No abstract

2D metal-semiconductor heterostructures based on transition metal dichalcogenides (TMDs) are considered as intriguing building blocks for various fields, such as contact engineering and high-frequency devices. Although, a series of p-n junctions utilizing semiconducting TMDs have been constructed hitherto, the realization of such a scheme using 2D metallic analogs has not been reported. Here, the synthesis of uniform monolayer metallic NbS on sapphire substrate with domain size reaching to a millimeter scale via a facile chemical vapor deposition (CVD) route is demonstrated. More importantly, the epitaxial growth of NbS -WS lateral metal-semiconductor heterostructures via a "two-step" CVD method is realized. Both the lateral and vertical NbS -WS heterostructures are achieved here. Transmission electron microscopy studies reveal a clear chemical modulation with distinct interfaces. Raman and photoluminescence maps confirm the precisely controlled spatial modulation of the as-grown NbS -WS heterostructures. The existence of the NbS -WS heterostructures is further manifested by electrical transport measurements. This work broadens the horizon of the in situ synthesis of TMD-based heterostructures and enlightens the possibility of applications based on 2D metal-semiconductor heterostructures.

Chinese Spelling Check (CSC) is a task to detect and correct spelling errors in Chinese natural language. Existing methods have made attempts to incorporate the similarity knowledge between Chinese characters. However, they take the similarity knowledge as either an external input resource or just heuristic rules. This paper proposes to incorporate phonological and visual similarity knowledge into language models for CSC via a specialized graph convolutional network (SpellGCN). The model builds a graph over the characters, and SpellGCN is learned to map this graph into a set of inter-dependent character classifiers. These classifiers are applied to the representations extracted by another network, such as BERT, enabling the whole network to be end-to-end trainable. Experiments 1 are conducted on three human-annotated datasets. Our method achieves superior performance against previous models by a large margin.