Interfacial-hybridization-modified Ir Ferromagnetism and Electronic Structure in LaMnO$_3$/SrIrO$_3$ Superlattices

Artificially fabricated 3$d$/5$d$ superlattices (SLs) involve both strong electron correlation and spin-orbit coupling in one material by means of interfacial 3$d$-5$d$ coupling, whose mechanism remains mostly unexplored. In this work we investigated the mechanism of interfacial coupling in LaMnO$_3$/SrIrO$_3$ SLs by several spectroscopic approaches. Hard x-ray absorption, magnetic circular dichroism and photoemission spectra evidence the systematic change of the Ir ferromagnetism and the electronic structure with the change of the SL repetition period. First-principles calculations further reveal the mechanism of the SL-period dependence of the interfacial electronic structure and the local properties of the Ir moments, confirming that the formation of Ir-Mn molecular orbital is responsible for the interfacial coupling effects. The SL-period dependence of the ratio between spin and orbital components of the Ir magnetic moments can be attributed to the realignment of electron spin during the formation of the interfacial molecular orbital. Our results clarify the nature of interfacial coupling in this prototypical 3$d$/5$d$ SL system and the conclusion will shed light on the study of other strongly correlated and spin-orbit coupled oxide hetero-interfaces.

to sustain magnetic orderings due to the large spatial extension of the 5d electrons [15].
While their well-investigated 3d TMO counterparts usually possess weak SOC, even though the strength of electron correlation is always sufficient to support magnetism.
Interfaces between dissimilar materials can provide an intriguing playground for manipulation of various physical properties [16][17][18][19][20]. Great improvement of thin-film fabrication techniques enables accurately controlled design of epitaxial TMO heterostructures and SLs with atomically abrupt interfaces. It appears to be a natural strategy that artificial 3d/5d TMO heterostructures or SLs are promising candidates to involve both significant electron correlation and SOC simultaneously. Pioneering research about 3d/5d SLs was triggered by investigation on SrIrO 3 /SrTiO 3 (SIO/STO) perovskite SLs [21], as a comparison with the Ruddlesden-Popper series iridates Sr n+1 Ir n O 3n+1 . Meanwhile, strong interfacial 3d-5d coupling was reported in La 1−x Sr x MnO 3 /SrIrO 3 (0<x <1, LSMO/SIO) SLs [22][23][24][25][26]. Emergent Ir ferromagnetic (FM) moments can be induced by the interfacial coupling with Mn FM moments, and in turn the magnetic properties of Mn layers can be significantly modified as well. Perpendicular magnetic anisotropy and concomitant anomalous Hall effect were observed in x = 1 SLs [22], and modulation of magnetic anisotropy in LSMO layers was also studied [23,24]. Recent reports claim that interfacial hybridization between Ir and Mn is responsible for the charge transfer in x = 1 SLs [27] and spectroscopic properties of x = 0.33 SLs [25].
Perovskite SLs are ideal systems for investigation of interfacial coupling mechanisms thanks to their high interface quality. The modification of electronic structure at the interfaces can lead to consequent change of magnetic properties. Research on the SL-period dependent evolution of the interfacial electronic structure will be informative for understanding the role of interfacial 3d-5d coupling to affect the Ir magnetism, which has not been systematically investigated so far. For this purpose, we fabricated LaMnO 3 /SrIrO 3 (LMO/SIO) SLs with different repetition periods. X-ray magnetic circular dichroism (XMCD) were employed to study how the SL period and the interfacial coupling can affect the properties of FM Ir moments. X-ray absorption spectra (XAS) and hard x-ray photoemission spectroscopy (HAXPES) were carried out to characterize the SL-period dependence of the electronic structure. A systematic SL-period dependent trend of the ratio between orbital and spin magnetic moments of Ir as well as the electronic structure were observed. First-principles calculations demonstrate a satisfactory consistency with the experimental results and reveal that the formation of the interfacial Ir-Mn molecular orbital associated with concomitant electronic-structure change is the pivotal mechanism behind this interfacial coupling. 24), (2,12) and (8,3), where a is counted in unit cells, SLaa in abbreviation) SL samples as well as LMO and SIO reference samples (24 unit cells) were fabricated by laser molecular beam epitaxy. A KrF excimer pulsed laser (λ=248 nm) with a repetition rate of 2 Hz and an energy density of ∼1.5 J/cm 2 was employed. The sample temperature and ambient oxygen pressure were controlled at 720 o C and 16 Pa, respectively.
