Direct Epitaxy of SnSe₂/SnSe Hetero-Bilayer with a Type-III Band Gap Alignment

Abstract

Van der Waals (vdW) heterostructures formed by stacking two distinct semiconductor monolayers have gained increasing research interest because of the various predicted and realized exotic phenomena that are absent in the corresponding monolayers. However, constructing such a vdW hetero-bilayer is very challenging and mostly relies on top-down mechanical methods. Here, we report a direct growth of an SnSe₂/SnSe hetero-bilayer by using molecular beam epitaxy (MBE), in which elaborate interface engineering is the key to success. Scanning tunneling microscopy (STM) characterization demonstrated the well-defined and uniform moiré patterns, indicating an atomic-scale clean and uniform SnSe₂/SnSe interface. In combination with first-principles density functional theory (DFT) calculations, we further unveiled a type-III band gap alignment between the SnSe₂ and SnSe monolayers. This work provides a new method for building vertical SnSe₂/SnSe hetero-bilayers and a novel platform for exploring functional devices based on the type-III band alignment.

1. Introduction

As thinned down to the atomic-layer limit, two-dimensional (2D) van der Waals (vdW) materials usually exhibit unique and tunable properties that are distinguishable from their bulk counterparts. Since the successful discovery of graphene in 2004 [1], 2D vdW materials have gained tremendous attention due to their great potential in various fields such as energy, environment, electronics, and optoelectronics, etc. [2,3,4,5,6,7]. Graphene hosts a Dirac-type energy band, and exhibits excellent electrical conductivity, high carrier mobility, high transmittance, and a wide spectral response range [8]. As the most remarkable example of transition metal dichalcogenides (TMDs), 2H-MoS₂ is a promising 2D vdW semiconductor with an indirect band gap (~0.88 eV) for bulk and a direct band gap (~1.71 eV) for the monolayer [9,10,11]. Superior to the transition metals, the group IVA metals are earth abundant, low cost, and environmentally friendly [12], and thus the group IVA chalcogenides offer a great opportunity for future applications. Group IVA chalcogenides are usually formed in two formats of MX and MX₂ (M = Ge, Sn or Pb, X = S, Se or Te), according to the different chemical valence of the metal atoms [13]. SnSe takes the black phosphorus structure and is a semiconductor with a direct band gap (~1.4 eV) for the monolayer layer, and an indirect band gap (~0.9–1.3 eV) for few layers [14,15]. On the other hand, SnSe₂ hosts a T phase sandwich atomic structure, and is an indirect band gap (~1.07 eV) semiconductor [16]. Compared with other van der Waals materials, SnSe₂ has a high electron mobility (233 cm²V⁻¹s⁻¹) and a low thermal conductivity (3.82 W m⁻¹K⁻¹) at room temperature [17].

When two different vdW monolayers are put in contact with each other, thus forming a heterojunction, more intriguing phenomena beyond the constituent monolayers can be induced. At the presence of a moiré field [18,19,20], the vdW heterostructures provide a new platform for the study of electron correlation, such as the Mott insulator, unconventional superconductor, fractional quantum Hall effect, and the excitonic insulator [21,22,23,24]. While the lateral heterojunction of SnSe₂ and SnSe has been experimentally realized by electron irradiation, thermally annealing, or laser heating that induces a SnSe₂-to-SnSe transition [25,26], it is still very challenging to construct a well-defined SnSe₂/SnSe bilayer.

The top-down method of mechanical exfoliation has been widely adopted to construct a vdW heterostructure, and molecular beam epitaxy is an ideal choice to achieve an atomic-scale clean surface and interface [27]. In this work, we tried to use interface engineering to grow a hetero-bilayer made of two distinct monolayers. By choosing NbSe₂ as the proper substrate, we successfully grew a high-quality SnSe₂/SnSe hetero-bilayer. Interestingly, it was found that the SnSe monolayer preferred to grow as the first layer, and subsequently, SnSe₂ as the second layer on the top of the SnSe monolayer. Scanning tunneling microscopy (STM) characterization demonstrated uniform moiré patterns originated from the lattice mismatch between the top SnSe₂ and bottom SnSe layers and thus, indicated an atomic-scale well-defined interface. First-principles density functional theory (DFT) calculations further unveiled that the band gap of the SnSe₂/SnSe hetero-bilayer forms a type-III band alignment. Our work paves the way for further research of promoting efficient electron–hole separation [28,29,30,31,32,33], and thus shows potential in various applications [34,35].

