# Fileset

[NCommun_17(2026)10841_Fukushima.pdf](https://mdr.nims.go.jp/filesets/ffa6b25f-56de-407b-a146-1ad81c3caab1/download)

## Creator

[Yuto Fukushima](https://orcid.org/0009-0004-5820-7001), [Hiroyuki S. Suzuki](https://orcid.org/0000-0001-8468-2632), Masayuki Ochi, Takuma Nishida, Toshiya Ikenobe, [Hiraku Saito](https://orcid.org/0009-0004-6657-0863), [Yuto Kinoshita](https://orcid.org/0000-0002-0785-0011), [Kaishu Kawaguchi](https://orcid.org/0000-0002-2211-9841), [Hiroaki Tanaka](https://orcid.org/0000-0002-5178-7777), Ayumi Harasawa, Takushi Iimori, [Koichiro Yaji](https://orcid.org/0000-0002-0721-1316), [Masashi Tokunaga](https://orcid.org/0000-0002-1401-9381), [Taro Nakajima](https://orcid.org/0000-0001-6557-5508), [Kiyohisa Tanaka](https://orcid.org/0000-0003-2779-5870), [Takahiro Morimoto](https://orcid.org/0000-0003-4825-7377), [Ryo Mori](https://orcid.org/0000-0002-8684-9421), [Takeshi Kondo](https://orcid.org/0000-0002-3912-5172)

## Rights

[Creative Commons BY Attribution 4.0 International](https://creativecommons.org/licenses/by/4.0/)

## Other metadata

[Direct observation of a spin-polarized surface Dirac gap in the antiferromagnetic topological insulator NdBi](https://mdr.nims.go.jp/datasets/bb44a2d8-d20a-4c0f-9f0b-18597f7d951d)

## Fulltext

Article https://doi.org/10.1038/s41467-026-77571-4Direct observation of a spin-polarizedsurface Dirac gap in the antiferromagnetictopological insulator NdBiYuto Fukushima 1, Hiroyuki S. Suzuki 1, Masayuki Ochi2,3, Takuma Nishida1,Toshiya Ikenobe1, Hiraku Saito 1, Yuto Kinoshita 1, Kaishu Kawaguchi 1,4,Hiroaki Tanaka 1, Ayumi Harasawa1, Takushi Iimori1, Koichiro Yaji 5,6,Masashi Tokunaga 1,7, Taro Nakajima 1,8,9, Kiyohisa Tanaka 10,11,Takahiro Morimoto 12,13, Ryo Mori 1,4 & Takeshi Kondo 1,7A gapped Dirac cone in antiferromagnetic topological insulators (AFM-TIs) isessential for realizing the quantized anomalous Hall effect, yet its intrinsicmagnitude remains controversial. In the prototypical AFM-TI MnBi2Te4, how-ever, the observed Dirac gap is significantly smaller than theoretical predic-tions or nearly absent, highlighting a long-standing discrepancy. Here wedemonstrate that NdBi hosts a spin-polarized topological surface Dirac conewith a sizable gap of 17 ± 2meV at the Dirac point, in excellent agreement withdensity functional theory predictions (12 meV). This direct observation isenabled by high-resolution 7 eV laser-based spin- and angle-resolved photo-emission spectroscopy, which unambiguously disentangles the spin-polarizedsurface Dirac cone from bulk-derived states. Furthermore, through tempera-ture dependence and controlled surface contamination, we establish a directlinkage between surfacemagnetism and the Dirac gap, demonstrating that thegap opens only in the presence of surface magnetic order. These results pro-videdefinitive spectroscopic evidence for an intrinsicDirac gap inAFM-TIs andestablish NdBi as a model platform to resolve the long-standing discrepancybetween experiment and theory in AFM-TIs.Topological insulators (TIs) realize a striking quantum state in whichthe bulk remains insulating while the surface hosts gapless Dirac fer-mions protected by time-reversal symmetry. Most TIs discovered todate have been realized in spin-orbit-coupled semiconductors withweak electronic correlations. More recently, strongly correlated TIshave attracted significant attention as a fertile ground for realizingexotic topological phases, including topological superconductors,topological Kondo insulators, and topological Mott insulators1–8.Antiferromagnetic topological insulators (AFM-TIs) provide acompelling platform for exploring correlated topological phases. InReceived: 14 September 2025Accepted: 27 August 2026Check for updates1Institute for Solid State Physics, The University of Tokyo, Kashiwa, Chiba, Japan. 2Department of Physics, The University of Osaka, Toyonaka, Osaka, Japan.3Forefront Research Center, The University of Osaka, Toyonaka, Osaka, Japan. 4Maryland Quantum Materials Center, Department of Physics, University ofMaryland, CollegePark,MD,USA. 5Center for BasicResearch onMaterials, National Institute forMaterials Science (NIMS), 3-13 Sakura, Tsukuba, Ibaraki, Japan.6Unprecedented-Scale Data Analytics Center, Tohoku University, 468-1 Aoba, Aramaki-Aza, Aoba-Ku, Sendai, Japan. 7Trans-scale Quantum Science Institute,The University of Tokyo, Tokyo, Japan. 8Institute of Materials Structure Science, High Energy Accelerator Research Organization, Tsukuba, Ibaraki, Japan.9RIKENCenter for EmergentMatter Science,Wako, Japan. 10SOKENDAI (TheGraduateUniversity for AdvancedStudies), Okazaki, Japan. 11UVSORSynchrotronFacility, Institute for Molecular Science, Okazaki, Japan. 12Department of Applied Physics, The University of Tokyo, Tokyo, Japan. 13Department of Physics,Kyoto University, Kyoto, Japan. e-mail: kondo1215@issp.u-tokyo.ac.jpNature Communications |        (2026) 17:10841 11234567890():,;1234567890():,;https://doi.org/10.1038/s41467-026-77571-4http://orcid.org/0009-0004-5820-7001http://orcid.org/0009-0004-5820-7001http://orcid.org/0009-0004-5820-7001http://orcid.org/0009-0004-5820-7001http://orcid.org/0009-0004-5820-7001http://orcid.org/0000-0001-8468-2632http://orcid.org/0000-0001-8468-2632http://orcid.org/0000-0001-8468-2632http://orcid.org/0000-0001-8468-2632http://orcid.org/0000-0001-8468-2632http://orcid.org/0009-0004-6657-0863http://orcid.org/0009-0004-6657-0863http://orcid.org/0009-0004-6657-0863http://orcid.org/0009-0004-6657-0863http://orcid.org/0009-0004-6657-0863http://orcid.org/0000-0002-0785-0011http://orcid.org/0000-0002-0785-0011http://orcid.org/0000-0002-0785-0011http://orcid.org/0000-0002-0785-0011http://orcid.org/0000-0002-0785-0011http://orcid.org/0000-0002-2211-9841http://orcid.org/0000-0002-2211-9841http://orcid.org/0000-0002-2211-9841http://orcid.org/0000-0002-2211-9841http://orcid.org/0000-0002-2211-9841http://orcid.org/0000-0002-5178-7777http://orcid.org/0000-0002-5178-7777http://orcid.org/0000-0002-5178-7777http://orcid.org/0000-0002-5178-7777http://orcid.org/0000-0002-5178-7777http://orcid.org/0000-0002-0721-1316http://orcid.org/0000-0002-0721-1316http://orcid.org/0000-0002-0721-1316http://orcid.org/0000-0002-0721-1316http://orcid.org/0000-0002-0721-1316http://orcid.org/0000-0002-1401-9381http://orcid.org/0000-0002-1401-9381http://orcid.org/0000-0002-1401-9381http://orcid.org/0000-0002-1401-9381http://orcid.org/0000-0002-1401-9381http://orcid.org/0000-0001-6557-5508http://orcid.org/0000-0001-6557-5508http://orcid.org/0000-0001-6557-5508http://orcid.org/0000-0001-6557-5508http://orcid.org/0000-0001-6557-5508http://orcid.org/0000-0003-2779-5870http://orcid.org/0000-0003-2779-5870http://orcid.org/0000-0003-2779-5870http://orcid.org/0000-0003-2779-5870http://orcid.org/0000-0003-2779-5870http://orcid.org/0000-0003-4825-7377http://orcid.org/0000-0003-4825-7377http://orcid.org/0000-0003-4825-7377http://orcid.org/0000-0003-4825-7377http://orcid.org/0000-0003-4825-7377http://orcid.org/0000-0002-8684-9421http://orcid.org/0000-0002-8684-9421http://orcid.org/0000-0002-8684-9421http://orcid.org/0000-0002-8684-9421http://orcid.org/0000-0002-8684-9421http://orcid.org/0000-0002-3912-5172http://orcid.org/0000-0002-3912-5172http://orcid.org/0000-0002-3912-5172http://orcid.org/0000-0002-3912-5172http://orcid.org/0000-0002-3912-5172http://crossmark.crossref.org/dialog/?doi=10.1038/s41467-026-77571-4&domain=pdfhttp://crossmark.crossref.org/dialog/?doi=10.1038/s41467-026-77571-4&domain=pdfhttp://crossmark.crossref.org/dialog/?doi=10.1038/s41467-026-77571-4&domain=pdfhttp://crossmark.crossref.org/dialog/?doi=10.1038/s41467-026-77571-4&domain=pdfmailto:kondo1215@issp.u-tokyo.ac.jpwww.nature.com/naturecommunicationsAFM-TIs, it is predicted that long-range magnetic order opens anenergy gap at the Dirac point of the topological surface state (Diracgap), leading to the emergence of one-dimensional chiral edge statesalong the crystal hinges9. The MnBi2nTe3n+1 family, the first identifiedAFM-TI10, has exhibited a variety of exotic responses, such as thequantum anomalous Hall effect and axion electrodynamics undermagnetic fields11–17.Despite these remarkable discoveries, a central challenge persists:the magnitude of the Dirac gap observed by angle-resolved photo-emission spectroscopy (ARPES) is frequently inconsistent with theo-retical predictions, and in many cases, absent10,18–28. Although earlystudies ofMnBi2Te4 reported a sizable gap exceeding 50meV10,20, laterinvestigations clarified that it primarily originates from bulk states,while the surface Dirac cone remains nearly gapless even under mag-netic order21–25. Furthermore, atomic-scale disorder and surfacereconstruction have been indicated to strongly suppress the surfaceDirac gap, rendering it highly sensitive to sample quality26–29. As aresult, conclusive ARPES evidence of an intrinsic surface Dirac gap inAFM-TIs has remained elusive, especially in MnBi2Te4. Such a dis-crepancy limits the high-temperature applicability of these materialsand obscures the understanding of the intricate interplay betweenbulkmagnetismand topological surface state (TSS), which is central tocorrelated topological physics.Recently, the type-A antiferromagnet NdBi with