Jiajie
Zhu
and
Udo
Schwingenschlögl
*
PSE Division, King Abdullah University of Science and Technology, Thuwal 23955-6900, Kingdom of Saudi Arabia. E-mail: udo.schwingenschlogl@kaust.edu.sa
First published on 17th March 2015
Many semiconducting substrates, such as GaS and MgBr2, have been explored for silicene. However, large lattice mismatches, complicated control of terminal layers and small band gaps are critical limiting factors. First-principles results on the stability and electronic properties of silicene on WSe2 show that the energy barriers for lateral translation between the two subsystems are very small due to weak van der Waals interactions. For the same reason, the Dirac physics of silicene is preserved. It turns out that the induced band gap is sufficient to withstand thermal fluctuations.
Although silicene has been successfully deposited on several metallic substrates, including ZrB2(0001),9 Ir(111)10 and Ag(111),11,12 the Dirac characteristics are destroyed because of hybridization of π bands with the substrate.13–15 On the other hand, semiconducting substrates have been investigated in order to avoid such strong interaction. For example, GaS nanosheets have been predicted to yield a Dirac cone with a 170 meV band gap, while the lattice mismatch of 7.5% casts doubts on the stability of this hybrid system.16 Kokott and coworkers have found for silicene on H(Cl)-passivated Si(111) substrate (lattice mismatch less than 1%) linearly dispersing π bands with a band gap of 3(56) meV.17,18 Similar results have been reported for H-passivated Si- and C-terminated SiC(0001).19,20 Although Dirac physics can be achieved on MgX2(0001) (X = Cl, Br and I) substrates without dangling surface bonds, which simplifies the synthesis, band gaps below 16 meV strongly limit the applicability.21 A band gap of 52 meV is induced by F-terminated CaF2(111), but control of the preparation process is very problematic.18
A band gap of some 100 meV is typically desirable to overcome thermal fluctuations at room temperature. Several approaches have been studied to open such a band gap in graphene, including external electric fields22,23 and multilayer stacking,24,25 which may also be useful for silicene. However, the experimentally accessible electric fields limit the band gap to about 30 meV.26,27 Moreover, adjacent Si layers are predicted to form clusters instead of bilayer silicene, which completely destroys any Dirac states because of the strong interaction.21,28 While adsorbed metal atoms are predicted to open a band gap in freestanding silicene, they may aggregate instead of maintaining a uniform distribution.29,30 WSe2 is a semiconductor with a hexagonal structure consisting of sandwich-like layers along the [001] direction without dangling bonds.31 It has minimal lattice mismatch with low-buckled silicene among the transition metal dichalcogenides, consequently overcoming the practical obstacles caused by related substrates with much larger lattice mismatch.32 Since the van der Waals interaction is expected to be stronger than for MgBr2 (more electrons are contributing to the dipole), the band gap of silicene on WSe2 is expected to be enhanced. Therefore, WSe2 appears to fulfill all criteria of a suitable substrate for silicene. We will investigate in this work the stability and electronic properties of this hybrid system and will demonstrate very encouraging results.
The in-plane lattice constant of WSe2(001) is found to be 3.313 Å and thus similar to the experimental and previous theoretical values of 3.282 Å and 3.31 Å, respectively.36 The lattice mismatch between 2 × 2 WSe2 and low-buckled silicene is only 0.6%. High-buckled silicene is also considered, since it has been prepared on MoS2.37 Furthermore, monolayer WSe2 has been obtained by mechanical exfoliation.38 We use a vacuum layer of 15 Å thickness to avoid unphysical interaction between images due to the periodic boundary conditions. Dipole corrections result only in tiny energy variations of less than 0.1 meV per atom. Low-buckled silicene on bilayer WSe2 is investigated for comparison with the monolayer case. In addition, S and Te atoms are substituted for Se (concentration 12.5%) to modify the stability of the hybrid system as well as the band gap of silicene.
