Magnetic topological insulators (MTIs) and Weyl semimetals have both received a great deal of attention in recent condensed matter physics research1,2,3. Particularly, the intrinsic MTI MnBi2Te4 has provided researchers with an ideal candidate to study the relationship between topological quantum states and magnetic phases4,5,6,7,8,9,10,11,12,13,14,15. Theoretical predictions16,17,18,19 and recent experimental results20 show that when its Mn local moment spins are aligned by an external magnetic field, the energy bands of bulk MnBi2Te4 have a single isolated pair of Weyl crossing points that are close to the Fermi level and therefore can be accessed by controlling the carrier-density.

In this work, we report on thickness-dependent magneto-transport studies of the mechanically-exfoliated Mn(Bi1−xSbx)2Te4 thin flakes with three different Sb concentrations. The Sb substitutions in the MnBi2Te4 parent compound (i) move the Fermi level of the bulk bands closer to the charge neutrality point (CNP)20,21,22, and (ii) modulate the Weyl point separation in momentum space as illustrated in Fig. 1a. We focus here on the Chern insulator states of the spin-moment aligned Mn(Bi1−xSbx)2Te4. We show that Sb substitution extends the surface gap regime to a wider thickness range by suppressing conduction from the trivial bulk bands. Thin films Mn(Bi1−xSbx)2Te4 provide a rich plethora of topologically distinct quasi-two-dimensional (2D) states that includes Chern insulator4,5,6,8,9,10 and axion insulator5,7,16,23 states. The application of external magnetic fields to Mn(Bi1−xSbx)2Te4, therefore, generates an interplay between Chern gaps and Landau levels (LLs) quantization, allowing us to study rich quantum Hall physics that has not yet been fully explored.

Fig. 1: Tunable Chern insulator states and transport properties.
figure 1

a DFT calculated band structures for spin-aligned Mn(Bi1−xSbx)2Te4 intrinsic magnetic topological insulators at different Sb substitution levels. The positions of the Weyl point (\({k}_{w}\)) for x = 0%, 25%, and 50% are shown as black, red, and blue nodes, respectively, in the figure. The \(C\) = 0 and 1 denote the Chern number outside and within the ±\({k}_{w}\), respectively. The solid and dashed lines represent band structure along the \(\Gamma -Z\) and \(\Gamma -\bar{z}\) lines, respectively. b Calculated magnetic exchange gap size as a function of film thickness for the Mn(Bi1−xSbx)2Te4 at different Sb substitution levels. The thickness ranges for \(C\) = 0, 1, 2, 3 at x = 25% are labeled in the figure. c Resistivity at charge neutrality point (\({\uprho}_{{{{{\mathrm{xx}}}}}}^{{{{{\mathrm{CNP}}}}}}\)) plot as a function of flake thickness for Mn(Bi1−xSbx)2Te4 at Sb concentrations of x = 0, 0.20 and 0.26. \({\uprho}_{{{{{\mathrm{xx}}}}}}^{{{{{\mathrm{CNP}}}}}}\) is defined as the gate-dependent resistivity peak measured at a temperature, T = 2 K, at zero magnetic field. The error bars in (c) are estimated from the uncertainty in the determination of the device geometry. d ρxx plot as a function of backgate voltage for a 18-SL Mn(Bi0.74Sb00.26)2Te4 device measured at temperature, T = 2 K, at zero magnetic field. The extracted field effect mobility (\({\mu }_{{FE}}\)) as a function of backgate voltage is plotted in the secondary y-axis on the right. The inset in (d) is the \({\mu }_{{FE}}\) mapping as a function of dual-gate voltages taken at 2 K. The hole–hole (h–h) and electron–electron (e–e) labels in the inset denote the carriers’ conduction of the top-bottom surfaces as controlled by top-bottom gate voltages.