The distance between the stoichiometric LMO or SIO targets and STO(001) single crystal substrates was set at 6 cm. The layer-by-layer growth of the SLs was guaranteed by monitoring the oscillation of the reflection high-energy electron diffraction signal (as reported in Ref. [26]). The sample structure is schematically displayed in Fig. 1(a). The crystal structure of the SLs was characterized by an x-ray diffractometer with Cu K α radiation (XRD, Rigaku RINT-2200). Basic magnetic properties of the SLs were characterized by a superconducting quantum interference device (SQUID, Quantum Design).
The Ir L edge XAS/XMCD measurements were conducted at BL39XU of SPring-8. A He-flowing cryostat was used to cool the samples to a lowest temperature of 30 K. In-plane magnetic field up to 2 T along the x-ray propagation was applied by an electromagnet. The Ir L 3,2 edge XAS/XMCD spectra were collected by standard helicity reversal technique [28] with a grazing incidence geometry (5.5 o incidence angle) and partial fluorescence yield (PFY) mode. For PFY detection of XAS/XMCD at the Ir L 3 and L 2 edges, Ir L α and L β emissions were collected and energy-analyzed respectively by a four-element silicon drift detector (Sirius 4, SGX Sensortech Inc.). The positive magnetic field direction is defined opposite to the x-ray propagation. HAXPES measurements were carried out at BL47XU of SPring-8.
The incidence angle of 7.94 keV hard x-ray was set at ∼1 o and the emitted photoelectrons were detected by a Scienta R-4000 electron energy analyzer, whose energy resolution was ∼280 meV.
First-principles calculations were carried out within the framework of density functional theory (DFT) [29,30] using the generalized gradient approximation (GGA) in the parameterization of Perdew-Burke-Ernzerhof (PBE) format exchange-correlation functional [31], as implemented in the Vienna ab initio Simulation Package (VASP) [32][33][34][35]. SOC is implemented in the projector augmented wave (PAW) method [36,37] which is based on a transformation that maps all electron wave functions to smooth pseudowave functions to describe the interaction between electrons and ions. The corresponding electronic configurations for each element are Sr: 4s4p5s; Ir: 5d6s; O: 2s2p; La: 5s5p6s5d; Mn: 3p3d4s. The cutoff energy is set to 500 eV. To account for strong correlation effects [38], we included the Hubbard correction U for Ir and Mn d states with U Ir 5d =2 eV and U Mn 3d =3 eV [27,39].
We used 4 × 4 × 4 K-points following the Monkhorst-Pack scheme in our systems. The convergence criterion for the electronic relaxation is 10 −6 eV. In this calculation, we relaxed the SL cell parameters and atomic positions with the in-plane lattice constant constrained to that of STO. The doubled unit cell has been used with the experimental in-plane lattice constant of STO, a=b=3.905× √ 2Å (see Fig. 7 in the Appendix A). Optimized SL structures were achieved when forces on all the atoms were <0.01 eV/Å. We calculated SL11, SL22, SL33 and SL44 rather than SL11, SL22 and SL88 investigated in our experiments since the supercell of SL88 is too large for DFT-based calculations.

RESULTS AND DISCUSSIONS
The XRD L scan spectra in (0 0 L) direction of the SLs are presented in Fig. 1(b).
SL satellite peaks can be clearly observed beside the (0 0 1) diffraction of the STO substrate, confirming the high quality of the SLs. In-plane magnetization-field (M -H) and magnetization-temperature (M -T ) curves shown in Fig. 1 [15].
To detect and comprehensively analyze the SL-period dependence of the Ir magnetism, states. No obvious peak shift induced by interfacial charge transfer between Ir and Mn was observed, which is consistent with previous reports [40]. Ir-Mn charge transfer will lead to deviation of the Ir valence from the nominal valence state of 4+ and add complexity to the investigation. So we chose pure LMO to construct the SLs since it is reported that with the increase of the Sr%, significant charge transfer between Ir and Mn will appear in LSMO/SIO SL system [22,27]. The XAS peak intensity slightly changes in different SLs, which suggests possible modification of the spin-orbital states of Ir and will be detailedly discussed in the latter parts.
