Introduction

Magnetic topological materials have emerged as a frontier in condensed matter physics, providing promising platforms for exploring exotic quantum phenomena and applications in topological spintronics1,2,3. For uncovering novel topological physics, successful fabrication of high-quality devices with quantized transport is prerequisite. As the first identified material possessing van der Waals characteristics, intrinsic magnetism, and nontrivial band topology simultaneously, MnBi2Te4 not only exhibits rich novel quantized phenomena when exfoliated down to few-layer limit4,5,6, but is also considered capable of addressing the disorder issue that is prevalent in magnetically doped topological insulators (TIs)7. The bulk crystal of MnBi2Te4 can be regarded as a stacking of Te–Bi–Te–Mn–Te–Bi–Te septuple layer (SL) along z-direction (Fig. 1a). A-type antiferromagnetic (AFM) structure with interlayer AFM order and intralayer ferromagnetic (FM) order forms below the Néel temperature (TN) ~ 25 K. When interacting with band topology, this layer-dependent magnetic ordering can give rise to a rich variety of topological quantum states and exotic magnetoelectric response8,9,10,11,12,13,14,15. In odd-SL MnBi2Te4 film, the gapped Dirac topological surface states due to the parallel surface magnetizations drive the system into the quantum anomalous Hall (QAH) state with 1D dissipationless chiral edge state transport8 (Fig. 1b). This manifests as a quantized Hall conductivity σxy = Ce2/h at zero-magnetic field (μ0H = 0), where C represents the Chern number, e is the electron charge, and h is the Planck constant. In even-SL MnBi2Te4, the opposite surface magnetizations result in vanishing σxy and lead to the axion insulator state characterized by a zero plateau at μ0H (ref. 10). Recent progresses in MnBi2Te4 have unveiled a plethora of novel topological phenomena, including the Möbius insulator16, layer Hall effect11, axion optical induction17, and quantum metric nonlinear transport18,19.

Fig. 1: Fabrication and optical contrast characterizations of few-layer MnBi2Te4 flakes.
figure 1

a Crystal structure of MnBi2Te4. b Schematic of the QAH effect in an odd-SL MnBi2Te4. The red arrows in a and b represent the magnetic moments of Mn in each layer. c Illustration of the device fabrication process. This method is developed based on the standard EBL process. By simply depositing a thin layer of AlOx on the MnBi2Te4 surface, the PMMA resist is isolated from the top surface. The high insulation and compactness of AlOx make it possible to fabricate Hall bar patterns while protecting the sample from chemical reagents. d Front and side views of a transport device. e Optical images of MnBi2Te4 flakes exfoliated from the same crystal. The top (bottom) panel compares the color change of MnBi2Te4 flakes without (with) AlOx capping before and after contact with PMMA, respectively. f Variation of Oc for selected spots along the line traces in flakes of different thicknesses in (e). Oc is defined as (Iflake - Isubstrate)/Isubstrate, where Iflake and Isubstrate are the intensity of MnBi2Te4 and substrate, respectively. g Statistical analysis of Oc of 47 MnBi2Te4 flakes with (red) and without (blue) AlOx capping layer. Different data points represent the Oc of different flakes after mechanical exfoliation and after contact with PMMA, respectively. The red and blue dashed line denote the Oc reduction by 0 and 20%, respectively, corresponding to no change and a decrease of effective thickness by one layer. The variability of the degree of Oc reduction among different flakes does not stem from measurement errors but rather arises from the inhomogeneities and sample quality fluctuations within the bulk crystal.