Mn(Bi1−xSbx)2Te4 thin film Weyl semimetals

According to density-functional-theory (DFT), the spin-aligned magnetic configuration of Mn(Bi1−xSbx)2Te4 is a simple type-I (or II depending on lattice parameters16) Weyl semimetal with Weyl points at \({\pm k}_{w}\) along the \(\Gamma -Z\) line. As shown in Fig. 1a, the distance between Weyl points \((2{k}_{w})\) decreases with Sb fraction x. The bulk Hall conductivity normalized per layer can be expressed as \(\frac{{\sigma }_{{xy}}^{3D}}{d}=\frac{{e}^{2}}{h}\frac{{k}_{w}d}{\pi }\), where \({k}_{w}\) is the position of Weyl point and \(\pi /d\) is the size of Brillouin zone along the \(\Gamma -Z\) line with d as the septuple layer (1SL ≈ 1.4 nm). The corresponding thin films can be viewed as quasi-2D crystals and are expected to have quantized anomalous Hall conductivities with Chern numbers (\(C\)) that increase by one when the film thickness increases by \(\triangle t \sim \pi /{k}_{w}\)18. For the case of Sb doping x ~ 25% as illustrated by the red curves in Fig. 1a, b, the position of the Weyl point (\({k}_{w}\, \approx \, 3\pi /25d\)) shifts closer to the \(\Gamma\) point than at x = 0 case, and thus corresponds to the expansion of the \(C\) = 1 state to larger film thicknesses. Figure 1b shows the theoretical thin film Chern gaps versus thickness obtained by fitting the bulk DFT bands to a simplified model18 as detailed in the Supplemental Note 1. The gaps close when a topological phase transition occurs between the different Chern numbers. As the Weyl semimetal state can exist only under the spin-aligned condition, our focus is thus on the Chern insulator states in the magnetic field induced spin-moment alignment phase, which can be observed more consistently in Mn(Bi1−xSbx)2Te4 thin films.

Electrical transport in Mn(Bi1−xSbx)2Te4 films

We first examine the low-temperature transport properties of the Mn(Bi1−xSbx)2Te4 films. Figure 1c plots the four-terminal resistivity (\({\uprho}_{{{{{\mathrm{xx}}}}}}^{{{{{\mathrm{CNP}}}}}}\)) measured at the CNP as a function of Mn(Bi1−xSbx)2Te4 thickness for different Sb substitution levels (x = 0, 0.20, and 0.26). We refer to a sample as surface-like when its conductivity is thermally-activated at low temperatures, suggesting the presence of a bulk energy gap. Samples classified as bulk-like have weaker temperature dependence and are presumed to have disorder-induced bulk states at all energies. The representative resistivity versus temperature curves for the Mn(Bi1−xSbx)2Te4 are included in Supplementary Figs. 5, 6 to distinguish the surface-like and bulk-like behaviors. The surface-like \({\uprho}_{{{{{\mathrm{xx}}}}}}^{{{{{\mathrm{CNP}}}}}}\) behavior persists to the largest thickness range for the Sb concentration x = 0.26 at T = 2 K while \({\uprho}_{{{{{\mathrm{xx}}}}}}^{{{{{\mathrm{CNP}}}}}}\) is bulk-like for thickness above 21-SLs. For x = 0.20, \({\uprho}_{{{{{\mathrm{xx}}}}}}^{{{{{\mathrm{CNP}}}}}}\) decreases abruptly at thicknesses above 12-SLs. This trend persists for MnBi2Te4. This trend is expected since MnBi2Te4 has bulk n-type doping5,24, while the substitution of Sb on the Bi sites can shift the Fermi level of the bulk band toward p-type doping, with x = 0.26 being closest to CNP20. Sb doping at higher concentrations x > 0.26 leads to excessive p-type doping and prevents access to CNP in thin flakes21. Moving the Fermi level by the Sb to Bi ratio can thus maximize the surface-like regime for probing their quantum transport properties.

Figure 1d shows a representative ρxx curve as a function of backgate voltage (Vbg) for an 18-SL Mn(Bi0.74Sb00.26)2Te4 device measured at a temperature of 2 K. The ambipolar gate-dependent ρxx suggests a bulk gap with an intrinsic surface state at the CNP (Vbg ~ +3 V). The surface carrier density can be tuned to either hole or electron transport by controlling the gate voltage. The field-effect mobility \({\mu }_{{{{{{{\rm{FE}}}}}}}}=\frac{1}{{C}_{g}}\frac{{{{{{\rm{d}}}}}}{G}_{{xx}}}{{{{{{\rm{d}}}}}}{V}_{g}}\), where \({C}_{{g}}\) is the gate capacitance (≈80 nF/cm2 for a ~ 30 nm mica dielectric), Gxx ( = 1/ρxx) is the four-terminal conductance, and Vg is the voltage applied through the graphite gate-electrode, is plotted in Fig. 1d. We see that the electron mobility increases with gate voltage (electron density), while the hole mobility responds weakly and remains small in the low gate voltage (hole density) regime. The mobility is more than one order of magnitude higher mobility for electrons compared to hole carriers. By applying dual-gate voltages, the top and bottom surface carrier densities can be modulated to achieve electron mobilities as high as 4000 cm2/Vs at a total carrier density of >5 × 1011 cm2.

