Tungsten Oxide Mediated Quasi-van der Waals Epitaxy of WS2 on Sapphire

Conventional epitaxy plays a crucial role in current state-of-the art semiconductor technology, as it provides a path for accurate control at the atomic scale of thin films and nanostructures, to be used as the building blocks in nanoelectronics, optoelectronics, sensors, etc. Four decades ago, the terms “van der Waals” (vdW) and “quasi-vdW (Q-vdW) epitaxy” were coined to explain the oriented growth of vdW layers on 2D and 3D substrates, respectively. The major difference with conventional epitaxy is the weaker interaction between the epi-layer and the epi-substrates. Indeed, research on Q-vdW epitaxial growth of transition metal dichalcogenides (TMDCs) has been intense, with oriented growth of atomically thin semiconductors on sapphire being one of the most studied systems. Nonetheless, there are some striking and not yet understood differences in the literature regarding the orientation registry between the epi-layers and epi-substrate and the interface chemistry. Here we study the growth of WS2 via a sequential exposure of the metal and the chalcogen precursors in a metal–organic chemical vapor deposition (MOCVD) system, introducing a metal-seeding step prior to the growth. The ability to control the delivery of the precursor made it possible to study the formation of a continuous and apparently ordered WO3 mono- or few-layer at the surface of a c-plane sapphire. Such an interfacial layer is shown to strongly influence the subsequent quasi-vdW epitaxial growth of the atomically thin semiconductor layers on sapphire. Hence, here we elucidate an epitaxial growth mechanism and demonstrate the robustness of the metal-seeding approach for the oriented formation of other TMDC layers. This work may enable the rational design of vdW and quasi-vdW epitaxial growth on different material systems.

: Additional HAADF STEM cross section images of the epi growth samples in which the WS2 (upper brighter layer), the WO3 (ordered on the sapphire surface) and the sapphire (bulk-bottom, the [11][12][13][14][15][16][17][18][19][20] direction towards the paper). Figure S9: Additional HAADF STEM cross section images of the epi growth samples in WS2free areas. The WO3 (ordered on the sapphire surface) and the sapphire (bulkbottom, the [11][12][13][14][15][16][17][18][19][20] direction towards the paper). Figure S10: Additional HAADF STEM cross section images of the epi growth samples in WS2free areas. The WO3 (ordered on the sapphire surface) and the sapphire (bulkbottom, the [1-100] direction towards the paper). The inset in (c) shows the atomic arrangement at the interface, which we obtained from averaging several unit cells of a highresolution image.   In both cases high contrast atoms can be observed at the sapphire surface with the same structure. At the interface with the NW, few well-ordered W atomic layers are seen (1)(2)(3)(4), after that the arrangement seems to break into a more disordered structure, (b). (d) EDS analysis, from left to right, HAADF image, elemental mapping with Al, W, O and Pt. Again the yellow arrow in the W mapping shows the presence of tungsten in the NW-free areas. Figure S14: XPS characterization of a sample after 5 mins metal seeding and 1 min growth, (a), and after 30 mins metal seeding, (b). In (a) a small tungsten sub-oxide is seen (in red) while its contribution is much pronounced after 30 mins metal seeding, (b). The reason for the latter probably being the presence of the amorphous nanowires which have mixed oxide stoichiometry.   (c) STEM-DF image taken after acquisition of the EELS data shows that the WS2 layer is completely damaged, white arrow. (d) NNMF leads to three factors, which are attributed to sapphire substrate, interface region (WOx) and carbon protection layer. A factor for WS2 is not found due to the rapid damaging. The spectra reveal the C-K and O-K edge and a small peak at the W-N2,3 edge position is observed in the interface region. (e) Spatial distribution of different factors agree well with expectation from the image. Scale bar is 7 nm.