Full size image

Although the QAH effect and axion insulator states have been observed in 5- and 6-SL MnBi2Te4, the temperature (T) below which quantization is realized remains much lower than its TN. A more formidable challenge arises from the exceptionally low yield of MnBi2Te4 film exhibiting quantized transport. Over the past five years, neither perfectly quantized9,12,13,15,20,21,22 nor zero plateau12,17,18,19,22 at μ0H = 0 can be consistently reproduced. The lack of quantization not only obstructs the discovery of new phenomena but also complicates the interpretation of available data. Possible reasons include various structural defects or impurity phases23,24,25,26,27, instability of surface electronic structures28,29,30,31, and weakened surface out-of-plane magnetic anisotropy (MA)31,32,33. Our recent studies combining optical contrast (Oc), transport, magneto-optical Kerr effect (MOKE) measurements revealed a substantial impact of fabrication on the properties of MnBi2Te4 devices22. The contact with photoresist not only reduces the Oc value during the fabrication process, but may also results in mismatched even-odd-layer-dependent magnetotransport22. The mechanism likely originates from the formation of a dead insulating layer on the MnBi2Te4 surface, which is caused by the change of surface band structure30,34,35. Over the past five years, developing a low-damage fabrication method to reproduce the QAH effect has become one of the most pressing tasks in the field of magnetic topological quantum materials and devices.

In this work, we optimize the fabrication process by depositing an AlOx protective layer on MnBi2Te4 top surface prior to the standard electron beam lithography (EBL). To ensure a good ohmic contact between the MnBi2Te4 and electrode, we implement an additional Ar ion etching step to selectively remove the AlOx at the electrode areas after the lithography process. By employing this method, we can fabricate transport devices with good electrical contact while using an insulating AlOx layer to fully isolate the photoresist Polymethyl Methacrylate (PMMA). Through optical measurement on a series of MnBi2Te4 thin flakes, we find that the fabrication issue induced by PMMA photoresist is largely mitigated. Most importantly, this simple idea overcomes the bottlenecks in the field of topological quantum materials over the past five years. We not only achieve the QAH effect in multiple MnBi2Te4 devices, but also reveal the key factors influencing the zero-magnetic-field quantization. Our work introduces a simple yet effective method for fabricating high-quality transport devices, paving the way for realizing the QAH effect and exploring more exotic topological quantum phenomena.

Results

Device fabrication and optical contrast

We got inspiration from previous experiments where those MnBi2Te4 devices exhibiting large anomalous Hall (AH) effect often had Al2O3 on the bottom of the flakes8,21,36, either as a substrate or a supporting layer. This implies that the contact with Al2O3 may help to improve the quality of MnBi2Te4 device. Combined with our recent finding of the detrimental effect of fabrication on the top surface of MnBi2Te4, we come up with a straightforward yet effective idea that by depositing an AlOx layer on top of MnBi2Te4 to achieve the QAH effect. Figure 1c shows the schematic of the fabrication process. First, we transferred thick MnBi2Te4 flakes from a bulk crystal to the substrate using a Scotch tape. We then employed the mechanical exfoliation method to obtain the flakes with target thickness. The one-to-one correspondence between Oc and thickness allows the rapid determination of SL number by optical method11. Subsequently, a 3-nm AlOx layer was deposited by thermal evaporation. We then adopted the standard EBL process to expose the designed Hall bar patterns. Next, the AlOx layer above the designed electrode areas was etched away by Ar ion etching, followed by the deposition of Cr/Au electrodes. To ensure the charge transport is not affected by the etching process in the contact region, we assessed the MnBi2Te4 thickness in the etched region by Oc and atomic force microscopy measurements (see Supplementary Figs. 1 and 2). Finally, a PMMA layer was coated for further protection. The details of the fabrication are described in the Methods section. Compared to the Al2O3-assisted exfoliation and stencil mask method8, our method is more straightforward and is based on the standard EBL process, which enables the fabrication of specific nano-devices with reduced sample size. Furthermore, since the AlOx is on the top surface, it not only offers an effective protection, but also can be used as a dielectric layer for top gate controllability.

