# Fileset

[MBSAug 2.pdf](https://mdr.nims.go.jp/filesets/d494f3e4-6f83-4b87-8b06-26dfffd5a24e/download)

## Creator

R. Fukushima, V. N. Antonov, M. M. Otrokov, [T. T. Sasaki](https://orcid.org/0000-0002-5952-7638), R. Akiyama, K. Sumida, K. Ishihara, S. Ichinokura, K. Tanaka, Y. Takeda, D. P. Salinas, S. V. Eremeev, E. V. Chulkov, A. Ernst, T. Hirahara

## Rights

©2024 American Physical Society[In Copyright](http://rightsstatements.org/vocab/InC/1.0/)

## Other metadata

[Direct evidence of induced magnetic moment in Se and the role of misplaced Mn in <math>  <mrow>    <msub>      <mi>MnBi</mi>      <mn>2</mn>    </msub>    <msub>      <mi>Se</mi>      <mn>4</mn>    </msub>  </mrow></math>-based intrinsic magnetic topological insulator heterostructures](https://mdr.nims.go.jp/datasets/2df46f28-43ef-4b69-8424-f818cd4ada3b)

## Fulltext

Direct evidence of induced magnetic moment in Se and the role of misplaced Mn in MnBi2Se4-based1intrinsic magnetic topological insulator heterostructures2R. Fukushima,1 V. N. Antonov,2 M. M. Otrokov,3 T. T. Sasaki,4 R. Akiyama,1 K. Sumida,5, ∗ K. Ishihara,1 S. Ichinokura,1 K.3Tanaka,6 Y. Takeda,5, † D. P. Salinas,7 S. V. Eremeev,8 E. V. Chulkov,9, 10, 11, 12, 13 A. Ernst,14, 2 and T. Hirahara1, ‡41Department of Physics, Tokyo Institute of Technology, Tokyo 152-8551, Japan52Institute for Theoretical Physics, Johannes Kepler University, A-4040 Linz, Austria63Instituto de Nanociencia y Materiales de Aragón (INMA),7CSIC-Universidad de Zaragoza, Zaragoza 50009, Spain84Research Center for Magnetic and Spintronic Materials,9National Institute for Materials Science, Tsukuba 305-0047, Japan105Materials Sciences Research Center, Japan Atomic Energy Agency, Sayo, Hyogo 679-5148, Japan116UVSOR III Synchrotron, Institute for Molecular Science, Okazaki 444-8585, Japan127ALBA Synchrotron Light Source, E-08290 Cerdanyola del Valles, Spain138Institute of Strength Physics and Materials Science, Tomsk, 634055, Russia149Donostia International Physics Center (DIPC), Paseo de Manuel Lardizabal,154, 20018 San Sebastián/Donostia, Basque Country, Spain1610Departamento de Física de Materiales, Facultad de Ciencias Químicas,17UPV/EHU, Apdo. 1072, 20080 San Sebastián, Basque Country, Spain1811Centro de Física de Materiales, CFM-MPC, Centro Mixto CSIC-UPV/EHU,19Apdo.1072, 20080 San Sebastián/Donostia, Basque Country, Spain2012Tomsk State University, Tomsk, 634050, Russia2113Saint Petersburg State University, Saint Petersburg, 198504, Russia2214Max Planck Institute of Microstructure Physics, D-06120 Halle (Saale), Germany23(Dated: August 2, 2024)24Intrinsic magnetic topological insulators, in which magnetism and topology are inherently combined, areexcellent systems to realize exotic phenomena such as the quantum anomalous Hall effect. However there aremany reports that show that the experimental samples are not so ideal and the effect of the unintentional disorderin these systems needs to be considered carefully. In this study, we investigate the role of misplaced magneticatoms as well as nonmagnetic elements in the intrinsic magnetic topological insulator heterostructures basedon MnBi2Se4 and Bi2Se3. We find that Mn atoms are not only placed at the central layer of the MnBi2Se4septuple layer (SL) but also intermix with Bi (antisite Mn) as well as reside in the van der Waals (vdW) gap.Through a detailed comparison between the experimental and theoretical X-ray magnetic circular dichroism(XMCD) spectra, we find that the antisite Mn is coupled ferromagnetically whereas the vdW Mn couple anti-ferromagnetically to the Mn in the central atomic plane of the SL. Furthermore, we detect a clear XMCD signalin nonmagnetic Se, providing unambiguous evidence of its magnetic interaction with Mn.I. INTRODUCTION25The interplay of magnetism and topological properties [1]26leads to exotic quantum phenomena like the quantized anoma-27lous Hall effect (QAHE) [2, 3], topological magnetoelectric28effect [4], or the half-integer quantum Hall effect [5]. Intrin-29sic magnetic topological insulators (TIs) such as MnBi2Te430(MBT) are materials which intrinsically possess both mag-31netic and topologically nontrivial properties. They are exper-32imentally realized both in thin films [6, 7] or in the bulk form33[8]. Even superlattices composed of magnetic TIs and non-34magnetic TIs have been fabricated [9].35The influence of the native defects, which are misplaced36Mn atoms, on the magnetic and electronic structure of the37compounds of the MBT family is being actively studied cur-38∗ Present address: Research Institute for Synchrotron Radiation Science,Hiroshima University, 2-313 Kagamiyama, Higashi-Hiroshima 739-0046,Japan† Deceased‡ hirahara@phys.titech.ac.jprently. Both macroscopic [10–12] and local [13] measure-39ments reveal that in MBT and MnSb2Te4 the Mn atoms in40the central layer of the septuple layer (SL) and those in the41Bi/Sb layers couple antiferromagnetically (AFM). Accord-42ing to the recent density functional theory (DFT) calcula-43tions [14], this magnetic structure may be responsible for44the unexpected reduction of the gap in the surface Dirac45cone (DC) of MBT, observed by angle-resolved photoe-46mission spectroscopy (ARPES) [8, 15–25]. The gap size47fluctuations across the surface have been visualized using48scanning tunneling spectroscopy (STS) for the surfaces of49the MnBi2−xSbxTe4 bulk single crystals [26] as well as the50molecular-beam epitaxy grown MBT [27, 28] and MnSb2Te451[29, 30] films. As far as the cousin compound MnBi2Se452(MBS) and heterostructures on its basis are concerned [6, 31,5332], the coupling of the Mn antisites to the main Mn sites54has not been studied yet. Especially, an element-specific as55well as a site-specific study that can directly correlate the lo-56cal magnetic properties of atoms residing at different places of57the sample in real space is not throughly conducted. Besides,58for intrinsic magnetic TI family, so far there has been no evi-59dence of induced magnetic moment in