Density-Functional Theory Calculations
Our density-functional theory (DFT) calculations were carried out using the VASP software. 1 Our basis set included plane waves with a kinetic energy cutoff of 600 eV. The electrons were treated within the projector-augmented wave method; 2 we explicitly treated the following as valence electrons: 2s and 2p of oxygen; 3s and 3p of aluminum; and 5d, 6s, and 6p of tungsten. We used the revised Perdew-Burke-Ernzerhof functional for solids (PBEsol) 3 to handle exchange and correlation effects.
Optimization of bulk alumina resulted in hexagonal lattice parameters of a = 4.773 Å and c = 13.008 Å, in agreement with previous DFT calculations and experimental results (in all our optimization calculations the resulting forces were below 0.02 eV/Å). Bulk tungsten oxide can crystallize in a variety of distorted perovskite phases; we performed structural optimizations starting from each of the 20 unit cells available at the Open Quantum Materials Database (OQMD) 4 and we obtained similar results to those in a recent study of the polymorphism of this material: 5 the most stable phase has P-1 symmetry, and it is, in our case, 86 meV per formula unit below the simple-cubic phase. A comparison of the metal atoms in the relaxed structures shows that while in Al2O3 the Al atoms of a layer occupy the sites of an hexagonal two-dimensional lattice of parameter 4.773 Å, in WO3 the W atoms of a layer are displaced from the hexagonal lattice points by up to tenths of an Å, so that first neighbors are between 5.083 Å and 5.677 Å apart.
We proceeded to simulate slabs of Al2O3 with W atoms deposited on them by taking a unit cell like the one of Figure 5(a) and allowing for vacuum at the top and bottom of it (we used the same 6 layers of oxygen and 12 layers of aluminum of the bulk, but we used a simulation cell with a c parameter at least twice as large as the bulk one to account for that vacuum next to the surfaces). On top (and, to keep symmetry, on bottom) of this symmetric aluminum-terminated slab we added one layer of W and one layer of O. For the tungsten layer, we tried the two possible arrangements of hexagonal symmetry compatible with the aluminum layers (one that continues the same stacking pattern, and one that creates a stacking fault). Regarding the oxygen layer added, we tried the two orientations present in alumina, and the two orientations present in WO3 along [111]. We tried stacking both the W layer on top of the aluminum, and the oxygen layer on top of the aluminum. In all, we started with 16 possible configurations, which after atomic relaxation (keeping the lattice vectors constant) evolved to 6 inequivalent configurations. The lowest-energy configuration among these is oxygen terminated, and its outermost metal layer contains full W occupancy; it is the one represented in Figure 5(c) of the main text. The rest of slab configurations described in the main text were done following a similar procedure, but correcting the energy results for the presence of a surface dipole when needed. To compute the separation energies, we subtracted from the energy of the Al2O3 slab with deposited tungsten the energy of the isolated Al2O3 slab and the energy of the adequate isolated set of layers of tungsten and oxygen.  20 Step-Edge-Guided Nucleation and Growth of Aligned WSe2 on Sapphire via a Layer-over-Layer Growth Mode

M-O CVD
Step-edge guided growth [11][12][13][14][15][16][17][18][19][20] ∥ [11][12][13][14][15][16][17][18][19][20] Figure S20: X-ray fluorescence spectrum collected in 2000 s in the vicinity of the W L lines from a WS2/WO3/ Al2O3(0001) sample using 17.44 Kev X-rays with an incident intensity of 8x10 7 photons /s and at a 6° incident angle. The 50 mm 2 Si drift-diode XRF detector was facing the sample and 10 mm away from the 20 mm 2 radiated spot. Under identical conditions, spectra were collected from the WO3 sample and a Si/Ge/Si quantum well calibrated standard. The count rates were corrected for background counts and detector dead-time. To convert XRF cps into atoms per nm 2 we used the 17.44 keV relative XRF cross section for Ge K of 2891 and W LofThese calculated values are based on an extension of Puri et al. 22 22 22 The WS2-free sample had 5.2 W / nm 2 , the WS2/WO3/sapphire sample had 10.6 W/ nm 2 . This would imply 5.4 W/ nm 2 in the WS2 layer, which is equivalent to 0.5 ML of WS2.