Surface Electronic Structure Engineering of Manganese Bismuth Tellurides Guided by Micro‐Focused Angle‐Resolved Photoemission

Modification of the electronic structure of quantum matter by ad atom deposition allows for directed fundamental design of electronic and magnetic properties. This concept is utilized in the present study in order to tune the surface electronic structure of magnetic topological insulators based on MnBi2Te4. The topological bands of these systems are typically strongly electron‐doped and hybridized with a manifold of surface states that place the salient topological states out of reach of electron transport and practical applications. In this study, micro‐focused angle‐resolved photoemission spectroscopy (microARPES) provides direct access to the termination‐dependent dispersion of MnBi2Te4 and MnBi4Te7 during in situ deposition of rubidium atoms. The resulting band structure changes are found to be highly complex, encompassing coverage‐dependent ambipolar doping effects, removal of surface state hybridization, and the collapse of a surface state band gap. In addition, doping‐dependent band bending is found to give rise to tunable quantum well states. This wide range of observed electronic structure modifications can provide new ways to exploit the topological states and the rich surface electronic structures of manganese bismuth tellurides.


Introduction
Intrinsic magnetic topological insulators based on manganesebismuth tellurides (MBTs) have come to prominence MBTs tend to cleave with microscopic domains of different surface terminations that are characterized by extremely rich surface electronic structures. Their energy-and momentum-dependent dispersion relations have been resolved by angle-resolved photoemission spectroscopy (ARPES) with micro-focused laser systems operating with photon energies below 7 eV, enabling observation of hybridized surface bands and Rashba effects that are intertwined with topological states. [1,2,12,[23][24][25][26][27][28] The measurements also reveal that the as-grown materials are strongly electron-doped, most likely due to interstitial defects that lead to a deficit of Mn atoms. [29] The doping typically places the topological bands at binding energies in the range between 200-300 meV, making it practically challenging to achieve electron transport via the topological state. It is therefore highly desirable to further develop platforms for tuning the chemical potential and surface electronic structure of MBTs while being able to probe the effects of such modifications in situ by micro-focused ARPES (mi-croARPES).
Substitution at the Bi sites of MBT crystals with Sb has been demonstrated as a viable method for shifting the chemical potential over a range of 400 meV between the bulk valence band (BVB) and the bulk conduction band (BCB) while maintaining the salient magnetic properties of MnBi 2 Te 4 . [4] Applying a similar strategy for the m = 1 heterostructure, i.e. MnBi 4 Te 7 , leads to more complex behaviors, as the Sb substitution alters the magnetic properties by making the ferromagnetic ground state more favorable than the antiferromagnetic, which is interesting in its own merit for tuning and inducing Weyl fermions or axion insulators. [30][31][32][33][34] A downside of Sb substitution is that the doping cannot be monitored in situ but requires synthesis of multiple different crystals, leading to large sample variability between different experiments. [35] An alternative strategy that solves this problem is based on dosing alkali metals, such as Li, K, Rb, or Cs, on the material surface while collecting ARPES spectra, which has been demonstrated on a range of semimetals [36][37][38] and semiconductors. [39][40][41][42] In the simplest approximation, the adsorbed alkali metals provide electron doping at the surface of the materials with the effect of merely shifting the chemical potential. This picture has proven much too simple in a recent study of alkali-driven electronic structure modifications of MnBi 2 Te 4 bulk and thin films, where surprising hole doping effects were observed. [7,10] The impact of alkali adsorption on the various possible surface terminations of MBT heterostructure compounds has not been investigated and a universal picture of the effect of alkali doping on the MBT family of materials is lacking. Such doping-dependent ARPES studies are complicated by the fact that surface adsorption strongly affects the surface potential of the materials, which in turn impacts the ARPES photoemission intensity through photon energy-dependent matrix element effects. It is therefore beneficial to employ photonenergy tunable sources in the extreme ultraviolet range that are complementary to the low-energy laser-based ARPES methods. Moreover, it is crucial to apply a micro-focused beam for ARPES to permit spatially distinguishing different surface terminations.

Results and Discussion
This study presents a systematic approach for tuning the surface electronic structure of MBTs following the outline in Figure 1a.
