Transition metal dichalcogenide (TMD) lateral heterostructures are a promising realization of atomically thin p-n junctions and optical devices. While experimental heterostructures show interfacial disorder across various length scales, the computational cost of first-principles modeling restricts analysis to near-pristine interfaces. We bridge this gap by combining a machine-learned tight-binding model with the recursive Green's functions technique for over 1300 explicitly disordered $\mathrm{Mo}{\mathrm{S}}_2\text{−}\mathrm{W}{\mathrm{S}}_2$ interface widths up to 40 nm. Using an information theoretic approach, we define an effective disorder width, $w_{\mathrm{eff}}$, to quantify the extent of on-lattice intermixing disorder. For a small disorder width, $w_{\mathrm{eff}}\le 6$ nm, we observe improvement in conduction band transparency compared to the pristine interfaces. Additionally, at higher energies in the conduction band, we find a transparency increase up to $3.66\times$, with improvements for $w_{\mathrm{eff}}$ up to 20 nm. We attribute this improvement to the opening of nonspecular transmission channels that relax transverse momentum conservation, enabling tunneling between otherwise mismatched valleys. Our results suggest that moderate interfacial intermixing can be used to optimize transmission across lateral interfaces.

Close

PrevNext

View figure in article

• Ballistic transport
• Disordered systems
• Transition metal dichalcogenides
• Green's function methods
• Tight-binding model