nonmagnetic elements,60mailto:hirahara@phys.titech.ac.jp2although many X-ray magnetic circular dichroism (XMCD)61measurements have been performed [8, 32–39]. This is in62contrast to the case of magnetic impurity-doped TIs [40–42].63Therefore in the present work, we study MBS and char-64acterize the Mn distribution of MBS / Bi2Se3 (BS, nonmag-65netic) and MBS / n quintuple layer (QL) BS / MBSBS het-66erostructures with scanning transmission electron microscopy67(STEM) at the atomic scale. Then, this information is cor-68related with the site-specific magnetic property of the sys-69tem, focusing on the misplaced Mn and Se atoms obtained70with element-specific XMCD measurements. We find that Mn71atoms are not only placed at the central layer of the MBS SL72but also intermix with Bi as well as reside in the van der Waals73(vdW) gap. By comparing the experimental and theoretical74XMCD spectra, it is revealed that the former two couple fer-75romagnetically (FM) whereas the vdW Mn couple AFM to76the Mn in the central atomic plane of the SL. This behavior77is different from the case of MBT and is also reproduced by78directly calculating the exchange coupling constant. Further-79more, we succeed in detecting a clear XMCD signal in one80of the nonmagnetic constituents of the heterostructures - Se,81providing unambiguous evidence of its magnetic interaction82with Mn.83II. METHODS84The heterostructure samples were prepared by molecular85beam epitaxy in ultrahigh vacuum (UHV) chambers equipped86with a reflection-high-energy electron diffraction (RHEED)87system. First, a clean Si(111)-7× 7 surface was prepared on88an n-type substrate by a cycle of resistive heat treatments. The897×7 surface was terminated with Bi which lead to the forma-90tion of the Si(111)-√3×√3 surface. Then Bi was deposited91on the√3×√3 surface at ∼200 ◦C in a Se-rich condition.92Such a procedure is reported to result in a smooth epitaxial93film formation with the stoichiometric ratio of Bi : Se = 2 : 3.94The grown Bi2Se3 films were annealed at ∼250 ◦C for 5 min-95utes. The thickness of the Bi2Se3 films in this work is ∼8 QL.96Finally, Mn was deposited on Bi2Se3 in a Se-rich condition97at ∼250 ◦C. In this process, Mn and Se intercalate into the98topmost QL of BS to form the MBSBS heterostructure. The991× 1 periodicity with the same lattice constant is maintained100during this process for the samples we have fabricated. Then101an additional (n+ 1) QL of Bi2Se3 was deposited on top of102the MBSBS, and then Mn and Se were intercalated to form103the MBS / n QL BS / MBSBS heterostructures (Fig. S1 (a)).104For the X-ray magnetic circular dichroism (XMCD) and105scanning transmission emission microscopy (STEM) mea-106surements, the fabricated samples were first characterized107with angle resolved photoemission spectroscopy (ARPES) at108room temperature. Then they were capped with 10 nm of Se109before taking them out of the UHV chamber.110For the XMCD measurements, the samples were annealed111at ∼250 ◦C to remove the capping layers prior to the measure-112ments. The X-ray absorption spectroscopy (XAS) and XMCD113measurements were performed at BL23SU of SPring-8 [43]114and at BL29 BOREAS of ALBA with the total-electron-yield115method [44].116Electron transparent specimens for STEM observations117were prepared by the standard lift-out technique using an118FEI Helios G4-UX dual-beam system. Probe abberation cor-119rected STEM, FEI TitanG2 80–200 microscope, was used.120Chemical compositions were measured by energy-dispersive121X-ray spectroscopy (EDS). EDS data was obtained for a1222.4×7.3 nm2 region with a beam size of 100×300 pm2 that123can resolve the layered structure of Mn, Bi and Te at ∼0.2 nm124spacing.125Electronic structure calculations were carried out within126the density functional theory (DFT) using the projector127augmented-wave (PAW) method [45] as implemented in the128VASP code [46, 47]. The Hamiltonian contained scalar rela-129tivistic corrections and the spin-orbit coupling was taken into130account by the second variation method [48]. The generalized131gradient approximation (GGA-PBE [49]) for the exchange-132correlation energy and the DFT-D3 van der Waals (vdW) func-133tional with Becke-Johnson damping [50] were applied. The134k-point mesh of 10×10×1 were used to sample the slab Bril-135louin zone. The Mn 3d-states were treated employing the136GGA+U approach [51] within the Dudarev scheme [52]. The137Ueff =U−J value for the Mn 3d-states was chosen to be equal138to 5.34 eV.139Exchange interactions were studied applying the magnetic140force theorem as it is implemented within the multiple scat-141tering theory [53, 54]. For that, the electronic structures of142MBT and MBS/BS were calculated using a self-consistent143Green’s function method within the density functional the-144ory [54, 55] within PBESol approximation to the exchange-145correlation functional [56]. At that, the Mn 3d-states were146treated employing the GGA+U approach [51], the U value147being equal to 3.5 eV for both MBT and MBSBS. Chemical148disorder was modeled by mixing two atomic species on the149same atomic site within the coherent potential approximation150(CPA) [57, 58].151Theoretical XAS and XMCD simulations have been per-152formed using a linear response approach as it is implemented153within an LMTO method [59].154To simulate the XAS and XMCD spectra as well as to cal-155culate the exchange coupling parameters the position of the156Mn atom in the vdW gap has been determined by means of157the total-energy calculations done using VASP. We have found158that the Mn atom prefers the tetrahedral vdW site (Fig. 1(f)),159being located practically within the vdW Se layer of the SL.160III. RESULTS AND DISCUSSIONS161First we discuss the atomic structure of the samples we have162fabricated. Figure 1(a) shows the STEM image of the het-163erostructure with n = 1 and this clearly indicates that the de-164signed structure is formed in this region. However in Fig. 1(b),165which is the STEM image