Atomically constructing a van der Waals heterostructure of CrTe2/Bi2Te3 by molecular beam epitaxy

A 2D heterostructure with proximity coupling of magnetism and topology can provide enthralling prospects for hosting new quantum states and exotic properties that are relevant to next-generation spintronic devices. Here, we synthesize a delicate van der Waals (vdW) heterostructure of CrTe2/Bi2Te3 at the atomic scale via molecular beam epitaxy. Low-temperature scanning tunneling microscopy/spectroscopy measurements are utilized to characterize the geometric and electronic properties of the CrTe2/Bi2Te3 heterostructure with a compressed vdW gap. Detailed structural analysis reveals complex interfacial structures with diversiform step heights and intriguing moiré patterns. The formation of the interface is ascribed to the embedded characteristics of CrTe2 and Bi2Te3 by sharing Te atomic layer upon interfacing, showing intercoupled features of electronic structure for CrTe2 and Bi2Te3. Our study demonstrates a possible approach to construct artificial heterostructures with different types of ordered states, which may be of use for achieving tunable interfacial Dzyaloshinsky–Moriya interactions and tailoring the functional building blocks in low dimensions.


Introduction
The explosive growth of 2D materials has given rise to a rich variety of long-range ordered quantum states that have been unprecedentedly uncovered in reduced dimensions, such as low-dimensional superconductivity [1,2], ferroelectricity * Authors to whom any correspondence should be addressed.
Original content from this work may be used under the terms of the Creative Commons Attribution 4.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI. [3], ferromagnetism [4,5] and antiferromagnetism [6]. The weak van der Waals (vdW) coupling further inspires intensive studies with an additional degree of freedom to artificially twist/stack/intercalate each 2D piece, giving rise to a series of exotic quantum phenomena, including chiral Majorana fermions [7], unconventional superconductivity [8], quantum anomalous Hall effect [9] and magnetic skyrmions [10]. In this light, 2D heterostructure of various ingredients can be greatly functionalized with modulated electronic properties while simultaneously possessing nontrivial topology, magnetism and correlated interaction. These strategies have been implemented in many heterostructures. For instance, unusual electronic states and the

Future perspectives
Constructing heterostructure is a powerful strategy to comprehensively utilize functionalized properties of individual constituents in 2D materials by selective and controllable manipulation of nontrivial topology, magnetism and correlated interaction. Recent advances have achieved many unprecedented quantum phenomena in superlattice and heterostructures, including unconventional superconductivity, quantum anomalous Hall effect, chiral Majorana fermions and magnetic skyrmions. Delicate molecular beam epitaxy offers a promising route to synthesize self-assembly and artificial structures with precise fabrication at atomic scale. Combined with the high-resolution real-space structural/electronic characterization at atomic level by scanning tunneling microscopy/spectroscopy (STM/STS), new ways will be visioned to explore the buried interface with hybridized interaction in the heterostructure. These unique advantages will facilitate the realization of proximity coupling, as well as the achievement of precise detection and tunability, in low-dimensional physics and quantum materials superconducting proximity effect of Bi(110) films can be modulated by NbSe 2 [11], and topological superconductivity with 1D Majorana edge modes was reported when the ferromagnetic CrBr 3 is coupled with a superconducting NbSe 2 substrate [12], quantum anomalous Hall effect can be realized in proximity-coupled (Zn,Cr)Te/(Bi,Sb) 2 Te 3 /(Zn,Cr)Te [13] and Cr:Sb 2 Te 3 /Cr 2 O 3 [14] heterostructures, as well as in twisted bilayer graphene aligned to hexagonal boron nitride [15], heavy fermions can artificially emerge in a 1T/1H-TaS 2 heterostructure by the Kondo coupling between localized magnetic moments and itinerant electrons [16], etc. Thus, artificial heterostructures offer a fantastic avenue to engineer 2D materials as building blocks with desired properties.