the Néel tem-perature (TN) of 24 K has been proposed as a second AFM-TI, and itstopological properties have been intensively investigated30,31. NdBibelongs to a class of low-carrier-density semimetals with a bandstructure featuringmultiple Dirac cones. The valence bands aremainlyderived from Bi 6p orbitals, forming two distinct hole pockets. Incontrast, the conduction bands originate from Nd 5d states and formelectronpockets near theXpoints32. Due to strong spin-orbit coupling,the band gap narrows systematically from NdP to NdSb30,31,33–37, cul-minating in a band inversion in NdBi. This inversion underlies thetopological phase transition in these materials.The presence of an odd number of Dirac cones along the �Γ– �Mdirection—one at �Γ and two at �M30—underscores the topological non-triviality of NdBi and motivates detailed spectroscopic investigations.Beyond topology, rare-earth monopnictides have drawn considerableattention due to their strongly correlated effects in the low-carrier-density regime. The small Fermi energy (EF) in these systems weakensThomas-Fermi screening, enabling long-range Coulomb interactions.As a result, the coupling between itinerant and localized f electronsbecomes a defining factor in their physical properties38; a notableexample is CeSb, which exhibits complex magnetic transitions knownas the “devil’s staircase”39–42. The sister compound NdBi, therefore,provides an ideal platform for studying exotic AFM-TI phases drivencooperatively by magnetism and strong electron correlations in a low-carrier-density system. However, the effect of electron correlation onthe topological properties, including the TSS, remains unclear.A recent ARPES study onNdBi reported a sizable energy gapat thetopological Dirac point as large as 125 meV below the Néeltemperature30. This feature was attributed to time-reversal symmetrybreaking induced by out-of-plane ferromagnetic order in the top sur-face layer (corresponding to the antiferromagnetic phase, Domain-C).In that work30, however, a large Dirac gap (~80meV) is also reportedeven for the side surface (corresponding to the antiferromagneticphase, Domain-A), where the Dirac point is expected to remain gaplessdue to symmetry protection. This ambiguity likely arises from thecomplex ARPES signals involving both surface and bulk states,reflecting a broader unresolved issue in AFM-TIs: the persistent dis-crepancy between experiment and theory, particularly regarding thenature and origin of Dirac gaps. Spin-resolved experiments that isolatethe topological surface state from bulk states are essential to resolvethis issue and fully understand the antiferromagnetic topologicalproperties of this system.In this work, we overcome these experimental limitations byemploying high-resolution laser-based ARPES combined with spin-resolved measurements, enabling clear separation of the TSS frombulk states and its direct comparison with ab initio calculations. Thisapproach provides evidence of the intrinsic surface Dirac gap of17 ± 2meV in NdBi that originates from the topological surface state, inagreement with theoretical predictions (12meV). Our results resolvethe persistent discrepancy between experiment and theory in AFM-TIsand reveal pronounced band reconstruction in the antiferromagneticphase, underscoring the essential role of inherent magnetism in sta-bilizing gapped surface states. Althoughmuch smaller than previouslyreported30, the intrinsic Dirac gap we revealed in NdBi is still largerthan the negligible gap in MnBi2nTe3n+121–25. These results not onlyresolve a central puzzle in AFM-TI research but also establish NdBi as apromising platform for topological device engineering.ResultsWe begin by presenting, in Fig. 1, the electronic structures around �Γabove (Fig. 1b–e) and below (Fig. 1f–i, 1j–m) TN obtained by ARPES anddensity functional theory (DFT) calculations. Below TN, the magneticmoments of Nd 4f electrons are ordered ferromagnetically within the(001) plane and stacked antiferromagnetically along the [001] direc-tion [Fig. 1f, j]. Therefore, three magnetic domains are created withantiferromagnetic (AFM) moments aligned along the a, b, and c axes,for which we label AFM-A, AFM-B, and AFM-C, respectively. Thesemagnetic domains can be distinguished from the symmetry of theFermi surface near the �Γ point. To reveal the electronic structure ofindividual magnetic domains, we employed ARPES with a focused 7 eVlaser beam (~50μm in spot size), which enables not only high-resolution measurements, but also magnetic-domain-selective obser-vations by spatially scanning the small beam spot across the (001)cleavage surface.The paramagnetic (PM) phase has three sheets of Fermi surfacesaround �Γ in the bulk state (Fig. 1c): one square extending in the �Γ– �Mdirection, another smaller square, and the smallest circular one at thecenter. Our DFT calculations (Fig. 1a, e) identify that these Fermi sur-faces correspond to the heavy hole, light hole, and electron pocket,respectively. In Fig. 1d, e, ARPES intensity maps for energy dispersionand Fermi surface are compared with surface calculations, whichcontain both bulk states projected onto the crystal surface andintrinsic surface states. In ARPES spectra, bulk bands with strong kzdispersion appear as a broad continuum. This is due to kz broadeningarising from the surface sensitivity of photoemission. The data andcalculations consistently show highly dispersive Dirac-like dispersioncharacterized by sharp-peak spectra. Although this Dirac-like band hasbeen interpreted as a topological surface Dirac cone in prior studies,its nature remains ambiguous and will be revisited in a later section.In AFM-A (Fig. 1f), themagneticmoments are ordered in the [100]direction, and the Brillouin zone (BZ) is folded back along kx, resultingin the appearance of two-fold Fermi surfaces in the bulk states [Fig. 1h,i]. In addition to the band folding effect, additional surface states withsharp spectra emerge and cross the Fermi level, consistent with pre-vious reports. In the DFT calculation, two extra branches are seenaround −0.3 eV [black arrow in Fig. 1i], which, however, were notobserved by ARPES likely due to the matrix-element effect. It is worthnoting that AFM-B shares the same electronic structure as AFM-A, butrotated by 90 degrees (see Supplementary Note 2).In AFM-C [Fig. 1j], the magnetic moment is aligned perpendicularto the cleavage plane, causing theBZ to foldback along kz. Unlike AFM-A andAFM-B, AFM-C retainsC4 symmetryover the cleavage plane evenunder the magnetic order. While the Fermi surface [Fig. 1l, m] hardlychanges from the PM phase, two surface bands (SS1, SS2) emerge, asalso reported in previous studies43. Our DFT calculations reproducethese surface states, supporting that the observed domain is indeedthe AFM-C domain.Article https://doi.org/10.1038/s41467-026-77571-4Nature Communications |        (2026) 17:10841 2www.nature.com/naturecommunicationsThe basic electronic structures presented above have beenreported before30,43–46. The main results of this work are the observa-tion of the Dirac gap and the intrinsic interplay between bulk andsurface states. Beforedelving into thedetails presented in Figs. 3 and4,we highlight a key challenge in Fig. 2: the difficulty in distinguishingsurface and bulk states. Overcoming this difficulty is crucial for iden-tifying a genuine TSS and, thus, the intrinsic Dirac gap in NdBi.In ARPES research, surface states are typically assigned to statesthat show no signs of kz dispersion. We found that in NdBi, some ofthe bulk bands exhibit negligible kz dispersion, and significantly, the2D-like bulk band displays an energy dispersion resembling a Diraccone. To clarify this, in Fig. 2a, we calculated the bulk bands of thePM phase and integrated them over the entire kz. Although thiscalculation does not include surface states, the obtained resultsshow Dirac-like bands consisting of sharp peak spectra. For com-parison, we performed the surface calculations in Fig. 2b, which notonly exhibit the bulk band projected onto the surface but also sur-face states. Our results indicate that the topological surface Diraccone emerges only in a narrow energy window within the band gapsof the Dirac-like bulk bands. Since the bulk band gap is only ~13meV,standard ARPES with a typical energy resolution of ~15meV canhardly distinguish the genuine surface Dirac cone from the bulkbands. As simulated in Fig. 2f, the surface Dirac cone can be identi-fied when the resolution reaches a level of 2.5meV (as employed inthis work).The above consideration suggests that unveiling the topologicalsurface Dirac cone requires high-resolution measurements and thecapability to distinguish between surface and bulk states, advancingbeyond previous studies30,31,36. Our laser-based ARPES technique pro-vides sufficiently high-resolution (ΔE = 2.5meV) to distinguish them(Fig. 2f). A further decisive approach is to use the spin-resolved ARPES.Because theTSS is spin-polarized,whereas thebulk states are not, spin-resolved ARPES could disentangle the complex interplay betweensurface and bulk states in NdBi.In Fig. 3a, we show the high-resolution ARPES image obtained inthe PM phase. A crossing point around −0.27 eV indicates the band asthe Dirac cone. Nevertheless, the energy distribution curve (EDC) at �Γ(red curve), overlaid in the image, exhibits a shoulder-like featurerather than a clean single peak. As discussed in Fig. 2, one needs toconsider the contribution of Dirac-like bulk states to the data.To resolve this, we performed spin-ARPES and plotted in Fig. 3gthe obtained spin-resolved EDCs at positive and negative momenta +kand −k (dotted lines in Fig. 3a) equidistant from �Γ. We found that theupper part of the Dirac cone exhibits spin flipping across kx =0Å−1. Incontrast, the bottom part of the band does not exhibit clear spinflipping, as it is dominated by the bulk band of the light hole band,which contributes significantly to the spectral intensity. These obser-vations conclude that a spin-polarized Dirac cone indeed exists. Amore detailed discussion of the bulk-state contribution to the mea-sured spin polarization is provided in Supplementary Note 9.