of the n = 3 designed sample, one166can find structures of n = 1, 3, and 4, showing that these sam-167ples can be inhomogeneous with regions of different n coex-168isting. Furthermore, areas where the MBS layers are absent as169well as regions with three SLs were also observed. A variety1703Si substrateSe cap7 (SL)75 (QL)5555555(a)Si substrateSe cap77775555 nm(b)5(c)(d)7 75 5 5 5 50 1 2 3 4 5 6 7Position (nm)MnSeBiIntensity (arb. units)vdWantisitec-SL0 2 4 6 8Energy (keV)Intensity (arb. units)(e) Se LBi MMn KvdW 5-5 αvdW 5-5 βvdW 5-7 γ5555(f)BiSeMn c-SLantisiteantisitevdWSe with mag. momentα β γ×6FIG. 1. (a, b) STEM image of the MBS / n QL BS / MBSBS sample for n = 1 (a) and 3 (b), respectively. (c) Close-up image of the n = 4region seen on the right of panel (b). (d) EDS mapping of (c), showing the chemical composition of the heterostructure. Mn can not only befound in the central layer of the SL (c-SL), but is intermixed with Bi (antisite Mn) as well as reside inside the vdW gap. The Mn spectrumhas been enhanced by a factor of six. (e) Energy-dispersive X-ray spectra at the vdW gap of different positions in the sample indicated in (c).Whereas the Mn peak is absent at the vdW gap between 2 QLs, a clear Mn peak can be detected at the SL-QL vdW gap. (f) Schematic drawingof the Mn distribution inside the MBSBS heterostructure. The arrows show the mutual alignments of the local magnetic moments at the threedifferent Mn sites, as deduced from the comparison of measured and calculated XAS and XMCD spectra. The Se layers with finite magneticmoment are also indicated.of different structures that is observed is shown in Figs. S1171(b)-(e) of the Supplementary Material [60]. Thus our STEM172measurements suggest that although the heterostructure sam-173ples are mostly the same as our original design of Fig. S1 (a),174other structures can coexist and one needs to take this into ac-175count when performing macroscopic measurements. We also176found that variation in n was larger for samples designed for177larger n. This fact is particularly important to discuss the band178structure of these samples as is scrutinized in Figs. S2 and S3.179Next, we concentrate on the actual atomic composition of180the heterostructures. Figure 1(c) shows the high-resolution181high-angle annular dark field STEM (HAADF-STEM) im-182age taken from the [110] direction of the n = 4 region. Fig-183ure 1(d) shows the results of Energy-dispersive X-ray spec-184troscopy (EDS) measurements. To emphasize the distribution185of Mn, the curve for Mn has been multiplied by a factor of six.186As anticipated from the original design, Mn mainly lies at the187center of the SL (c-SL). The position of Bi and Se seems to188be also the same as the designed structure. However, a de-189tailed inspection shows that the width of the Mn peak shown190in Fig. 1(d) is broad and not only limited to the center of the191SL but extends into the adjacent layers. Particularly, it seems192that Mn can intermix in the Bi layers which we will call “an-193tisite Mn”. Furthermore, the Mn signal still seems to be larger194than the background intensity even further away from the cen-195ter of the SL, extending to the vdW gap between adjacent Se196atoms. To verify this characteristic more vividly, Fig. 1(e)197shows the EDS spectra at the vdW gap at three different po-198sitions of Fig. 1(c). While peaks that correspond to Se L and199Bi M transitions can be identified in all the spectra shown,200the Mn K peak is only detected at the SL-QL vdW gap. This201clearly shows that Mn atoms can reside even in the vdW gap202of the heterostructures of MBS and BS. Although magnetic203atoms have been known to reside in the vdW gaps for doped204samples [64, 65], to the best of our knowledge, this is the first205experimental observation in the intrinsic magnetic TIs. Fig-206ure 1(f) summarizes the present findings. Ideally, Mn should207only reside at the center of the MBS SL, but experimentally it208can intermix with Bi as well as reside in the vdW gap.209As discussed above, it is known that the misplaced Mn210atoms alter the magnetic property of the intrinsic magnetic211TI. Therefore, to clarify the magnetic interaction between212the different Mn sites, we performed XMCD measurements.213Figure 2(a) shows the X-ray absorption spectroscopy (XAS)214spectra taken at 6 K with a magnetic field of 5 T applied per-215pendicular to the sample at the Mn L edge for the MBS / 7 QL216BS / MBSBS sample. µ+ and µ− correspond to the spec-217trum obtained with left and right-handed circularly polarized218photons, respectively. The corresponding XMCD spectrum is219also shown and a clear signal is detected both at the L3 and L2220edges. The XMCD intensity has been deduced by normaliz-221ing µ+−µ− with the magnitude of the peak intensity at the L3222edge [the difference of the values of the averaged XAS spec-223trum at 635 eV (background), and at 640 eV (peak position)].224We now compare the averaged XAS and XMCD data with225theory to verify the magnetism of Mn at different sites. As226shown in Fig. S4, the shape of the XMCD spectra did not227change significantly for different heterostructure samples as228well as for different measurement conditions. This is proba-229bly because the spot size in the XMCD measurements is ∼200230µm and regions with different n coexist in all the samples as231well as the fact that the concentration of the misplaced Mn232is nearly the same since the same sample fabrication proce-233dure is employed. Thus we performed the calculation of the234XAS and XMCD spectra for a single septuple layer MBS and235compared to the experimental data. Figures 2(b)-(d) show the236XAS and XMCD spectra for the Mn in the central atomic237plane of the SL (b), at the Bi site (c), and in the vdW gap2384630 640 650 660Intensity (arb. units)MBS / 7QL BS / MBSBS (a) µ+ XASµ- XASXMCD Photon enegy (eV)L3L2T = 6 K, µ0H = 5 T Mn0-0.1-0.2630 640 650 660Photon enegy (eV)Intensity (arb. units)0(b) c-SL Mn630 640 650 660Photon enegy (eV)0(c) antisite Mn 630 640 650 660Photon enegy (eV)0(d) vdW Mn 630 640 650 660Photon enegy (eV)expcal630 640 650 660Photon enegy (eV)expcal0(e) (f) XASXMCD XASXMCD XASXMCD XASXMCD FIG. 2. (a) X-ray absorption spectra (XAS) measured at 6 K for a circularly polarized incident light when a magnetic field of 5 T was appliedalong the sample surface-normal direction for the MBS / 7 QL BS/ MBSBS heterostructure at the Mn L edge. µ+ and µ− correspond to thespectrum obtained with left and right-handed circularly polarized photons, respectively. The corresponding XMCD spectra is also shown.