Seeking magnetic skyrmions, local swirls with topologically nontrivial chiral spin textures, is a typical example of taking full advantage of the hybrid systems [17]. To generate and stabilize the noncollinear Dzyaloshinskii-Moriya interaction (DMI), a decisive factor to realize skyrmions that combines the strong spin-orbit coupling (SOC) with broken inversion symmetry at an interface [18], various kinds of artificial heterostructures are theoretically proposed and experimentally fabricated, including ferromagnet/heavy-metal [19], ferromagnet/topological insulator [20], topological insulator/ magnetic insulator [21] and heavy-metal/magnetic insulator [22]. To possess the essential characteristic of large SOC, the prototypical topological insulator Bi 2 Te 3 with nontrivial surface states is commonly selected to mediate DMI. In addition, Cr x Te y compounds, a newly discovered family of 2D vdW magnets that manifest fruitful structural phases and magnetic properties at stable states [23][24][25] have recently been widely considered to favor strong interfacial DMI and the formation of skyrmions. Néel-type skyrmion lattice was predicted in the CrTe 2 /WTe 2 bilayer [26], and giant interfacial topological Hall effect was achieved to confirm the topological spin texture of skyrmions in both CrTe 2 /Bi 2 Te 3 and Cr 2 Te 3 /Bi 2 Te 3 hybrid systems [27,28]. Strikingly, it was argued that the existence of Bi bilayer nanosheets intercalated at the interface of CrTe 2 and Bi 2 Te 3 should be indispensable to the formation of magnetic skyrmions [28][29][30]. Such a prerequisite for the interface is subtly coincident with another skyrmion heterostructure of Mn(Bi,Sb) 2 Te 4 sandwiched by (Bi,Sb) 2 Te 3 , where the thickness of the (Bi,Sb) 2 Te 3 spacer layer is essential to control the coupling between the gapped topological surface states in the two Mn(Bi,Sb) 2 Te 4 layers to stabilize the skyrmion formation [31]. This crucially calls for a thorough understanding of the interfacial structure in topological insulator/ magnet heterostructure.
In this study, we demonstrate the achievement of vdW epitaxial growth for a high-quality CrTe 2 /Bi 2 Te 3 heterostructure by molecular beam epitaxy (MBE). The CrTe 2 /Bi 2 Te 3 heterostructure processes all the essential ingredients for the formation of magnetic skyrmions: (i) inversion symmetry breaking, (ii) large SOC, and (iii) the entanglement of magnetism and strong SOC. Moreover, the superiority of heterostructures also lies in the advantages of the fabrication aspect. CrTe 2 and Bi 2 Te 3 are both simple telluride compounds sharing similar hexagonal surface lattice symmetry and the vdW interlayer coupling will facilitate reduced dimension with controlled thickness, the atomically sharp interface and flexibility in stacking order. These all favor the generation of strong interfacial DMI and topological spin textures in real space. In a step-by-step way, the atomic morphologies with sharp interface structure are demonstrated by low-temperature STM/STS measurements. Intriguingly, compared to the bulk CrTe 2 , our CrTe 2 layer shows a ∼28% compressed vdW gap with respect to the underlying Bi 2 Te 3 , which is only half that of the single-monolayer (ML) CrTe 2 /graphene films. Specifically, we observe a 3.5 nm moiré superstructure with a step height of 0.44 nm, but a 0.40 nm step on the same CrTe 2 terrace without moiré patterns. Meanwhile, the real-space STS distribution shows a smooth evolution when crossing the step boundaries from the Bi 2 Te 3 to the CrTe 2 segments, exhibiting evident interfacial states at the step boundaries. Based on the high Te chemical potential growth conditions and shared Te atomic layers, we propose a schematically structural model of CrTe 2 pieces embedded in Bi 2 Te 3 films, reasonably reproducing the observed 40 pm difference in height, as well as the emergence of moiré periodicity. Our work not only shields light from constructing 2D heterostructures with different functionalities between topology and magnetism that are interlayer coupled, but also provides a versatile material platform for exploring magnetic skyrmions with tunable DMI and topological properties at the interface.

Methods
The CrTe 2 /Bi 2 Te 3 heterostructure was fabricated through a standard MBE growth process in an ultrahigh vacuum chamber at a base pressure of 5 × 10 −10 Torr. Prior to growth, a uniform bilayer graphene substrate was obtained as previously reported [32]. The SiC substrate was first degassed at 600 • C for 3 h, then annealed at 950 • C under a Si flux for ten cycles, and further flashed to 1400 • C for 10 min. The high purity of Bi (99.999%) and Te (99.999%) sources was co-evaporated from two homemade thermal effusions with a large Te:Bi flux ratio of ∼15. The graphene-covered SiC substrate was held at 260 • C during the growth, which was monitored with an infrared spectrometer with an emissivity of 0.9. After growth, the Bi 2 Te 3 films were post annealed at the growth temperature for 5 min with a Te flux to improve the crystalline quality, which was also checked by STM measurements before capping the CrTe 2 layers. The CrTe 2 films were further grown onto the Bi 2 Te 3 films by co-evaporating highpurity Cr (99.995%) and Te (99.999%) atoms with a Te:Cr flux ratio of ∼20 while the substrate was kept at 300 • C.