-2.0-1.5-1.0-0.500.5Momentum-0.2 0 0.2kx (1/Å)-0.4-0.3-0.2-0.10E - EF (eV)-0.2 0 0.2kx (1/Å)-0.4-0.3-0.2-0.10E - EF (eV)-0.2 0 0.2kx (1/Å)-0.2 0 0.2kx (1/Å)-0.100.1k y (1/Å)-0.100.1k y (1/Å)E - EF (eV)-0.2 0 0.2kx (1/Å)-0.4-0.3-0.2-0.10E - EF (eV)-0.2 0 0.2kx (1/Å)-0.100.1k y (1/Å)X XXΓΓ MBi 6pNd 5dabΓ XL KWHeavy HoleLight Hole Band Inversioncfgh i l mjk(001)ΓMMSS1SS2SS1SS2BiNdabchνParamagnetic Phase (PM)Antiferromagnetic Phase: Domain-A (AFM-A) Antiferromagnetic Phase: Domain-C (AFM-C)d eWFig. 1 | Electronic structure ofNdBi acrossmagnetic phases anddomains. aBulkband structure of NdBi in the paramagnetic (PM) phase, highlighting the bandinversion between Bi 6p and Nd 5d orbitals. b Crystal structure of NdBi, with thecleavage plane shown in green. c Schematic of the bulk and surface Brillouin zones(BZs). d Fermi surface map and ARPES intensity plot of the band dispersion along�Γ-- �M of NdBi measured by ARPES in the paramagnetic phase. e Fermi surface andband dispersion along �Γ-- �M calculated using density functional theory (DFT).fMagnetic structure of the AFM-A domain. g Folded BZ due to magnetic orderingalong [100]. h Fermi surface map and band dispersion along �Γ-- �M measured byARPES. i Corresponding DFT results. j–m Same as (f–i), but for the AFM-C domain.Yellow lines in the (g, k) represent the folded BZ. Two surface states (SS1 and SS2)are resolved in the AFM-C phase (l, m).Article https://doi.org/10.1038/s41467-026-77571-4Nature Communications |        (2026) 17:10841 3www.nature.com/naturecommunicationsThe spectral relationship between the surface Dirac cone and thebulk states in the AFM-C domain (Fig. 3d) is essentially the same as inthe PM phase (Fig. 3a); while the upper bands correspond to the sur-face state, the lower bands consist of both surface and bulk states thatcan be distinguished only in a narrow energy-momentum window.However, a clear difference fromthePMphase is observed in the shapeof the surface band. Intriguingly, the upper Dirac cone, which wasV-shaped in the PMphase, transformed into anM-shaped dispersion inthe AFM-C domain. As revealed later (Fig. 4a), theM-shaped dispersionis the product of hybridization between the original V-shaped upperDirac cone and a hole-like surface band, which emerges under anti-ferromagnetism. The features of these bands are further examined byspin-resolved ARPES in Fig. 3h. The upper cone clearly exhibits spinpolarization that reverses across �Γ, while the lower cone, which almostoverlapswith bulk states, is not clearly resolved. Themagnitudeof spinpolarization in the upper cone is approximately 20% and graduallydecreases with increasing kx, as reproduced by our spin-dependentDFT calculations [Fig. 3i]. This behavior can be attributed to the non-spin-polarized SS1, which hybridizes with the TSS, giving rise to theM-shaped dispersion.Most importantly, we found a two-peak structure in the EDC at �Γ(red curve overlaid in Fig. 3d), as marked by the black arrows, inaddition to the intense peak of the bulk band. These two peaksindicate the opening of a gap at the Dirac point. The curvature plotin Fig. 3e offers a clearer view of the gap opening in the AFM-C.Compared with the PM phase (Fig. 3b), the lower Dirac conebecomes better discernible (Fig. 3e), as the bulk band shifts down-ward in the AFM-C domain. In these experiments, we used an ARPESanalyzer with a deflector, which allows a precise determination ofthe �Γ point. In addition, the low-energy photons of the 7 eV laser weuse enable us to minimize the angle integration for each EDCdetection, thus achieving measurements with high momentumresolution. Under this carefully controlled experimental setup, weconsistently obtained the same gap size (17 ± 2meV) from severaldifferent samples.Thegapsize (17 ± 2meV) identifiedhere for the surfaceDirac coneat �Γ is an order of magnitude smaller than the previously reportedvalue (~125meV)30,31. This discrepancy likely stems frommisidentifyingthe Dirac-like bulk bands shown in Fig. 2 as surface states. This issuearises not only from limited energy resolution but also from matrixelement effects, which are sensitive to the photon energy used inARPES. To examine this, wemeasured the photon-energy dependenceof the band-dispersion signals in NdBi at a synchrotron facility spe-cialized in low-photon-energy experiments (Fig. 4). At a low photonenergy (hν) of 8 eV (Fig. 4c), close to hν = 7 eV in our laser-ARPES, thesurface states (TSS, SS1, and SS2) are clearly observed with highintensity. In contrast, at a higher hν of 10 eV (Fig. 4d), these surfacestates almost entirely vanish, and the spectral weight is dominated bybulk bands (Fig. 4a, b). This photon-energy dependence is a con-sequence of matrix-element effects, by which lower photon energiesmore effectively excite the TSS in NdBi. This implies that the large gap(~125meV) reported in previous studies at hν = 60 eV30,31 likely corre-sponds to a bulk band gap rather than the topological surfaceDirac gap.Notably, similar matrix-element effects have been reported inMnBi2Te4, which is well established as an AFM-TI21,22,47. The TSS in thatsystem is also clearly resolved only at low photon energies, whereas athigher photon energies (e.g., above ~20 eV), the ARPES spectra aredominated by bulk bands. This can lead to an apparent large gap ori-ginating frombulk states beingmisinterpreted as the topologicalDiracgap, even though the intrinsic TSS gap is much smaller. The closeanalogy betweenMnBi2Te4 andNdBi underscores that low-hνARPES is-0.02 0 0.02kx (1/Å)-0.02 0 0.02kx (1/Å)-0.30-0.28-0.26-0.24-0.22E - EF (eV)-0.2 0 0.2kx (1/Å)-0.4-0.3-0.2-0.10E - EF (eV)-0.2 0 0.2kx (1/Å)-0.4-0.3-0.2-0.10E - EF (eV)-0.02 0 0.02kx (1/Å)-0.02 0 0.02kx (1/Å)-0.30-0.28-0.26-0.24-0.22E - EF (eV)ecaMΓMb d fΔE = 15 meV ΔE = 2.5 meVThe bulk calculationThe surface calculationMΓM ΔE = 15 meV ΔE = 2.5 meVTSSNd 5dBi 6pBulkBulkBulkBulkPMPMFig. 2 | Energy resolution effect on the bulk band and the surface Dirac conein NdBi. a, bDFT-calculated band dispersion of paramagnetic phase along the �Γ-- �Mdirection, showing the spectral weight of bulk (a) and surface (b) states. c, dSimulated ARPES images showing the impact of energy resolution on the visibilityof the bulk band gap marked by a dotted rectangle. The calculations in (a, b) areconvoluted with energy resolutions of 15meV in (c, d), and 2.5meV in (e, f),respectively. The topological surface state (TSS) is only clearly resolved in thesurface calculation when using high energy resolution (2.5meV), as shown in (f).Article https://doi.org/10.1038/s41467-026-77571-4Nature Communications |        (2026) 17:10841 4www.nature.com/naturecommunicationsvery effective for directly accessing the intrinsic topological surfacestates in AFM-TIs.Magnetic transitions in antiferromagnets are typically viewed assimple on-off switches between PM and ordered phases. The richnessof their magnetic behavior can also be unveiled by the intermediate-temperature evolution. In particular, tracing how the M-shaped sur-face band (Fig. 3d) gradually develops as the interaction between theoriginal TSS (the surface Dirac cone) and the emerging surface statebelow TN offers new insight into the temperature-driven dynamicinterplay between topology and magnetism. Our high-resolutionARPES measurements uniquely revealed this previously unexploredtemperature evolution.Figure 5a exhibits the temperature evolution of bands in a widerenergy window for Fig. 3d. Two surface states (SS1 and SS2), which areabsent in the PM phase, abruptly emerge just below TN. As the tem-peraturedecreases, both the surface states shift significantly upward inenergy, while the energy position of the Dirac cone remains unchan-ged. In particular, SS1 with a hole-like dispersion intersects the upperDirac cone, and the band hybridization occurs at these intersections,leading to the formation of the M-shaped dispersion. The magnitudeof the hybridization gap (ΔHyb) increases as the temperature decreases.The gap opens isotropically, as shown in Fig. 5d, which presents themomentum-resolved gap distribution in the two-dimensional (kx, ky)plane at 21 K and 13 K. The complete temperature-dependent datasetfrom 25K to 13 K in 1 K steps is provided as Supplementary Movie 1.The band shift with temperature and