(b-d) Calculated XAS and XMCD spectra for the Mn at the central layer in the SL (b), the Mn intermixed with Bi (antisite Mn) (c), and the Mnin the vdW gap (d), respectively. (e) Comparison of the experimental and calculated XAS specta. The calculated spectrum is the convolutionof the spectra shown in (b), (c), (d) with a ratio of 7 : 2 : 1. (f) Comparison of the experimental and calculated XMCD spectra. The calculatedspectrum is the convolution of the spectra shown in (b), (c), (d) with a ratio of 7 : 2 : -1.(d), respectively. The Mn valence in these sites is +2, +2,239and +3, respectively. For the antisite Mn and Mn at the vdW240gap, the Mn portion was set at 10 %. It can be easily noticed241that the experimental data in Fig. 2(a) cannot be reproduced242by considering c-SL Mn alone (Fig. 2(b)) and one needs to243consider Mn at different sites. To be more specific, the for-244mer can only show a single peak for the L2 edge, whereas in245the experiment there are clearly two peaks. Quantitatively, we246notice that the energy position of the largest XMCD signal is247not the same for different Mn sites, as indicated by the dotted248lines in Fig. 2(b)-(d).249We have tried to convolute the calculated spectra of the250three different Mn sites and reproduce the experimental XAS251and XMCD curves, as shown in Figs. 2(e) and (f). We could252not obtain a perfect match, but the overall consistency was253good when the ratio between the three Mn components of254Figs. 2(b)-(d) was 6-7 : 3-2 : 1 for the XAS spectra (Fig. S5)255[66]. The spectra shown in Fig. 2(e) is the case for a ratio256of 7 : 2 : 1, whereas it is 7 : 2 : -1 in the XMCD spectrum257in Fig. 2(f). The meaning of the plus (minus) sign is that the258magnetic coupling is FM (AFM). Comparison of the experi-259mental and convoluted theoretical XMCD spectra for various260magnetic coupling scenarios is shown in Fig. S5. The impor-261tant conclusion from this analysis is that the antisite Mn is262coupled FM to the c-SL Mn whereas the vdW Mn is AFM263coupled to the former two (Fig. 1(f)). This is in contrary to264the case of MBT, where the c-SL Mn and antisite Mn were265shown to couple AFM and can diminish the DC gap in the266band dispersion [14].267To verify if this conclusion can be reproduced by a different268approach, we have calculated the Heisenberg exchange cou-269pling constants directly using the magnetic force theorem for270MBT and MBS/BS, as shown in Fig. 3. The exchange inter-271actions of the c-SL Mn with antisite Mn atoms in MBT and272MBSBS, as well as with the vdW Mn atoms in MBSBS, are273shown. Note that the patterns of the J0 j(r0 j) dependence for274the two systems are different because there are no Mn atoms275in the vdW gap in the MBT case. It can be seen from the276figure that in both systems the J01 parameters are positive, in-277dicating FM coupling between the nearest neighbors inside278the c-SL Mn layers. However, for the interactions between279the c-SL and antisite Mn atoms the opposite signs of J0 j are280revealed. While in MBT these parameters are negative (J02281and J03), indicating the AFM coupling in agreement with ex-282periment [13, 67], in MBSBS they are positive (J02 and J04),283meaning the FM coupling. Moreover, for MBSBS the AFM284coupling between the c-SL Mn and that in the vdW gap is re-285vealed, as both J03 and J05 are negative. Thus, the results of286the magnetic force theorem calculations for the c-SL Mn cou-287pling to the antisite and vdW Mn are in agreement with the288conclusions drawn from the fitting of the experimental XMCD289curves by the calculated ones.290To elucidate why the signs of the exchange integrals for291the couplings with antisites in MBSBS are opposite to those2925FIG. 3. Calculated Heisenberg exchange coupling constants J0 j forthe Mn-Mn pair interactions as a function of the distance rMn(c-SL)-Mn(j) for MBT (blue circles) and MBSBS (red circles). Theinteractions with the atoms from the neighboring SL blocks are notshows. As indicated in the legend, the c-SL - antisite interactions aredescribed by J02 and J03 in MBT, while in MBSBS they correspondto J02 and J04. This is because in MBSBS there Mn atoms in the vdWgap, their couplings to c-SL Mn being described by J03 and J05.of MBT, we have further analyzed the electronic densities of293states. As it can be seen in Fig. 4, the hybridization of the294antisite Mn 3d states with the Bi and Se states in MBSBS is295stronger than that with Bi and Te states in MBT. The stronger296hybridization can be explained by shorter interatomic dis-297tances in MBSBS that, as a result, enhances indirect double-298exchange interaction between the local magnetic moments.299DOS(arb.units)-6 -5 -4 -3 -2 -1 0 1 2E - EF (eV)MnBi2Te4Total w/o Mnc-SL Mnantisite Mn-6 -5 -4 -3 -2 -1 0 1 2E - EF (eV)MnBi2Se4/Bi2Se3Total w/o Mnc-SL Mnantisite MnFIG. 