Figure 1d depicts a schematic of a Hall bar device covered with AlOx capping layer and its cross-sectional view. To investigate the influence of AlOx on MnBi2Te4, we first compare the optical properties of MnBi2Te4 flakes with varied thicknesses, which were exfoliated from the same single crystal (Supplementary Fig. 3 for the Oc change in each step). The optical images in Fig. 1e clearly suggest that for flakes without AlOx (up panel), the colors of all four regions change significantly before and after contact with PMMA. In contrast, those regimes with AlOx (down panel) do not exhibit noticeable change during the same process. To further explore the effect of AlOx quantitatively, we extract their Oc values of the selected spots along the line traces in Fig. 1e and compare their variations directly. As shown in Fig. 1f, significant reductions of Oc in all the four areas without AlOx are observed. In contrast, Oc remains nearly unchanged for the five areas with AlOx capping layers. To eliminate the influences of device quality fluctuations on our observation, we compare the Oc values of 47 MnBi2Te4 exfoliated from the same crystal (Fig. 1g). Different data points represent the Oc of different flakes after exfoliation and after contact with PMMA. The distribution of Oc falls well into two parts (red and blue dots). According to the one-to-one correspondence between Oc and thickness11, the decrease of Oc indicates the reduction of effective thickness during fabrication. The red and blue dashed lines represent the Oc reduction by 0 and 20 %, respectively, which corresponds to unchanged thickness and a decrease of thickness by one SL. For the flakes without AlOx layer, regardless of their initial Oc values, most of the samples display a pronounced decrease in Oc after fabrication. Notably, the degree of Oc change exhibits certain variability among different samples. This behavior does not stem from measurement errors but rather originates from the inhomogeneities and sample quality fluctuations within the MnBi2Te4 crystal. Such sample-dependent sensitivity to fabrication has also been observed in previous experiments22. These results clearly demonstrate that AlOx can effectively mitigates the damages caused by PMMA22.

Statistical analysis of the influence of AlOx on transport properties

In magnetic topological systems, the AH effect typically results from three mechanisms: intrinsic Berry curvature Ω(k), skew-scattering, and side-jump. In the transport of MnBi2Te4, due to defects or impurity phases, all the three mechanisms could contribute to the AH effect20. However, in an ideal quantized Hall system, the transverse transport should be dominated by Ω(k) in momentum (k) space37. Theoretically, σxy can be calculated by integrating Ω over k, as expressed by:

$${\sigma }_{{xy}}=-\frac{{e}^{2}}{2\pi h}\int \Omega \left({{\bf{k}}}\right){d}^{2}{{\bf{k}}}$$

When the Fermi level (EF) is tuned into the magnetic exchange gap, the integral of Ω equals the C number multiplied by 2π, resulting in the quantization of σxy at e2/h. To investigate the influence of AlOx on the intrinsic AH effect, we measured the transport properties of 17 odd-SL MnBi2Te4 devices. All the data shown in the main text was obtained at the charge neutral point (CNP) unless otherwise specified. Prior to this, we measured the current–voltage curve of the AlOx layer to exclude its contribution to transport (Supplementary Fig. 4). Figure 2a–d shows the μ0H dependence of σxy and σxx for two 7-SL devices exfoliated from the same thick flake on the same tape. Both devices show quantized σxy at the high μ0H Chern insulator state (C = −1). However, their AH effects at zero field exhibit dramatically different behaviors. For device #1 without AlOx (Fig. 2a), σxy almost vanishes at the low μ0H AFM regime. Such behavior is consistent with our previous observation that fabrication process can damage the top surface, leading to a reduction of effective thickness by one SL22. However, for the device #9 fabricated by the current method, a large σxy accompanied with a square-shaped hysteresis is observed. The insets show the schematic distribution of the topological surface states wave functions (green) for devices without and with AlOx. To further demonstrate the influence of AlOx on transport more clearly, we display the gate voltage (Vg) dependent σxy and σxx at μ0H = −8 T and 0 for the two devices, as displayed in Fig. 2b, d. For device #1, σxy smoothly crosses zero with increasing Vg, indicating a reduced gap size during the fabrication process22. In contrast, device #9 shows a wide plateau in σxy in the same Vg range of the Chern insulator, indicating an incipient QAH state in the AFM state.