the associated evolution ofthe hybridization gap are shown in Fig. 5b, which plots the band dis-persions determined around the intersection between SS1 and thesurfaceDirac cone. To establish a direct link between the hybridizationgap and the magnitude of the magnetic order, we additionally per-formed neutron scattering measurements. In particular, we measuredthe temperature dependence of the integrated intensity of the 110magnetic Bragg reflection I110, whose square rootffiffiffiffiffiffiffiI110pis proportionalto themagneticmoment ofNd (Fig. 5c). As demonstrated in Fig. 5d, wefind excellent agreement between the temperature evolutions offfiffiffiffiffiffiffiI110p(reflecting the magnetic order parameter) and the experimentallyobserved hybridization gap ΔHyb. This provides compelling evidencethat the gap opening originates from band hybridization induced bythe exchange interaction associated with the magnetic order. Neutronscattering is a bulk-sensitive probe. Therefore, the observation of ahybridization gap between the surfacebands indicates that the surfacemagnetismevolves in tandemwith the bulkmagnetic order, at least on-0.05 0 0.05kx (1/Å)-0.3-0.2E - EF (eV)-0.05 0 0.05kx (1/Å)-0.05 0 0.05kx (1/Å)-0.05 0 0.05kx (1/Å)-0.3-0.2E - EF (eV)-0.05 0 0.05kx (1/Å)-0.05 0 0.05kx (1/Å)Int. (arb. units)-0.4 -0.2E - EF (eV)-0.05 0 0.05kx (1/Å)-0.3-0.2E - EF (eV)0.040-0.04kx (1/Å)-0.3-0.2E - EF (eV)Int.-0.5 0SYInt.a cbd e fg h i-k+kPMAFM-CPM AFM-C-k +kDFTTSSSS1TSSBulkBulkMΓM MΓM MΓM0.5 -0.2 0SY0.2Fig. 3 | Observation of a gapped surface Dirac cone entangled from thebulk band. a High-resolution ARPES image along the kx direction (ky =0Å−1)measured in the paramagnetic phase (T = 44K). The energy distribution curve(EDC) at �Γ is overlaid in red.bCurvature plot of the ARPESdata shown in (a). cDFT-calculated band structure in the paramagnetic phase. d–f Same as a–c but for theAFM-C domain (T = 13 K), showing the emergenceof anM-shaped surface band andthe opening of a Dirac gap. The EDC in d shows two peaks at �Γ (black arrows),indicating the gap. g Spin-resolved EDC spectra at +k and −k positions (as indicatedin a), showing opposite spin polarizations. Red and blue lines represent up anddown spin components, respectively.h Spin-resolved band dispersionmeasured inthe AFM-C domain. iCorresponding spin-polarizedDFTcalculation, confirming thespin texture of the gapped surface Dirac cone.Article https://doi.org/10.1038/s41467-026-77571-4Nature Communications |        (2026) 17:10841 5www.nature.com/naturecommunicationsa freshly cleaved surface. This result also suggests that the hybridiza-tion gap provides a sensitive probe of the surface magnetism.The temperature evolution of surface states below TN—specifi-cally, the emergence and energy shift of SS1 and its hybridization withthe topological Dirac cone—can be understood as follows. In NdBi,AFM ordering folds the bulk bands, and a bulk gap opens at themagnetic zone boundary where the original and folded bulk bandsintersect andmutually hybridize via the exchange interaction. In AFM-C, the bulkband folding occurs along the kzdirection, thereby openingthe corresponding bulk gap around �Γ, which overlaps the original bulkgap in the PM phase. In general, surface states tend to appear insidebulk band gaps. This gap increases as temperature decreases, reflect-ing the growth of the magnetic moment and the concomitantenhancement of the exchange interaction. Likewise, SS1 and SS2 ofNdBi emerge within this bulk gap that opens below TN. As the gapwidens, SS1 shifts upward in energy. Concurrently, SS1 intersects thesurface Dirac cone, thus a hybridization gap ΔHyb of these surfacebands opens to form theM-shape dispersion. This hybridization arisesfrom exchange interactions driven by surface magnetism, whichincreases as the bulkmagnetism intensifies at lower temperatures. Theemergence of the M-shaped surface band in the AFM phase is repro-duced by our DFT calculations (Supplementary Note 6).Our DFT calculations reveal a pronounced reconstruction of thebulk bands associated with the AFM phase transition (SupplementaryNote 6), leading to a substantial enlargement of the bulk band gap.Importantly, this provides a natural explanation for the sizable gap of~125meV reported for the AFM-C domain in Ref. 30. A similarly largegap is observed in our synchrotron-ARPESmeasurements, inwhich thespectra are dominated by bulk contributions due to matrix-elementeffects, thereby demonstrating that such large gaps originate frombulk bands rather than from topological surface states.A large Dirac gap (~80meV) was also reported at �M in AFM-Adomain30, despite the theoretical expectation that the Dirac pointremains gapless due to symmetry protection. According to our DFTcalculations, this apparent inconsistency can be consistently resolvedby recognizing that the reported gap does not reflect an intrinsicsurface Dirac gap, but instead arises from bulk-band contributions(Supplementary Note 7).Since the early days of topological insulator research, surfacecontamination has been widely employed to probe perturbativeeffects on TSS48–54. This approach is particularly important for AFM-TIs, as studies of the MnBi2nTe3n+1 family suggest that surface con-tamination may be a key factor underlying the conflicting reports onthe presence or absence of a surface Dirac gap. In the case of NdBi,this method is expected to provide deeper insights, as the hybridi-zation gap can serve as an indicator of surface magnetism. Therelationship between the Dirac gap and surface magnetism cantherefore be identified through simultaneous observation of thehybridization gap.In our study, we exploited natural surface contamination. Wefound a drastic change in both the hybridization and Dirac gaps evenunder an ultra-high vacuum of ~3 × 10−11 torr in the measurementchamber. Figure 5f displays the time evolution of the ARPES image forNdBi in the energy range capturing both the hybridization gap and theDirac gap. Over time, the spectral quality, as evaluated by peak width,remained almost constant, indicating that surface degradation is notsevere during the four-hour measurement. We found that the energyposition of SS1 (ESS1) remained unchanged over time. However, theΔHyb vanished four hours after the sample was cleaved (black arrows inFig. 5f). This is attributed to the loss of surfacemagnetism. Our resultssuggest that the surface magnetic order is more fragile than the sur-face electronic structure and can be suppressed even when the banddispersion remains well-defined, with the spectral sharpness andhence the surface quality, nearly unchanged.Very intriguingly, we found that theDirac gap, clearly observed ona freshly cleaved surface, disappeared after three hours. This is evidentin the EDC spectra at �Γ (Fig. 5f): the double-peak structure associatedwith the gapped Dirac conemerges into a single peak after four hours,consistent with the Dirac gap closure. We emphasize again that thespectral peak width remained nearly constant throughout these mea-surements. This rules out surface-degradation-induced spectralbroadening as the cause and indicates that these gap closures areintrinsic. TheDiracgapclosure is coincidentwith thehybridizationgapclosure, thus validating that the Dirac gap originates from surfacemagnetism.The time-dependent evolution of the band structure is schema-tically illustrated in Fig. 5g. The surface Dirac cone originates from thebulk band topology, as in the PM phase (left panel). The bulkmagneticorder (or bulk exchange interactions) generates SS1, which intersectswith the surface Dirac cone. On a freshly cleaved surface, the surfacemagnetic order aligns with the bulk magnetic order. The associatedsurface exchange interactions give rise to two types of gaps (middlepanel): (i) a hybridization gap between the surface Dirac cone and SS1,resulting in an M-shaped surface band, and (ii) the Dirac gap, whichopens due to time-reversal symmetry breaking induced by surfacemagnetism. Over time, residual gas gradually accumulates on thecrystal surface. The surface Dirac cone and SS1 remain present sincethe surface contamination (right panel) does not affect either the bulkband topology or the bulk magnetism. Nonetheless, the magneticorder is lost at the surface, as supported by the disappearance of thehybridization gap ΔHyb, and the Dirac gap closes.Our DFT calculations reveal that surface states in the AFM-Cdomain are more localized on the crystal surface than those in the PMphase (Supplementary Note 8). While the decay length of the TSS intothe bulk is approximately 17 nm in the PM state, it is reduced to 2–4nmin the AFM-C domain. This reduction occurs because a large bulk gap,opened by bulk magnetism, significantly decouples the TSS from thebulk states in the AFMphase.When the TSS becomesmore localized atthe surface, it experiences the surface magnetism more strongly.Therefore, forming the gapped Dirac cone in NdBi requires both bulk-0.05 0 0.05kx (1/Å)-0.4-0.3-0.2-0.10E-EF(eV)Γ MMΓ MMa bcBulkBulkSS1SS2-0.05 0 0.05kx (1/Å)-0.4-0.3-0.2-0.10E-E F(eV)minmaxdTSSSS1SS28 eV 10 eVTSSFig. 