In situ Rb deposition on the surfaces of bulk crystals of the  basic building blocks MnBi 2 Te 4 and Bi 2 Te 3 as well as the two  possible surface terminations of MnBi 4 Te 7 is tracked by ARPES  with microscale spatial resolution at the new SGM4 beamline  for micro-and nanoARPES at the ASTRID2 light source, Aarhus University. [43] Photons with tunable energy over a range of 14-150 eV are obtained from an undulator insertion device and selected using a spherical grating monochromator (SGM-type). An intermediate beam focus with a diameter on the order of 60 μm is achieved using two separate plane ellipse mirrors placed after the monochromator. The beam is then demagnified on a capillary optic [Sigray Inc.], which is placed approximately 7.8 mm from the sample surface. The capillary produces a minimum lateral spotsize with a diameter of 3.1 μm on the sample, as determined from knife edge measurements. The optic is achromatic and highly transmissive as it operates essentially as an elliptical mirror, enabling photon energy-tunable microARPES over the full energy range of the beamline. Photoemitted electrons from the microscopic spot on the sample are collected using a SPECS Phoibos 150 scanning angle lens (SAL) analyzer, such that angle-resolved spectra can be measured both parallel and perpendicular to the analyzer slit without sample rotation. The capillary and sample can be aligned and scanned with nanometer precision using a piezo-based manipulator with 11 degrees of freedom [SmarAct]. A rendering of the microARPES measurement geometry is presented in Figure 1b.
As a starting point, the electronic structures of bulk MnBi 2 Te 4 and Mn(Sb d Bi (1 − d) ) 2 Te 4 with a nominal mole fraction given by d = 0.3 are considered to establish the doping-dependent low-energy spectra of MBTs in our measurement configuration. Figure 1c presents an optical micrograph image of the Mn(Sb d Bi (1 − d) ) 2 Te 4 crystal glued to the sample holder. The same crystal has been cleaved in the microARPES ultra-high vacuum end-station and scanned under the microscopic beam using a photon energy of 100 eV, which enables mapping of the spatiallydependent photoemission intensity of the Te 4d and Sb 4d core level and low-energy BVB and BCB regions. The (x, y)-dependent intensity in Figure 1d is a projection of the (E, k)-region on the detector containing the Te 4d core level, which produces a map of the entire crystal in correspondence with the optical image. Each point on the map contains a core level spectrum as shown for two representative areas in Figure 1e. The variation of intensity in the map is caused by a combination of insufficiently cleaved areas and core level shifts that move the peak intensity out of the (E, k)-integration region of the map. Shifts of up to 100 meV are observed as shown via the red and blue curves in Figure 1e. The corresponding low-energy spectra and Fermi surfaces are presented in Figure S1, which exhibit consistent doping effects. These observations highlight the local inhomogeneity of Sb substitution and resulting doping at the probed surface, which necessitates the use of microARPES to resolve the electronic structure.
The low-energy dispersion and Fermi surface from the area marked by a red circle in Figure 1d are compared with those of a bulk MnBi 2 Te 4 crystal in Figures 1f-g. The sharp "v"-shaped bands of MnBi 2 Te 4 correspond to the BVB and BCB, which display an energy gap with a peak-to-peak separation of 150 meV. The topological state is not visible in the gap at the photon energy of 21 eV used here due to strong matrix element effects. [24] A pair of faint parabolic bands are visible near the Fermi level, E F , which have previously been described on the basis of laserbased ARPES as a pair of Rashba-type surface bands that are strongly hybridized with the topological state. [28] A six-fold set of lobes appear in the Fermi surface from these states, in addition to near-circular features from BCB states. The Sb substituted crystal simply displays the "v"-shaped BVB around E F and a point-like Fermi surface, providing an ideal situation for electron transport or optical pump-probe experiments involving the topological state. However, given the spatial doping inhomogeneity, additional means to fine control the doping would be required in a hypothetical transport experiment involving such a sample. In the following, we consider Rb deposition on pristine MBTs as an alternative method to tune the low-energy band structure.
