Titanium d xy ferromagnetism at the LaAlO 3 /SrTiO 3 interface J.-S. Lee 1,*, Y. W. Xie 2, H. K. Sato 3, C. Bell 3, Y. Hikita 3, H. Y. Hwang 2,3, C.-C. Kao 1 1 Stanford Synchrotron Radiation Lightsource, SLAC National Accelerator Laboratory, Menlo Park, California 94025, USA 2 Geballe Laboratory for Advanced Materials, Department of Applied Physics, Stanford University, Stanford, California 94305, USA 3 Stanford Institute for Materials and Energy Sciences, SLAC National Accelerator Laboratory, Menlo Park, California 94025, USA * email: jslee@slac.stanford.edu (J.L) NATURE MATERIALS www.nature.com/naturematerials 1
2 S1. Probing for magnetic impurities: As discussed in the manuscript, we explored the possible presence of magnetic impurities in the samples. In general, ferromagnetic 3d-transition metals (i.e., Cr, Mn, Fe, Co, and Ni) are the most likely candidates. For this purpose, we employed XAS measurements (an element specific technique). The scanned energy range is from 560 ev to 890 ev; the presence of magnetic impurities in a sample can be detected via characteristic 2p 3d dipole transitions. The sample shown here is LAO(3.3 UC)/STO, also used in the XMCD measurement in the manuscript. Figure S1 shows the XAS spectra at Cr, Mn, Fe, Co, and Ni L 2,3 -edges. Note that the signals (upper panel in the figure) via total electron yield (TEY) were normalized by I 0 signals (lower panel in the figure). No spectral features are observed at each L 2,3 edges, except in the case of Ni where it shows a dip near the energy positions corresponding to Ni absorption edges. If such behavior is due to the sample, the spectral feature should be rather a peak; we confirmed that the dip is due to variation of I 0. These measurements exclude that the presence of 3d magnetic impurities in the probing volume of our sample within the detection limit (~10 12 atoms/cm 2 ) of our experiment. S2. Estimating the magnitude of the interfacial Ti moment: In a thin LAO/STO heterostructure, the Ti L-edge absorption signal comes from both the interfacial Ti ions and those Ti extending into the STO substrate. Here, we estimate the Ti magnetic moment assuming all the magnetic Ti ions are located in the first unit cell (~ 0.4 nm) of interfacial STO. To do this, we first consider the total electron escape profile which is described as [S1]: I x) = It exp( x/ x ). ( 0 Here x 0 is the escape depth of secondary electrons in the material. For 3d metals, x 0 is measured to be around 2.5 nm. In some previous reports [S2, S3], ~ 5 nm was used for perovskite 2 NATURE MATERIALS www.nature.com/naturematerials
SUPPLEMENTARY INFORMATION 3 transition metal oxides. We used an escape depth x 0 = 4 nm, measured in another experimental study of the same LAO/STO system [S4]. Based on this parameter, the ratio between the TEY, I(x), to the total excited electron intensity (I t ) is shown in Figure S2. If we choose 10% as a cutoff for the TEY signal, then the Ti TEY measurement would be sensitive to a ~7.8 nm interfacial region of the entire STO substrate, in addition to the 1.2 nm of the LAO that is above the STO layer (i.e., a total probing depth of ~10 nm; this can be crosschecked using the Ti L-edge spectra as a function of LAO-thickness). Thus, based on this first order estimation, the contribution to the measured Ti XMCD should be primarily coming from a 7.8 nm layer in the STO below the LAO. To estimate the magnetic moment of this 7.8 nm layer, the detailed information about the profile of the magnetic moment is needed. Using the sum rule, the magnetic moment has been estimated to be about 0.01 µ B /Ti. However, estimating the exact value of the moment entails a large error since the exact electron number of the Ti valence state is not known. Nevertheless, to give an approximate description, a profile with exponential decay is assumed; m( x) = m0 exp[ ( x 1.2) / x'); 1.2 nm x 10 nm, with x' chosen as 0.4 nm, the single unit cell thickness of STO just at the interface. Then the TEY intensity from the Ti L-edge could be estimated as: I = I t exp( 1.2 / 4) exp[ ( x 1.2) / 4 dx. 10 1 ] 1.2 The TEY intensity coming from the ferromagnetically polarized Ti could be estimated as: 10 I 2 = I t exp( 1.2 / 4) exp[ ( x 1.2) / 4]exp[ ( x 1.2) / 0.4] dx. 1.2 NATURE MATERIALS www.nature.com/naturematerials 3