4 | Suppressionof surface state ARPES signals at higherphotonenergies. a,b Schematic band dispersions representing both surface states (TSS, SS1, and SS2)and bulk states (a), and only bulk states (b), corresponding to c and d, respectively.c, d ARPES intensity maps along the �Γ-- �Mdirectionmeasured at photon energies of8 eV (c) and 10 eV (d). While both surface and bulk bands are observed at the lowerphoton energy (8 eV), matrix-element effects strongly suppress the surface-stateintensity at the higher photon energy (10 eV), dominantly leaving only the signalsfrom the bulk states, as illustrated in (a, b), respectively.Article https://doi.org/10.1038/s41467-026-77571-4Nature Communications |        (2026) 17:10841 6www.nature.com/naturecommunicationsmagnetism, which decouples the TSS from the bulk, and surfacemagnetism, which couples directly to the localized TSS. Our experi-mental data in Fig. 4 indeed demonstrate that the Dirac gap is highlysensitive to the presence or absence of surface magnetic order,opening or closing accordingly.DiscussionThe observed exchange gap in the Dirac cone of NdBi (17 ± 2meV) is inexcellent agreement with our DFT calculations (12meV). Importantly,examination of the hybridization gap ΔHyb confirmed that this closematch between theory and experiments occurs only when the surfacemagnetism is ideally aligned in the out-of-plane direction. Notably, theintrinsic gap we identified in NdBi is significantly larger than the van-ishingly small gap observed in MnBi2nTe3n+121–25. These results willprovide an important clue toward the ultimate resolution of a centralpuzzle in AFM-TI research: the theory-reality mismatch.In MnBi2Te4, a relatively small gap (<3meV) limits the observa-tion of the quantum anomalous Hall effect (QAHE) to ultra-low tem-peratures (~1.4 K)13. In contrast, the substantially larger gapobserved inNdBi (17 ± 2meV) suggests the possibility of realizingQAHE at elevatedtemperatures on the order of 10K. Although further materials opti-mization will be required, this result points to a promising routetoward higher-temperature QAHE.We further note that thin-film fabrication offers a promisingapproach toward device realization. Quantum confinement in thinfilms is expected to drive a semimetal-to-semiconductor transitionwhile preserving both magnetism and band topology55,56. In addition,electrostatic gating could enable tunability of the chemical potential0 20 40 60-0.1 0 0.1kx (1/Å)-0.1 0 0.1kx (1/Å)-0.1 0 0.1kx (1/Å)-0.1 0 0.1kx (1/Å)-0.1 0 0.1kx (1/Å)-0.4-0.3-0.2-0.10E-EF(eV)-0.30-0.25-0.20-0.15E-EF( eV)0.060.040.020kx (1/Å)-0.05 0 0.05kx (1/Å)-0.3-0.2-0.1E-EF(eV)-0.05 0 0.05kx (1/Å)-0.05 0 0.05kx (1/Å)15 K-0.05 0 0.05kx (1/Å)86420Int.(x103/2s ec.)-56 -54ω (deg.)2.6 K28 K1.00.50√ �11 0(arb. units)252015Temperature (K)40200ΔHyb(meV)ARPESNeutron13 K25 K 22 K 19 K 16 KabfΓ kxkyθθAFM-CPara.13 K22 KΔHyb30 min. 3 hour 4 hour13 K21 K2 hourBVBESS1SS1SS2d ecPM AFM-C AFM-C(4 hour)gapSS1TSSBulk17 meVgΔHybΔHyb(meV)TSSΔHyb110M MΓFig. 5 | Evolutionof thehybridizationgapandDirac conewith temperature andsurface contamination. a Temperature evolution of ARPES images from 25K(paramagnetic phase) to 13 K (AFM-C domain), showing the emergence of surfacestates (SS1 and SS2) and the development of the hybridization gap (ΔHyb).b Temperature dependence of the Dirac cone and SS1, determined from the EDCpeak positions. For the EDC fittings, we employed a Gaussian function combinedwith a linear background. No constraints were imposed on the fitting parameters.The gray dotted line traces the TSS in the PMphase. c Temperature dependence ofthe 110 magnetic reflection peak measured by neutron scattering. The horizontalaxis ω denotes the sample rotation angle. d Comparison of the temperaturedependenceof the hybridization gap observed by ARPES (ΔHyb, left) and the squareroot of the integrated intensity of the 110 magnetic Bragg reflection (ffiffiffiffiffiffiffiI110p, right).ΔHyb is defined as the energy difference between the top and bottom of the para-bolic bands in (b). e Angular dependence of ΔHyb measured at 21K and 13 K. Theangle θ is defined from the �Γ-- �M direction in the (kx, ky) plane, as illustrated in theinset. f ARPES images measured at 15 K at different time intervals after cleavage,showing the degradation of the Dirac and hybridization gaps due to surface con-tamination. EDCs at �Γ are overlaid. g Schematic illustration of the electronicstructure in the paramagnetic phase, in the AFM-C domain immediately aftercleavage, and in the AFM-C domain after 4 h, highlighting the evolution of thesurface Dirac gap (17meV), hybridization gap (ΔHyb), and surface state SS1.Article https://doi.org/10.1038/s41467-026-77571-4Nature Communications |        (2026) 17:10841 7www.nature.com/naturecommunicationsand enhance the effective gap, offering a viable strategy to optimizeQAHE conditions.Finally, we point out that previous studies on magnetic topolo-gical insulators in the MnBi2nTe3n+1 family treat magnetism as a per-turbation that opens a Dirac gap to an effective Hamiltonian thatdescribes nonmagnetic states18,24,57. In contrast, our study reveals thatmagnetism in NdBi intricately intertwines with both bulk and surfacestates. Such magnetism is beyond the scope of a simple perturbativeframework for understanding the magnetic topological states. Theneed to go beyond the perturbative regime arises from the stronginterplay between localized 4f electrons and itinerant carriers—a qua-litatively different situation from that inMnBi2nTe3n+1, wheremagneticproperties are perturbatively induced bymagnetic layers composed ofMn atoms with 3d orbitals that are located in proximity to topologicalBi2Te3 layers. These findings for the second-generation AFM-TI (NdBi)mark a significant step beyond the first-generation AFM-TI(MnBi2nTe3n+1) and provide valuable guidance for the future designand discovery of antiferromagnetic topological insulators among awide variety of magnets.MethodsCalculationsFirst-principles calculations were performed using the WIEN2ksoftware58,59. We used the modified Becke–Johnson (BJ) potential60.The maximum modulus for the reciprocal lattice vectors Kmax waschosen so that RKmax = 8 where the muffin-tin radius R for Nd and Biatomswas 2.5 Bohr.We used a 19 × 19 × 19 k-meshwith the inclusion ofthe spin-orbit coupling. The experimental lattice constant a = 6.41Åwas taken from Ref. 61.For the paramagnetic phase, we adopt the open-core treatmentfor the Nd-f orbitals. For the spin-polarized phases, we adopt theDFT +U method62,63 with U = 6.3 eV and J =0.7 eV for the Nd-f orbitalsas adopted in the previous study45. We constructed the maximallylocalizedWannier functions64,65 for the Nd-d and Bi-p orbitals from thecalculated band structure using the Wannier90 software66. The semi-infinite-slab tight-binding models constructed from these Wannierfunctions were used for calculating the surface spectra in the waydescribed in Ref. 67.Single crystal growthNdBi single crystals were grown by the Bridgmanmethodwith a sealedtungsten crucible and a high-frequency induction furnace. High-purityNd (5N) andBi (6N)metalswith the respective composition ratioswereused as starting materials. The obtained samples were characterizedby X-ray diffraction patterns.ARPES measurementsAll the high-resolution laser-ARPES measurements were performed atthe Institute for Solid State Physics, the University of Tokyo, using twodifferent setups with 6.994 eV lasers68,69. The domain-selective mea-surements (Fig. 1) and the observation of the Dirac cone (Fig. 3) wereperformed using an ARPES system equipped with a ScientaOmicronDA30L spectrometer. Spin-resolved ARPES (SARPES) measurementswere conducted with VLEED-type spin-detectors installed on the Sci-entaOmicron DA30L spectrometer. We used an effective Shermanfunction of Seff =0.27. The VLEED-type spin detector operates by per-iodically switching the target magnetization during data acquisition,which effectively removes instrumental asymmetry. The energy(angular) resolution was set to 2.5 meV (0.1∘) and 21 meV (1.4∘) forARPES and SARPES, respectively.To investigate the temperature dependence of spectra in detail,we employed another ARPES system equipped with a Scienta R4000analyzer. The excellent temperature tunability in the sample manip-ulator enabled temperature scans in fine steps of 1 K whilemaintainingthe deviation within ±0.1 K42. The scans were first performed fromlower to higher temperatures, followed by another temperature scanfrom higher to lower temperatures to ensure data reliability.For temperature-dependent measurements, the samples werecleaved at the lowest temperature (13 K), whereas for other measure-ments, cleavage was performed at themeasurement temperature. Thebeam spot sizewas approximately 50μm,which is significantly smallerthan the typical magnetic domain size (>200μm, SupplementaryNote 2), ensuring that themeasurements predominantlyprobe a singlemagnetic domain. In addition, a total of 38 samples were cleaved andmeasured for ARPES. Although the data quality varies from sample tosample, themain results presented in themanuscript were found to behighly reproducible. We note that NdBi has a rock-salt crystal struc-ture, for which a single crystallographic termination is expected uponcleavage.The synchrotron-ARPES experiments with tunable low photonenergies were