Computational Study of Interfaces between 2D MoS2 and Surroundings

The extracted mobilities from MoS2 devices are usually much lower than the theoretical value. Without understanding the origin of mobility killers (surrounding dipoles/traps) and the effects of substrate passivation treatments, it is difficult to improve transport properties. This work presents a study of the interfaces between MoS2 and its surroundings (substrates/dielectrics) using density functional theory (DFT). Introduction: Two-dimensional (2D) transition metal dichalcogenide material MoS2 has been proposed as one of the promising candidates for next-generation electronics 1, 2 due to its unique properties, such as pristine surfaces, ultra-small thickness and uniform bandgap (1.8 eV). However, the low mobility extracted from MoS2 devices greatly limits the current as well as the switching speed. Though high-K dielectrics strengthen the confinement of electric flux and reduce Coulomb scattering , however, in the microscopic view, the interaction between MoS2 and any surrounding materials (substrates/gate dielectrics) is unfortunately inevitable and still remains unclear, especially the origin of dipoles and traps that affect the mobility. In this work, using DFT, we comprehensively study the interfaces between MoS2 and surrounding materials. Various materials are examined, including the commonly used 3D SiO2, Al2O3, HfO2, 2D hexagonal BN (h-BN) and 1D poly(methyl methacrylate) (PMMA). Impact of different substrate passivation treatments is also discussed. Methodology: The interface unit cells are periodical in x and y direction and separated by vacuum in z direction 4 (Fig.1a). For 3D surrounding materials, the thicknesses are ≥10 Å and the most widely adopted crystalline phases are chosen – I-42d β-cristobalite (Fig.1b) , r-3c α-Al2O3 (Fig.1c) and P21/c m-HfO2 (Fig.1d). The bottom dangling O atoms are H-terminated and are fixed at bulk locations to reduce size effects , while other atoms are allowed to relax. Numerical DFT calculations are performed using Atomistix ToolKit 7 with Local Density Approximation 8 and double-ζ polarized basis set. The Brillouin zones are sampled by 4×8×1 k-points for SiO2 or HfO2 interfaces, 4×4×1 for Al2O3 and 8×8×1 for PMMA or h-BN. This methodology employed 9 has already shown good agreement with experimental results 10, . Results: (1) As shown in Fig. 1e-g, among the 3D substrate/dielectric materials, Al2O3 has the largest dipole depth (Δz) from surfaces. Doping effects (Fermi potential eΦF) are sorted by Al2O3 < SiO2< HfO2 (Fig.2a-d), indicating MoS2-Al2O3 interface forms the least numbers of doping dipoles. Hence, among the 3D materials, Al2O3 is recommended due to its greatest potential to preserve the carrier mobility in MoS2. (2) In MoS2-HfO2 system, MoS2 suffers from shallow dipoles in HfO2 (Fig. 1g) and high doping (Fig.2d). Moreover, overlap states (Fig.2e) form high-effective-mass bands in the bandgap (Fig.2f), which act as traps, thereby reducing mobility. Hence, dipoles and traps in HfO2 offset its benefits of high permittivity. (3) A proper passivation treatment of bulk substrates is a necessity. At defective SiO2 surface, interface bonds are formed between SiO2 and MoS2 (Fig.3a,b), which undermine the MoS2 band structure (Fig.3c,d), leading to interface traps. Moreover, E-k distortion is observed in the original Ec/Ev (Fig.3c) resulting in the increase of effective masses and thereby, in reduction of mobility. HF-dipping treatment induces methyl atomic groups at SiO2 surface , and reduces charge transfer by enlarging Δz (Fig.3e,f). Hence, such treatment can reduce the formation of dipoles and weaken the scattering from dipoles and traps in SiO2. (4) h-BN (Fig.4a) and PMMA (Fig.4b-d) do not dope MoS2 (Fig.4e). Interface with h-BN has the largest dipole (multipole) depth (Fig.4f), but interface with PMMA has very shallow dipoles (multipoles) (Fig.4g), indicating that dipole scattering from h-BN is weak while PMMA may induce strong scattering in MoS2 thereby degrading its mobility. Nevertheless, the low permittivity of h-BN limits its use for boosting the mobility. Hence, it can be predicted that using h-BN as a buffer layer between high-K materials and MoS2 may offer the advantage of both high permittivity of high-K material and interface of h-BN. Summary: The systematic results obtained via ab-initio DFT simulations in this work suggest that the mobility-degrading effects of surrounding dipoles/traps on monolayer MoS2 devices can be suppressed by selecting the optimal substrate or gate dielectric material (Al2O3 or h-BN+high-K) and by passivation treatment of substrate (by HF), which can help optimize the transport properties of MoS2 under embedded conditions. Moreover, the significance of the methodology is apparent for a broad range of 2D materials. Fig.1: (a) 3D view of a relaxed unit cell. (b-d) side and bottom views of monolayer (1L) MoS2 on H-passivated (b) SiO2, (c) Al2O3 and (d) HfO2. (e-g) Electron differential density (EDD) contours of (b-d). Hydroxyl (OH) atomic groups form dipoles (circled) in dielectrics. Color bar in (e) is common for all contours. Fig.2: Partial density of states (PDOS) of 1L MoS2 (a) without (w/o) surrounding and (b-d) with (w/) passivated (b) SiO2, (c) Al2O3 and (d) HfO2. (e) PDOS curves of trap states in the first 3 layers of bulk surrounding materials. (f) band structure of original MoS2 and distorted band structure of MoS2 with passivated HfO2. Fig.3: (a) relaxed unit cell of 1L MoS2 on SiO2 with unpassivated surface. (b) electron density isosurface (surface representing points with constant electron density value 0.8 Å) of (a). S and O atoms are found to be covered by the isosurface indicating formation of S-O bonds. (c) band structure of (a); (d) PDOS of MoS2 in (a). (e, f) relaxed unit cells of 1L MoS2 on passivated SiO2 terminated by H and CH3 at different positions. Fig.4: (a) 3D and top view of the relaxed unit cell consisting of interface between 1L MoS2 and h-BN. (b-d) views of the relaxed unit cells with 1L MoS2 and PMMA with different PMMA orientations. Inset figure in (d) shows the repeating unit and molecular formula of PMMA. (e) PDOS of MoS2 with h-BN and PMMA. Since MoS2 in samples (b-d) show similar PDOS, only PDOS of MoS2 in (d) is shown. (f, g) EDD of (a, d). ─────────── 1 B. Radisavljevic, et al., Nature Nanotech., 6 (2011). 2 S. Bertolazzi, et al., ACS Nano, 7, 4 (2013). 3 D. Jena, et al., Phys. Rev. L, 98,136805 (2007). 4 J. Kang, et al., Phys. Rev. X, 4, 031005 (2014). 5 D. E. Jiang, et al., Phys. Rev. B, 72, 165410 (2005). 6 T.-R. Shan, et al., Phys. Rev. B, 83, 115327 (2011). 7 Atomistix ToolKit v. 12.8.2, QuantumWise A/S. 8 J. P. Perdew, et al., Phys. Rev. B, 23, 5048 (1981). 9 J. Kang, et al., IEDM 2012, 407-410. 10 W. Liu, et al., IEDM 2013, 499-502. 11 J. Kang, et al., Appl. Phys. Lett., 104, 093106 (2014). 12 K. Nagashio, et al., IEDM 2010, 564-567.