Fig. 2: Comparison of transport behaviors for devices obtained by different fabrication methods.
figure 2

a Transport behaviors at CNP for a 7-SL device without AlOx covering layer. Due to fabrication effects, the surface state shifts down to the second SL (inset), and the hysteresis of σxy near zero field almost disappears. b Variation of σxy and σxx as a function of Vg at μ0H = −8 T and 0, respectively. The black dashed line represents the Vg window where σxy plateaus coexist under zero and high μ0H conditions, respectively. c Transport behavior of a 7-SL device exfoliated from a MnBi2Te4 flake on the same tape, but with an AlOx layer deposited during the fabrication process. The large hysteresis indicates excellent protection of device performance. The inset illustrates that the topological surface state remains predominantly distributed on the outermost surface due to the protection of AlOx. d In the same Vg range of high field Chern insulator state (marked by black dashed line), σxy at μ0H = 0 exhibits a broad plateau during sweeping Vg. e, f μ0H-dependent σxy at T = 1.5 K for 17 odd-SL devices. The only difference during their fabrication lies in whether the surface was deposited with AlOx. All the devices exhibit quantized σxy at high μ0H, as marked by the black dashed lines. g Summarized σxy at μ0H = 0 and ∆σxy/∆H values at the plateau transition for the 17 devices. Devices with AlOx capping layer (blue) generally show a larger AH effect than those without AlOx (red). The inset displays the histograms of AH effect distribution of the 17 devices. The size of each bin on the σxy-axis is 0.1 e2/h.

Full size image

For 2D materials, the transport properties of thin flakes are inevitably influenced by the fluctuations of device qualities. Previous experiments have suggested that MnBi2Te4 exhibits sample-dependent properties, even for flakes prepared from the same crystal8,24,25. Therefore, it is challenging to expect any fabrication method to guarantee the production of perfect QAH devices with 100 % certainty. To demonstrate that the enhancement of the AH effect arises from AlOx, we studied the transport properties of 17 odd-number-SL MnBi2Te4 devices, with the data presented in Fig. 2e, f. These devices were numbered based on the increasing order of their σxy at μ0H = 0. Apart from devices #2, #4, #6, and #8 that were acquired from crystal #1, all the other 13 devices were obtained from crystal #2. All the devices exhibit σxy = e2/h at the high μ0H Chern insulator state, indicating the overall high quality of our devices. However, these devices show dramatically different behaviors in their AH effect. Remarkably, the 9 devices without AlOx layer exhibit small σxy at μ0H = 0. And some devices even exhibit almost indiscernible hysteresis. In sharp contrast, all the other 8 devices with AlOx capping layer manifest large σxy and square-shaped hysteresis, with two devices almost quantized at e2/h. Figure 2g summarizes the σxy at μ0H = 0 of these devices. To largely avoid any artificial trend, we adopted the strategy in previous statistical studies of MnBi2Te4 crystals25 by sorting the σxy from smallest to largest. In the inset of Fig. 2g, we also present the histograms of their σxy distribution. The size of each bin on the σxy-axis is 0.1 e2/h. For the devices without AlOx, their σxy values are distributed within the range of 0 to 0.3 e2/h. In contrast, for those devices with AlOx, their σxy distribution displays a clear shift towards higher values. Notably, despite not all devices with AlOx exhibit the QAH effect, their AH effects have already surpassed the values for most MnBi2Te4 devices in literatures9,11,12,13,20,22. In Fig. 2g, we also summarized the values of ∆σxy/∆H of the 17 devices (blue points), which represent the sharpness of magnetic transition. The variation of ∆σxy/∆H aligns with the trend of σxy. These results undoubtedly demonstrate that AlOx plays a significant role in enhancing the AH effect.