S doping | Pristine | Te doping | |
---|---|---|---|
b sil (Å) | 0.51 | 0.51 | 0.49 |
d sil-sub (Å) | 3.19 | 3.20 | 3.03 |
a (Å) | 3.829 | 3.842 | 3.861 |
E (meV) | 112 | 122 | 135 |
ΔH (eV) | 0.01 | 0.32 | |
d W-dop (Å) | 2.42 | 2.54 | 2.73 |
The variation of the total energy of the hybrid system as a function of the lateral shift along the [100] direction with respect to the unshifted structure is illustrated in Fig. 2. Note that the [100] and [010] directions are equivalent due to the three-fold rotational symmetry of the lattice. The unshifted structure in addition has mirror symmetry and the translational periodicity is half of the cell size due to the 2 × 2 WSe2 supercell. The maximal translation energy barrier is found to be less than 0.4 meV per atom, which reflects a high uncertainty in the structure of the hybrid system at room temperature as well as below. The binding energy of the two components is also addressed in Fig. 2, showing a variation of less than 2 meV per Si atom with distinct valleys at lateral shifts of ±0.25 along the [100] direction.
Fig. 2 Total energy per atom with respect to the unshifted structure and binding energy per Si atom for low-buckled silicene on WSe2 as a function of the lateral shift along the [100] direction. |
We next substitute S or Te for Se to modify the interaction at the interface, see Fig. 1. Different configurations of the dopant atoms ranging from clusters to a homogeneous distribution are compared. The latter turns out to be energetically favorable due to small local distortions around the dopant atoms, which are located on top of the Si atoms for 1/6 and 1/3 lateral shifts, respectively, and occupy hollow and bridge sites for 0 and 1/2 lateral shifts. The corresponding formation energies are 0.01 and 0.32 eV (Table 1). The W–S (2.42 Å) and W–Te (2.73 Å) bond lengths deviate from the W–Se bond length (2.54 Å) as expected from the ionic radii (1.70, 1.84 and 2.07 Å for S, Se and Te, respectively). Smaller (3.829 Å) and larger (3.861 Å) in-plane lattice constants are predicted for S and Te doping. In addition, the distance between silicene and the substrate remaines similar for S doping (3.19 Å) but is significantly reduced for Te doping (3.03 Å), see Table 1. For S doping a ±1/3 lateral shift is energetically favorable, see Fig. 2, whereas the unshifted structure now has the highest total energy. The resistance against translation is enhanced substantially as compared to the pristine case, which is a consequence of stronger variations of the binding energy. Under Te doping the unshifted structure is most stable and the maximal translation energy barrier is enhanced to 7 meV, which corresponds to a binding energy difference of 19 meV per Si atom (including 2 meV of relaxation energy from the substrate).
Fig. 3 shows densities of states without shift and with a lateral shift of 1/3 (most and least stable configurations for different dopings). The top of the valence band and the bottom of the conduction band are dominated by Si 3p states. Hybridization between the substrate and Si 3p states extends closer to the valence band maximum for Te doping than for S doping due to the smaller distance to the silicene. Moreover, larger hybridizations between the S/Te p and Si 3p states are characteristic for the unshifted structures. The band structures in Fig. 4 show the Dirac cone, as expected, at the Γ point. While alkali metal intercalation can also restore the Dirac cone of double-layered silicene, the present system is less difficult to handle experimentally.40 Without lateral shift we obtain band gaps of 0.30, 0.32 and 0.34 eV for S doped, pristine and Te doped WSe2. The slight increase reflects the growing interaction at the interface, compare Table 1. Spin–orbit coupling reduces the band gap of the pristine system without lateral shift by 2.8 meV due to spin splitting (2.0 meV in the valence band and 1.5 meV in the conduction band), which is negligible as compared to the original value. On the other hand, the band gap is close to zero for the shifted structures because of weaker interaction: in the unshifted case S/Se/Te occupies a hollow site and thus has six Si neighbours, whereas in the shifted case it occupies a top site with only a single interaction partner. Although Se vacancies are inevitable in monolayer WSe2,41 they only slightly reduce the band gap, for example by 0.06 eV for undoped WSe2 without lateral shift (one Se vacancy in a 4 × 4 supercell), and leave the principal features of the band structure unaffected.
