arXiv subjects

Fan Zhang

Publications and source records attributed to Fan Zhang.

At least 847 records · Page 47Linked to original sources

arXiv2018

Memristor-based Deep Convolution Neural Network: A Case Study

In this paper, we firstly introduce a method to efficiently implement large-scale high-dimensional convolution with realistic memristor-based circuit components. An experiment verified simulator is adapted for accurate prediction of analog crossbar behavior. An improved conversion algorithm is developed to convert convolution kernels to memristor-based circuits, which minimizes the error with consideration of the data and kernel patterns in CNNs. With circuit simulation for all convolution layers in ResNet-20, we found that 8-bit ADC/DAC is necessary to preserve software level classification accuracy.

cs.NE↗

arXiv2018

A photonic interface of chiral cavity quantum electrodynamics

Cavity quantum electrodynamics studies light-matter interactions at single quanta level. Chiral photon-emitter coupling in photonic structures is characterized as unidirectional propagation locked by the local polarization of light. However, to realize the strong interaction of photon-emitter with direction-locked propagation, which will bring novel applications in nonreciprocal quantum devices, has not been reported yet. Here, we propose the coupled photonic crystal and metallic nanoparticle structure, where through strong local field with high helicity, the rate of circularly polarized photons emitting into photonic crystal waveguide is one order larger than that without the nanoparticle and the linewidth of Rabi splitting spectra is about one-tenth of that with only the nanoparticle, both with $\sim$ 95$\%$ photons propagating unidirectionally, which can be utilized in directional quantum light sources.

physics.optics↗

arXiv2018

TreeSegNet: Adaptive Tree CNNs for Subdecimeter Aerial Image Segmentation

For the task of subdecimeter aerial imagery segmentation, fine-grained semantic segmentation results are usually difficult to obtain because of complex remote sensing content and optical conditions. Recently, convolutional neural networks (CNNs) have shown outstanding performance on this task. Although many deep neural network structures and techniques have been applied to improve the accuracy, few have paid attention to better differentiating the easily confused classes. In this paper, we propose TreeSegNet which adopts an adaptive network to increase the classification rate at the pixelwise level. Specifically, based on the infrastructure of DeepUNet, a Tree-CNN block in which each node represents a ResNeXt unit is constructed adaptively according to the confusion matrix and the proposed TreeCutting algorithm. By transporting feature maps through concatenating connections, the Tree-CNN block fuses multiscale features and learns best weights for the model. In experiments on the ISPRS 2D semantic labeling Potsdam dataset, the results obtained by TreeSegNet are better than those of other published state-of-the-art methods. Detailed comparison and analysis show that the improvement brought by the adaptive Tree-CNN block is significant.

cs.CV↗

arXiv2018

Topological negative refraction of surface acoustic waves in a Weyl phononic crystal

Reflection and refraction occur at interface between two different media. These two fundamental phenomena form the basis of fabricating various wave components. Specifically, refraction, dubbed positive refraction nowadays, appears in the opposite side of the interface normal with respect to the incidence. Negative refraction, emerging in the same side by contrast, has been observed in artificial materials1-5 following a prediction by Veslago6, which has stimulated many fascinating applications such as super-resolution imaging7. Here we report the first discovery of negative refraction of the topological surface arc states of Weyl crystals, realized for airborne sound in a novel woodpile phononic crystal. The interfaces are one-dimensional edges that separate different crystal facets. By tailoring the surface terminations of such a Weyl phononic crystal, open equifrequency contours of surface acoustic waves can be delicately designed to produce the negative refraction, to contrast the positive counterpart realized in the same sample. Strikingly different from the conventional interfacial phenomena, the unwanted reflection can be made forbidden by exploiting the open nature of the surface equifrequency contours, which is a topologically protected surface hallmark of Weyl crystals8-12.

cond-mat.mtrl-sci↗

arXiv2018

Observation of acoustic Landau quantization and quantum-Hall-like edge states

Many intriguing phenomena occur for electrons under strong magnetic fields. Recently, it was proposed that an appropriate strain texture in graphene can induce a synthetic gauge field, in which the electrons behave like in a real magnetic field. This opened the door to control quantum transport by mechanical means and to explore unprecedented physics in high-field regime. Such studies have been achieved in molecular and photonic lattices. Here we report the first experimental realization of giant uniform pseudomagnetic field in acoustics by introducing a simple uniaxial deformation to acoustic graphene. Benefited from the controllability of our macroscopic platform, we observe the acoustic Landau levels in frequency-resolved spectroscopy and their spatial localization in pressure-field distributions. We further visualize the quantum-Hall-like edge states (connected to the zeroth Landau level), which have been elusive before owing to the challenge in creating large-area uniform pseudomagnetic fields. These results, highly consistent with our full-wave simulations, establish a complete framework for artificial structures under constant pseudomagnetic fields. Our findings, conceptually novel in acoustics, may offer new opportunities to manipulate sound.