performed at the BL7U beamline of UVSOR, equippedwith an MBS A-1 analyzer. The energy resolution was better than15meV, and the angular resolution was 0.3∘. The synchrotron data(Figs. 4 and Supplementary S5) exhibit a slight distortion in the bandshape due to a minor imperfection in the analyzer lens setting. Whilethis affects the precise determination of the band dispersion andprevents a fair comparison across different photon energies, it doesnot affect the key observation that the surface-state intensity isstrongly suppressed at 10 eV due to matrix element effects.In all these ARPES experiments, we used the linear polarizationwith the electric field aligned in the incident plane. The crystals werecleaved in situ at the measurement temperatures. The base pressureduring the ARPES measurements was better than 1 × 10−8 Pa.Neutron scattering experimentNeutron scattering experiments were conducted using the POlarizedNeutron Triple-Axis triple-axis spectrometer PONTA installed at the5G beamhole in the Japan Research Reactor 3 (JRR-3) of Japan AtomicEnergy Agency70. The spectrometer was operated in a two-axis dif-fraction mode. An incident neutron beam with the energy of34.05meV was obtained using a pyrolytic graphite monochromator.A single crystal sample with a volume of 22.52mm3 was oriented inthe (H, K, 0) horizontal scattering plane and loaded in a clamp-typeuniaxial-stress cell71,72. A uniaxial stress of 25MPa was applied ontothe (001) surfaces of the sample, so that the magnetic domains withthe magnetic moments parallel to the [001] direction were selectedat low temperatures. The cell was attached to the cold head of aclosed-cycle 4He refrigerator and cooled to the lowest temperatureof 2.3 K. Further experimental details are provided in Supplemen-tary Note 4.Data availabilityThe data that support the findings of this study are available in fig-share, “https://doi.org/10.6084/m9.figshare.32920193”.References1. Fu, L. & Kane, C. L. Superconducting proximity effect andMajoranafermions at the surface of a topological insulator. Phys. Rev. Lett.100, 096407 https://doi.org/10.1103/PhysRevLett.100.096407 (2008).2. Dzero, M., Sun, K., Galitski, V. & Coleman, P. Topological Kondoinsulators. Phys. Rev. Lett. 104, 106408 https://doi.org/10.1103/PhysRevLett.104.106408 (2010).3. Miyazaki, H., Hajiri, T., Ito, T., Kunii, S. & Kimura, S.-I. Momentum-dependent hybridization gap and dispersive in-gap state of theKondo semiconductor SmB6. Phys. Rev. B Condens. Matter Mater.Phys https://doi.org/10.1103/PhysRevB.86.075105 (2012).4. Alexandrov, V., Coleman, P. & Erten, O. Kondo breakdown intopological Kondo insulators. Phys. Rev. Lett. 114, 177202 https://doi.org/10.1103/PhysRevLett.114.177202 (2015).Article https://doi.org/10.1038/s41467-026-77571-4Nature Communications |        (2026) 17:10841 8https://doi.org/10.6084/m9.figshare.32920193https://doi.org/10.1103/PhysRevLett.100.096407https://doi.org/10.1103/PhysRevLett.100.096407https://doi.org/10.1103/PhysRevLett.100.096407https://doi.org/10.1103/PhysRevLett.104.106408https://doi.org/10.1103/PhysRevLett.104.106408https://doi.org/10.1103/PhysRevLett.104.106408https://doi.org/10.1103/PhysRevB.86.075105https://doi.org/10.1103/PhysRevB.86.075105https://doi.org/10.1103/PhysRevLett.114.177202https://doi.org/10.1103/PhysRevLett.114.177202https://doi.org/10.1103/PhysRevLett.114.177202www.nature.com/naturecommunications5. Ohtsubo, Y. et al. Non-trivial surface states of samarium hexaborideat the (111) surface.Nat. Commun. 10, 2298 https://doi.org/10.1038/s41467-019-10353-3 (2019).6. Weng, H., Zhao, J., Wang, Z., Fang, Z. & Dai, X. Topological crys-talline Kondo insulator in mixed valence ytterbium borides. Phys.Rev. Lett. 112, 016403 https://doi.org/10.1103/PhysRevLett.112.016403 (2014).7. Pesin, D. & Balents, L. Mott physics and band topology in materialswith strong spin-orbit interaction. Nat. Phys. 6, 376–381 https://doi.org/10.1038/nphys1606 (2010).8. Yamaji, Y. & Imada, M. Mott physics on helical edges of two-dimensional topological insulators. Phys. Rev. B.83, 205122 https://doi.org/10.1103/PhysRevB.83.205122 (2011).9. Mong, R. S. K., Essin, A. M. & Moore, J. E. Antiferromagnetictopological insulators. Phys. Rev. B: Condens. Matter Mater.Phys. 81, 245209 https://doi.org/10.1103/PhysRevB.81.245209 (2010).10. Otrokov, M. M. et al. Prediction and observation of an anti-ferromagnetic topological insulator. Nature 576, 416–422 https://doi.org/10.1038/s41586-019-1840-9 (2019).11. Chang, C.-Z. et al. Experimental observation of the quantumanomalous Hall effect in a magnetic topological insulator. Science340, 167–170 https://doi.org/10.1126/science.1234414 (2013).12. Chang, C.-Z. et al. High-precision realization of robust quantumanomalous Hall state in a hard ferromagnetic topological insulator.Nat. Mater. 14, 473–477 https://doi.org/10.1038/nmat4204 (2015).13. Deng, Y. et al. Quantum anomalous Hall effect in intrinsic magnetictopological insulator MnBi2Te4. Science 367, 895–900 https://doi.org/10.1126/science.aax8156 (2020).14. Ge, J. et al. High-Chern-number and high-temperature quantumHall effect without Landau levels. Natl. Sci. Rev. 7, 1280–1287https://doi.org/10.1093/nsr/nwaa089 (2020).15. Liu, C. et al. Robust axion insulator and Chern insulator phases in atwo-dimensional antiferromagnetic topological insulator. Nat.Mater. 19, 522–527 https://doi.org/10.1038/s41563-019-0573-3 (2020).16. Zhao, Y. & Liu, Q. Routes to realize the axion-insulator phase inMnBi2Te4(Bi2Te3)n family. Appl. Phys. Lett https://pubs.aip.org/aip/apl/article/119/6/060502/41929 (2021).17. Lin, W. et al. Direct visualization of edge state in even-layerMnBi2Te4 at zeromagnetic field.Nat. Commun. 13, 7714 https://doi.org/10.1038/s41467-022-35482-0 (2022).18. Zhang, D. et al. Topological axion states in the magnetic insulatorMnBi2Te4 with the quantized magnetoelectric effect. Phys. Rev.Lett. 122, 206401 https://doi.org/10.1103/physrevlett.122.206401 (2019).19. Li, J. et al. Intrinsicmagnetic topological insulators in van derWaalslayered MnBi2Te4-family materials. Sci. Adv. 5, eaaw5685 https://doi.org/10.1126/sciadv.aaw5685 (2019).20. Rienks, E. D. L. et al. Largemagnetic gap at the Dirac point in Bi2Te3/MnBi2Te4 heterostructures. Nature 576, 423–428 https://doi.org/10.1038/s41586-019-1826-7 (2019).21. Chen, Y. J. et al. Topological electronic structure and its tempera-ture evolution in antiferromagnetic topological insulator MnBi2Te4.Phys. Rev. X 9, 041040 https://doi.org/10.1103/PhysRevX.9.041040 (2019).22. Hao, Y. -J. et al. Gapless surface Dirac cone in antiferromagnetictopological insulator MnBi2Te4. Phys. Rev. X. 9, 041038 https://doi.org/10.1103/PhysRevX.9.041038 (2019).23. Li, H. et al. Dirac surface states in intrinsic magnetic topologicalinsulators EuSn2As2 and MnBi2nTe3n+1. Phys. Rev. X. 9, 041039https://doi.org/10.1103/PhysRevX.9.041039 (2019).24. Nevola, D. et al. Coexistence of surface ferromagnetism and agapless topological state in MnBi2Te4. Phys. Rev. Lett. 125, 117205https://doi.org/10.1103/PhysRevLett.125.117205 (2020).25. Swatek, P. et al. Gapless Dirac surface states in the anti-ferromagnetic topological insulator MnBi2Te4. Phys. Rev. B. 101,161109 https://doi.org/10.1103/PhysRevB.101.161109 (2020).26. Garnica, M. et al. Native point defects and their implications for theDirac point gap at MnBi2Te4(0001). Npj Quantum Mater. 7,1–9 https://doi.org/10.1038/s41535-021-00414-6 (2022).27. Li, S. et al. Progress on the antiferromagnetic topological insulatorMnBi2Te4. Natl. Sci. Rev. 11, nwac296 https://doi.org/10.1093/nsr/nwac296 (2024).28. Yang, S. & Ye, Y. Intrinsic and defect-related magnetism ofMnBi2Te4(Bi2Te3)n family from the bulk to two-dimensional limit. 2dMater https://doi.org/10.1088/2053-1583/ada044 (2024).29. Tan, H. & Yan, B. Distinctmagnetic gaps between antiferromagneticand ferromagnetic orders driven by surface defects in the topolo-gicalmagnetMnBi2Te4. Phys. Rev. Lett. 130, 126702https://doi.org/10.1103/physrevlett.130.126702 (2023).30. Honma, A. et al. Antiferromagnetic topological insulator withselectively gappedDirac cones.Nat. Commun. 14, 7396https://doi.org/10.1038/s41467-023-42782-6 (2023).31. Honma, A. et al. Universal role of combined symmetry for the pro-tection of the Dirac cone in antiferromagnetic topological insula-tors. Phys. Rev. B. 110, 115152 https://doi.org/10.1103/PhysRevB.110.115152 (2024).32. Hasegawa, A. Fermi surface of LaSb and LaBi. J. Phys. Soc. Jpn. 54,677–684 https://doi.org/10.1143/JPSJ.54.677 (1985).33. Kuroda, K. et al. Experimental determination of the topologicalphase diagram in cerium monopnictides. Phys. Rev. Lett. 120,086402 https://doi.org/10.1103/PhysRevLett.120.086402 (2018).34. Duan, X. et al. Tunable electronic structure and topological prop-erties of LnPn (Ln =Ce, Pr, Sm,Gd, Yb; Pn = Sb, Bi).Commun. Phys. 1,1–7 https://doi.org/10.1038/s42005-018-0074-8 (2018).35. Lou, R. et al. Evidenceof topological insulator state in the semimetalLaBi. Phys. Rev. B Condens. Matter. 95, 115140 https://doi.org/10.1103/PhysRevB.95.115140 (2017).36. Nayak, J. et al. Multiple Dirac cones at the surface of the topologicalmetal LaBi. Nat. Commun. 