The effect of in situ Rb deposition on MnBi 2 Te 4 is tracked via energy distribution curves (EDCs) extracted at normal emission and plotted against Rb coverage, as shown in Figure 2a. Calibration of a complete Rb monolayer (ML) is determined to be the point at which the Rb 3d core level peak position stops increasing linearly and the linewidth remains constant, as presented in Figure S2 (Supporting Information). Note that 0 ML indicates the instant where the shutter is opened and the surface is exposed to Rb atoms. Data shown before 0 ML represent a baseline for the situation without adsorbed Rb atoms. Representative snapshots of the (E, k)-dependent dispersion at different Rb coverages are shown in Figure 2b. Initially, the BVB and BCB edges shift rigidly down by a maximum of around 25 meV at 0.2 ML, signifying mi-nuscule electron-doping by the adsorbed Rb atoms. Upon further increase of the Rb coverage, the bands surprisingly shift back toward E F . At 0.5 ML, the BVB and BCB edges are back to their original position. Beyond 0.8 ML, two features start to emerge in the bulk gap. They are most clearly seen at a coverage of 1.25 ML in Figure 2b, where they are marked by blue arrows. A possible explanation of these features is that they reflect emerging photoemission intensity from the robust topological bands as the adsorbed Rb atoms destroy the remaining surface states and their hybridization with the topological bands. The changed surface potential due to the Rb adatoms also influences the photoemission matrix elements, potentially causing the emergence of intensity from the topological states at our photon energy of 21 eV. Intriguingly, a peak-to-peak separation of 45 meV is determined between the features, which could hint at a doping-dependent gapped crossing between the topological bands. [35] An alternative explanation involves Rb-induced surface band bending and the appearance of quantum well states of the BVB and BCB. [44] However, this appears rather unlikely given the reduced amount of doping. The entire Rb deposition sequence can be viewed as divided into two stages, as marked by green and purple arrows in Figure 2a: An initial electron-doping stage with rigid energy shifts in the dilute limit, followed by hole-doping and dramatic electronic structure changes in the dense limit. The hole-doping stage is unexpected as one would normally anticipate alkali metals to merely donate electrons at the surface. The experiment is repeated on the quintuple layer building block of the MBTs, i.e. Bi 2 Te 3 , in order to gain further insights on the Rb coverage-dependent effects and their impact on the lowenergy dispersion of these materials. Two stages of doping following a similar sequence as for MnBi 2 Te 4 are observed but with more striking changes to the low-energy dispersion as shown in Figure 2c,d. Pristine Bi 2 Te 3 exhibits linear bands dispersing to E F , which correspond to the topological surface state. The bands cross just below the edge of the broad "M"-shaped continuum of states that reflect the BVB. [45,46] Deposition of Rb initially leads to strong electron-doping with bands shifting 230 meV to higher binding energies concomitant with the appearance of several sharp parabolic states near E F . Additionally, several "M"-shaped replicas of the BVB appear, as marked by blue arrows on the 0.27 ML spectrum in Figure 2d. These features are interpreted as quantum well states induced by strong surface band bending, as described in similar experiments on Bi 2 Se 3 . [47] As the Rb coverage is increased beyond 0.5 ML, the bands shift back toward E F , the BCB occupation strongly decreases and the quantum well states disappear. At an extreme Rb coverage of 2.34 ML, the topological surface state is observed with a modified dispersion compared to the pristine case, and the topological bands exhibit a crossing above the BVB edge. The behaviors observed at extreme Rb coverages on both MnBi 2 Te 4 and Bi 2 Te 3 are indicative of a complex interaction between the adsorbed Rb and the surface lay-ers of the materials, possibly hinting at Rb intercalation in the van der Waals gap or alloying between the materials.
The surface character of the Rb-induced states in MnBi 2 Te 4 can be further verified by a flat dispersion with photon energy. In order to establish a comparison, we first present the photon energy dependent dispersion of pristine MnBi 2 Te 4 at normal emission in Figure 3a. The dominant features are the dispersing BCB and BVB edges, which have been highlighted by dashed red curves. Figure 3b presents ARPES spectra at three pertinent photon energies, providing ideal matrix elements for visualizing the topological surface state (h = 15.0 eV), and revealing the maximum (h = 17.3 eV) and minimum (h = 20.6 eV) of the bulk bandgap. These key features are marked by green and red arrows, respectively. Our observations are fully in line with previous synchrotron studies of MnBi 2 Te 4 . [4] At an Rb coverage of 1.85 ML, deep in the second stage of doping, we primarily observe nondispersing states with photon energy, as shown in Figure 3c,d and highlighted by blue arrows. Note that the near-vertical features in the intensity around photon energies of 19 and 24 eV in Figure 3c stem from core level photoemission. The lack of dispersion is a strong indication for the surface-derived nature of the states at high Rb coverages.