Fig. 3 Density of states of low-buckled silicene on (a, b) S doped, (c, d) pristine and (e, f) Te doped WSe2 without lateral shift (left row) and with a 1/3 lateral shift (right row). |
Fig. 4 Band structure of low-buckled silicene on (a, d) S doped, (b, e) pristine and (c, f) Te doped WSe2 without lateral shift (upper row) and with a 1/3 lateral shift (lower row). |
The position of the Dirac point (middle point between the π and π* bands) with respect to the Fermi level and the corresponding band gap are illustrated in Fig. 5. Doping is found to have hardly any effect on both these quantities. On the other hand, the band gaps are reduced to almost zero for the 1/6 and 1/3 lateral shifts, which are structurally similar with the W and Se or dopant atoms on top of the Si atoms. Since the translation energy barrier is quite small for the pristine substrate, thermal fluctations even at low temperature can modify the electronic properties dramatically, ranging from metallic to semiconducting characters. On the contrary, stable metallic and semiconducting silicene is obtained by S and Te doping, respectively. In each case the Dirac point is located less than 0.16 eV above the Fermi level, reflecting weakly p-doped silicene. The position of the Dirac point follows the same trend as the band gap, which suggests that the value of the band gap is determined by the amount of charge redistribution between silicene and substrate. Indeed, the internal electric field created at the interface breaks the symmetry of the two silicene sublattices, which opens the band gap.
Fig. 5 Position of the Dirac point with respect to the Fermi level and band gap as a function of the lateral shift along the [100] direction for low-buckled silicene. |
The charge density difference Δρ = ρhyb + ρsil + ρsub (ρhyb, ρsil and ρsub being the charge densities of the hybrid system, silicene and substrate, respectively) due to interaction of silicene with WSe2 is shown in Fig. 6 for the cases without and with a lateral 1/3 shift. Both ρsil and ρsub are calculated with the same parameters as ρhyb. From left to right Fig. 6 shows an enhanced charge density redistribution at the interface of the two subsystems for the unshifted structure (upper row), reflecting the growing interaction (the number of valence electrons grows along the series S–Se–Te). Charge also shifts within the silicene sheet from the lower sublattice to the upper sublattice. Under the lateral shift (lower row) the charge density redistribution at the interface is substantially weaker.
The structure of high-buckled silicene on WSe2 is shown in Fig. 7 for different locations of the Si atoms, namely on top of W, Se and the hollow site. The first case turns out to be energetically favorable by 10 and 3 meV per atom with respect to the other cases, respectively. The Si buckling height is found to be 1.3 Å, which is smaller than in the case of high-buckled silicene on the MoS2 substrate (2.0 Å) due to a reduced lattice compression.37 The larger binding energy (195 meV) as compared to low-buckled silicene (122 meV) is a consequence of the smaller distance to the substrate (2.55 Å instead of 3.20 Å). This strong interaction destroys the Dirac cone, see Fig. 7(d), in contrast to the low-buckled system and similar to previous results on the MoS2 substrate.37,42
Using WSe2 as a support will open access to the unique properties of silicene, which so far could not be utilized as they were always perturbed by the substrate. Of particular interest are the unusual performances of silicene-based spintronic devices8 and the potential to obtain ultra high speed (THz frequency range) field-effect transistors. Such field-effect transistors show a high on/off current ratio43 but require a complicated preparation procedure.44 All this is only possible if the fundamental shortcomings of the current substrates are resolved. On the one hand, the Dirac cone has to be preserved, which excludes many potential candidates. On the other hand, the lattice mismatch has to be small to avoid strain effects on silicene, the termination must not cause complications, and the band gap must be intrinsically large enough to operate at standard temperatures. All these criteria can be achieved by utilizing WSe2 to exploit the intrinsic properties of silicene, thus avoiding, for example, the large supply voltage required when adatoms are used to open the band gap.45 It should be noted that silicene already has been grown epitaxially on MoS2,37 which is isostructural to WSe2 and has very similar physical and chemical properties, so that the same preparation method can be expected to be successful.
This journal is © The Royal Society of Chemistry 2015 |