Introduction

Perovskite manganite R1-xAxMnO3 (R = rare earth element; A = alkaline metal) has attracted long standing attention because of its fascinating properties related to the correlations between spin, charge, and orbital degrees of freedom. In 1994, the discovery of colossal magneto-resistance in ferromagnetic La0.67Ca0.33MnO3 thin films activated the potential applications of perovskite oxides on magnetic recording media1. However, there are still many issues to resolve before the practical integration of perovskite oxides into semiconductor devices. They include developing an understanding of the band structure and controlling the band gap. Similar to conventional semiconductors, the electronic structure of R1-xAxMnO3 can be tuned using either hole doping (x > 0)2,3,4,5,6,7,8 or internal strain7,8. Unlike conventional semiconductors, the transport mechanism of R1-xAxMnO3 involves strong electron-electron correlations, leading to complex transport properties different from the typical metallic or insulating behavior. Strong on-site repulsion (Hubbard U-term) between 3d electrons may cause an integer-filled 3dn configuration, resulting in a Mott-insulator. Before turning into a Mott-insulator, the degree of hybridization of O - 2p and Mn - 3d bands may be modified to cover a wide range of transport properties from the charge-transfer type (such as AMnO3) to Mott-Hubbard type (such as in RMnO3)9,10,11,12. Based on the Mott-Hubbard type theory, the charge fluctuations of between dn(i)dn(j) and dn-1(i)dn+1(j) states are strongly suppressed by high exchange energies, with i and j being different transition-metal sites. Additionally, the unique characteristics of the intermedium state open a new channel to tune the gap energy via ionic valence. Many reports suggest that the magnetic and electrical properties of mixed-valence R1-xAxMnO3 compounds could be modified either by changing the doping level, x, or by inserting various rare earth elements13,14,15,16,17. Accordingly, a reduction of the effective ionic size reff [reff = (1 − x)rRare + xrAlkai] would enhance the local deformation of MnO6 octahedron and narrow the effective width of the eg band. An increment of the ionic radii mismatch between R and A could also induce quenched lattice disorder16, which influences the stability of the ferromagnetic/orbital phase. Here, the x-effect is more complex since the variation of x could simultaneously change both the carrier concentration and the effective ionic size. Therefore, a simple approach to study the pure ionic effect is to fix the doping concentration. To design a functional read-write device, the required energies of switching between different electronic states are crucial. Overall, the correlation between crystal structure, electronic energy state, the electrical transport and magnetic property in manganite oxides is not yet fully understood14,15,16,17.

Although many papers report the magnetic and electronic properties of R1-xAxMnO318,19,20, they mostly report on bulks and much less often on films21,22,23,24. For future applications of manganite films, a detailed characterization of the film properties is required. Particularly for the narrow bandwidth manganites with R = Pr, Sm and Nd, the competing magnetic and electronic states often reside in close energetic proximity, and therefore their physical properties are sensitive to atomic replacement24. In thin film form, the interfacial strain could play an important role in addition to the atomic replacement. In this work, x is fixed as 0.33 to study the effects of R/A replacement on the properties of R1-xAxMnO3 thin films. We report a systematic analysis on the electronic structures of R0.67A0.33MnO3 (R = La, Sm; A = Sr, Ca) thin films with X-ray absorption spectroscopy (XAS) and X-ray photoelectron spectroscopy (XPS). Using La0.67Sr0.33MnO3 (LSMO) as a reference compound, we find that replacing a La-ion with a Sm-ion is more effective than replacing Sr with Ca to modify the band structure, regardless of the similar effective ionic sizes in La0.67Ca0.33MnO3 (LCMO) and Sm0.67Sr0.33MnO3 (SSMO). This phenomenon has been observed in bulk samples and explained by the quenched disorder16, but it could be more (or less) profound in thin film form due to the additional interfacial strain. To our knowledge, this article is the first report discussing the strain effect on the electronic structure of SSMO thin film. The mixed phase of ferromagnetic and antiferromagnetic states makes the properties of SSMO more sensitive to changes in crystal structure20. The significant difference of physical properties between bulk and thin film has important implications for device design.

Results

θ-2θ XRD patterns of LSMO, LCMO, and SSMO thin films of 10 nm thickness are plotted in Fig. 1. There is no trace of any impurity peaks in XRD patterns. Only the (00 l) peaks of STO and the films are observed, indicating that all films are c-oriented along the surface-normal direction. Lattice parameters of films are calculated based on the (00 l) diffraction peaks. The obtained c-lattice parameter of orthorhombic structure is 7.69 Å for LSMO, 7.63 Å for LCMO, and 7.56 Å for SSMO films (all shorter than their bulk values), suggesting a tensile interfacial strain on the planes of all three films. However, the reciprocal space mapping data of XRD is desirable, which could provide information on in-plane lattice constants.