2. Results and Discussion

Figure 1a,b illustrate the atomic structures of SnSe and SnSe₂, respectively. SnSe has an orthorhombic structure (Space group: $Pnma$, a = 4.15 Å and b = 4.44 Å), and SnSe₂ has a hexagonal structure (Space group: $P3m1$, a = b = 3.87 Å). The SnSe monolayer exhibits as a puckered honeycomb lattice and is composed of alternating Sn and Se atomic lines [36,37,38,39]. Unlike SnSe, the SnSe₂ monolayer has a sandwich-like triple atomic layered structure, in which the Sn atoms are arranged in a hexagonal lattice and sandwiched by the top and bottom Se atomic layers [40,41,42].

The epitaxial growth of chalcogenide monolayers is usually carried out under a rich chalcogen environment, and the flux ratio of chalcogen to metal is far beyond the equilibrium condition. As a result, SnSe₂ is more preferable to grow than SnSe since it is the most thermally stable structure. On the other hand, the growth of a SnSe monolayer is more difficult without considering the substrate effect. To verify the influence of substrate, we firstly grew the samples on a bilayer graphene (BLG)/SiC substrate that provides a very weak vdW interaction and negligible lattice stress effect to the epilayer. Figure 1c,d shows the scanning tunneling microscopy (STM) images of the grown film on the BLG/SiC substrate. We found that the SnSe₂ monolayer is the only structure formed directly on the BLG/SiC substrate. The measured step height, as extracted from the line-cut profile in the inset to Figure 1c, is ~7.0 Å, larger than the out-plane lattice constant of the bulk, indicating a weak vdW interaction.

We next chose NbSe₂ as the substrate, which also has a hexagonal lattice structure (space group: P63/mmc, a = b = 3.44 Å), but stronger vdW interactions than bilayer graphene. Figure 1e shows the STM topographic image of the epilayers grown on the NbSe₂ substrate. The epilayer is composed of both first and second layers. Surprisingly, the first layer hosts tetragonal symmetry regardless of the hexagonal symmetry underneath the NbSe₂ substrate, while the second layer hosts hexagonal symmetry, as indicated by the atomic-resolution images in Figure 1f,g. The measured lattice parameters for the first layer are a = ~4.24 Å and b = ~4.26 Å, in agreement with that of SnSe [43]. The lattice parameter for the second layer is a = b = ~3.9 Å, consistent with that of SnSe₂ as well [44]. As shown in the inset to Figure 1e, the line-scan profile measurement shows that the height of the island is ~4.6 Å for the first layer and ~5.5 Å for the second layer. Thus, all the experimental observations suggest that the first layer is a SnSe monolayer and the second layer is a SnSe₂ monolayer, respectively. Notably, as shown in Figure 1f, the one-dimensional atomic chains on the top of the SnSe are possibly the Sn ad-atoms, which might be due to the lattice anisotropy in the two directions of SnSe. No such defects were found on the surface of the second SnSe₂ layer.

The success of hetero-epitaxy owes to the competition between thermodynamic control and the substrate’s strain effect. The first layer’s growth is mainly controlled by the substrate, since the lattice parameter along the diagonal direction for NbSe₂ is very close to that of SnSe, as shown in Figure 1h. Our previous studies [45,46,47] have demonstrated that the substrate with lattice constants close to the epilayer can apply a slight strain during epitaxy, and well control the structure of the epilayer. The second layer’s growth is thermodynamically controlled since the lattice parameter is very different between SnSe and SnSe₂, thus the strain effect does not apply. Our DFT results show that the interlayer binding energy of the SnSe/NbSe₂ bilayer is ~440 meV/Ų, which is significantly higher than that of the SnSe₂/NbSe₂ layer (336 meV/Ų). This result suggests that the growth of the SnSe monolayer on the surface of NbSe₂ is energetically more stable to some extent.

To explore the electronic structure of the SnSe₂/SnSe bilayer, we measured the differential conductance dI/dV spectra on the first SnSe and second SnSe₂ layers, as shown in Figure S1 and Figure 1i. To make a direct comparison, we also measured the dI/dV spectra on the SnSe₂ layer grown directly on the BLG/SiC substrate. As shown in the inset to Figure 1i, the STS spectra taken on the SnSe₂ monolayer on BLG/SiC shows a band gap of ~1.30 eV, comparable with the freestanding SnSe₂ monolayer. Thus, one can tell that the interface with the BLG/SiC substrate has no impact on the electronic structure of the epitaxial SnSe₂ monolayer. Surprisingly, in contrast to the SnSe₂ monolayer on the BLG/SiC substrate, the STS spectrum taken on the SnSe₂/SnSe hetero-bilayer shows finite intensities in the whole bias range, and no band gap is observed, which exhibits metallic behavior. Considering the insulating band gap of ~0.9 eV of the first SnSe layer (Figure S1), the influence from NbSe₂-induced metallic DOS can be ruled out. This is likely attributed to the band-alignment between SnSe₂ and SnSe.