Low-temperature STM/STS measurements were performed on a Unisoku 1500 STM system operating at 4.5 K, if not specifically noted [33]. The normal W tip by electrochemical etching was cleaned by e-beam heating to ∼1800 • C to remove oxides. Prior to STM measurements, the W tip was calibrated on Ag(111)/Si islands. All topographic images were taken in a constant-current mode, and the tunneling dI/dV spectra and conductance mappings were acquired using a standard lock-in technique at 983 Hz with a modulation voltage amplitude of 1% of the set bias voltage. The topographic images were processed using software from WSxM and Gwyddion.

Figure 1(a)
is an atomic structural model of CrTe 2 /Bi 2 Te 3 heterostructure from a side view. Bi 2 Te 3 crystallizes in a tetradymite-type structure with the space group viewed in a layered structure. It stacks in a sequence of Te-Bi-Te-Bi-Te atomic layers along the crystallographic c direction, forming a quintuple layer (QL) with a height of 1 nm. As a result, there are two inequivalent Te and Bi lattice sites, respectively [34]. On the other hand, CrTe 2 is in a trigonal layered structure with the space group of R3m1. Each Cr cation is centered by a Te octahedron with a 1 T structure, analogous to other TMDs [35]. Although bulk CrTe 2 crystals and thin films of 1-20 ML had already been identified in a ferromagnetic ground state with a Curie temperature above room temperature [36][37][38], its single-layer film grown on graphene substrate demonstrated an intriguing antiferromagnetic order and fieldinduced spin reorientation by real-space observation of spinpolarized STM measurements [39]. The mutability of magnetism in CrTe 2 films indicates that some external perturbations of charge transfer, strain, film thickness and interlayer coupling may bring about new issues regarding the heterostructure [40]. Here, due to the inert graphene substrate and weak vdW bonds of both Bi 2 Te 3 and CrTe 2 layers, the stacking combinations and film thicknesses of CrTe 2 /Bi 2 Te 3 heterostructure can be well controlled by adjusting the MBE growth conditions. Figure 1(b) is a typical STM image of a bare Bi 2 Te 3 film with a nominal thickness of ∼2.6 QL. The surface is atomically flat and smooth, with terraces as large as ∼100 nm, which terminate in a triangular Te-1 × 1 lattice with an in-plane lattice constant of 0.44 nm, as shown in figure 1(c). All edges of the Bi 2 Te 3 islands can be well determined as 120 • -orientated, conforming to their three-fold crystal symmetry. We record the evolution of dI/dV spectra near the Fermi level (E F ) by STS measurements for various thicknesses of Bi 2 Te 3 films in figure 1(d), which is proportional to the local density of states (LDOS). It is apparent that the STS spectrum exhibits a rather flat minimum around 300 meV below E F for 1 and 2 QL, indicating the intrinsic bulk band insulator of Bi 2 Te 3 [7]. As the thickness increases up to 3 QL, a 'V'-shaped LDOS minimum is formed near −140 meV, within the energy range between the bulk valence band maximum and conduction band minimum, suggesting the appearance of the topological surface state by the Dirac-cone band structure. The Dirac point shifts upwards −120 meV for the 4 QL, associated with the apparent quantum well states in both valence and conduction bands as well. This topological transition of the surface states in Bi 2 Te 3 ultrathin films at 3 QL is well consistent with previous studies [7,41]. Here, E F lies inside the bulk gap for all thicknesses, indicating that the surface carrier density dominates rather than the bulk contribution.
After depositing CrTe 2 layers, the morphology shown in figure 1(e) still remains atomically continuous with little changed surface roughness, implying that the interface of the heterostructure should be atomically sharp. Figure 1(f) presents the atomic resolution STM image of the CrTe 2 surface, displaying a triangular lattice with an in-plane constant of 0.39 nm. Moreover, two 60 • rotated domains exist, within which a clear 2 × 1 superstructure is obviously observed. As elucidated in our previous work in [39], this 2 × 1 periodicity is closely related to its antiferromagnetic spin texture, and can be regarded as a hallmark of the formation of CrTe 2 films. This rules out the possibility of CrTe 3 and other Cr x Te y compounds in our samples [42], also confirming the high crystalline quality of the grown heterostructure. Since the exposed Bi 2 Te 3 and capped CrTe 2 surfaces can coexist, which is verified by the atomic resolution, we compare the tunneling spectrum of the underneath Bi 2 Te 3 and topmost CrTe 2 layers in figure 1(g). We find that Bi 2 Te 3 exhibits similar DOS before and after the deposition of CrTe 2 films. The Dirac point further shifts to −80 meV, approaching E F , indicating that the CrTe 2 layer effectively serves as a p-doping source to Bi 2 Te 3 [43]. This can be attributed to the consumption of Te layers in Bi 2 Te 3 during the growth of CrTe 2 , which will be discussed later. In addition, the STS spectrum of the CrTe 2 surface on a large energy scale also bears a striking resemblance to the CrTe 2 /graphene samples [39,42], suggesting the flexibility of retaining the properties of separate segments in CrTe 2 /Bi 2 Te 3 heterostructure by vdW epitaxial synthesis.