\({{{{{\boldsymbol{C}}}}}}\) = 1 state in Mn(Bi1−xSbx)2Te4 films

The magnetic field-dependent transport properties of Mn(Bi1−xSbx)2Te4 films with the Sb concentrations x = 0, 0.20, and 0.26 for a variety of thicknesses were studied. When a perpendicular magnetic field is applied, Mn(Bi1−xSbx)2Te4 undergoes the spin–flop and spin–flip transitions20,22 from the antiferromagnetic (AFM) to the canted, and finally to the aligned spin-moment configurations. To compare the magnetic field dependence of the Mn(Bi1−xSbx)2Te4 film with different Sb substitutions, we plot the color maps of ρyx as functions of gate voltage and magnetic field for the 8-SL MnBi2Te4, 10-SL Mn(Bi0.8Sb0.2)2Te4, and 21-SL Mn(Bi0.74Sb0.26)2Te4, respectively, in Fig. 2a–c. The spin-flop and spin-flip transition fields, as determined from the kinks in their ρxx and ρyx versus magnetic field curves, happen at magnetic fields of ~±2–3 T and ~±7 T, respectively, are observed in all the samples. The magnetic transition fields identified by the color line marks depicted in Fig. 2a–c agree with their parent bulk compounds20 and theoretically calculated values25. Line profiles of ρxx and ρyx curves for the 8-SL MnBi2Te4, 10-SL Mn(Bi0.8Sb0.2)2Te4, and 21-SL Mn(Bi0.74Sb0.26)2Te4 are plotted in Fig. 2d–f. The ρyx increases sharply with magnetic field in the canted antiferromagnetic (CAFM) phase, and saturates at ~h/e2 as the thin film is driven into the FM phase by magnetic field. The suppression of ρxx in the FM phase further confirms the development of \(C\) = 1 state in all three samples. The interpretation is supported by the gate-dependent ρxx and ρyx curves measured at magnetic field of 9 T for the three respective samples in Supplementary Figs. 7a, 8, and 10, respectively, where the ρyx plateau and ρxx minimum can be seen. Despite not being fully quantized, the 21-SL Mn(Bi0.74Sb0.26)2Te4 film exhibits all the features of the \(C\) = 1 state. The anomalous Hall loop at low magnetic field for the 8-SL MnBi2Te4 and 10-SL Mn(Bi0.8Sb0.2)2Te4 could be due to the uncompensated surface magnetization11 or antiferromagnetic domain walls26, whereas no zero field hysteretic behavior observed in the 21-SL Mn(Bi0.74Sb0.26)2Te4 film. The observed \(C\) = 1 states in all three Mn(Bi1−xSbx)2Te4 films with the different Sb substitutions confirm that their spin-alignment configuration is in the topological phase with a Chern insulator gap.

Fig. 2: Magnetic field induced quantization at different Sb substitution levels.
figure 2

