Discrete nature of thickness and flat basal planes of two-dimensional (2D) nanostructures display unique diffraction features. Their origin was uncovered by a new analysis method of powder X-ray diffraction, which reveals thickness and lattice orientation of the 2D nanostructures. Results indicate necessity of adoption of a different unit cell from the corresponding bulk crystal with the same internal atomic packing. For CdSe 2D nanostructures with zinc blende atomic packing, pseudotetragonal lattices are adequate, instead of face-centered cubic.

Zinc-blende CdS nanoplatelets with atomically flat and very large {100} basal planes terminated solely by one type of element (either Cd or S atoms) are synthesized. Optical spectroscopy, X-ray diffraction, X-ray absorption, and transmission electron microscopy confirm that the surface structures of newly developed S-terminated CdS nanoplatelets are at least as well-defined as the original Cd-terminated nanoplatelets. Band gaps of the nanoplatelets are found to depend on not only the quantum-confined dimension (thickness) but also the nature of the surface termination. The facet structure dictates the packing of the ligands (carboxylate for Cd-terminated nanoplatelets and alkyl for S-terminated nanoplatelets), which causes a difference in the lattice strain and significantly affects the optical spectral width. Experimental and theoretical results reveal that engineering the exciton spatial distribution by the tailored synthesis of semiconductor nanocrystals with a precisely controlled surface structure is fully possible, which should open a new door for delivering the long-promised potential of semiconductor nanocrystals.

This paper proposes a new graph convolutional neural network architecture based on a depth-based representation of graph structure deriving from quantum walks, which we refer to as the quantum-based subgraph convolutional neural network (QS-CNNs). This new architecture captures both the global topological structure and the local connectivity structure within a graph. Specifically, we commence by establishing a family of K-layer expansion subgraphs for each vertex of a graph by quantum walks, which captures the global topological arrangement information for substructures contained within a graph. We then design a set of fixed-size convolution filters over the subgraphs, which helps to characterise multi-scale patterns residing in the data. The idea is to apply convolution filters sliding over the entire set of subgraphs rooted at a vertex to extract the local features analogous to the standard convolution operation on grid data. Experiments on eight graph-structured datasets demonstrate that QS-CNNs architecture is capable of outperforming fourteen state-of-the-art methods for the tasks of node classification and graph classification.

Network representation learning (NRL) aims to map vertices of a network into a low-dimensional space which preserves the network structure and its inherent properties. Most existing methods for network representation adopt shallow models which have relatively limited capacity to capture highly nonlinear network structures, resulting in sub-optimal network representations. Therefore, it is nontrivial to explore how to effectively capture highly nonlinear network structure and preserve the global and local structure in NRL. To solve this problem, in this pap er we prop ose a new graph convolutional autoencoder architecture based on a depth-based representation of graph structure, referred to as the depth-based subgraph convolutional autoencoder (DS-CAE), which integrates both the global topological and local connectivity structures within a graph. Our idea is to first decompose a graph into a family of K-layer expansion subgraphs rooted at each vertex aimed at better capturing long-range vertex inter-dependencies. Then a set of convolution filters slide over the entire sets of subgraphs of a vertex to extract the local structural connectivity information. This is analogous to the standard convolution operation on grid data. In contrast to most existing models for unsupervised learning on graph-structured data, our model can capture highly non-linear structure by simultaneously integrating node features and network structure into network representation learning. This significantly Many naturally occurring problems can be represented using an underlying graph or network structure, such as natural language processing [1], computer vision [2], or social network analysis [3]. It is therefore crucial to accurately extract useful information from a network. One promising strategy is to map vertices of a network into a low-dimensional space, i.e. learn vector representations for each vertex, with the goal of encoding meaningful information conveyed by the graph in the learned embedding space. As a result, the learned representation can be used directly with existing machine learning methods to perform various network analysis tasks, such as vertex classification [3] and clustering [4].As u c c e s s f u ln e t w o r kr e p r e s e n t a t i o nl e a r n i n gm o d e ls h o u l de x h i b i tt w o desirable properties: a) a High non-linearity: As [5] stated, the structure and inherent properties of a network are highly non-linear. b) Structurepreserving [6]: The learned representation should preserve both the global topological arrangement information (e.g. long-range dependencies) and the local connectivity structure of a network.Traditional metho ds for graph dimensionality reduction such as Lo cal Linear Embedding (LLE) [7], Laplacian eigenmap [8] and IsoMAP [9,10] are based on a graph affinity matrix decomposition, which treats the leading eigenvectors as a representation. However, the time complexity of these methods is at least quadratic in the number of graph nodes, which makes them not applicable to large-scale netw...

We have systematically investigated the geometrical structures, relative stabilities and electronic properties of small bimetallic AunBe (n = 1, 2, . . . , 8) clusters using a density functional method at BP86 level. The optimized geometries reveal that the impurity beryllium atom dramatically affects the structures of the Aun clusters. The averaged binding energies, fragmentation energies, second-order difference of energies, the highest occupied-lowest unoccupied molecular orbital energy gaps and chemical hardness are investigated. All of them exhibit a pronounced odd-even alternation, manifesting that the clusters with even number of gold atoms possess relatively higher stabilities. Especially, the linear Au 2 Be cluster is magic cluster with the most stable chemical stability. According to the natural population analysis, it is found that charge-transferring direction between Au atom and Be atom changes at the size of n = 4.