The fabrication of high-quality devices enables us to compare the influences of magnetic properties on transport. Figure 3a–c shows the μ0H-dependent σxx and σxy for three devices obtained from the same crystal. Fortunately, for devices #11 and #16, they were obtained on the same substrate during one cleaving process. It enables us to further explore the influences of AlOx on MnBi2Te4 flake while preserving the consistency of the devices. Overall, the three devices manifest consistent transport behaviors, with their main differences being the values of σxx and σxy at μ0H = 0. However, the sharpness of the plateau transition, which reflects the magnetic flipping process, differs dramatically. For device #11, the σxy and σxx at μ0H = 0 are 0.5 and 1.1 e2/h, respectively, and the transition is relatively gentle. For device #16, although the value of σxx does not change, σxy is significantly improved to 0.96 e2/h, comparable to the value in previous report8. In addition, the plateau transition is also sharper than that of device #11. Device #17 completely enters the QAH state, with σxy reaching e2/h and σxx dropping to 0. Because the QAH effect in magnetic TIs originates from the exchange field between local moments and electron spins38. The out-of-plane magnetic order plays a crucial role in the μ0H-dependent transport behaviors. Therefore, the improved quantization along with the sharp σxy transition suggests that device #17 is likely to have a stronger out-of-plane MA.

Fig. 3: Scaling relation between σxy and σxx of the intrinsic AH effect.
figure 3

a–c μ0H-dependent σxy and σxx for three 7-SL devices exfoliated from the same crystal. Devices #11 and #16 were obtained simultaneously in one cleaving process on the same substrate, with the former one undergoing an extra AlOx deposition process, having a thickness of 33 nm. d–f Evolution of σxy with σxx during the cooling process. With the formation of AFM order as lowering Ts, the scaling relation between σxy and σxx at different Vgs gradually collapses into one single curve, and σxy saturates at e2/h. The σxx independent behavior reflects the typical Berry curvature-dominated mechanism of the AH effect. g Schematics of the distribution of Berry curvature. From top to bottom, as the AFM order is tuned to the FM order, the out-of-plane component of the total magnetization is enhanced. As a result, the exchange gap increases, and the Berry curvature exhibits greater robustness against thermal fluctuations.

Full size image

The scaling relation between σxy and σxx may further help us understand the role of AlOx in enhancing the QAH effect. Figure 3d–f displays the variation of σxy as a function of σxx during the cooling process under different μ0H and Vgs. As the AFM order strengthens at low Ts, σxy begins to exhibit behavior independent of σxx and gradually approaches quantization, which is of typical the scaling behavior of the intrinsic AH effect dominated by Ω(k) (ref. 37). Upon increasing μ0H, the device undergoes AFM, canted AFM, and finally enters FM state, accompanied by σxy saturating at e2/h at higher Ts. For device #11 with relatively weaker out-of-plane magnetic order, the exchange gap is expected to be smaller in the AFM state, thermal fluctuations could more easily smear out the role of Ω(k) (top in Fig. 3g). Therefore, complete quantization appears only when all moments are parallelly aligned because the gap is overall positively correlated with magnetization. However, for the device #17 with stronger out-of-plane order, the larger gap allows for σxy reaching quantization even in the AFM state despite a small net moment. The influence of different magnetic configurations on exchange gap and Ω(k) is illustrated in Fig. 3g. Ω(k) can be interpreted as an effective μ0H in k space acting on electrons. The red and blue represent the distribution of Ω(k) of opposite sign in conduction and valence band, respectively.