cond-mat.mes-hall↗

arXiv2018

High-speed Tracking with Multi-kernel Correlation Filters

Correlation filter (CF) based trackers are currently ranked top in terms of their performances. Nevertheless, only some of them, such as KCF~\cite{henriques15} and MKCF~\cite{tangm15}, are able to exploit the powerful discriminability of non-linear kernels. Although MKCF achieves more powerful discriminability than KCF through introducing multi-kernel learning (MKL) into KCF, its improvement over KCF is quite limited and its computational burden increases significantly in comparison with KCF. In this paper, we will introduce the MKL into KCF in a different way than MKCF. We reformulate the MKL version of CF objective function with its upper bound, alleviating the negative mutual interference of different kernels significantly. Our novel MKCF tracker, MKCFup, outperforms KCF and MKCF with large margins and can still work at very high fps. Extensive experiments on public datasets show that our method is superior to state-of-the-art algorithms for target objects of small move at very high speed.

cs.CV↗

arXiv2018

Topological Triply Degenerate Points Induced by Spin-Tensor-Momentum Couplings

The recent discovery of triply degenerate points (TDPs) in topological materials has opened a new perspective toward the realization of novel quasiparticles without counterparts in quantum field theory. The emergence of such protected nodes is often attributed to spin-vector-momentum couplings. Here we show that the interplay between spin-tensor- and spin-vector-momentum couplings can induce three types of TDPs, classified by different monopole charges ($\mathcal{C}=\pm 2,\pm 1,0$). A Zeeman field can lift them into Weyl points with distinct numbers and charges. Different TDPs of the same type are connected by intriguing Fermi arcs at surfaces, and transitions between different types are accompanied by level crossings along high-symmetry lines. We further propose an experimental scheme to realize such TDPs in cold-atom optical lattices. Our results provide a framework for studying spin-tensor-momentum coupling-induced TDPs and other exotic quasiparticles.

cond-mat.mes-hall↗

arXiv2018

Intrinsic valley Hall transport in atomically thin MoS2

Electrons hopping in two-dimensional honeycomb lattices possess a valley degree of freedom in addition to charge and spin. In the absence of inversion symmetry, these systems were predicted to exhibit opposite Hall effects for electrons from different valleys. Such valley Hall effects have been achieved only by extrinsic means, such as substrate coupling, dual gating, and light illuminating. Here, we report the first observation of intrinsic valley Hall transport without any extrinsic symmetry breaking in the non-centrosymmetric monolayer and trilayer MoS2, evidenced by considerable nonlocal resistance that scales cubically with local resistance. Such a hallmark survives even at room temperature with a valley diffusion length at micron scale. By contrast, no valley Hall signal is observed in the centrosymmetric bilayer MoS2. Our work elucidates the topological quantum origin of valley Hall effects and marks a significant step towards the purely electrical control of valley degree of freedom in topological valleytronics.

cond-mat.mes-hall↗

arXiv2018

Flavor Symmetry and Ferroelectric Nematics in Transition Metal Dichalcogenides

Recent magneto-transport experiments have provided compelling evidence for the presence of an energetically isolated threefold Q-valley degeneracy in few-layer transition metal dichalcogenides. We study the flavor SU(3) symmetry breaking when each Landau level triplet is one-third filled or empty and predict that a pure flavor nematic phase and a flavorless charge-density-wave phase will occur respectively below and above a critical magnetic field. Surprisingly, electrons carry flavor-dependent electric dipole moments even at zero magnetic field, rendering the nematics ferroelectric, allowing electric-field manipulation of the flavors, and leading to the concept of flavortronics.

cond-mat.mes-hall↗

arXiv2018

Parameter-free $\ell_p$-Box Decoding of LDPC Codes

The Alternating Direction Method of Multipliers (ADMM) decoding of Low Density Parity Check (LDPC) codes has received many attentions due to its excellent performance at the error floor region. In this paper, we develop a parameter-free decoder based on Linear Program (LP) decoding by replacing the binary constraint with the intersection of a box and an $\ell_p$ sphere. An efficient $\ell_2$-box ADMM is designed to handle this model in a distributed fashion. Numerical experiments demonstrate that our decoder attains better adaptability to different Signal-to-Noise Ratio and channels.