8, 13942 https://doi.org/10.1038/ncomms13942 (2017).37. Li, P. et al. Tunable electronic structure and surface states in rare-earth monobismuthides with partially filled f shell. Phys. Rev. BCondens. Matter. 98, 085103 https://doi.org/10.1103/PhysRevB.98.085103 (2018).38. Kasuya, T. Physics in low carrier strongly correlated systems: kondoinsulator magnetic polaron and high Tc. Physica B Condens. Matter215, 88–98 https://doi.org/10.1016/0921-4526(95)90122-E (1995).39. Rossat-Mignod, J. et al. Phase diagram and magnetic structures ofCeSb. Phys. Rev. 16, 440–461 https://doi.org/10.1103/PhysRevB.16.440 (1977).40. Heer,H., Furrer, A., Halg,W. & Vogt, O. Neutron spectroscopy in theceriummonopnictides. J. Phys. C: Solid State Phys. 12, 5207 https://doi.org/10.1088/0022-3719/12/23/025 (1979).41. Rossat-Mignod, J. et al. Magnetic properties of cerium monopnic-tides. J. Magn. Magn. Mater. 31-34, 398–404 https://doi.org/10.1016/0304-8853(83)90295-0 (1983).42. Kuroda, K. et al. Devil’s staircase transition of the electronic struc-tures in CeSb. Nat. Commun. 11, 2888 https://doi.org/10.1038/s41467-020-16707-6 (2020).43. Kushnirenko, Y. et al. Directional effects of antiferromagneticordering on the electronic structure in NdSb. Phys. Rev. B. 108,115102 https://doi.org/10.1103/PhysRevB.108.115102 (2023).44. Schrunk, B. et al. Emergence of Fermi arcs due tomagnetic splittingin an antiferromagnet. Nature 603, 610–615 https://doi.org/10.1038/s41586-022-04412-x (2022).45. Wang, L.-L. et al. Unconventional surface state pairs in a high-symmetry lattice with anti-ferromagnetic band-folding. Commun.Phys. 6, 1–9 https://doi.org/10.1038/s42005-023-01180-6 (2023).Article https://doi.org/10.1038/s41467-026-77571-4Nature Communications |        (2026) 17:10841 9https://doi.org/10.1038/s41467-019-10353-3https://doi.org/10.1038/s41467-019-10353-3https://doi.org/10.1038/s41467-019-10353-3https://doi.org/10.1103/PhysRevLett.112.016403https://doi.org/10.1103/PhysRevLett.112.016403https://doi.org/10.1103/PhysRevLett.112.016403https://doi.org/10.1038/nphys1606https://doi.org/10.1038/nphys1606https://doi.org/10.1038/nphys1606https://doi.org/10.1103/PhysRevB.83.205122https://doi.org/10.1103/PhysRevB.83.205122https://doi.org/10.1103/PhysRevB.83.205122https://doi.org/10.1103/PhysRevB.81.245209https://doi.org/10.1103/PhysRevB.81.245209https://doi.org/10.1103/PhysRevB.81.245209https://doi.org/10.1038/s41586-019-1840-9https://doi.org/10.1038/s41586-019-1840-9https://doi.org/10.1038/s41586-019-1840-9https://doi.org/10.1126/science.1234414https://doi.org/10.1126/science.1234414https://doi.org/10.1038/nmat4204https://doi.org/10.1038/nmat4204https://doi.org/10.1126/science.aax8156https://doi.org/10.1126/science.aax8156https://doi.org/10.1126/science.aax8156https://doi.org/10.1093/nsr/nwaa089https://doi.org/10.1093/nsr/nwaa089https://doi.org/10.1038/s41563-019-0573-3https://doi.org/10.1038/s41563-019-0573-3https://doi.org/10.1038/s41563-019-0573-3https://pubs.aip.org/aip/apl/article/119/6/060502/41929https://pubs.aip.org/aip/apl/article/119/6/060502/41929https://pubs.aip.org/aip/apl/article/119/6/060502/41929https://doi.org/10.1038/s41467-022-35482-0https://doi.org/10.1038/s41467-022-35482-0https://doi.org/10.1038/s41467-022-35482-0https://doi.org/10.1103/physrevlett.122.206401https://doi.org/10.1103/physrevlett.122.206401https://doi.org/10.1103/physrevlett.122.206401https://doi.org/10.1126/sciadv.aaw5685https://doi.org/10.1126/sciadv.aaw5685https://doi.org/10.1126/sciadv.aaw5685https://doi.org/10.1038/s41586-019-1826-7https://doi.org/10.1038/s41586-019-1826-7https://doi.org/10.1038/s41586-019-1826-7https://doi.org/10.1103/PhysRevX.9.041040https://doi.org/10.1103/PhysRevX.9.041040https://doi.org/10.1103/PhysRevX.9.041040https://doi.org/10.1103/PhysRevX.9.041038https://doi.org/10.1103/PhysRevX.9.041038https://doi.org/10.1103/PhysRevX.9.041038https://doi.org/10.1103/PhysRevX.9.041039https://doi.org/10.1103/PhysRevX.9.041039https://doi.org/10.1103/PhysRevLett.125.117205https://doi.org/10.1103/PhysRevLett.125.117205https://doi.org/10.1103/PhysRevB.101.161109https://doi.org/10.1103/PhysRevB.101.161109https://doi.org/10.1038/s41535-021-00414-6https://doi.org/10.1038/s41535-021-00414-6https://doi.org/10.1093/nsr/nwac296https://doi.org/10.1093/nsr/nwac296https://doi.org/10.1093/nsr/nwac296https://doi.org/10.1088/2053-1583/ada044https://doi.org/10.1088/2053-1583/ada044https://doi.org/10.1103/physrevlett.130.126702https://doi.org/10.1103/physrevlett.130.126702https://doi.org/10.1103/physrevlett.130.126702https://doi.org/10.1038/s41467-023-42782-6https://doi.org/10.1038/s41467-023-42782-6https://doi.org/10.1038/s41467-023-42782-6https://doi.org/10.1103/PhysRevB.110.115152https://doi.org/10.1103/PhysRevB.110.115152https://doi.org/10.1103/PhysRevB.110.115152https://doi.org/10.1143/JPSJ.54.677https://doi.org/10.1143/JPSJ.54.677https://doi.org/10.1103/PhysRevLett.120.086402https://doi.org/10.1103/PhysRevLett.120.086402https://doi.org/10.1038/s42005-018-0074-8https://doi.org/10.1038/s42005-018-0074-8https://doi.org/10.1103/PhysRevB.95.115140https://doi.org/10.1103/PhysRevB.95.115140https://doi.org/10.1103/PhysRevB.95.115140https://doi.org/10.1038/ncomms13942https://doi.org/10.1038/ncomms13942https://doi.org/10.1038/ncomms13942https://doi.org/10.1103/PhysRevB.98.085103https://doi.org/10.1103/PhysRevB.98.085103https://doi.org/10.1103/PhysRevB.98.085103https://doi.org/10.1016/0921-4526(95)90122-Ehttps://doi.org/10.1016/0921-4526(95)90122-Ehttps://doi.org/10.1103/PhysRevB.16.440https://doi.org/10.1103/PhysRevB.16.440https://doi.org/10.1103/PhysRevB.16.440https://doi.org/10.1088/0022-3719/12/23/025https://doi.org/10.1088/0022-3719/12/23/025https://doi.org/10.1088/0022-3719/12/23/025https://doi.org/10.1016/0304-8853(83)90295-0https://doi.org/10.1016/0304-8853(83)90295-0https://doi.org/10.1016/0304-8853(83)90295-0https://doi.org/10.1038/s41467-020-16707-6https://doi.org/10.1038/s41467-020-16707-6https://doi.org/10.1038/s41467-020-16707-6https://doi.org/10.1103/PhysRevB.108.115102https://doi.org/10.1103/PhysRevB.108.115102https://doi.org/10.1038/s41586-022-04412-xhttps://doi.org/10.1038/s41586-022-04412-xhttps://doi.org/10.1038/s41586-022-04412-xhttps://doi.org/10.1038/s42005-023-01180-6https://doi.org/10.1038/s42005-023-01180-6www.nature.com/naturecommunications46. Yamamoto, R. et al. Spin splitting in the surface electronic structureof antiferromagnet NdBi. Phys. Rev. Res. 7, L022005 https://doi.org/10.1103/PhysRevResearch.7.L022005 (2025).47. Chen, B. et al. Intrinsicmagnetic topological insulator phases in theSb-doped MnBi2Te4 bulks and thin flakes. Nat. Commun. 10, 4469https://doi.org/10.1038/s41467-019-12485-y (2019).48. Kong, D. et al. Rapid surface oxidation as a source of surfacedegradation factor for Bi2Se3. ACS Nano 5, 4698–4703 https://doi.org/10.1021/nn200556h (2011).49. Benia,H.M., Lin,C., Kern, K. &Ast,C. R. Reactive chemical dopingofthe Bi2Se3 topological insulator.Phys. Rev. Lett. 107, 177602https://doi.org/10.1103/physrevlett.107.177602 (2011).50. Wang, X., Bian, G., Miller, T. & Chiang, T.-C. Fragility of surfacestates and robustness of topological order in Bi2Se3 against oxida-tion. Phys. Rev. Lett. 108, 096404 https://doi.org/10.1103/physrevlett.108.096404 (2012).51. Yang, J. et al. Robust topological states in Bi2Se3 against surfaceoxidation. J. Phys. Chem. C Nanomater. Interfaces 124, 6253–6259https://doi.org/10.1021/acs.jpcc.0c00458 (2020).52. Zhu, Z.-H. et al. Rashba spin-splitting control at the surface of thetopological insulator Bi2Se3. Phys. Rev. Lett. 107, 186405 https://doi.org/10.1103/physrevlett.107.186405 (2011).53. King, P. D. C. et al. Large tunable Rashba spin splitting of a two-dimensional electron gas in Bi2Se3. Phys. Rev. Lett. 107, 096802https://doi.org/10.1103/physrevlett.107.096802 (2011).54. Valla, T., Pan, Z.-H., Gardner, D., Lee, Y. S. & Chu, S. Photoemissionspectroscopy of magnetic and nonmagnetic impurities on thesurface of the Bi2Se3 topological insulator. Phys. Rev. Lett. 108,117601 https://doi.org/10.1103/physrevlett.108.117601 (2012).55. Ho, D. Q., Hu, R., To, D. Q., Bryant, G. W. & Janotti, A. Emergingnontrivial topology in ultrathin films of rare-earth pnictides. ACSNano 17, 20991–20998 https://doi.org/10.1021/acsnano.3c03307 (2023).56. Inoue, H. et al. Band engineering of a magnetic thin film rare-earthmonopnictide: a platform for high Chern number. Phys. Rev. Mater.3, 101202 https://doi.org/10.1103/PhysRevMaterials.3.101202 (2019).57. Yan, C. et al. Origins of electronic bands in the antiferromagnetictopological insulator MnBi2Te4. Phys. Rev. B. 104, L041102 https://doi.org/10.1103/physrevb.104.l041102 (2021).58. Blaha, P. et al. WIEN2k, an augmented plane wave + local orbitalsprogram for calculating crystal properties. (Karlheinz Schwarz,Vienna University of Technology, Austria, 2018).59. Blaha, P. et al. WIEN2k: an APW+lo program for calculating theproperties of solids. J. Chem. Phys. 152, 074101 https://doi.org/10.1063/1.5143061 (2020).60. Tran, F. & Blaha, P. Accurate band gaps of semiconductors andinsulators with a semilocal exchange-correlation potential. Phys.Rev. Lett. 102, 226401 https://doi.org/10.1103/PhysRevLett.102.226401 (2009).61. Abdusalyamova, M. N., Rakhmatov, O. I. & Shokirov, K. S. Someproperties of monobismuthides of cerium subgroup REM. Russ.Metall. 