4. Calculated density of states (DOS) of MBT (left) and MB-SBS (right). The green curves show the sums of the projected DOSsof all Bi and Se atoms, while the red and blue ones – the projectedDOSs of c-SL Mn and antisite Mn, respectively. The calculations aremade taking spin-orbit coupling into account.Now we will try to unveil the role of the nonmagnetic ele-300ments in these systems. Since measuring small XMCD signals301for nonmagnetic elements is known to be extremely difficult302[37, 68, 69], we have performed careful XAS/XMCD mea-303surements at 6 K with a magnetic field of 10 T applied perpen-304dicular to the sample at the Se L edge for the MBS / 7 QL BS /305MBSBS sample, as shown in Fig. 5(a). The XMCD intensity306has been deduced by normalizing µ+ − µ− with the magni-307tude of the peak intensity at the L3 edge [the difference of the308value of the averaged XAS spectrum at 1431 eV (background)309and 1447 eV (peak)]. One can notice that finite XMCD sig-310nals arise at the L3 and L2 absorption edges. It seems that the311peak structure is complex and the signal for both edges con-312tain a pair of positive and negative peaks. To show that this313signal is not an artifact, we show the Se XMCD spectra for314different samples as well as the spectrum measured at differ-315ent conditions in a separate facility in Fig. 5(b). Qualitatively,316we can say that the prominent pair peak structure for differ-317ent samples are the same and are sure that this signal is a real318signal of Se magnetization. Compared to the results of simi-319lar measurements performed at the same beamline in SPring-3208 (Ref. [68]) or ALBA (Ref. [69]) that report the absence of321XMCD signals at the Se L edge, we are sure that the signals322observed are finite and not an artifact. We emphasize that this323kind of XMCD signal in nonmagnetic elements has been re-324ported for Heusler alloys [70] or magnetic impurity-doped TIs325[40–42]. However, this is the first example of a clear detection326of the magnetic moment of a nonmagnetic element in intrinsic327magnetic TIs, directly showing the magnetic interaction with328the Mn layer. We also note that a very unclear XMCD signal329was detected at the Bi N edge (4d → 6p) as shown in Fig. S6330and the reason for this maybe that the peak signal of the XAS331spectra itself is quite weak in Bi.332By carefully comparing the XAS and XMCD spectra, one333notices that the peak at lower photon energy in the XMCD334spectrum appears prior to the main absorption for both the335L3 and L2 edges, as colored in green. These pre-edge peaks336are opposite in sign with the main peaks colored in red and337blue and moreover, their intensity is nearly the same order338of magnitude as the main peaks. In addition, the sign of the339XMCD signal is the opposite between the Mn and Se for the340main peaks, thus suggesting that the Mn and Se are AFM cou-341pled, consistent with what has been theoretically predicted in342Ref. [6].343Figure 6(a) compares the experimental XAS data subtracted344by the background with the calculation for the Se L edge [70].345The Se in this case corresponds to the atoms in the layer ad-346jacent to central plane of the SL (i.e. Mn) and not those com-347posing the vdW gap, as shown in Fig. 1(f). Although the cal-348culation shows fine features not observed in the experiment,349we can say that the two are consistent concerning the largest350peaks. Figure 6(b) shows the comparison between the exper-351imental and calculated XMCD signals and again the consis-352tency between the two is good, further reinforcing the fact353that the experimentally observed signal at the Se L edge is not354an artifact. However, the pre-edge signal is somewhat weaker355for the calculated spectrum.356Since the main peak of the L edge should correspond to the3572p → 4d transition, it is possible that the pre-edge peak which358is at ∼ 2 eV smaller photon energy with opposite sign, in-359clude the contribution from the 2p → 4s transition, since the360change of the angular momentum in the transition is the op-36161420 1460 1500Intensity (arb. units)MBS / 7QL BS / MBSBS (a) µ+ XASµ- XASXMCD = µ+- µ-  Photon enegy (eV)L3L2T = 6 Kµ0H = 10 T Se0-10-310-3(b)   n = 2n = 7n = 15n = 1MBSBSL3L2T = 6 K, µ0H = 10 T T = 4 K, µ0H = 6 T Photon enegy (eV)1420 1460 1500 XMCD Intensity (arb. units)0-10-310-30-10-310-30-10-310-30-10-310-30-10-310-3FIG. 5. (a) X-ray absorption spectra (XAS) measured at 6 K for a circularly polarized incident light when a magnetic field of 10 T was appliedalong the sample surface-normal direction for the MBS / 7 QL BS/ MBSBS heterostructure at the Se L edge. The corresponding XMCDspectra is also shown, indicating the clear detection of the Se magnetization. The red and blue main peaks likely correspond to the 2p → 4dtransition and the green pre-edge peaks correspond to the 2p → 4s transition. (b) Comparison of the XMCD spectra between the MBS / n QLBS / MBSBS heterostructures for n = 1,2,7,15 and that of MBSBS.Intensity (arb. units)1420 1460 1500Photon enegy (eV)exp-BGcal L3XAScal L21420 1460 1500Photon enegy (eV)expcalXMCD (a) (b) Intensity (arb. units)XAS Intensity (arb. units)(c) (d) XMCD Intensity (arb. units)0 20 40Relative photon enegy (eV)cal L3s DOS×10d DOScal L3s spin pol.×4d spin pol.000DOSDOS difference0 20 40Relative photon enegy (eV)L3L2FIG. 6. (a,b) Comparison of the experimental XAS (a) and XMCD(b) spectra with the calculation at the Se L edge. (c) Comparison ofthe calculated XAS spectrum of the Se L3 edge with the partial DOSof Se 4s and 4d orbitals. (d) Comparison of the calculated XMCDspectrum of the Se L3 edge with the spin polarization (difference ofthe partial DOS of the majority and minority state) of Se 4s and 4dorbitals.posite between 2p → 4d (+1) and 2p → 4s (−1). To test the362above hypothesis, we first compare the calculated XAS spec-363tra with the calculated spin-integrated unoccupied partial den-364sity of state (PDOS) of the Se orbitals in the single MBS SL.365Figure 6(c) shows the comparison between the XAS spectrum366and the PDOS of Se 4s and 4d for the L3 edge. The energy367position has been shifted so that the spectral features of the 4d368DOS and XAS will coincide with each other. One can notice369that the DOS of the s orbitals is much smaller than the d or-370bitals, and furthermore, the peak position of the s orbital is at371a smaller energy than the peaks for the d orbital. Figure 6(d)372compares the calculated XMCD spectra with the DOS differ-373ence between the majority and minority spin for the s and d374orbitals. The DOS difference, which