The XMCD signal in Fig and -0.0024 µ B /atom, for SL11, SL22 and SL88 respectively. With the assumption of the negligible charge transfer between Ir and Mn according to the previous theoretical report [40], Ir has a nominal valence state of 4+ in our SL system, so the number of 5d holes is estimated Since Ir moments often exhibit canted AFM ordering in perovskite iridates [21,25,45], the net Ir moment size depends on both the absolute size of the local Ir moment and the canting angle between moments in different AFM sublattices. Hence the net moment size evaluated by XMCD sum-rules analysis varies in different systems ( Fig. 3(a)). While the m o /m se ratio can be compared among different systems ( Fig. 3(b)) and reflects the local The m o /m s ratio of ideal J eff =1/2 model is 2 [11], as indicated in Fig. 3 can make the originally antiparallel electron spins of Ir align parallel. This point will be discussed in detail in the following parts.
To understand this SL-period dependent behaviors of the Ir magnetic moments, the electronic structure of the SLs should be investigated in detail. Fig. 4(a) shows schematic orbital energy level of the LMO/SIO SL system [27,46]. The octahedral crystal field splits both the Ir 5d and Mn 3d levels into e g and t 2g states. According to the relaxed crystal

structures in our DFT calculations, the O-Ir-O bond is compressed while O-Mn-O bond
is elongated along the c-axis in the SLs. Therefore in SIO layer the 3z 2 − r 2 orbital lies above the x 2 − y 2 orbital and vice versa in LMO layer due to the Jahn-Teller effect. The 3z 2 − r 2 orbitals of Ir and Mn can hybridize with each other along the c-axis and form molecular orbitals [25,27]. The formation of the molecular orbitals can be visualized by the calculated charge density difference in Fig. 4(b). Electrons are spatially redistributed from Ir/Mn atoms to the interfacial region near the O atoms. The physical picture of the interfacial molecular orbital can also be evidenced by the partial density of states (PDOS), as shown in Fig. 5. With the decrease of SL period, the 3z 2 − r 2 orbital (mainly located above 1 eV) of Ir obviously shifts to higher energy above the Fermi level (E F ), while the 3z 2 − r 2 orbital of Mn exhibits a trend of PDOS redistribution from higher to lower energy.
The SL-period dependence of PDOS is consistent with the formation of molecular orbital that anti-bonding molecular orbital is mainly contributed by Ir and lies at higher energy than the original 3z 2 − r 2 orbital of Ir, while bonding molecular orbital is mainly contributed by Mn and lies at lower energy than the original 3z 2 − r 2 orbital of Mn, as schematically shown in Fig. 4(a).
Experimental results also show some clues to understand the SL-period dependent electronic structure of the SLs. Details in XAS at the Ir L edge can provide information about the unoccupied Ir 5d states. By taking a closer look at the white-line regions of the XAS ( Fig. 6(a,b)), the white-line intensity at the L 2 edge (I L 2 ) obviously increases with the decrease of the SL period, while the white-line intensity of the L 3 edge (I L 3 ) keeps nearly constant and slightly decreases for SL11. This variation of the white-line intensity induces the systematic change of the branching ratio (BR) and the expectation value of SOC operator L · S (Fig. 6(c)). Here BR = I L 3 /I L 2 = (2 + r)/(1 − r), r = L · S /n h [47] and n h = 5.
It can be observed that the BR of SL11 is 4.61, which is similar to that of previous reported (LSMO) 1 /(SIO) 1 SL (BR ≈ 4.4) [25]. The BR and L · S value systematically increase with the SL period. Since the L 2 edge corresponds to the electric dipole transition from 2p 1/2 to 5d 3/2 states while the L 3 edge corresponds to the electric dipole transition from 2p 3/2 to 5d 5/2 states, the decrease of BR indicates less occupation of 5d 3/2 states and/or more occupation of 5d 5/2 states. In perovskite iridates, the octahedral crystal field splits the 5d 5/2 (J 5/2 ) states into e g states and J eff =1/2 states, and J eff =3/2 states originate from the atomic 5d 3/2 (J 3/2 ) states [11,44]. The change of BR indicates that the occupation of J eff =3/2 states is decreased by the interfacial coupling and the occupation of e g states or J eff =1/2 increases when the interfacial coupling is present.