We observed clear and well-defined moiré patterns on the second SnSe₂ layer, which is a characteristic feature of the hetero stacking of two monolayers with distinct lattice symmetries and suggests an atomic-scale clean and uniform SnSe₂/SnSe interface. Generally, when two vdW monolayers with different symmetry or a twist angle are stacked together, a long periodic moiré pattern is formed [48,49]. To unveil the stacking order of the SnSe₂/SnSe heterostructure, we further quantitatively analyzed the period and symmetry of the moiré pattern. As identified in Figure 1g, there exist a few different domains. Most of these moiré patterns are actually equivalent, with the orientations rotated ~120° or ~90° with each other. Figure 2a shows the atomic-resolution image of a typical moiré pattern, and Figure 2b its fast Fourier transform (FFT) image. In Figure 2b, the vectors of the lattice point of the second-layer SnSe₂ and the moiré pattern are marked by yellow and red circles, respectively. According to the vectors of the second layer SnSe₂ and the moiré pattern, the vectors of the first-layer SnSe can be deduced to be a = 4.22 Å, b = 4.26 Å, consistent with the direct measurement on the bare SnSe layer. The stacking structure for this moiré pattern is then schematically drawn in Figure 2c, and the twist angle θ is calculated to be ~12° in the reciprocal space. To further verify the stacking order, we performed a simulation by adopting the experimentally obtained lattice constants and twist angle, and the simulated result, as shown in Figure 2d, perfectly reproduces the STM image in Figure 2a.

To better understand the observed dI/dV spectra of the SnSe₂/SnSe hetero-bilayer, we performed first-principles calculations using the HSE06 hybrid functional. For single-layer structures, the optimized lattice constants are a = 4.29 Å, b = 4.38 Å for SnSe, and a = 3.86 Å for SnSe₂. The calculated band gaps are 1.38 eV for SnSe and 1.42 eV for SnSe₂, as illustrated in the band structure plots shown in Figure 3a and Figure 3b, respectively. These results are in good agreement with experimental measurements as well as previous theoretical calculations. Previous studies have demonstrated that in 2D material heterojunction systems, the relatively weak van der Waals (vdW) interactions between layers result in their energy bands being primarily aligned according to their respective vacuum energy levels [50,51,52,53,54]. On this basis, we calculated the band profiles of individual SnSe and SnSe₂ monolayers relative to the vacuum level. As schematically shown in Figure 3c, the conduction band minimum (CBM) of the SnSe₂ monolayer is approximately 0.58 eV below the valence band maximum (VBM) of the SnSe monolayer, strongly suggesting that the SnSe₂/SnSe vdW hetero-bilayer is highly likely to form a type-III band alignment.

Next, we constructed SnSe₂/SnSe hetero-bilayer superstructures to examine their interlayer band alignment from first-principles calculation. Based on the experimentally observed moiré structure of the SnSe₂/SnSe hetero-bilayer presented above, we generated the bilayer heterostructure, as shown in Figure 3d. Each periodic unit consists of 28 SnSe₂ and 20 SnSe primitive units, with in-plane strains of approximately 1.6% along both basis vectors and an inter-basis angle deviation of less than 1°. We explored various interlayer stacking registries and included the relaxation of both the size and shape of the supercell in the structural optimization. The relaxed supercell of the most stable SnSe₂/SnSe hetero-bilayer structure has two in-plane basis vector lengths of 29.84 Å and 13.29 Å, which are approximately 1% smaller than those of the pristine SnSe₂ lattice and about 2% shorter than those of the pristine SnSe lattice, while the angle between the in-plane basis vectors remains nearly identical to that in the pristine SnSe₂ lattice. The optimized interlayer spacing between the lowest Se atom in the top SnSe₂ layer and the highest Sn atom in the bottom SnSe layer was approximately 2.6 Å. The difference between the average vertical coordinates of all bottom-layer Se atoms in the SnSe₂ layer and all top-layer Sn atoms in the SnSe layer was approximately 3.0 Å. We further investigated the electronic properties of the relaxed SnSe₂/SnSe hetero-bilayer. Figure 3e illustrates the layer-projected density of states (DOS) of the heterostructure. The results show that the Fermi level of the system resides within the conduction band of SnSe₂ and the valence band of SnSe, clearly revealing a type-III band alignment between the two layers. This significantly differs from most of the previous reports that concluded type-II band gap alignment [55,56,57]. This discrepancy might be attributed to the band edge shift due to the interfacial strain and charge transfer from the NbSe₂ substrate, pushing the alignment from type-II to type-III. This is in fact similar to the transition from type-II to type-III band gap alignment in the InAs/GaSb system [58].