To further elucidate the interaction between Rb and the surface of MnBi 2 Te 4 , measurements of core level spectra as a function of Rb coverage are performed. Surveys obtained with a photon energy of 145 eV before and after deposition are shown in Figure 4a, providing an overview of accessible core levels in the experiment. During Rb deposition, the intensity of the Bi 5d and Te 4d peaks decreases linearly while the Rb 3d peak intensity grows steadily with a single doublet shifting in binding energy, as seen in Figure 4b. These monotonic behaviors of the intensity in combination with the absence of an additional set of Rb 3d components rule out intercalation of Rb in the van der Waals gap of the crystal. [47] In the dense limit, a side-peak appears on the Bi 5d core level, as demarcated by grey arrows in Figure 4b. Extracted Bi 5d and Te 4d core level shifts are presented together with the BVB shift as a function of Rb coverage in Figure 4c. All shifts follow similar trends and are of comparable magnitudes, indicating that the binding energy positions of the peaks are mainly determined by doping. [48] The Bi 5d side-peaks appear to split out from the main set of peaks with an increasing energy separation between main and side-peaks in the dense limit. This behavior is indicative of alloy formation, [7] similarly as observed when dosing Ti on Bi 2 Se 3 , [49] which results in hole doping and dramatic changes of the surface electronic structure. The hole-doping effect could also be explained by the formation of a surface metallic film by the adsorbed Rb that leads to a depolarization of the surface dipole. [50] In this case, one would expect an increase in the asymmetry of the Te 4d core level lineshape, however, this is found to decrease with coverage as shown in Figure S2 (Supporting Information) , which speaks for the formation of an alloy. These results emphasize that alkali doping is a complex process in general. Even in the early stages of doping, it is possible that a chemical bond is formed between the alkali and the surface layer of the material, which has previously been shown to lead to unexpected band gap sizes in alkali-doped semiconductors. [40,41,51] Having assessed the impact of Rb adsorption on the separate MnBi 2 Te 4 and Bi 2 Te 3 systems, the surface electronic structure changes of the corresponding terminations on the MnBi 4 Te 7 heterostructure are now examined. The two terminations are distinguished by their unique surface electronic states. To optimally bring out the details of their complex band structures, we have surveyed the low-energy spectra as a function of photon energy, as shown in Figure S3 (Supporting Information). Starting with the Rb dosing sequence on the Bi 2 Te 3 termination, the evolution of the surface electronic structure is tracked via EDCs in Figure 5a. Two stages of doping are again identified with an initial downwards shift of the bands due to electron-doping, followed by hole-doping and a significant reduction of photoemission intensity. It is noted that the evolution of the intensity during the second stage exhibits similar behavior on the MnBi 2 Te 4 termination, which is discussed in further detail below and in Figure S4   of surface bands that cross E F (blue dashed curves) as well as BVB and BCB states (orange dashed curves). [2,18,23,26,27] Upon Rb deposition, the surface states around E F and the hybridized surface bands shift by up to 150 meV toward higher binding energies while the bulk bands merely exhibit shifts of 50 meV, as seen in Figure 5c,d. This causes the surface hybridization gap to close, giving rise to a region of increased intensity around -0.38 eV. The doping-dependent behavior and modification of surface electronic structure seen here are highly similar to a previous observation of photoinduced filling of the surface hybridization gap on the Bi 2 Te 3 termination of MnBi 8 Te 13 , [13] suggesting that this gap is extremely fragile toward external stimuli that alter the carrier concentration.
Turning toward the MnBi 2 Te 4 termination of MnBi 4 Te 7 , the coverage-dependent EDCs at normal emission are shown in Figure 6a, which again reveals the trend of electron-doping followed by hole-doping. ARPES spectra are presented in Figure 6b for the clean surface, the maximally electron-doped surface at 0.34 ML as well as two spectra in the second stage with coverages of 0.74 ML and 1.74 ML. Fermi surfaces are displayed in Figure 6c for the pristine sample and in the case of a coverage of 1.74 ML. In the pristine case, the main distinguishing feature of the MnBi 2 Te 4 termination is a surface-induced Rashbalike state, which forms the outer circle in the Fermi surface. [18] The feature is marked by pink arrows in Figure 6b-c. Similarly as in bulk MnBi 2 Te 4 , the topological surface state is situated in the bulk gap around 300 meV below E F with its photoemission intensity strongly suppressed above photon energies of 16 eV. [18] The main effect of the adsorbed Rb on this termination appears to be a removal of the surface bands, including the Rashba-like state, which have completely vanished at a coverage of 0.74 ML. Intriguingly, intensity emerges in the region of the bulk gap when the coverage is further increased. This is exemplified by the spectrum at 1.74 ML, which reveals the intensity to stem from a pair of crossing bands, as seen via the blue arrow in Figure 6b. Spectra at high coverages exhibit emerging photoemission intensity from topological bands on both terminations of MnBi 4 Te 7 as shown in Figure S4 (Supporting Information), which is consistent with the surface-specific bands having been destroyed by the reaction between Rb and the surface layers. This suggests that the remaining features are the BVB and BCB in addition to a set of crossing topological bands.