Gate voltage-independent magnetism

Next, we investigate the influence of Vg on the magnetic properties of MnBi2Te4 devices. Previous studies have revealed different Vg-dependent magnetism on magnetically doped TIs39,40,41. The electrical control of van der Waals magnetism has also attracted wide attention. As the first layered topological antiferromagnet, it remains unclear whether Vg can exert similar effects. Figure 4a, b displays the μ0H-dependent σxy and σxx for device #16 at various Ts. The hysteresis vanishes at around T = 21 K, accompanied by the disappearance of σxx peaks. To quantitatively investigate the changes in the AFM state, we extract the coercive field (Hc) values at different Vgs and plot them as a function of T with an offset of 0.25 T (Fig. 4c). The Hc dependence of T can be well described by the power law ~(1 − T/TN)β, where β represents the critical exponent. We notice that Vg has almost negligible effect on the AFM order. TN remains a constant at ~21.3 K and β maintains at ~0.52. Similar results were also observed in previous neutron diffraction on MnBi2Te4 bulk crystal and reflectance magneto-circular dichroism (RMCD) measurement on exfoliated thin flakes14,42. Our experiments further point that this critical behavior cannot be tuned by a bottom Vg. Figure 4e shows the colormap of σxy as a function of Vg and μ0H. It clearly shows that Hc is independent of Vg, further supporting the Vg-independent magnetism in odd-SL MnBi2Te4 device. Reproducible results obtained from another two 7-SL MnBi2Te4 with and without AlOx capping layer are documented in Supplementary Figs. 5 and 6.

Fig. 4: Transport and magnetic properties tuned by Vg.
figure 4

a, b μ0H-dependent σxy and σxx at the CNP for device #16 at various Ts. The hysteresis and the double peak in σxx disappear at around T = 21 K. The black triangles mark the position of Hc at different Ts. c Hc extracted from the field sweep data as a function of T at varied Vgs. The solid squares are the data points. The black lines are the data fittings in the form of ~(1 − T/TN)β. The red dashed lines represent the position of Hc = 0 for each curve. d Summarized TN and β from the fittings as a function of Vg. The blue and red dashed lines represent the average positions of the β and TN values for different Vgs. TN and β are found to be ~21.2 K and 0.52, respectively, both of which are independent of Vg. e Colormap of σxy in the parameter space of μ0H and Vg. The boundary between blue and orange region marks the Vg-independent Hc.

Full size image

Discussion

Finally, we discuss the possible mechanisms underlying the enhancement of AH effect. In our previous research, we found that the coating of PMMA during the EBL process reduces the Oc of MnBi2Te4, leading to a reduction of effective thickness22. AlOx serves as an effective barrier by isolating the surface from direct contact with the resist, thus providing a substantial protection for MnBi2Te4. However, in the history of 2D materials, the most widely used and effective capping layer for protection is h-BN, rather than AlOx. In fact, previous experiments have suggested that the oxidation process may alter the intrinsic properties of MnBi2Te film28, therefore, using an AlOx capping layer to protect MnBi2Te is unconventional (Supplementary Fig. 7 for the aging effect). Furthermore, beyond employing a protective layer, a shadow mask method can also be employed to avoid direct contact with PMMA resist. However, many experiments have demonstrated that even employing these methods13,18,20,21, the AH effect remains non-quantized. Interestingly, a comparison of recent transport experiments in MnBi2Te4 reveals that regardless of the different device preparation or electrode deposition methods, all MnBi2Te4 devices exhibiting pronounced AH effect have one surface in contact with Al2O3 (refs. 8,21,36). In these experiments, Al2O3 is positioned under MnBi2Te4 and does not provide any protection to the top surface. Hence, the simple protective role is insufficient to explain the close correlation between large AH effect and AlOx in current experiments. It naturally raises the question of whether AlOx may play an additional role beyond protection.