Figure 1
figure 1

θ−2θ XRD diffraction patterns for LSMO, LCMO, and SSMO films, following the sequence from bottom to top.

Full size image

A cross sectional FETEM image for STO/LSMO single layer is shown in Fig. 2(a). A smooth interface between the substrate and the film is demonstrated with a clear image of long-range ordered atoms in the LSMO film as well as in those of LCMO and SSMO. The arrow points to the direction of surface-normal films. Figures 2(b–d) are the patterns of nano-beam electron diffraction (NBD) for LSMO, LCMO, and SSMO films, respectively. d002 and d110 mark the distance between the incident beam T and diffraction spots of (002) and (110), respectively. The ratio of d002 and d110 is estimated to be about 1, consistent with the database of orthorhombic structures with the space group of Pnma25,26,27.

Figure 2
figure 2

(a) Cross-sectional HRTEM image of (001) STO/LSMO. The nano-beam diffraction patterns for (b) LSMO, (c) LCMO and (d) SSMO.

Full size image

Curves of resistivity (ρ) vs. temperature (T) from 300 to 30 K for three films are plotted in Fig. 3(a). An insulator-to-metal transition temperature (TIM) is defined at the position of the peak (135 K) for LCMO and SSMO films, while the metal-like behavior of LSMO film indicates its TIM is higher than 300 K. The value of ρ at 300 K for SSMO film is 355 mΩ-cm, which is 150 times larger than that of LSMO (2.3 mΩ-cm) and 6 times larger than LCMO (56 mΩ-cm). The T-dependent normalized magnetization M/M(10 K) from 300 to 10 K is plotted in Fig. 3(b). The paramagnetic-ferromagnetic transition temperature Tc is defined as the crossing point of an extracted line of the paramagnetic state and the x-axis in the curves of M(T) as indicated with dash lines; it is determined as 331 K for LSMO (see the inset), 126 K for LCMO, and 79 K for SSMO. The values of TIM and Tc are very close for LSMO but different by 56 K for SSMO, indicating the double-exchange model may not be applicable in SSMO. In addition, the magnetic moments obtained at 10 K for these three films are very different. It is 2.7 μB/Mn for LSMO, 1.2 μB/Mn for LCMO, and 0.32 μB/Mn for SSMO.

Figure 3
figure 3

ρ(T) curves (a) and the normalized M(T) curves (b) for LSMO (■), LCMO (○), and SSMO (▲) film. The inset of (b) is the M(T) curve of LSMO film in the range of 250 K~350 K.

Full size image

Figure 4 shows the absorption spectra of Mn 2p for all three films, along with those for bulk MnO, SmMnO3, and MnO2 as the reference spectra of Mn2+, Mn3+, and Mn4+. Accordingly, the energies of the L3-edge are 639.9 eV for Mn2+, 641.8 eV for Mn3+, and 643.5 eV for Mn4+. The peak positions of all three films are located at around 642 eV, which is between the energies of Mn3+ and Mn4+. Thus, the results of absorption spectra confirm a constant mixed-valence of the Mn ion for all three films.

Figure 4
figure 4

Mn L2,3-edge absorption spectra for the polycrystalline samples of MnO (black line), SmMnO3 (red line) and MnO2 (blue line), and for the film samples of LCMO (green line), LSMO (brown line), and SSMO (cyan line).

Full size image

The XPS and XAS data are combined to display the whole range of electronic structure from −11 to 11 eV, as shown in Fig. 5. We set the fermi energy to zero; therefore, negative energies fall in the valence band and positive energies in the conduction band. The zero energy is calibrated with the carbon element and the MnO on the film’s surface for XPS and XAS, respectively. Both the XPS and XAS spectra are fitted with multiple Gaussian functions to determine the energy level of each band. The spectrum within the range of −3 eV and 6 eV is attributed to the hybridization of O 2p and Mn 3d-electrons, in which the energy structures can be decomposed into five orbital states, \({t}_{2g}^{2}(\,\uparrow \,),\,\,{t}_{2g}^{3}(\,\uparrow \,),\,{e}_{g}^{1}(\,\uparrow \,),\,{t}_{2g}^{4}(\,\downarrow \,)\) and eg(↓), plotted as green curves in Fig. 5. The energy levels of (Pr,Nd)0.7(Sr,Ca)0.3MnO3 films are used as references28 for the band assignment and marked as vertical lines at the bottom of Fig. 5.