corresponds to the spin375polarization of Se, is very small for both the s and d orbitals376but finite values can be found at energies that are close to the377peak positions in the XMCD spectrum. We can definitely say378that the main peak originates from the spin polarization of the3794d states. For the pre-edge peak, although it is difficult to say380that the spin-polarization of the s orbitals is giving the main381contribution, its contribution should be larger than the main382peak considering the larger DOS shown in Fig. 6(c). Thus we383conclude that the main feature of the XMCD spectra of Se384can be basically understood by the PDOS of the Se orbitals385and the slight discrepancy between the experimental and the-386oretical XMCD spectra should originate from the difference387in the actual contribution of the s orbitals.388IV. CONCLUSION389In summary, we performed STEM and XMCD measure-390ments on MBS / n QL BS / MBSBS heterostructures and391found that the Mn atoms are not placed only in the central392SL of MBS, but intermix with Bi as well as reside in the393vdW gap. By comparing the experimentally measured XMCD394spectra with theory, we find that the c-SL Mn and the antisite395Mn are are coupled ferromagnetically, whereas the vdW Mn396are most likely coupled antiferromagnetically with the former397two, which is different from the case of MBT. We also found398clear evidence of the magnetic interaction of the Mn and Se399from the detection of XMCD signal at the Se L edge. These400results suggest the importance of identifying the magnetism of401each elements at different environments in the intrinsic mag-402netic TIs.4037ACKNOWLEDGEMENT404We thank P. Gargiani and M. Valvidares for their assistance405in the XMCD experiments. T.H. acknowledges the support406by Grants-In-Aid from Japan Society for the Promotion of407Science (Nos. 18H03877 and 22H00293), the Murata Sci-408ence Foundation (No. H30-084), the Asahi Glass Foundation,409the Iketani Science and Technology Foundation (0321083-410A), and Support for Tokyo Tech Advanced Researchers.411M.M.O. acknowledges the support by MCIN/ AEI /10.13039/412501100011033/ (Grant PID2022-138210NB-I00) and "ERDF413A way of making Europe", by Ayuda CEX2023-001286-S fi-414nanciada por MICIU/AEI/10.13039/501100011033, as well415as MCIN with funding from European Union NextGener-416ationEU (PRTR-C17.I1) promoted by the Government of417Aragon. S.V.E. acknowledge support from the Government418research assignment for ISPMS SB RAS, project FWRW-4192022-0001. E.V.C. acknowledges Saint-Petersburg State Uni-420versity for a research project 95442847.421The ARPES measurements were performed under the422UVSOR proposal Nos. 21-681, 21-867, 22IMS6661,423and 22IMS6856. The XMCD measurements were per-424formed at JAEA beamline BL-23SU in SPring-8 (Proposal425Nos. 2020A3843, 2021A3843, and 2021B3843) and BL-42629 BOREAS in ALBA (Proposal No. 2023027296). The427work at SPring-8 was performed under the Shared Use Pro-428gram of JAEA Facilities (Proposal Nos. 2020A-E16, 2021A-429E19, and 2021B-E16) with the approval of Nanotechnol-430ogy Platform project supported by the Ministry of Edu-431cation, Culture, Sports, Science and Technology (Proposal432Nos. JPMXP09A20AE0016, JPMXP09A21AE0017, and JP-433MXP09A21AE0036).434[1] Y. Tokura, K. Yasuda, and A. Tsukazaki, Nat. Rev. Phys. 1, 126435(2019).436[2] F. D. M. Haldane, Phys. Rev. Lett. 61, 2015 (1988).437[3] C. Z. Chang, J. Zhang, X. Feng, J. Shen, Z. Zhang, M. Guo, K.438Li, Y. Ou, P. Wei, L. L. Wang, Z. Q. Ji, Y. Feng, S. Ji, X. Chen,439J. Jia, X. Dai, Z. Fang, S. C. Zhang, K. He, Y. Wang, L. Lu, X.440C. Ma, and Q. K. Xue, Science 340, 167 (2013).441[4] X.-L. Qi, T. L. Hughes, and S.-C. Zhang, Phys. Rev. B 78,442195424 (2008).443[5] M. Mogi, Y. Okamura, M. Kawamura, R. Yoshimi, K. Yasuda,444A. Tsukazaki, K. S. Takahashi, T. Morimoto, N. Nagaosa, M.445Kawasaki, Y. Takahashi, and Y. Tokura, Nat. Phys. 18, 390446(2022).447[6] T. Hirahara, S. V. Eremeev, T. Shirasawa, Y. Okuyama, T. Kubo,448R. Nakanishi, R. Akiyama, A. Takayama, T. Hajiri, S. Ideta, M.449Matsunami, K. Sumida, K. Miyamoto, Y. Takagi, K. Tanaka,450T. Okuda, T. Yokoyama, S. Kimura, S. Hasegawa, and E. V.451Chulkov, Nano Lett. 17, 3493 (2017).452[7] T. Hirahara, M. M. Otrokov, T. T. Sasaki, K. Sumida, Y. Tomo-453hiro, S. Kusaka, Y. Okuyama, S. Ichinokura, M. Kobayashi, Y.454Takeda, K. Amemiya, T. Shirasawa, S. Ideta, K. Miyamoto, K.455Tanaka, S. Kuroda, T. Okuda, K. Hono, S. V. Eremeev, and E.456V. Chulkov, Nat. Comm. 11, 4821 (2020).457[8] M. M. Otrokov, I. I. Klimovskikh, H. Bentmann, D. Estyunin,458A. Zeugner, Z. S. Aliev, S. Gaß, A. U. B. Wolter, A. V. Korol-459eva, A. M. Shikin, M. Blanco-Rey, M. Hoffmann, I. P. Rusinov,460A. Yu. Vyazovskaya, S. V. Eremeev, Yu. M. Koroteev, V. M.461Kuznetsov, F. Freyse, J. Sánchez-Barriga, I. R. Amiraslanov, M.462B. Babanly, N. T. Mamedov, N. A. Abdullayev, V. N. Zverev,463A. Alfonsov, V. Kataev, B. Büchner, E. F. Schwier, S. Kumar,464A. Kimura, L. Petaccia, G. D. Santo, R. C. Vidal, S. Schatz, K.465Kißner, M. Ünzelmann, C. H. Min, S. Moser, T. R. F. Peixoto,466F. Reinert, A. Ernst, P. M. Echenique, A. Isaeva, and E. V.467Chulkov, Nature 576, 416 (2019).468[9] I. I. Klimovskikh, M. M. Otrokov, D. Estyunin, S. V. Eremeev,469S. O. Filnov, A. Koroleva, E. Shevchenko, V. Voroshnin, A.470G. Rybkin, I. P. Rusinov, M. Blanco-Rey, M. Hoffmann, Z. S.471Aliev, M. B. Babanly, I. R. Amiraslanov, N. A. Abdullayev,472V. N. Zverev, A. Kimura, O. E. Tereshchenko, K. A. Kokh, L.473Petaccia, G. D. Santo, A. Ernst, P. M. Echenique, N. T. Mame-474dov, A. M. Shikin, and E. V. Chulkov, npj Quantum Mat. 5, 54475(2020).476[10] Y. Liu, L.-L. Wang, Q. Zheng, Z. Huang, X. Wang, M. Chi, Y.477Wu, B.C. Chakoumakos, M.A. McGuire, B.C. Sales, W. Wu, J.478Yan, Phys Rev X 11, 021033 (2021).479[11] Y. Lai, L. Ke, J. Yan, R. D. McDonald, and R. J. McQueeney,480Phys Rev B 103, 184429 (2021).481[12] S. X. M. Riberolles, Q. Zhang, Elijah Gordon, N. P. Butch,482Liqin Ke, J.