Color maps of ρyx as a function of magnetic field and gate voltage for (a) MnBi2Te4, (b) Mn(Bi0.8Sb0.2)2Te4, and (c) Mn(Bi0.74Sb0.26)2Te4 devices at flake thickness of 8-SLs, 10-SLs, and 21-SLs, respectively, measured at temperature of 2 K. The red and blue lines mark the spin–flop and spin–flip transition fields, respectively, for each sample. The AFM, CAFM, and FM denote the antiferromagnetic, canted antiferromagnetic, and ferromagnetic phases, respectively. The ρxx and ρyx line profiles in (d), (e), and (f) as a function of magnetic field are extracted from the color maps for the respective devices in (a), (b), and (c) at the gate voltages indicated by the blue arrows. The respective device images are inserted in (df). The color maps present the raw data without antisymmetrization. The ρxx and ρyx line profiles in (d)–(f) are symmetrized and antisymmetrized, respectively, with respect to the magnetic field. The high resistive ρyx (truncated region) at low magnetic field in (a)–(c) is due to the mixing from the magnetoresistance. The extracted ρxx (black rhombus) and ρyx (blue circle) values at the maximum ρyx as a function of flake thickness for (g) MnBi2Te4, (h) Mn(Bi0.8Sb0.2)2Te4, and (i) Mn(Bi0.74Sb0.26)2Te4 devices measured at magnetic field of 9 T. The blue color shades in (g)–(i) denote the thickness range where the \(C\) = 1 state is observed in the spin-aligned state.

To further evaluate the thickness-dependence of the \(C\) = 1 state, we plot ρxx and ρyx values as a function of thickness, with the gate voltage tuned to the ρyx maximum for each thickness, at magnetic field of 9 T. Our primary observation is that, despite their similar field-dependent quantization behavior, the \(C\) = 1 state prolongs to the larger thickness range with the Sb substitutions as shown in Fig. 2g–i. This trend is consistent with our DFT calculations, which indicate the shift of the Weyl point position by Sb doping. In Fig. 2g, the observation of \(C\) = 1 quantum Hall states up to 8-SL MnBi2Te4 agrees with our calculations (Fig. 1b) and the literature10. The thickness limit for the \(C\) = 1 state extends to 11-SLs for the Mn(Bi0.8Sb0.2)2Te4 as shown in Fig. 2h. The 16-SL Mn(Bi0.8Sb0.2)2Te4 (Supplementary Fig. 9) shows the absence of \(C\) = 1 state at magnetic field up to 18 T, presumably due to the excessive bulk conduction channels in this sample. In Fig. 2i, we show the substantially wider thickness range of the \(C\) = 1 state resolved for the Mn(Bi0.74Sb0.26)2Te4. Although this can be somehow related to the extension of the surface-like regime in thin films by the optimal Sb substitutions, according to our calculations in Fig. 1b, the Mn(Bi0.74Sb0.26)2Te4 films at the given film thickness range should lie in the higher Chern number states.

Dual-gate tuning of Chern states

To further identify the Chern insulator states in our thicker Mn(Bi0.74Sb0.26)2Te4, we perform a detailed analysis for these devices in a dual-gating platform. Figure 3a–f compare the dual-gate maps of longitudinal conductivity (σxx) and Hall resistivity (ρyx) for the Mn(Bi0.74Sb00.26)2Te4 devices at the thickness of 21, 18, and 14-SLs, respectively, measured at 9 T. As shown in Fig. 3a and b, the 21-SL Mn(Bi0.74Sb00.26)2Te4 reveal a clear \(C\) = 1 plateau in the dual-gate maps. This is also indicated by the σxyxx) versus Vbg line profiles, as shown in Fig. 3g. In this device, we observe no other Chern states develop near the CNP besides the \(C\) = 1 state at the highest accessible magnetic field of 9 T. Careful tracking of the \(C\) = 1 state in magnetic field reveals its formation at Fermi energy slightly below the CNP as detailed in Supplementary Figs. 10 and 11. Similar behavior was also observed in the 18-SL Mn(Bi0.74Sb00.26)2Te4 (Supplementary Figs. 12, 13).

Fig. 3: Tunable Chern states by dual-gating.
figure 3

Color maps of σxx and ρyx as a function of dual-gate voltages for Mn(Bi0.74Sb00.26)2Te4 at flake thicknesses of (a, b) 21-SLs, (c, d) 18-SLs, and (e, f) 14-SLs, respectively, measured at temperature of 2 K and magnetic field of 9 T. The white dashed lines in the color maps trace the boundaries of the quantization plateaus with the respective quantum states indexed in the ρyx maps. Line profiles of σxx and σxy versus backgate voltage curves swept across the charge neutrality as indicated by the black arrows in the color maps for the (g) 21-SLs, (h) 18-SLs, and (i) 14-SLs Mn(Bi0.74Sb00.26)2Te4. The σxy plateaus and the corresponding σxx minima (indicated by the black arrows) are indexed to the Chern numbers in (g)–(i). Vertical dashed lines in (g)–(i) mark the backgate voltages corresponding to the \({\uprho}_{{{{{\mathrm{xx}}}}}}^{{{{{\mathrm{CNP}}}}}}\) determined at zero magnetic field. The surface band structures in (g)–(i) are sketched to illustrate the LL spectra observed for the respective thicknesses at 9 T. The red shades in (g)–(i) denote the Chern states forming near the CNP.