Our scaling relation studies imply that the enhancement of perpendicular magnetic order may be crucial for the QAH effect. Based on our experimental data and the results in previous studies, we discuss the potential additional roles that AlOx may play. A conceivable scenario is the electric field enhanced magnetism at the AlOx/MnBi2Te4 interface43,44. However, this scenario can be largely excluded because our Vg-dependent experiments demonstrate that Vg has a small influence on magnetism. Another possible scenario is the enhanced perpendicular magnetic anisotropy (PMA) by AlOx. In spintronics, many experiments have demonstrated that depositing amorphous oxide (such as AlOx, MgO, TaOx, HfOx) can substantially increase the interfacial PMA at the interface between oxides and magnetic materials45,46,47. Therefore, it is naturally expected that the AlOx layer strengthens the interfacial PMA of MnBi2Te4, which in turn enhances the AH effect. In fact, there has been theoretical calculations suggesting that the MA in monolayer MnBi2Te4 is weak due to the weak p-d hybridization between Mn and Te (ref. 48). Later, inelastic neutron scattering pointed that for MnBi2Te4 crystals, the MA is enhanced by the interlayer two-ion anisotropy32. However, due to the absence of neighboring layers, the MA of the surface is still weak. These results naturally explain why, in all current MnBi2Te4 experiments exhibiting a large AH effect, the device must have at least one surface in contact with AlOx.

To validate our conjectures, we conducted cryogenic magnetic force microscopy (MFM) measurement to directly visualize the magnetic properties across different regions of the same 7-SL MnBi2Te4. As anticipated, the region with AlOx manifests a much stronger magnetic signal compared to the region without AlOx (Supplementary Fig. 8). Moreover, to further explore the influence of AlOx on QAH effect, we compared the magnetic hysteresis loops of two fully quantized devices with single-sided and double-sided AlOx contacts. Interestingly, the device with both top and bottom surfaces in contact with AlOx shows a larger Hc (see Supplementary Fig. 9 for details). In magnetic materials, Hc is proportional to the strength of PMA49. The larger Hc in device with double-sided AlOx aligns with the finding that AlOx can enhance the interfacial PMA45,46,47. In addition to PMA, a recent calculation has also suggested that bringing MnBi2Te4 surface close to a polar insulator can modify the surface potential, which is helpful for the QAH effect30. As a polar insulator, Al2O3 may also play an additional role in enhancing the QAH effect. It is worth noting that although all current experiments support the scenario that AlOx likely contributes to magnetism, the microscopic mechanism remains inadequately understood owing to the challenges in directly measuring the interfacial magnetism. Further studies are required to elucidate the exact mechanisms.

In summary, we report the successful realization of the QAH effect in MnBi2Te4 devices capped with AlOx by employing a revised fabrication method based on the standard EBL. By simply depositing an AlOx layer on top of MnBi2Te4, we observe a substantial enhancement of the AH effect, ultimately achieving quantization. Our experiments resolve a longstanding challenge in the field of magnetic topological materials, paving the way for fabricating high-quality devices and investigating the intricate interplay between nontrivial topology and 2D magnetism. Recently, novel transport phenomena unavailable in previous QAH systems have already been observed in 7-SL MnBi2Te4 with current configuration, and the enhancement of surface magnetism by AlOx is considered crucial to explaining these new phenomena50. This simple yet effective method is not only significant for fundamental studies, but also lays the groundwork for creating novel topological spintronic devices3,43,47. It is important to note that the current exploration on utilizing AlOx to achieve the QAH effect is still in the initial stage. More refined control of the AlOx growth parameters45,46 and the interface of oxide/MnBi2Te4 would further optimize the QAH effect, which remains a promising topic for future studies.

Methods

Crystal growth

High-quality MnBi2Te4 single crystals were synthesized by directly mixing Bi2Te3 and MnTe with a ratio of 1:1 in a vacuum-sealed silica ampoule. For crystal #1, the mixture was first heated up to 700 °C, and then slowly cooled down to 591 °C, followed by a long period of annealing process. The phase and crystal structure were examined by X-ray diffraction on a PANalytical Empyrean diffractometer with Cu Kα radiation. For crystal #2, a small amount of Te was added to the mixture, with the ratio between Bi2Te3, MnTe, and Te modified to 1:1:0.2. The ampoule was then slowly heated to 900 °C and maintained at this temperature for 1 h. Subsequently, it was cooled down to 700 °C, holding for 1 hour and then gradually cooled to 585 °C and maintained for 12 days. After the annealing, the ampoule was quenched in water to avoid phase impurities. Apart from devices #2, #4, #6, and #8, all other 13 devices were obtained from crystal #2. The four devices have already represented the samples showing the largest AH conductivity in the devices prepared from crystal #1.