Notably, the offset between the energy bands of SnSe₂ and SnSe in the heterostructure is smaller than that estimated from above the band profiles of the isolated monolayers. This discrepancy arises because, in the type-III band alignment, some electrons from the valence band of one-layer transfer to the conduction band of the other layer. These charge transfers induce a built-in counter electric field that effectively reduces the offset between the energy bands. To analyze the electron transfer between the two layers in real space, we calculated the difference between the total electron density of the heterostructure and the sum of the total electron densities of the isolated layers, where the latter were considered in the same structures as they were in the heterostructure. Figure 3f presents the electron density difference integrated over the 2D a-b plane and plotted along the vertical c axis. The positive regions indicate electron accumulation, while the negative regions indicate electron depletion upon the formation of the heterostructure from the two isolated layers. It can be seen that a portion of electrons indeed transfer from the lower SnSe layer to the upper SnSe₂ layer, aligning with the DOS results shown in Figure 3e. We further quantitatively estimated the number of electrons transferred between the layers, specifically, using the position where the electron density difference equals zero, as indicated by the red dots in Figure 3f, as the boundary between the two layers. Such a boundary plane is essentially consistent with the one determined by the Bader analysis method, where the boundary is defined by the plane of minimal total charge at the interface region [59].

It was predicted by first-principles calculations that the SnSe monolayer exhibits intrinsic ferroelectricity, whereas the SnSe bulk does not due to its centrosymmetric symmetry [60]. On the other hand, neither the SnSe2 monolayer nor the bulk exhibit ferroelectricity [61]. In our SnSe2/SnSe heterojunction, various factors such as charge transfer, interlayer coupling, and strain effect are expected to simultaneously occur at the interface. The combination of these effects may play a role in tuning the ferroelectricity, thus making the SnSe2/SnSe/NbSe2 heterojunction a promising platform to explore exotic ferroelectric properties.

3. Conclusions

Taking SnSe₂ and SnSe as an example, we demonstrated the success of the direct epitaxy of vdW hetero-bilayers, in which careful interface engineering plays an important role. The obtained hetero-bilayer exhibits an atomic-scale clean and uniform interface, and thus well-defined moiré patterns, showing potential in the exploration of exotic moiré physics. Furthermore, the verified type-III band gap alignment in the SnSe₂/SnSe hetero-bilayer is also expected to promote effective electron–hole separation and find applications in related fields.

4. Experimental Details

The epitaxy of the SnSe₂/SnSe hetero-bilayer on the NbSe₂ substrate was carried out in a molecular beam epitaxy (MBE) chamber under ultra-high vacuum (UHV, base pressure of 1 × 10⁻¹⁰ Torr). Before loading the sample to ultrahigh vacuum, a silicon wafer was attached underneath the NbSe₂ single crystal, for the purpose of direct current heating within an ultrahigh vacuum. The NbSe₂ single crystal as a substrate was in situ cleaved in the MBE chamber. The optimized temperature window for the SnSe₂/SnSe growth is between ~200 °C and ~220 °C. High-purity Sn and Se sources are evaporated from standard Knudsen-type effusion cells on the NbSe₂ substrate, giving rise to a direct growth of the SnSe₂/SnSe hetero-bilayer. During MBE growth, a Se-rich atmosphere is adopted to compensate for the loss of volatile Se molecules. The flux ratio of Sn:Se was set to ~1:10 by keeping the temperatures of the Sn and Se effusion cells at 800 °C and 85 °C, respectively. A growth rate of 30 min per bilayer was achieved. Post-growth in situ annealing was performed at the growth temperature for a duration of 20 min. The sample was then transferred to a low temperature scanning tunneling microscope (USM1500, Unisoku Co., Osaka, Japan) for scanning tunneling microscopy/spectroscopy characterization at ~77 K.

5. Computational Details

The first-principles density functional theory calculations were performed using the Vienna Ab Initio Simulation Package [62]. Valence electrons were treated using the projector-augmented wave (PAW) method [63,64]. The exchange and correlation functional was treated using the Perdew–Burke–Ernzerhof (PBE) [65] version of generalized gradient approximation (GGA) for structural relaxations and total energy calculations. To achieve more accurate band structures and band alignment, we employed the Heyd–Scuseria–Ernzerhof (HSE06) hybrid functional [66]. To model the freestanding 2D layers, the supercells contain a vacuum region of more than 15 Å. In the SnSe₂/SnSe hetero-bilayer calculations, we included the van der Waals corrections as parameterized in the semiempirical DFT-D3 method [67]. For k-point sampling, a Γ-centered 12 × 12 × 1 Monkhorst-Pack [68] k-mesh was used for the calculations of the single unit cells of monolayer SnSe and SnSe₂, and a k-mesh of similar density was employed for the SnSe₂/SnSe hetero-bilayer calculations. A consistent plane-wave energy cutoff of 300 eV was applied, and atomic structures were optimized until all forces were below 0.01 eV/Å.