Figure 5
figure 5

XPS (left-hand side from zero) and XAS spectra (right-hand side from zero) of (a) LSMO, (b) LCMO and (c) SSMO with respect to the patterns from top to bottom. The simulated spectra are drawn by black lines and the observed data are represented by open circle “○”. The simulated spectra are fitted by Gauss functions and displayed as several features drawn by green lines. The five t2g and eg states of Mn 3d electrons are indicated at the bottom. The vertical lines at the bottom are the reference energy states obtained from ref. 25.

Full size image

Apart from the orbital energy levels, three energies are extracted as the band gap (Eg), charge transfer energy Ecf, and the magnetic exchange energy (Eex). Eg mimics the band gap in semiconductors and is determined by the energy separation between the two midpoints of the pre-edge feature near EF, marked with two dashed lines as Un+1 and Un-1 in Fig. 5(a–c). The value of Eg is estimated at 2.64 eV for LSMO, 3.03 eV for LCMO, and 4.36 eV for SSMO. Ecf is determined by the energy separation between \({e}_{g}^{1}(\,\uparrow \,)\) and \(\,{t}_{2g}^{3}\)(↑) states and estimated at 2.15 eV, 2.25 eV, and 3.09 eV for LSMO, LCMO, and SSMO, respectively. Eex is determined by the energy difference between \({t}_{2g}^{4}(\,\downarrow \,)\) and \(\,{t}_{2g}^{3}(\,\uparrow \,)\), which is 2.96 eV, 3.01 eV, and 3.91 eV for LSMO, LCMO, and SSMO, respectively.

Discussion

From the XPS and XAS data, we conclude that the band structures of LSMO, LCMO, and SSMO are different due to the modification of the hybridization between Mn 3d and O 2p states. The obtained three energies Eg, Ecf, and Eex, as well as the orbital energy levels, are listed in Table 1 for LSMO, LCMO, and SSMO films. The energy levels of un-doped bulk samples of LaMnO3 and CaMnO3 are also listed in Table 1 for reference29,30,31,32,33,34. Compared with LSMO, significant increases in Eg, Ecf, and Eex is observed in SSMO, in contrast with a minor change in LCMO. It is interesting to note that the band gaps of LSMO and LCMO differ by 15% with a change of reff by 0.04 Å, while the band gap in SSMO is larger than that of LSMO by 65% with a change of reff by 0.05 Å. This implies that the value of reff may not be the prime factor influencing the degree of hybridization between O 2p and Mn 3d band in these films. The distinction between SSMO and the other two samples is reflected from the ρ(T) behaviors of three films. As seen from Fig. 3(a), the ρ-value of SSMO below the transition temperature increases with decreasing temperature, suggesting the ground state of SSMO is an insulating state. Conversely, the ρ-value of LSMO and LCMO below the transition temperature reduces with decreasing temperature, consistent with the nature of metallic state. In addition, the values of TIM (insulator-metal transition) and Tc (magnetic transition) deviate by 50 K in SSMO film, while these two values are very close in LSMO and LCMO films.

Table 1 Various physical parameters including the effective radius of A-ions rA, orbital states \(({t}_{2g}^{2},\,{t}_{2g}^{3},\,{e}_{g}^{1},\,{t}_{2g}^{4},\,{e}_{g}^{2},\,and\,{e}_{g}^{3})\), band gap (Eg), exchange energy (Eex), and charge fluctuation energy (Ecf) for LSMO, LCMO, and SSMO films with the parent compounds LaMnO3 and CaMnO3 listed as references.
Full size table

Does the unusual behavior of SSMO film relate to its interfacial strain effect as reflected by the change of c-lattice parameter? It is noted that the c-lattice parameter is also sensitive to the chemical composition, e.g. oxygen or cation deficiency. To avoid an oxygen deficiency, the film was post-annealed in an oxygen atmosphere after growth (as described in the following section). To understand the chemical composition of oxide film, we previously performed the atomic-scale interfacial studies on our Nd0.35Sr0.65MnO3 films and the results proved a well-defined interface with a correct cation composition35,36. Therefore, we attribute the changes of c-lattice parameter to the interfacial strain. The interfacial strain ε of SSMO film is compared to those of LSMO and LCMO films. Theoretically, the ε value [defined as (√2 * aSTO − abulk)/abulk] between STO substrate (a = b = c = 3.906 Å) and LSMO bulk (pseudocubic, a = b = 5.472 Å) is 0.95% with respect to both the a- and b-axes. For LCMO (a = 5.460 Å and b = 5.476 Å) on STO, ε is around 1.17% along the a-axis and 0.88% along the b-axis. For SSMO (a = 5.418 Å and b = 5.430 Å) on STO, ε is 1.95% along the a-axis and 1.73% along the b-axes. To separate the strain effect from the size effect, we have measured the values of TC and TIM for SSMO bulk, (the data are not shown here). Compared with the properties of SSMO bulk, the difference between Tc and TIM of SSMO films (79 K and 136 K, respectively) indeed are much higher than those of the bulk (85 K and 67 K, respectively). It is evident that the tensile strain has a strong influence on the electrical transition but less on the magnetic transition of SSMO films.

