No abstract

to strong quantum confinement effect, electrons and holes are tightly bonded together in the TMDs, which is extensively harmful for generation and separation of free carriers. [12,13] Recently, thin-layered TMDs can be vertically stacked to create a van der Waals (vdW) heterojunction with type-II band alignment, providing spatial segregation of photogenerated electrons and holes to different layers. [14][15][16][17][18][19][20][21][22][23][24] For instance, n-type MoX 2 monolayer, where the X is sulfur (S) or selenium (Se), can be vertically stacked on p-type WX 2 monolayer to create vertical type-II p-n junction, and then drives vertical separation of holes and electrons. [18][19][20][21] Furthermore, interlayer excitons are generated in type-II vdW heterojunction, which are promising for applications in optoelectronic devices, photovoltaic applications and spin-valleytronic device, etc. [16][17][18][19][20] So far, most of TMD's vertical heterojunctions have been limited to the crystalline structure of the layered materials. [25] Many nonlayered materials based on group IV, III-V, or II-VI semiconductors also exhibit attractive photocatalytic, solar cells and optical detectors properties, etc. [25][26][27][28][29][30][31] Combination of such nonlayered functional semiconductors (e.g., CdS or PbS) with layered TMDs Van der Waals heterojunctions of 2D layered semiconductors and nonlayered technological important II-V semiconductors provide unprecedented opportunities to engineer exciton and carrier dynamics in 2D optoelectronic devices. However, fabrication of such artificial heterojunctions with type-II band alignment structure and realization of interlayer excitons is challenging. Here, CdS-MoS 2 type-II heterojunctions vertically stacked with few layered MoS 2 and ultrathin CdS film are reported. Steady-state spectroscopy and time-resolved photoluminescence spectroscopy are used to study exciton and carrier dynamics in these heterojunctions. The surface states of the ultrathin CdS film caused by dangling bonds mediate interlayer exciton emission located at 753 nm in a CdS-bilayer MoS 2 heterojunction via charge transition between the MoS 2 indirect band and the CdS valence band. As a contrast, the surface states of CdS impede the recombination of interlayer excitons in a CdS-monolayer MoS 2 heterojunction. These results are helpful for development of high-performance ultrathin optoelectronic and energy devices including light emission diodes, solar cells, and photodetectors. 2D Semiconductors2D transition metal dichalcogenides (TMDs) semiconductors hold great potentials in ultrathin optoelectronic devices, including light-emitting diodes, transistors, and detectors, etc. [1][2][3][4][5][6][7][8][9][10][11] Performance of these optoelectronics devices is highly determined by free carrier dynamics of TMDs from genera-

No abstract

Spiking Neural Networks (SNNs) have been attached great importance due to their biological plausibility and high energy-efficiency on neuromorphic chips. As these chips are usually resource-constrained, the compression of SNNs is thus crucial along the road of practical use of SNNs. Most existing methods directly apply pruning approaches in artificial neural networks (ANNs) to SNNs, which ignore the difference between ANNs and SNNs, thus limiting the performance of the pruned SNNs. Besides, these methods are only suitable for shallow SNNs. In this paper, inspired by synaptogenesis and synapse elimination in the neural system, we propose gradient rewiring (Grad R), a joint learning algorithm of connectivity and weight for SNNs, that enables us to seamlessly optimize network structure without retraining. Our key innovation is to redefine the gradient to a new synaptic parameter, allowing better exploration of network structures by taking full advantage of the competition between pruning and regrowth of connections. The experimental results show that the proposed method achieves minimal loss of SNNs' performance on MNIST and CIFAR-10 datasets so far. Moreover, it reaches a ~3.5% accuracy loss under unprecedented 0.73% connectivity, which reveals remarkable structure refining capability in SNNs. Our work suggests that there exists extremely high redundancy in deep SNNs. Our codes are available at https://github.com/Yanqi-Chen/Gradient-Rewiring.