-Q. Yan, and R. J. McQueeney, Phys. Rev. B 104,483064401 (2021).484[13] M. Sahoo, I. Onuorah, L. Folkers, E. Chulkov, M. Otrokov,485Z. Aliev, I. Amiraslanov, B. Büchner, L. Corredor, Ch.486Wang, Z. Salman, A. Isaeva, R. De Renzi, and G. Allodi,487arXiv:2402.06340.488[14] M. Garnica, M. M. Otrokov, P. C. Aguilar, I. I. Klimovskikh,489D. Estyunin, Z. S. Aliev, I. R. Amiraslanov, N. A. Abdullayev,490V. N. Zverev, M. B. Babanly, N. T. Mamedov, A. M. Shikin, A.491Arnau, A. L. V. de Parga, E. V. Chulkov, and R. Miranda, npj492Quan. Mat. 7, 7 (2022).493[15] E. D. L. Rienks, S. Wimmer, J. Sánchez-Barriga, O. Caha, P.494S. Mandal, J. Růžička, A. Ney, H. Steiner, V. V. Volobuev, H.495Groiss, M. Albu, G. Kothleitner, J. Michalička, S. A. Khan, J.496Minár, H. Ebert, G. Bauer, F. Freyse, A. Varykhalov, O. Rader,497and G. Springholz, Nature 576, 423 (2019).498[16] Y. Gong, J. Guo, J. Li, K. Zhu, M. Liao, X. Liu, Q. Zhang, L.499Gu, L. Tang, X. Feng, D. Zhang, W. Li, C. Song, L. Wang, P.500Yu, X. Chen, Y. Wang, H. Yao, W. Duan, Y. Xu, S.-C. Zhang,501X. Ma, Q.-K. Xue, and K. He, Chin. Phys. Lett. 36, 076801502(2019).503[17] J. Wu, F. Liu, M. Sasase, K. Ienaga, Y. Obata, R. Yukawa, K.504Horiba, H. Kumigashira, S. Okuma, T. Inoshita, H. Hosono,505Sci. Adv. 5, eaax9989 (2019).506[18] Y.-J. Hao, P. Liu, Y. Feng, X.-M. Ma, E. F. Schwier, M. Arita,507S. Kumar, C. Hu, R. Lu, M. Zeng, Y. Wang, Z. Hao, H.-Y. Sun,508K. Zhang, J. Mei, N. Ni, L. Wu, K. Shimada, C. Chen, Q. Liu,509and C. Liu, Phys. Rev. X 9, 041038 (2019).510[19] H. Li, S.-Y. Gao, S.-F. Duan, Y.-F. Xu, K.-J. Zhu, S.-J. Tian,511J.-C. Gao, W.-H. Fan, Z.-C. Rao, J.-R. Huang, J.-J. Li, D.-Y.512Yan, Z.-T. Liu, W.-L. Liu, Y.-B. Huang, Y.-L. Li, Y. Liu, G.-B.513Zhang, P. Zhang, T. Kondo, S. Shin, H.-C. Lei, Y.-G. Shi, W.-514T. Zhang, H.-M. Weng, T. Qian, and H. Ding, Phys. Rev. X 9,515041039 (2019).516[20] Y. J. Chen, L. X. Xu, J. H. Li, Y. W. Li, H. Y. Wang, C. F. Zhang,517H. Li, Y. Wu, A. J. Liang, C. Chen, S. W. Jung, C. Cacho, Y.5188H. Mao, S. Liu, M. X. Wang, Y. F. Guo, Y. Xu, Z. K. Liu, L. X.519Yang, and Y. L. Chen, Phys. Rev. X 9, 041040 (2019).520[21] B. Chen, F. Fei, D. Zhang, B. Zhang, W. Liu, S. Zhang, P.521Wang, B. Wei, Y. Zhang, Z. Zuo, J. Guo, Q. Liu, Z. Wang,522X. Wu, J. Zong, X. Xie, W. Chen, Z. Sun, S. Wang, Y. Zhang,523M. Zhang, X. Wang, F. Song, H. Zhang, D. Shen, and B. Wang,524Nat. Comm. 10, 4469 (2019).525[22] P. Swatek, Y. Wu, L.-L. Wang, K. Lee, B. Schrunk, J. Yan, and526A. Kaminski, Phys. Rev. B 101, 161109(R) (2020).527[23] S. H. Lee, Y. Zhu, Y. Wang, L. Miao, T. Pillsbury, H. Yi, S.528Kempinger, J. Hu, C. A. Heikes, P. Quarterman, W. Ratcliff,529J. A. Borchers, H. Zhang, X. Ke, D. Graf, N. Alem, C.-Z.530CHANG, N. Samarth, and Z. Mao, Phys. Rev. Research 1,531012011(R) (2019).532[24] A. M. Shikin, T. P. Makarova, A. V. Eryzhenkov, D. Y. Usa-533chov, D. A. Estyunin, D. A. Glazkova, I. I. Klimovskikh, A. G.534Rybkin, and A. V. Tarasov, Physica B 645, 414443 (2023).535[25] A. M. Shikin, D. A. Estyunin, N. L. Zaitsev, D. Glazkova, I.536I. Klimovskikh, S. O. Filnov, A. G. Rybkin, E. F. Schwier, S.537Kumar, A. Kimura, N. Mamedov, Z. Aliev, M. B. Babanly, K.538Kokh, O. E. Tereshchenko, M. M. Otrokov, E. V. Chulkov, K. A.539Zvezdin, and A. K. Zvezdin, Phys. Rev. B 104, 115168 (2021).540[26] F. Lüpke, M. Kolmer, J. Yan, H. Chang, P. Vilmercati, H.H.541Weitering, W. Ko, A.-P. Li, Communications Materials 4, 82542(2023).543[27] M. Liu, C. Lei, H. Kim, Y. Li, L. Frammolino, J. Yan, A.H.544Macdonald, C.K. Shih, Proc. Natl. Acad. Sci. U.S.A. 119,545e2207681119 (2022).546[28] Q. Li, I. Di Bernardo, J. Maniatis, D. McEwen, A. Dominguez-547Celorrio, M.T.H. Bhuiyan, M. Zhao, A. Tadich, L. Watson, B.548Lowe, and others, Advanced Materials 36, 2312004 (2024).549[29] S. Wimmer, J. Sanchez-Barriga, P. Kuppers, A. Ney, E.550Schierle, F. Freyse, O. Caha, J. Michalicka, M. Liebmann,551D. Primetzhofer, M. Hoffmann, A. Ernst, M. M. Otrokov, G.552Bihlmayer, E. Weschke, B. Lake, E. V. Chulkov, M. Morgen-553stern, G. Bauer, G. Springholz and O. Rader, Adv. Mater. 33,5542102935 (2021).555[30] P. Küppers, J. Zenner, S. Wimmer, G. Springholz, O. Rader, M.556Liebmann, M. Morgenstern, arXiv preprint arXiv:2202.11540557(2022).558[31] R.C. Walko, T. Zhu, A.J. Bishop, R.K. Kawakami, J.A. Gupta,559Physica E: Low-dimensional Systems and Nanostructures 143,560115391 (2022).561[32] T. Zhu, A. J. Bishop, T. Zhou, M. Zhu, D. J. O ’Hara, A. A.562Baker, S. Cheng, R. C. Walko, J. J. Repicky, T. Liu, J. A. Gupta,563C. M. Jozwiak, E. Rotenberg, J. Hwang, I. Zutic, and R. K.564Kawakami, Nano Lett. 21, 5083 (2021).565[33] A. M. Shikin, D. A. Estyunin, I. I. Klimovskikh, S. O. Filnov, E.566F. Schwier, S. Kumar, K. Miyamoto, T. Okuda, A. Kimura, K.567Kuroda, K. Yaji, S. Shin, Y. Takeda, Y. Saitoh, Z. S. Aliev, N. T.568Mamedov, I. R. Amiraslanov, M. B. Babanly, M. M. Otrokov,569S. V. Eremeev, and E. V. Chulkov, Sci. Rep. 10, 13226 (2020).570[34] P. Kagerer, C. I. Fornari, S. Buchberger, T. Tschirner, L. Veyrat,571M. Kamp, A. V. Tcakaev, V. Zabolotnyy, S. L. Morelhão, B.572Geldiyev, S. Müller, A. Fedorov, E. Rienks, P. Gargiani, M.573Valvidares, L. C. Folkers, A. Isaeva, B. Büchner, V. Hinkov,574R. Claessen, H. Bentmann, and F. Reinert, Phys. Rev. Res. 5,575L022019 (2023).576[35] P. Kagerer, C. I. Fornari, S. Buchberger, S. L. Morelhão, R. C.577Vidal, A. Tcakaev, V. Zabolotnyy, E. Weschke, V. Hinkov, M.578Kamp, B. Büchner, A. Isaeva, H. Bentmann, and F. Reinert, J.579Appl. Phys. 128, 135303 (2020).580[36] R. C. Vidal, A. Zeugner, J. I. Facio, R. Ray, M. H. Haghighi,581A. U. B. Wolter, L. T. C. Bohorquez, F. Caglieris, S. Moser,582T. Figgemeier, T. R. F. Peixoto, H. B. Vasili, M. Valvidares, S.583Jung, C. Cacho, A. Alfonsov, K. Mehlawat, V. Kataev, C. Hess,584M. Richter, B. Büchner, J. van den Brink, M. Ruck, F. Reinert,585H. Bentmann, and A. Isaeva, Phys. Rev. X 9, 041065 (2019).586[37] A. Tcakaev, B. Rubrecht, J. I. Facio, V. B. Zabolotnyy, L. T.587Corredor, L. C. Folkers, E. Kochetkova, T. R. F. Peixoto, P.588Kagerer, S. Heinze, H. Bentmann, R. J. Green, P. Gargiani, M.589Valvidares, E. Weschke, M. W. Haverkort, F. Reinert, J. van den590Brink, B. Büchner, A. U. B. Wolter, A. Isaeva, and V. Hinkov,591Adv. Sci. 10, 2203239 (2023).592[38] S. Wimmer, J. Sánchez-Barriga, P. Küppers, A. Ney, E.593Schierle, F. Freyse, O. Caha, J. Michalička, M. Liebmann,594D. Primetzhofer, M. Hoffman, A. Ernst, M. M. Otrokov, G.595Bihlmayer, E. Weschke, B. Lake, E. V. Chulkov, M. Morgen-596stern, G. Bauer, G. Springholz, and O. Rader, Adv. Mater. 