Interestingly, the appearance of persistent topological bands within 200 meV of E F seems to be a common feature between bulk MnBi 2 Te 4 , Bi 2 Te 3 and their terminations on MnBi 4 Te 7 at extreme Rb coverages. A peculiarity of MnBi 2 Te 4 is the relatively insignificant changes occurring during stage 1 of doping compared to the other systems. Additionally, maximum electron-doping of MnBi 2 Te 4 occurs around an Rb coverage of 0.2 ML, which is lower than observed in the other materials. A key reason for these differences is that we mainly observe the bulk states of MnBi 2 Te 4 during stage 1 of doping and these are less sensitive to the adsorbed Rb than the surface-derived bands that dominate the ARPES spectra of Bi 2 Te 3 and MnBi 4 Te 7 . Furthermore, it is possible that the adsorbate structures of Rb can vary between the surfaces, giving rise to different doping behaviors with coverage. The dramatic modifications of surface electronic structures with alkali adsorption during stage 2 of doping likely impact the magnetic properties of MnBi 2 Te 4 and MnBi 4 Te 7 , which will be interesting to probe across the Néel temperature with high-resolution laser-based microARPES operating in the sub-10 K temperature regime, as recently demonstrated on the pristine systems. [12]

Conclusion
In conclusion, the surface electronic structures of the MBT building blocks MnBi 2 Te 4 and Bi 2 Te 3 and their terminations on MnBi 4 Te 7 were resolved in situ using microARPES at the new SGM4 beamline at the ASTRID2 light source during deposition of the alkali metal Rb. In all situations, two stages of doping are observed with an initial electron-doping effect, which is then followed by a hole-doping effect. Core level measurements indicate that significant alloying between Rb and the surface layers occurs at high coverages, which further drives dramatic band structure changes, including the removal of surface states and hybridization effects. On the Bi 2 Te 3 termination of MnBi 4 Te 7 , a closing of the surface hybridization gap is induced before the surface electronic structure organizes into a minimal configuration with just the BVB and BCB remaining in addition to the robust topological bands. Further studies of the effects of alkali deposition on MBTs may additionally reveal the nature of the surviving topological bands and what role the alloy plays. The examples presented here underline the extreme tunability and sensitivity of the surface electronic states of the MBTs toward external stimuli, which could potentially be exploited for further developing the quantum transport phenomena in functioning MBT devices.

Experimental Section
Crystal Growth:: Single crystals were grown using the Bi 2 Te 3 flux method, as previously reported in Ref. [14] for MnBi 2 Te 4 and MnBi 4 Te 7 and Ref. [45] for Bi 2 Te 3 . To synthesize crystals of Mn(Sb d Bi (1 − d) ) 2 Te 4 , the elemental starting materials of Mn, Sb, Bi, and Te were mixed and sealed in a vacuum quartz tube. The tube was subsequently heated to 900°C, slowly cooled to 601°C, and then canted to remove the flux.
Photoemission Measurements:: The microARPES measurements were performed at the SGM4 beamline of the ASTRID2 synchrotron radiation source at Aarhus University, Denmark. [43] The MnBi 2 Te 4 , MnBi 4 Te 7 , and Bi 2 Te 3 crystals were cleaved in situ and measured at a temperature of 120 K in a base pressure better than 1 · 10 −10 mbar. The Mn(Sb d Bi (1 − d) ) 2 Te 4 samples were cleaved in situ at room temperature and a base pressure better than 5 · 10 −8 mbar. ARPES and core level spectra were collected using a SPECS Phoibos 150 SAL analyzer. Measurements of the (E, k x , k y )dependent photoemission intensity were performed using the scanning angle lens feature of the analyzer while the sample position was held fixed. The applied photon energies were in the range of 16-21 eV for the ARPES spectra, while the core level measurements were performed with 145 eV, except in Figure 1e where 100 eV was applied. The light polarization was predominantly linear horizontal. The energy-and angular resolution were better than 20 meV and 0.1°, respectively. The second-derivative ARPES intensity was produced using the method described in Ref. [52]. Rb adatoms were deposited from a SAES alkali getter during measurements. The Rb coverage was estimated by comparing the Te 4d and Rb 3d core level areas before and after a deposition cycle. The point at which the peak position of the Rb 3d core level stopped shifting linearly and the linewidth remained constant is taken as an indication for an ordered structure corresponding to 1 ML, as shown in Figure S2 (Supporting Information). The ratio of the area growth rates for the Te 4d and Rb 3d core levels at this coverage was then used as calibration for calculating the Rb coverage from the change of the Te 4d area at each dosing step.
Statistical Analysis:: The raw ARPES spectra were transformed from angle to momentum coordinates, and a Fermi level correction was applied based on reference spectra of polycrystalline tantalum. Core level spectra were fitted with a Doniach-Šunjić function with a linear background at every point of the Rb deposition series. The data was plotted and analyzed using WaveMetrics IGOR Pro 7 software.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.