The 18-SL Mn(Bi0.74Sb00.26)2Te4 device with higher electron mobility (Fig. 1d) shows more quantization features at 9 T. With the dual-gating structure, we can access these quantum Hall states by tracing the boundaries of these states with dashed lines in the dual-gate maps. The color map of σxx versus dual-gate voltages depicted in Fig. 3c shows multiple minima corresponding to the different quantized ρyx plateaus as indexed in the dual-gate map in Fig. 3d, showing the different Chern numbers. The σxyxx) versus Vbg line profiles depicted in Fig. 3h reveal the well-developed \(C\) = 3 plateau and the developing \(C\) = 1 and 5 quantization states. Tracking the Chern state development at the lower magnetic field reveals an additional \(C\) = 2 state in the CAFM phase (Supplementary Fig. 14). The existence of the \(C\) = 2 state in the CAFM phase is also verified by the ρxx and ρyx dual-gate maps and the flow diagram in (σxx, σxy) parameter space swept at magnetic field of 6 T as shown in Supplementary Figs. 12, 14, respectively. Also, we note that the \(C\) = 2 state coincides with the \({\uprho}_{{{{{\mathrm{xx}}}}}}^{{{{{\mathrm{CNP}}}}}}\) at the zero magnetic field.

We further analyze the quantization states in the 14-SL Mn(Bi0.74Sb00.26)2Te4. In addition to the \(C\) = 1 state, the dual-gate maps in Fig. 3e, f show a series of oscillatory σxx minima and ρyx plateaus, respectively, corresponding to the different quantum Hall states develop at 9 T when tuning the dual-gate voltages. The linecuts of σxx and σxy as a function of backgate voltage at magnetic field of 9 T, as depicted in Fig. 3i, reveal the quantum Hall plateaus with \(C\) = 0, 1, 3, 5, etc. The color maps of ρxx and ρyx as functions of magnetic field and backgate voltage (Supplementary Fig. 15) resolve the fan diagram of Landau levels at odd integer fillings, where the ρxx minima for each filling factor can be traced down linearly to a gate voltage at zero magnetic field corresponding to the CNP. Such a feature is similar to the surface states’ LL fan diagram in non-magnetic topological insulators27,28. Different from the 18-SL and 21-SL Mn(Bi0.74Sb00.26)2Te4, the \({\uprho}_{{{{{\mathrm{xx}}}}}}^{{{{{\mathrm{CNP}}}}}}\) obtained at zero magnetic field sits between \(C\) = 0 and 1 states at high magnetic field for the 14-SL Mn(Bi0.74Sb00.26)2Te4 (Supplementary Fig. 16). The schematic diagrams inserted in Fig. 3g–i illustrates the surface band structures with different Chern states resolved in the respective thicknesses of the Mn(Bi0.74Sb00.26)2Te4 films.

Fig. 4: Spectra of anomalous and non-anomalous Landau levels.
figure 4

Landau level fan diagrams and filling factors of an 18-SL Mn(Bi0.75Sb0.25)2Te4 film. a The Landau level structures for the case of a non-magnetic thin film (\({J}_{S}\) = 0). The n ≠ 0 non-anomalous Landau levels are plotted with black curves, while the n = 0 anomalous Landau levels whose energies are independent of magnetic field are plotted with blue and red curves. The red curves distinguish anomalous Landau levels that are localized at the surface. b Band energies at 2D wavevector \({{{{{\bf{k}}}}}}=0\) at zero magnetic field versus the same-layer exchange splitting, \({J}_{S}\). Spin up (down) states distributed in the bulk are labeled with orange (blue) color, and the bold purple (red) curve is for the spin up (down) state localized at the thin film surfaces. The Chern numbers of the anomalous LLs for down (up)-spins are indexed in (b). c, d Landau level structures for thin film with an aligned moment spin configuration at \({J}_{S}\) = 30 and 34 meV, respectively, labeled in (b) with black vertical dashed lines. Strong quantum Hall states occur when only anomalous Landau levels are close the Fermi level. The orange arrows in (c) and (d) denote the region where the anomalous Landau levels developed at high magnetic fields.