1 (1988).62. Anisimov, V. I., Solovyev, I. V., Korotin, M. A., Czyzyk, M. T. &Sawatzky, G. A. Density-functional theory and NiO photoemissionspectra. Phys. Rev. BCondens.Matter.48, 16929–16934 https://doi.org/10.1103/physrevb.48.16929 (1993).63. Liechtenstein, A. I., Anisimov, V. I. & Zaanen, J. Density-functionaltheory and strong interactions: orbital ordering in Mott-Hubbardinsulators. Phys. Rev. B Condens. Matter. 52, R5467–R5470 https://doi.org/10.1103/physrevb.52.r5467 (1995).64. Marzari, N. & Vanderbilt, D. Maximally localized generalized Wan-nier functions for composite energy bands. Phys. Rev. B Condens.Matter 56, 12847–12865 https://doi.org/10.1103/PhysRevB.56.12847 (1997).65. Souza, I., Marzari, N. & Vanderbilt, D. Maximally localized Wannierfunctions for entangled energybands. Phys. Rev. BCondens.Matter65, 035109 https://doi.org/10.1103/PhysRevB.65.035109 (2001).66. Pizzi, G. et al. Wannier90 as a community code: new features andapplications. J. Phys. Condens. Matter 32, 165902 https://doi.org/10.1088/1361-648X/ab51ff (2020).67. Sancho, M. P. L., Sancho, J. M. L., Sancho, J. M. L. & Rubio, J. Highlyconvergent schemes for the calculation of bulk and surface Greenfunctions. J. Phys. F: Met. Phys. 15, 851 https://doi.org/10.1088/0305-4608/15/4/009 (1985).68. Shimojima, T., Okazaki, K. & Shin, S. Low-temperature and high-energy-resolution laser photoemission spectroscopy. J. Phys. Soc.Jpn. 7, 072001 https://doi.org/10.7566/JPSJ.84.072001 (2015).69. Yaji, K. et al. High-resolution three-dimensional spin- and angle-resolved photoelectron spectrometer using vacuum ultravioletlaser light. Rev. Sci. Instrum. 87, 053111 https://doi.org/10.1063/1.4948738 (2016).70. Nakajima, T. et al. Polarized and unpolarized neutron scattering formagnetic materials at the triple-axis spectrometer PONTA in JRR-3.J. Phys. Soc. Jpn 93, 091002 https://doi.org/10.7566/jpsj.93.091002 (2024).71. Aso, N. et al. A uniaxial pressure cell for neutron diffraction inves-tigation and its use in studying the single-crystalline Sr3Ru2O7compound. J. Phys. Condens. Matter 17, S3025–S3028 https://doi.org/10.1088/0953-8984/17/40/005 (2005).72. Nakajima, T. et al. Spin-wave spectrum in “single-domain”magneticground state of triangular lattice antiferromagnet CuFeO2. J. Phys.Soc. Jpn. 80, 014714 https://doi.org/10.1143/jpsj.80.014714 (2011).AcknowledgmentsThe sample preparation in this work was carried out using a glovebox inthe facilities of the Materials Design and Characterization Laboratory atthe Institute for Solid State Physics, the University of Tokyo. A part of thiswork was conducted at the BL7U of UVSOR Synchrotron Facility, Insti-tute for Molecular Science (IMS program 24IMS6656 and 25IMS6663).The neutron scattering experiment at PONTA in JRR-3 was carried outalong the proposals (No. 24401), and was partly supported by theNeutron Science Laboratory of the Institute for Solid State Physics, theUniversity of Tokyo.Author contributionsT.K. conceived and designed the project. Y.F. performed the ARPESexperimentswith the help of T.Nishida, K.K., H.T., A.H., T.Iimori, K.Y., andK.T., and analyzed the data under the supervision of R.M. and T.K. H.S.S.fabricated the samples. M.O. performed the DFT calculations, and Y.F.,T.K., M.O., and T.M. interpreted the data. T.Ikenobe performed thetransport experiments. Y.F., H.S., and T.Nakajima performed the neutronexperiments. Y.K. performed the polarized microscopy experimentsunder the supervision of M.T. All authors discussed the results andcontributed to their interpretation. Y.F. and T.K. wrote the manuscriptwith input from all authors. T.K. supervised the project.FundingR.M. discloses support for the research of this work from The Universityof TokyoEdgeCapital Partners (UTEC)-University of Tokyo FutureSocietyInitiative (UTokyo FSI) Research Grant Program, and JST PRESTO GrantNo. JPMJPR24LA. T.K. discloses support for the research of this workfrom the JSPS KAKENHI [Grant Numbers: JP25H01250, JP25H01246, andJP26H02014], MEXT Q-LEAP [Grant Number: JPMXS0118068681], TheAsahi Glass Foundation, The Murata Science Foundation, TheMitsubishiFoundation, and Toray Science Foundation.Competing interestsThe authors declare no competing interests.Article https://doi.org/10.1038/s41467-026-77571-4Nature Communications |        (2026) 17:10841 10https://doi.org/10.1103/PhysRevResearch.7.L022005https://doi.org/10.1103/PhysRevResearch.7.L022005https://doi.org/10.1103/PhysRevResearch.7.L022005https://doi.org/10.1038/s41467-019-12485-yhttps://doi.org/10.1038/s41467-019-12485-yhttps://doi.org/10.1021/nn200556hhttps://doi.org/10.1021/nn200556hhttps://doi.org/10.1021/nn200556hhttps://doi.org/10.1103/physrevlett.107.177602https://doi.org/10.1103/physrevlett.107.177602https://doi.org/10.1103/physrevlett.107.177602https://doi.org/10.1103/physrevlett.108.096404https://doi.org/10.1103/physrevlett.108.096404https://doi.org/10.1103/physrevlett.108.096404https://doi.org/10.1021/acs.jpcc.0c00458https://doi.org/10.1021/acs.jpcc.0c00458https://doi.org/10.1103/physrevlett.107.186405https://doi.org/10.1103/physrevlett.107.186405https://doi.org/10.1103/physrevlett.107.186405https://doi.org/10.1103/physrevlett.107.096802https://doi.org/10.1103/physrevlett.107.096802https://doi.org/10.1103/physrevlett.108.117601https://doi.org/10.1103/physrevlett.108.117601https://doi.org/10.1021/acsnano.3c03307https://doi.org/10.1021/acsnano.3c03307https://doi.org/10.1021/acsnano.3c03307https://doi.org/10.1103/PhysRevMaterials.3.101202https://doi.org/10.1103/PhysRevMaterials.3.101202https://doi.org/10.1103/physrevb.104.l041102https://doi.org/10.1103/physrevb.104.l041102https://doi.org/10.1103/physrevb.104.l041102https://doi.org/10.1063/1.5143061https://doi.org/10.1063/1.5143061https://doi.org/10.1063/1.5143061https://doi.org/10.1103/PhysRevLett.102.226401https://doi.org/10.1103/PhysRevLett.102.226401https://doi.org/10.1103/PhysRevLett.102.226401https://doi.org/10.1103/physrevb.48.16929https://doi.org/10.1103/physrevb.48.16929https://doi.org/10.1103/physrevb.48.16929https://doi.org/10.1103/physrevb.52.r5467https://doi.org/10.1103/physrevb.52.r5467https://doi.org/10.1103/physrevb.52.r5467https://doi.org/10.1103/PhysRevB.56.12847https://doi.org/10.1103/PhysRevB.56.12847https://doi.org/10.1103/PhysRevB.56.12847https://doi.org/10.1103/PhysRevB.65.035109https://doi.org/10.1103/PhysRevB.65.035109https://doi.org/10.1088/1361-648X/ab51ffhttps://doi.org/10.1088/1361-648X/ab51ffhttps://doi.org/10.1088/1361-648X/ab51ffhttps://doi.org/10.1088/0305-4608/15/4/009https://doi.org/10.1088/0305-4608/15/4/009https://doi.org/10.1088/0305-4608/15/4/009https://doi.org/10.7566/JPSJ.84.072001https://doi.org/10.7566/JPSJ.84.072001https://doi.org/10.1063/1.4948738https://doi.org/10.1063/1.4948738https://doi.org/10.1063/1.4948738https://doi.org/10.7566/jpsj.93.091002https://doi.org/10.7566/jpsj.93.091002https://doi.org/10.7566/jpsj.93.091002https://doi.org/10.1088/0953-8984/17/40/005https://doi.org/10.1088/0953-8984/17/40/005https://doi.org/10.1088/0953-8984/17/40/005https://doi.org/10.1143/jpsj.80.014714https://doi.org/10.1143/jpsj.80.014714www.nature.com/naturecommunicationsAdditional informationSupplementary information The online version containssupplementary material available athttps://doi.org/10.1038/s41467-026-77571-4.Correspondence and requests for materials should be addressed toTakeshi Kondo.Peer review information Nature Communications thanks Maria CarmenAsensio and the other anonymous reviewer(s) for their contribution tothe peer review of this work. A peer review file is available.Reprints and permissions information is available athttp://www.nature.com/reprintsPublisher’s note Springer Nature remains neutral with regard to jur-isdictional claims in published maps and institutional affiliations.Open Access This article is licensed under a Creative CommonsAttribution 4.0 International License, which permits use, sharing,adaptation, distribution and reproduction in any medium or format, aslong as you give appropriate credit to the original author(s) and thesource, provide a link to the Creative Commons licence, and indicate ifchanges were made. The images or other third party material in thisarticle are included in the article's Creative Commons licence, unlessindicated otherwise in a credit line to the material. If material is notincluded in the article's Creative Commons licence and your intendeduse is not permitted by statutory regulation or exceeds the permitteduse, you will need to obtain permission directly from the copyrightholder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.© The Author(s) 2026Article https://doi.org/10.1038/s41467-026-77571-4Nature Communications |        (2026) 17:10841 11https://doi.org/10.1038/s41467-026-77571-4http://www.nature.com/reprintshttp://creativecommons.org/licenses/by/4.0/http://creativecommons.org/licenses/by/4.0/www.nature.com/naturecommunications Direct observation of a spin-polarized surface Dirac gap in the antiferromagnetic topological insulator NdBi Introduction Results Discussion Methods Calculations Single crystal growth ARPES measurements Neutron scattering experiment Data availability References Acknowledgments Author contributions Funding Competing interests Additional information