Device fabrication

MnBi2Te4 flakes were mechanically exfoliated onto 285 nm thick SiO2/Si substrate by using the Scotch tape method in an Ar-filled glovebox with O2 and H2O level lower than 0.1 ppm. Initially, the substrate was thoroughly cleaned with acetone, isopropanol, and deionized water. Then the surface of SiO2/Si was treated with air plasma at ~125 Pa for 3 min. The tape-covered substrate was heated up to 60 °C for 3 min to facilitate smooth exfoliation of the single crystals into flakes. Micrometer-sized thin flakes can be obtained by mechanically exfoliation on thick flakes for several times. The thickness was identified by optical contrast measurement in the glovebox immediately after exfoliation. After the target flakes were obtained, a 3-nm aluminum was deposited onto the surface using a thermal evaporator with a deposition rate 0.04 nm/s under a vacuum better than 4 × 10−4 Pa. Oxygen was then introduced into the chamber, and the aluminum layer was oxidized for 5 min at a pressure of 2 × 10−2 Pa. For device #11, an extra deposition process with longer time was employed to compare the influence of different AlOx parameter on transport. During this process, an additional 30-nm AlOx was deposited under a controlled oxygen environment at a pressure of 2 × 10−2 Pa.

To assess the effect of PMMA on the MnBi2Te4 samples, 270 nm thick PMMA was spin-coated onto the samples in an Ar-filled glovebox at a controlled speed of 4000 round/min. The samples were then heated at 60 °C for 7 min and left to stabilize in the glovebox for 24 h. Subsequently, the samples were than immersed in acetone for 20 min, rinsed with acetone followed by isopropanol, and their optical contrasts were measured immediately after the removal of PMMA. Standard EBL was employed on MnBi2Te4 samples to pattern the Hall bar structure. The oxidized aluminum was first etched from the sample surface using an Ar ion milling machine at a pressure of 2 × 10−4 Torr for 75 s. Cr/Au electrodes (3/50 nm) were then deposited using a thermal evaporator connected to a glovebox. Following this, the samples were again spin-coated with PMMA adopting the same parameters as before for further protection.

Transport measurement

Standard four probe transport measurements for devices #1 to #16 were carried out in a cryostat with the lowest T of ~1.5 K and an out-of-plane magnetic field up to ~8 T. The longitudinal and Hall voltages were acquired simultaneously via two lock-in amplifiers with an AC current (100 nA, 13 Hz) generated by a Keithley 6221 current source meter. For device #17, the transport was performed in a dilution refrigerator with AC current excitation of 10 nA at 13 Hz. To correct the geometrical misalignment, both the longitudinal and Hall signals were symmetrized and antisymmetrized with respect to the magnetic field. The back-gate voltage was applied by a Keithley 2400 source meter through the SiO2/Si substrate.

MFM measurement

Cryogenic MFM experiments were conducted in a commercial atomic force microscope (atto-AFM) equipped with commercial cantilevers (spring constant k ≈ 2.8 N/m and resonance frequency ≈75.8 kHz) in a closed-cycle helium cryostat. An out-of-plane magnetic field was applied using a superconducting magnet. MFM images were taken in a constant height mode with lift height of ~200 nm. The MFM signal, the change of cantilever resonance frequency, is proportional to the gradient of out-of-plane stray field. Electrostatic interaction was minimized by balancing the tip-surface potential difference.