Then, I 1 /I 2 = 3.5582/0.3648 = 9.7538. Therefore, the magnetic moment at the interface can be estimated to be 0.0975 µ B /Ti (9.7538 0.01 µ B /Ti). We note that this estimate will change with the value of x'. S3. Thickness dependent XAS at the Ti L 2 -edge: In the manuscript, we examined in detail the XAS features around the Ti L 3 -edge, using a 2-dimensional map (LAO-thickness vs. photon energy). Here, we show that the spectral features at the Ti L 2 -edge are similar (Fig. S3). With different thicknesses of LAO, a change in photon energy of the e g level is negligible. In contrast, the t 2g level shows an energy shift with increasing LAO thickness, for which the interfacial Ti is relatively enhanced. The t 2g level moves to a lower photon energy, by an amount estimated to be about 40 mev. From these findings, we confirm similar spectral features at both L 2,3 -edges with the same corresponding implications as discussed in the main text. S4. Polarization dependence of XAS at the Ti L 2,3 -edges: To determine the distortion of the TiO 6 octahedron in STO, we performed polarization-dependent Ti L 2,3 XAS measurements as shown in Fig. S4. The polarization directions of the linearly polarized x-rays (98% polarized) are tuned by rotating the x-ray incident angle, with 90 o and 30 o incident corresponding to complete in-plane (E ab : E//a or E//b) and out-of-plane (E c : E//c) polarized components, respectively (see the inset of the figure). Generally, any anisotropic orbital distribution can be estimated by the asymmetric x-ray linear dichroism (XLD) intensity [S5]. At both Ti L 2,3 -edges, we found strong polarization effects, showing features corresponding to both t 2g and e g orbitals, which implies an electronic (orbital) anisotropy in the TiO 6 octahedron. 4 NATURE MATERIALS www.nature.com/naturematerials
SUPPLEMENTARY INFORMATION The overall shape of these polarization-dependent XAS spectra is similar with Salluzzo et al. s previous XLD measurements [S6]. However, the dichroic spectrum of Fig S4 is not completely identical to the previous work: here the integrated XLD is slightly positive, because at each peak the positive part of the derivative-like shape is much stronger than the negative one, whereas in the reference the integral is very close to zero. A nonzero XLD integral indicates the presence of Ti 3+ state, which is totally consistent with the detection of Ti local magnetic moments. Moreover, to quantify this distortion, we fit the XLD spectra, using the cluster model calculations. The estimated results are summarized in the manuscript in Fig. 3. LIST OF REFERENCES [S1] Nakajima, R. et al. Electron-yield saturation effects in L-edge x-ray magnetic circular dichroism spectra of Fe, Co, and Ni. Phys. Rev. B 59, 6421-6429 (1999). [S2] Park, J.-H. et al. Magnetic properties at surface boundary of a half-metallic ferromagnet La 0.7 Sr 0.3 MnO 3. Phys. Rev. Lett. 81, 1953-1956 (1998). [S3] Kavich, J. J. et al. Nanoscale suppression of magnetization at atomically assembled manganite interfaces: XMCD and XRMS measurements. Phys. Rev. B 76, 014410 (2007). [S4] Sing, M. et al. Profiling the interface electron gas of LaAlO 3 /SrTiO 3 heterostructures with hard x-ray photoelectron spectroscopy. Phys. Rev. Lett. 102, 176805 (2009). [S5] Huang. D. J. et al. Orbital ordering in La 0.5 Sr 1.5 MnO 4 studied by soft x-ray linear dichroism. Phys. Rev. Lett. 92, 087202 (2004). NATURE MATERIALS www.nature.com/naturematerials 5
[S6] Salluzzo, M. et al. Orbital reconstruction and the two-dimensional electron gas at the LaAlO 3 /SrTiO 3 interface. Phys. Rev. Lett. 102, 166804 (2009). FIGURE CAPTIONS Figure S1 XAS spectra of the LAO(3.3 UC)/STO heterostructure. Spectra (upper panel) are normalized by the I 0 signal (lower panel). Both I 0 and the sample signals were acquired by recording the total electron yield (TEY). Figure S2 TEY intensity profile for the LAO(3.3 UC)/STO heterostructure. Figure S3 2-dimentional map of the XAS spectra around the Ti L 2 -region (left panel). The white dashed lines indicates the energy position of the thinnest structure, LAO(1 UC)/STO. The enlarged t 2g level is shown in the right panel. The black dashed line indicates the energy position of the t 2g level. Figure S4 Polarization dependent XAS spectrum and XLD at the Ti L 2,3 -edges. The inset shows a schematic picture of the experimental configuration. 6 NATURE MATERIALS www.nature.com/naturematerials
SUPPLEMENTARY INFORMATION Intensity (arb. units) 2.5 2.0 1.5 1.0x10 5 5.0x10 4 XAS measured by TEY Cr Mn Fe Co Ni I 0 signal measured by TEY 560 600 640 680 720 760 800 840 880 Photon Energy (ev) Figure S1: Lee et al. NATURE MATERIALS www.nature.com/naturematerials 7
1.0 LAO STO substrate 0.8 Interface STO(1 UC) 0.6 Probing depth sensitivity (I/I 0 ) I: TEY, I t : total excited electron I/I t 0.4 Bulk STO 0.2 Cut off below 1/(3e)=0.1225 0.0 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 Depth (nm) Figure S2: Lee et al. 467 466 0.9902 0.7613 463.2 t 2g level (expanded) 0.6700 0.4826 Photon Energy (ev) 465 464 463 462 461 e g weak strong Interface dominance t 2g 0.5025 0.2438-0.01500 462.8 462.4 462.0 461.6 461.2 ~40 mev shift 0.2338-0.01500 5 10 15 20 25 LAO thickness (UC) 5 10 15 20 LAO thickness (UC) Figure S3: Lee et al. 8 NATURE MATERIALS www.nature.com/naturematerials
SUPPLEMENTARY INFORMATION X-ray polarization E ab Intensity (arb. units) 3 2 1 0 Ti L 3 -edge Ti L 2 -edge t 2g e g t 2g e g E ab : E // aˆ (or bˆ) : E // cˆ 452 456 460 464 468 472 E c E c XLD = Ec E ab aˆ (or bˆ) ĉ b ˆ (or aˆ) Photon Energy (ev) Figure S4: Lee et al. NATURE MATERIALS www.nature.com/naturematerials 9