Zooming into the CrTe 2 surface, pronounced periodic patterns can be clearly resolved, associated with atomically sharp boundaries that separate CrTe 2 and Bi 2 Te 3 , as displayed in figures 1(e) and 2(a). From the atomically resolved STM image, the moiré periodicity is measured as 3.50 ± 0.02 nm, which suitably matches nine times that of the CrTe 2 lattice constant (0.389 × 9 = 3.501 nm) and eight times that of Bi 2 Te 3 (0.438 × 8 = 3.504 nm). This is well in accordance with our simulated pattern of 3.5 nm periodicity by overlaying atomic structures of Bi 2 Te 3 and CrTe 2 , as compared in figure 2(b). This agreement also validates that the terminated Te atoms of both Bi 2 Te 3 and CrTe 2 should be aligned along the same direction without any twisted angle [10]. In addition to the morphologies, the 3.5 nm moiré pattern can also be reflected by the electronic structure. In figure 2(e), we plot the spatial STS distributions along the moiré structure, and examine the conductance spectra at different sites with atomic precision. Explicitly, an electronic modulation of 3.5 nm is resolved, agreeing well with the moiré periodicity. This demonstrates the ability to tune the local electronic structure as a function of atomic registry in CrTe 2 /Bi 2 Te 3 heterostructure.
Surprisingly, we find that the apparent step height of CrTe 2 films is about 0.44 ± 0.05 nm, which is largely reduced compared to the CrTe 2 /graphene films or the bulk value of 0.62 nm. The height is also strongly sensitive to different applied biases, as shown in figures 2(c) and (d). Generally, there are three aspects in the atomic height of CrTe 2 /Bi 2 Te 3 that are different to 1 ML CrTe 2 /graphene, which ranges from 0.88-1.05 nm [39]. (i) The former manifests an averaged 29% reduction compared to the bulk, while the latter is 58% larger than the bulk value. (ii) The evolution of the apparent height in 1 ML CrTe 2 /Bi 2 Te 3 can be roughly classified as a slightly larger value for the fulfilled states and a smaller height for the unoccupied ones, with little change for the negative (positive) biases. However, the height of 1 ML CrTe 2 /graphene shows a monotone decline with increasing bias voltages. (iii) We also observe a global 0.04 nm lower terrace than the 0.44 nm CrTe 2 above Bi 2 Te 3 , accompanied by the disappearance of moiré patterns, as compared in figures 2(a) and (c). To check the difference between the regions with and without moiré patterns, we acquire STM images on the same CrTe 2 surface across their boundaries at different biases in figures 2(f) and (g). The atomically resolved topographies clearly verify that the Te atoms are uninterrupted in a perfect triangular lattice around the borders, while the 0.04 nm corrugation remains.
To gain more insight into the CrTe 2 -Bi 2 Te 3 interface, we perform detailed spatially dependent dI/dV measurements across the step boundaries between the Bi 2 Te 3 and CrTe 2 terraces. As shown in figures 3(a) and (b), the STS around the boundry exhibits strong spectrum shifts that are distinct from both Bi 2 Te 3 and CrTe 2 , an indicator for the formation of interfacial states. Interestingly, Bi 2 Te 3 and CrTe 2 all show an overall V-shaped landscape with a conductance minimum near E F , which is different to the isolated Bi 2 Te 3 and CrTe 2 cases in figure 1(g). Simultaneously, the STS line cut shows a smooth transition from Bi 2 Te 3 and CrTe 2 , with the crossover region distributed around the step boundaries. The similarity of STS between Bi 2 Te 3 and CrTe 2 suggests that they should be intercoupled with each other and electronically hybridized at the interface, which conversely modifies the structural morphologies of the CrTe 2 -Bi 2 Te 3 interface with changing step height. The common Te-terminated surface, clean structure and transitional electronic properties at the interface suggest that the topological surface states of Bi 2 Te 3 are permitted to extend into the magnetic CrTe 2 layers, facilitating efficient magnetic exchange coupling and the hybridization of Bi p-orbitals and Cr d-orbitals to greatly enhance the DMI. Recent theoretical simulations on a similar CrTe 2 /Bi 2 Te 3 [27] and Cr 2 Te 3 /Bi 2 Te 3 [28] heterostructure have demonstrated the formation of Ne´eltextured magnetic skyrmions at the interface. Moreover, the enhanced DMI strength is calculated to be 36 pJ m −1 for a 6 nm CrTe 2 layer, and the skyrmion size in CrTe 2 /Bi 2 Te 3 heterostructure is estimated to be a maximum value of 34 nm [27].