We conducted bulk-sensitive HAXPES measurements to further characterize the valence band structure of these SLs (Fig. 6(d)). It can be clearly observed in the inset of Fig. 6(d) that a feature at a binding energy of ∼1.2 eV is enhanced with the decrease of the SL period, which should be the interfacial-coupling-enhanced occupation of the bonding molecular orbital. Simultaneously, the intensity near the E F decreases with the decrease of the SL period, showing clear evidence that some density of states (DOS) near the E F was transferred to deeper levels due to the Ir-Mn interfacial coupling.
Enhanced occupation of the e g states rather than J eff =3/2 states induced by the interfacial coupling is more likely due to the following reasons. As displayed in the schematic in Fig. 4(a), the bonding molecular orbital appears below the E F and changes the relative occupation of different orbitals. In maganite/iridate SL systems, rather than an ideal J eff =1/2 scheme, mixed occupation of both J eff =1/2 and J eff =3/2 states can often occur [23]. When the molecular orbital is formed by interfacial coupling, some of the J eff =1/2 and J eff =3/2 electrons of Ir near the E F will be transferred into the bonding molecular orbital. In particular, the electron transfer from J eff =3/2 states to the molecular orbital will lead to the change of BR and consequently the local property change of the Ir moments. One may argue that when the SL period increases, the change of Ir 5d bandwidth induced by the dimensionality of SIO layers may also account for the electron redistribution among J eff =3/2 and J eff =1/2 states and give rise to the change of BR. However, as the SL period increases, the enhanced Ir-Ir hopping will result in simultaneous increase of the bandwidth of both J eff =3/2 and J eff =1/2 states. Since J eff =3/2 states are nearly fully occupied, the center of J eff =3/2 states lies deep below the E F . While the center of J eff =1/2 states lies close to the E F , as schematically shown in Fig. 4(a). With the center of mass of the states fixed, widening of the J eff =3/2 states should lead to more DOS above E F and less occupation of itself, which does not agree with the change of XAS white-line intensity. Consequently, the experimentally observed BR change should be mainly attributed to the electron transfer between J eff =3/2 states and the e g states, but not simply the redistribution of electrons within the t 2g (J eff =3/2 and J eff =1/2) states.
Up to now, we illustrated the effect of interfacial Ir-Mn coupling on the electronic structure in LMO/SIO SL system. Based on this, we can discuss the consequent effect on the m o /m s(e) ratio. Since J eff =3/2 states originate from the atomic 5d 3/2 (J 3/2 ) states [11], whose spin and orbital moments are antiparallel. While J eff =1/2 and e g states originate from the atomic 5d 5/2 (J 5/2 ) states [11], whose spin and orbital moments are parallel. When Ir-Mn interfacial coupling transfers some electrons from J eff =3/2 states to e g states, the SOC sign of these electrons are effectively changed. Spin and orbital components of the Ir moments become more parallel. As a result, we observed smaller m o /m s(e) ratio in SLs with shorter SL period, in which the interfacial coupling is more efficient. From another angle of view, the consistent trend of m o /m s ratio in DFT calculations further reveals that m s is more sensitive to the interfacial coupling than m o . This is due to the fact that spin is directly carried by the redistributed electrons from t 2g states to e g states. The shorter SL period, the more electrons transferred from Ir t 2g to 3z 2 − r 2 molecular orbitals, and the more Ir t 2g spins which are originally antiparallel to the total Ir spin tend to reverse its direction and align parallel to the total Ir spin due to the Hund's coupling [27]. On the other hand, the m o is relatively robust to the interfacial coupling, which drives the m o /m s ratio to be smaller when the SL period decreases. reported experimental benchmarks [27,48,49].

APPENDIX B
To get a feeling about the spatial range that the interfacial coupling can influence, we averaged the magnetic moments obtained by sum-rules analysis to each interface. As shown in Fig. 8(a), the size of Ir magnetic moment per interface increases with the SL period, which indicates that the effect of the interfacial coupling may not be restricted only in the unit cells adjacent to the interface, especially by comparing SL22 and SL88. On the other hand, the difference between SL11 and SL22/SL88 can be attributed to other factors. As depicted in Fig. 8