Anomalous Landau levels in magnetic topological insulators

To interpret the rich Chern insulator states observed in our Mn(Bi0.74Sb0.26)2Te4 films, we calculate their LLs spectra using a simplified model18, in which the 2D massive Dirac cones are coupled by tunneling within and between the compound’s septuple-layer building blocks. For explanation purposes, we first present in Fig. 4a the LLs spectrum for a non-magnetic TI thin film29 which has no exchange coupling. The degenerate n = 0 LLs (red curves) lie at the Fermi level due to the non-trivial Berry’s phase30 in addition to the typical n ≠ 0 bulk LLs (black curves). When the exchange coupling is introduced, the quasi-2D films have discrete finite-length-chain hopping eigenstates, which have no spin-orbit coupling but are spin-split by exchange interactions with the aligned Mn spins. A series of spin-polarized anomalous LLs (red and blue lines) with magnetic field-independent energies emerge from the \({\vec{k}}_{\perp }=0\) states, as shown in Fig. 4c, d. The derivation of anomalous LLs bands from a generalized Su–Schrieffer–Heeger (SSH) model (Supplementary Note 1) and their wavefunction distributions can be found in Supplementary Fig. 1. The Chern numbers of the anomalous LLs are determined by \(({N}_{E < {E}_{F}}-{N}_{E > {E}_{F}})/2\) where \({N}_{E < {E}_{F}}\) (\({N}_{E > {E}_{F}}\)) is the total number of subbands below (above) Fermi level. Noted that for convenience in calculations, we ignore the magnetic transitions and assume the spin-aligned phase for all magnetic field ranges. This assumption enables the determination of the actual \(C\) due to the band topology in zero magnetic field and the development of anomalous LLs under high magnetic field. For example, the \(C\) = 2 can be identified from the LLs spectra in Fig. 4c, d by tracing down to the zero magnetic field limit. The \(C\) = 2 can be further verified by the \(C\) calculated from the bulk Hall conductivity at the same layer thickness of the Mn(Bi0.75Sb0.25)2Te4 films (Fig. 1b). Under high magnetic field, the non-anomalous LLs with n ≠ 0 indices (black curves in Fig. 4c, d) in conduction-valence pairs move further away from the Fermi level as magnetic fields strengthen. This leaves an interval of carrier density in which only the n = 0 anomalous LLs (red and blue lines in Fig. 4c, d) are present close to the Fermi level as indicated by the yellow arrows, and the gaps between these levels are large enough to support Hall quantization that is robust against disorder. The energy gaps of the anomalous LLs depend on the n ≠ 0 non-anomalous LLs and thus are magnetic field dependence.

A general picture of the relationship between the Chern numbers and exchange coupling (\({J}_{S}\)) is illustrated in Fig. 4b where the surface and bulk spin-splitting bands are plotted. In the case of \({J}_{S}=0\), the number of subbands above and below the Fermi level equals (\({N}_{E < {E}_{F}}={N}_{E > {E}_{F}}\)), and thus the Chern number is 0. When the exchange field is turned on (\({J}_{S}\, \ne \, 0\)), the anomalous LLs are spin-polarized with two nearly degenerate surface anomalous LLs labeled represented by the green and red curves in Fig. 4b. The filling factor at the Fermi level becomes 1 (−1) for down-spin (up-spin) as the \({N}_{E < {E}_{F}}\) and \({N}_{E > {E}_{F}}\) are now differed by one. When the crossings between up-spin (yellow curves) and down-spin (blue curves) happen at the Fermi level by further increasing \({J}_{S}\), the filling factor will increase by one. Subsequently, the Chern quantization is therefore expected to be observable over a range of filling factors magnitudes centered on 1+ the number of \({\vec{k}}_{\perp }=0\) crossings that occur between up and down spins as indexed in Fig. 4b. When the magnetic field is reversed, both the spin of the anomalous LLs and the sign of the exchange coupling between the local moments and the Dirac electrons are reversed, therefore the sign of the filling factor will follow the sign of the magnetic field. The dependency between the spin of the anomalous LLs and the sign of the exchange coupling further infers that the anomalous LLs are spin-polarized.