Although there have been several reports on the growth of Cr x Te y /Bi 2 Te 3 heterostructure by the MBE method [27,28,30], the compositive phase and detailed interface structure may be diverse from each other, where the growth temperature is argued to play a key role in forming the specific heterostructure. Generally, lower substrate temperature is favorable for forming Cr 2 Te 3 phase, while a higher substrate temperature is more propitious to obtaining CrTe 2 phase. This is well consistent with our previous study, which found that CrTe 3 can be gradually transformed into CrTe 2 by further in situ vacuum annealing [42]. On the other hand, due to the high volatility of Te atoms at elevated temperature, the effective flux of the Te source should be decreased and become insufficient to sustain the Te-rich environment. Moreover, Cr atoms have stronger reducibility than Bi atoms, making it possible for Cr atoms to extract Te atoms from Bi 2 Te 3 layer once the Te is in short supply, or Cr can even replace Bi atoms, thus forming a new Cr 2 Te 3 layer. Considering the three-and fiveatomic-layered structure of CrTe 2 and Bi 2 Te 3 , respectively, both sharing the common Te layers on both sides of the vdW gap, we thus construct a possible structural model by stacking a 5 ML CrTe 2 embedded in a 3 QL Bi 2 Te 3 film, as illustrated in figure 3(c). We speculate that, during the growth, the prophase of existing Te layers of Bi 2 Te 3 can also serve as a Te source to help form the adhered CrTe 2 layers. This means that the adjacent CrTe 2 and Bi 2 Te 3 will share the same Te atoms, especially when their Te atoms are very close, as marked by red dashed circles. Accordingly, the Bi 2 Te 3 film becomes deficient in Te atoms and more p-doped [43], consistent with the shift of the Dirac point in figure 3(b). In addition, the locations of vdW gaps in CrTe 2 and Bi 2 Te 3 become mismatched laterally, and the step height significantly deviates from the bulk value. In figure 3(c), the embedded structure can produce both 0.44 and 0.40 nm atomic heights between the CrTe 2 and Bi 2 Te 3 terraces, reasonably agreeing with our STM observations in figure 2. Moreover, the variation of moiré patterns on the same CrTe 2 terrace with a slight height difference can now be well understood. When the heterostructure interface is formed between both Te layers of CrTe 2 and Bi 2 Te 3 , the step is 40 pm higher with moiré periodicity. Otherwise, this moiré superstructure would be invisible when the Te layer of CrTe 2 interacts with the Bi layer of Bi 2 Te 3 . In this light, our structural and electronic observations of the CrTe 2 /Bi 2 Te 3 heterostructure directly reflect a much more contractive vdW gap at the interface, with distinct electronic properties from the CrTe 2 /graphene case, where the vdW gap is largely expended [39]. It is worth noting that other possibilities may exist, except for the above five layers of CrTe 2 , whereas the commonly observed compressed steps of overlaid CrTe 2 layers imply that there may be many other combinations of sharing Te layers between CrTe 2 and Bi 2 Te 3 , in accordance with our embedded scenario.

Conclusion
We study the vdW epitaxial growth of two 2D materials, CrTe 2 and Bi 2 Te 3 , using the MBE technique. We unravel the detailed morphological and electronic structures of the CrTe 2 /Bi 2 Te 3 heterostructure with the emergence of moiré patterns, which are strongly dependent on the step height of the CrTe 2 layer. The STS also shows a spectroscopic similarity between CrTe 2 and Bi 2 Te 3 with a crossover in the interfacial region. We provide an embedded structural model with a shared Te atomic layer at the interface to form a mismatched and compressed interface in the CrTe 2 /Bi 2 Te 3 heterostructure. Our results manifest a promising route to tailor unprecedented quantum states with versatile tunability by interface engineering in 2D vdW heterostructures.