the electron hopping along these two directions. Note that the

pseudocubic lattice constant of NGO is close to that of LSMO, so the overall strain is small (~ −0.3%).

DSO has an orthorhombic structure similar to that of NGO with lattice constants of 5.442 Å (along [001]o), 5.719 Å (along [010]o) and 7.904 Å (along [001]o) (ref. 15). The (110)o surface has

pseudocubic lattices of a=3.952 Å along [001]o/[100]pc and b =3.947 Å along [1–10]o/[010]pc directions (Fig. 2b), resulting in ~1.9% overall tensile strain on LSMO with ~0.13% in-plane

anisotropy. Theoretical calculation has shown that the in-plane Mn–O–Mn bond angle does not change much for c/ab. Both the XAS study and first principles calculations suggest that the

anisotropy is triggered by the preferred occupancy of the O 2px orbitals. Why does this only occur in films under large tensile strain is an open question and further study is needed.

High-quality epitaxial LSMO thin films are grown by pulsed laser deposition on atomically smooth (001)pc-oriented LAO, (110)o-oriented NGO, (001)c-oriented STO and (110)o-oriented DSO

single-crystal substrates. The laser pulse (248 nm) energy density was ~2 J cm−2 and the repetition rate was 3 Hz. The growth was carried out under 200 mTorr oxygen partial pressure at 800 

°C and the growth rate was ~0.8 nm min−1.

The samples were investigated by high-resolution XRD using the X-ray demonstration and development beamline at SSLS (Singapore Synchrotron Light Source). The diffractometer is the Huber

4-circle system 90000-0216/0, with high-precision 0.0001° step size for omega and 2θ scans. The storage ring, Helios 2, was running at 700 MeV with typical stored electron beam current of

300 mA. Incident X-ray beam was conditioned to select Cu Kα1 radiation equivalent (8.048 keV in photon energy) by a Si (111) channel-cut monochromator and toroidal focusing mirror, blocked

to be 1.00 mm high in vertical direction and 3.00 mm wide in horizontal direction by a slit system. Such set-up yielded incident X-ray beam with about 0.006° in vertical divergence. The

detector slit was adjusted to be 1.00 mm high and 3.00 mm wide. The distance between detector slit and sample centre is about 575 mm. The data for HR-XRD RSM were collected under the

framework of (hkl) coordinates with precision better than 1/1,000 reciprocal lattice unit. The typical counting time was 1 s for every step of the scans for the mappings.

In-plane transport property of the films was investigated using a PPMS (physical properties measurement system, Quantum Design) at temperatures ranging from 10 to 350 K. Pt electrodes with a

0.4 mm × 0.8 mm were deposited on the LSMO film. Two perpendicular electrode sets (eight electrical leads) were wire-bonded with Ohmic contacts to the two channels (ch1 and ch2) of the PPMS

dc resistivity puck. Using this set-up, we measured the temperature dependence of resistivity along the two in-plane directions simultaneously.

XAS measurements at Mn L-edge and O K-edge were obtained in total electron yield acquisition mode by recording sample current as a function of photon energy of incident X-rays at the SINS

beamline of SSLS in a ultrahigh vacuum chamber with a base pressure of 1 × 10−10 mbar24. Linearly polarized X-ray beam with a degree of polarization better than 90% and energy resolution

better than 200 meV was employed for the XAS measurements. The incidence angle (θ) of X-rays with respect to the sample surface plane was varied by rotating the polar angle of the sample.

All near edge X-ray absorption fine structure (NEXAFS) spectra were normalized by the incident X-ray intensity (I0) monitored by a refocusing mirror to eliminate the fluctuation in the

incident X-ray intensity.

Density functional calculations were performed using Perdew-Burke-Ernzerhof parametrization of the generalized gradient approximation (GGA-PBE) to exchange-correlation potential in the

density-functional theory as implemented in the VASP code. The projector augmented wave (PAW) pseudopotential was used, where La 5s25p65d16s2, Sr 4s24p65s2, Mn 3p63d54s2 and O 2s22p4 were

treated as valence electrons. We used a La4Sr2Mn6O18 supercell (1 × 3 × 2) to model the La–Sr disordered LSMO alloy (Fig. 5a). All calculations were performed with an energy cutoff of 600 eV

for the plane wave expansion. Atomic relaxations of the supercell are performed using a 6 × 3 × 3 Monkhorst–Pack grid for k-points sampling and atomic positions are converged until the

Hellmann–Feynman forces on each atom are