We also note that since the position of Weyl points \({k}_{w}\) depends on the exchange splitting18, which may be reduced by antisite defects31 leading to a change in the Chern numbers and the gaps in zero magnetic field spin alignment phase (Supplementary Fig. 2). To further elaborate on this effect, we examine two cases of \({J}_{S}\) = 30 and 34 meV for the 18-SL Mn(Bi0.75Sb0.25)2Te4 film in Fig. 4c, d, respectively. As shown in their LL spectra, the \({J}_{S}\) = 30 and 34 meV exhibit very different \(C\) = 2 Chern insulator gaps at zero magnetic field. Nevertheless, the well-spaced anomalous LLs dominate over a finite region of filling factor at high magnetic field does not vary too much with the \({J}_{S}\).

Finally, we compare our experimental results for the Mn(Bi0.74Sb0.26)2Te4 films with the calculated LLs structure at a similar thickness and Sb doping level. We plot in Supplementary Fig. 17 the calculated LL gaps at different filling factors as a function of magnetic field for the Mn(Bi0.75Sb0.25)2Te4 films at different thicknesses. The LL gap size of each filling factor is determined either by gaps between anomalous LLs or by gaps between anomalous (n = 0) and non-anomalous (n ≠ 0) LLs. Under a strong magnetic field, the calculated filling factor of \(C\) = 1 exhibits the largest LL gap in all three thicknesses of Mn(Bi0.75Sb0.25)2Te4 film. This explains the experimental observation where the \(C\) = 1 state was observed over the wide thickness range in our Mn(Bi0.74Sb00.26)2Te4 film. However, one noticeable feature in the case of larger film thickness is that the \(C\) = 1 filling shifts below the Fermi level and the surface gap at the Fermi level change to the filling factor of higher Chern number \(C\) = 2 (refer to Supplementary Figs. 4 and 17 for details). This is consistent with the observation in our 18-SL Mn(Bi0.74Sb00.26)2Te4 film, indicating that the 18-SL is in a higher Chern number \(C\) = 2 state. Moreover, the calculated gap size for the \(C\) = 3 state exceeds the \(C\) = 2 gap at high magnetic field, which explains the well-developed \(C\) = 3 state at 9 T at this film thickness.

Whereas for 14-SL Mn(Bi0.74Sb00.26)2Te4 film, both of our experimental observation and calculations suggest that the 14-SL Mn(Bi0.74Sb00.26)2Te4 film in the \(C\) = 1 state (Supplementary Figs. 3 and 17) with the higher filling factors of \(C\) = 3 and 5 states assigned to the non-anomalous band LLs. The calculations infer an anomalous LLs \(C\) = 2 state can develop at the higher magnetic field. To verify this, we performed measurements at a high magnetic field up to 18 T for the 11-SL and 10-SL Mn(Bi0.8Sb00.2)2Te4 films, both with the \(C\) = 1 state at the Fermi level, as depicted in Supplementary Fig. 18. Similar to the 14-SL Mn(Bi0.74Sb00.26)2Te4 film, the quantization steps of \(C\) = 0, 1, and 3 states can be observed in the 11-SL and 10-SL Mn(Bi0.8Sb0.2)2Te4 at a magnetic field of 10 T. While the additional \(C\) = 2 plateau starts to develop at higher magnetic field of >14 T in both samples. The phase boundaries traced by the red dashed lines in Supplementary Fig. 18(b) and (e) covering the \(C\) = 1 and 2 plateaus are consistent with our theoretical picture of the Chern insulator \(C\) = 1 state and anomalous LLs \(C\) = 2 forming at high magnetic field as the ordinary n ≠ 0 bulk band LLs of \(C\) = 3 and above states move away from Fermi level. Our results can thus support the existence of the spin-polarized anomalous LLs in Mn(Bi1−xSbx)2Te4 thin film Weyl semimetals, which emerges at larger film thickness or strong magnetic field near the Fermi level.