33,597e2102935 (2021).598[39] A. Tcakaev, B. Rubrecht, J. I. Facio, V. B. Zabolotnyy, L. T.599Corredor, L. C. Folkers, E. Kochetkova, T. R. F. Peixoto, P.600Kagerer, S. Heinze, H. Bentmann, R. J. Green, P. Gargiani, M.601Valvidares, E. Weschke, M. W. Haverkort, F. Reinert, J. van den602Brink, B. Büchner, A. U. B. Wolter, A. Isaeva, and V. Hinkov,603Adv. Sci. 10, 2203239 (2023).604[40] M. Ye, T. Xu, G. Li, S. Qiao, Y. Takeda, Y. Saitoh, S.-Y. Zhu,605M. Nurmamat, K. Sumida, Y. Ishida, S. Shin, and A. Kimura,606Phys. Rev. B 99, 144413 (2019).607[41] M. Ye, W. Li, S. Zhu, Y. Takeda, Y. Saitoh, J. Wang, H. Pan, M.608Nurmamat, K. Sumida, F. Ji, Z. Liu, H. Yang, Z. Liu, D. Shen,609A. Kimura, S. Qiao, and X. Xie, Nat. Comm. 6, 8913 (2015).610[42] M. F. Islam, C. M. Canali, A. Pertsova, A. Balatsky, S. K. Ma-611hatha, C. Carbone, A. Barla, K. A. Kokh, O. E. Tereshchenko,612E. Jiménez, N. B. Brookes, P. Gargiani, M. Valvidares, S.613Schatz, T. R. F. Peixoto, H. Bentmann, F. Reinert, J. Jung, T.614Bathon, K. Fauth, M. Bode, and P. Sessi, Phys. Rev. B 97,615155429 (2018).616[43] Y. Saitoh, Y. Fukuda, Y. Takeda, H. Yamagami, S. Takahashi,617Y. Asano, T. Hara, K. Shirasawa, M. Takeuchi, T. Tanaka, and618H. Kitamura, J. Synchrotron Radiat. 19, 388 (2012).619[44] A. Barla, J. Nicolás, D. Cocco, S. M. Valvidares, J. Herrero-620Martín, P. Gargiani, J. Moldes, C. Ruget, E. Pellegrin and S.621Ferrer, J. Synchrotron Radiat. 23, 1507–1517 (2016).622[45] P. E. Blöchl, Phys. Rev. B 50, 17953 (1994).623[46] G. Kresse and J. Furthmüller, Phys. Rev. B 54, 11169 (1996).624[47] G. Kresse and D. J Joubert, Phys. Rev. B 59, 1758 (1999).625[48] D. D. Koelling and B. N. Harmon, Journal of Physics C: Solid626State Physics 10, 3107 (1977).627[49] J. P. Perdew, K. Burke and M. Ernzerhof, Phys. Rev. Lett. 77,6283865 (1996).629[50] S. Grimme, S. Ehrlich and L. Goerigk, Journal of Computa-630tional Chemistry 32, 1456 (2011).631[51] V. I. Anisimov, J. Zaanen and O. K. Andersen, Phys. Rev. B 44,632943 (1991).633[52] S. L. Dudarev, G. A. Botton, S. Y. Savrasov, C. J. Humphreys634and A. P. Sutton, Phys. Rev. B 57, 1505 (1998).635[53] A. I. Liechtenstein, M. I. Katsnelson, V. P. Antropov, V. A.636Gubanov, Jour. Mag. Mag. Mat. 67, 65 (1987).637[54] M. Hoffmann, A. Ernst, W. Hergert, V. N. Antonov, W. A.638Adeagbo, R. M. Geilhufe, H. B. Hamed, Physica Status Solidi639(b) 257, 1900671 (2020).640[55] M. Geilhufe, S. Achilles, M. A. Köbis, M. Arnold, I. Mertig, W.641Hergert, A. Ernst, Jour. Phys.: Cond. Mat. 27, 435202 (2015).642[56] J. P. Perdew, A. Ruzsinszky, G. I. Csonka, O. A. Vydrov, G. E.643Scuseria, L. A. Constantin, X. Zhou, and K. Burke, Phys. Rev.644Lett. 100, 136406 (2008).645[57] P. Soven, Phys. Rev. 156, 809 (1967).6469[58] B. L. Gyorffy, Phys. Rev. B 5, 2382 (1972).647[59] V. N. Antonov, B. N. Harmon, A. N. Yaresko, Electronic struc-648ture and magneto-optical properties of solids, Springer (2004).649[60] See the supplementary information at ... for details of the650sample atomic structure, the band dispersion measured by651ARPES in comparison to DFT calculations, and the details652of the XMCD analysis concerning the Mn L edge, and the653XMCD spectra for Bi. The Supplementary Material also con-654tains Refs. [61–63].655[61] S. Kimura, T. Ito, M. Sakai, E. Nakamura, N. Kondo, T.656Horigome, K. Hayashi, M. Hosaka, M. Katoh, T. Goto, T.657Ejima, K. Soda, Rev. Sci. Instrum. 81, 053104 (2010).658[62] Y. Sakamoto, T. Hirahara, H. Miyazaki, S. Kimura, and S.659Hasegawa, Phys. Rev. B 81, 165432 (2010).660[63] Y. Zhang, K. He, C.-Z. Chang, C.-L. Song, L.-L. Wang, X.661Chen, J.-F. Jia, Z. Fang, X. Dai, W.-Y. Shan, S.-Q. Shen, Q.662Niu, X.-L. Qi, S.-C. Zhang, X.-C. Ma, and Q.-K. Xue, Nat.663Phys. 6, 584 (2010).664[64] M. M. Yee, Z.-H. Zhu, A. Soumyanarayanan, Y. He, C.-L.665Song, E. Pomjakushina, Z. Salman, A. Kanigel, K. Segawa, Y.666Ando, and J. E. Hoffman, Phys. Rev. B 91, 161306 (R) (2015).667[65] J. Růžička, O. Caha, V. Holý, H. Steiner, V. Volobuiev, A. Ney,668G. Bauer, T. Duchoň, K. Veltruská, I. Khalakhan, V. Matolín, E.669F. Schwier, H. Iwasawa, K. Shimada, and G. Springholz, New670Jour. Phys. 17, 013028 (2015).671[66] The reason for the absence of a perfect match may be that the672XMCD calculations have been done in a single-particle pic-673ture while in reality, there should be many-body effects as well674as clustering of the Mn atoms in the experiment. Furthermore,675theory assumes broadening of the peak structure with finite life-676time, and this is not easy to predict perfectly. This is particularly677significant at ∼644 eV in Fig. 2 (e); whereas only a shoulder678structure is seen for the calculation, it shows up as a clear peak679in the experimental XAS data.680Even when a more sophisticated calculation method is em-681ployed for a pure system, the theoretical and experimental682XMCD spectra cannot be perfectly matched, eg. Phys. Rev.683Lett. 132, 176701 (2024).684[67] Y. Lai, L. Ke, J. Yan, R. D. McDonald, and R. J. McQueeney,685Phys. Rev. B 103, 184429 (2021).686[68] K. Sumida, Y. Takeda, S. Kusaka, K. Kobayashi, and T. Hira-687hara, Phys. Rev. Mat. 6, 014006 (2022).688[69] A. I. Figueroa, F. Bonell, M. G. Cuxart, M. Valvidares, P. Gar-689giani, G. van der Laan, A. Mugarza, and S. O. Valenzuela, Phys.690Rev. Lett. 125, 226801 (2020).691[70] T. Yoshikawa, V. N. Antonov, T. Kono, M. Kakoki, K. Sum-692ida, K. Miyamoto, Y. Takeda, Y. Saitoh, K. Goto, Y. Sakuraba,693K. Hono, A. Ernst, and A. Kimura, Phys. Rev. B 102, 064428694(2020).695