According to a previous report on the strain effects of La1-xBaxMnO3 thin films22, the strain effect on the Curie temperature (Tc) was depending on x. The interfacial compressive strain reduced Tc for x = 0.3 and 0.33. This phenomenon was attributed to the competition of strain-induced modification of Mn-O bond length and eg orbital stability. In Eu0.7Sr0.3MnO3 thin films23, both the tensile and the compressive strain reduced Tc, which was correlated to the decrease of saturation magnetization. In contrary, our result of strain effect on Tc for SSMO film is different from those of La0.67Ba0.33MnO3 and Eu0.7Sr0.3MnO3. The tensile strain reduces the effective magnetic moment by one order of magnitude but the Tc only by 6 K. And TMI is greatly enhanced by near 70 K in SSMO film in comparison with bulk. Considering that a strong A-site disorder in Sm0.7Ba0.3MnO3 bulk could induce a huge separation of TMI and Tc16, it is possible that the anisotropic tensile strain enhances the A-site disorder in SSMO and thus raises TMI. Since the magnetic transition is dominated by the major phase of ferromagnetic domains, it is not affected by strain as much as the electrical and electronic properties are.

In summary, the electrical, magnetic, and electronic properties of three optimally hole-doped Perovskite manganite films are systematically investigated. The effects of interfacial strain are compared with that of atomic replacement. The temperature-dependent resistivity data show that the ground state of SSMO film is insulating while it is metallic for LSMO and LCMO films, suggesting the transport mechanism in SSMO film is different from other two samples. In addition, the strain effect on the insulator-metal temperature of SSMO film is much more significant compared with that of LSMO and LCMO films. The energy levels of \({t}_{2g}^{2}(\,\uparrow \,),\,{t}_{2g}^{3}(\,\uparrow \,),\,{e}_{g}^{1}(\,\uparrow \,),{t}_{2g}^{4}(\,\downarrow \,),\) and eg(↓) are identified within the range of the hybridization band. Accordingly, the values of the band gap, charge fluctuation energy, and magnetic exchange energy (Eg, Ecf, Eex) are obtained as (2.6 eV, 2.1 eV, 2.9 eV) for LSMO film, (3.0 eV, 2.2 eV, 3.0 eV) for LCMO film, and (4.3 eV, 3.0 eV, 3.9 eV) for SSMO film. A large band gap of 4.3 eV in SSMO film is obtained, which may be beneficial to the wide band gap applications.

Materials and Methods

La0.67Sr0.33MnO3 (LSMO), La0.67Ca0.33MnO3 (LCMO), and Sm0.67Sr0.33MnO3 (SSMO) thin films with a fixed thickness of 10 nm were synthesized and denoted as LSMO, LCMO, and SSMO, respectively. Three films were deposited on a (001) SrTiO3 (STO) single crystal using a 248 nm KrF excimer pulsed laser system, at 800 °C with an oxygen pressure of 100 mtorr. After as-grown, the samples were cooled down to 400 °C and in-situ annealed in 760 Torr oxygen atmosphere for 1 hour, then again cooled to room temperature. The growth rate is 0.04 Å/sec, allowing one to control the thickness of each film at the nanometer level. The crystalline structure was identified at room temperature with a Cu Kα1 (λ = 1.5406 Å) monochromatic x-ray diffraction (XRD) system. The microstructure and interface property were examined using the Field Emission Transmission Electron Microscope (FETEM, FEI Tecnai G2 F20). The temperature (T) dependence of resistivity (ρ) was measured using a standard four-probe method in a closed-cycle refrigerator with an input current ~ 10 μA. Keithley 220 and 182 were used as the current source and voltage meter, respectively. Magnetization (M) as a function of T was obtained with SQUID-VSM instrument (Quantum Design) under the condition of field cooling to 10 K at an in-plane external magnetic field of 100 Oe. The electronic structures were extracted from the XAS and XPS spectra at room temperature. A MnO polycrystalline sample and the carbon element on the film’s surface were referred as an energy calibration standard for absorption and photoelectron spectrum, respectively. The XAS measurements at the O K-edge and Mn L-edge were performed at the beamline 20A1 of the National Synchrotron Radiation Research Center in Hsinchu, Taiwan. The energy resolution was about 0.08 eV at O K-edges and 0.11 eV at Mn L2,3-edge, recorded in total electron yield. The slit widths were 20 μm by 20 μm. The XPS measurements were carried out using a spectrometer equipped with a Al Kα source (hv = 1486.6 eV) and a standard energy analyzer.