In summary, we studied the magnetoelectrical transport of the intrinsic MTI Mn(Bi1−xSbx)2Te4 for x = 0, 0.20, and 0.26 by probing Chern quantization states and their relationship with the flake thickness and Sb concentrations. We identified the thickness ranges for surface-like insulating and bulk-like metallic transport regimes. The thickness-dependent Hall conductivities, particularly for the \(C\) = 1 Chern insulator state in the Mn(Bi1−xSbx)2Te4, show a correlation with the separation of Weyl points as described by our theoretical models, indicating that Mn(Bi1−xSbx)2Te4 behaves as thin film Weyl semimetals. Our transport results for different Sb concentrations highlight the importance of the Weyl point separation in the spin-aligned magnetic phase to the Hall quantization. Moreover, we showed that the Mn(Bi1−xSbx)2Te4 at larger film thickness and strong magnetic field can give rise to the unusual quantization sequence and the intriguing anomalous LLs near the Fermi level. Our work illustrates the complexity of the intertwined topological surface states and ferromagnetism in Landau quantization and thus can serve as a guide to bridge the gap between the 2D Chern insulators and 3D Weyl semimetals.



Mn(Bi1−xSbx)2Te4 bulk crystals at different Sb doping levels were grown by a self-flux growth method20,24. Variable thicknesses of Mn(Bi1−xSbx)2Te4 thin flakes were exfoliated from the parent bulk crystals and then transferred into the heterostructures of graphite/muscovite mica sandwiched layers using a micromanipulator transfer stage. The graphite and muscovite mica layers were subsequently transferred onto Si/SiO2 substrate using polypropylene carbonate, and followed by an annealing process in argon gas to clean the polymer residues. The graphite/muscovite mica layers serve as the gate-electrode/dielectric layers. The Mn(Bi1−xSbx)2Te4 exfoliation and transfer processes were performed in an argon gas-filled glovebox with O2 and H2O levels <1 ppm and <0.1 ppm, respectively, to prevent oxidation in thin flakes. We fabricated the Mn(Bi1−xSbx)2Te4 devices into the Hall bar configuration using a standard electron beam lithography process and metal deposition with Cr/Au (20 nm/60 nm) as the contact electrodes using a CHA Solution electron beam evaporator. The Mn(Bi1−xSbx)2Te4 flakes were protected by polymethyl methacrylate (PMMA) while transporting for lithography and metal deposition processes.


Low-temperature magnetotransport measurements were performed in a Quantum Design Physical Properties Measurement System (PPMS) in helium-4 circulation (2–300K) and magnetic field up to 9 T. Two synchronized Stanford Research SR830 lock-in amplifiers at a frequency of 5–8 Hz were used to measure the longitudinal and Hall resistances concurrently on the Mn(Bi1−xSbx)2Te4 devices. The devices were typically sourced with a small AC excitation current of 20–100 nA. Two Keithley 2400 source meters were utilized to source DC gate voltages separately to the top and bottom gate electrodes. Magnetotransport measurements at high magnetic field were carried out in a helium-3 variable temperature insert at a base temperature of 0.4 K and magnetic field up to 18 T based at the National High Magnetic Field laboratory.

Theoretical calculations

DFT calculations were performed using Vienna Ab initio Simulation Package (VASP)32 in which Generalized gradient approximations (GGA) of Perdew-Burke-Ernzerhof (PBE)33 have been adopted for exchange-correlation potential. On-site correlation on the Mn-3d states is treated by performing DFT + U calculations34 with U–J as 5.34 eV. The global break condition for the electronic SC-loop is set to be 10−7 eV and the cutoff energy for the plane wave basis set is 600 eV during the self-consistent (SC) calculations. A 9 × 9 × 6 Gamma-centered k-point integration grid was employed with Gaussian broadening factors as 50 meV. In the calculations of bulk Mn(Bi1−xSbx)2Te4, supercells of 2 × 2 × 1 unit cell were used to model the doping density of Sb atoms with densities of 0, 25%, 50%, 75%, and 100%. The calculations of Landau levels are based on the coupled Dirac cone model illustrated in the supplemental material, with the parameters estimated from the DFT calculations.