cs.IT↗

arXiv2018

Atomically Resolved Observation of Continuous Interfaces between As-grown MoS2 Monolayer and WS2/MoS2 Heterobilayer on SiO2

Van der Waals (vdW) heterostructures synthesized through the chemical vapor deposition (CVD) method allow creation and tuning of intriguing electronic and optical properties of two- dimensional (2D) materials, the knowledge of which is critical for a wide range of potential applications. Here we report our scanning tunneling microscopy/spectroscopy (STM/STS) study of as-grown MoS2 monolayer and WS2/MoS2 heterobilayer on SiO2. The heterobilayer appears smoother than the MoS2 monolayer, with root mean square (RMS) roughness of 0.230 +- 0.021 nm in the former and 0.329 +- 0.033 nm in the latter. For the first time, to our knowledge, we directly observed a continuous interface between the MoS2 monolayer and the top layer of the heterobilayer with atomic resolution. This finding contrasts to the previously reported open edges in the top layer of the heterobilayer. Our STS results and density functional theory (DFT) calculations revealed the band gaps of the heterobilayer and the MoS2 monolayer.

cond-mat.mes-hall↗

arXiv2018

A simple field function for solving complex and dynamic fluid-solid system on Cartesian grid

In this paper, a simple field function is presented for facilitating the solution of complex and dynamic fluid-solid systems on Cartesian grids with interface-resolved fluid-fluid, fluid-solid, and solid-solid interactions. For a Cartesian-grid-discretized computational domain segmented by a set of solid bodies, this field function explicitly tracks each subdomain with multiple resolved interfacial node layers. As a result, the presented field function enables low-memory-cost multidomain node mapping, efficient node remapping, fast collision detection, and expedient surface force integration. Implementation algorithms for the field function and its described functionalities are also presented. Equipped with a deterministic multibody collision model, numerical experiments involving fluid-solid systems with flow conditions ranging from subsonic to supersonic states are conducted to validate and illustrate the applicability of the proposed field function.

physics.comp-ph↗

arXiv2018

An immersed boundary method for solving compressible flow with arbitrarily irregular and moving geometry

In this paper, a novel immersed boundary method is developed, validated, and applied. Through devising a second-order three-step flow reconstruction scheme, the proposed method is able to enforce the Dirichlet, Neumann, Robin, and Cauchy boundary conditions in a straightforward and consistent manner. Equipped with a fluid-solid coupling framework that integrates high-order temporal and spatial discretization schemes, numerical experiments concerning flow involving stationary and moving objects, convex and concave geometries, no-slip and slip wall boundary conditions, as well as subsonic and supersonic motions are conducted to validate the method. It is demonstrated that the proposed method can provide efficient, accurate, and robust boundary treatment for solving flow with arbitrarily irregular and moving geometries on Cartesian grids.

physics.flu-dyn↗

arXiv2018

Topological valleytronics: Brought to light

The topological valley Hall effect was predicted as a consequence of the bulk topology of electronic systems. Now it has been observed in photonic crystals, showing that both topology and valley are innate to classical as well as quantum systems. The discovery of counterflow in opposite valleys not only demonstrates that the valley can be a novel carrier of information and energy --- particularly valuable for systems without charge and spin --- but also exemplifies that symmetry-protected band topology can be universal to both quantum and classical systems.

cond-mat.mes-hall↗

arXiv2018

High-Temperature Majorana Corner States

Majorana bound states often occur at the end of 1D topological superconductor or at the $π$ Josephson junction mediated by a helical edge state. Validated by a new bulk invariant and an intuitive edge argument, we show the emergence of one Majorana Kramers pair at each corner of a square-shaped 2D topological insulator proximitized by an $s_\pm$-wave (e.g., Fe-based) superconductor. We obtain a phase diagram that emphasizes the roles of bulk parameters and edge orientations. We propose several experimental realizations in lattice-matched candidate materials. Our scheme offers a high-temperature platform for exploring higher-order non-Abelian quasiparticles.

cond-mat.mes-hall↗

arXiv2018

SeqFace: Make full use of sequence information for face recognition

Deep convolutional neural networks (CNNs) have greatly improved the Face Recognition (FR) performance in recent years. Almost all CNNs in FR are trained on the carefully labeled datasets containing plenty of identities. However, such high-quality datasets are very expensive to collect, which restricts many researchers to achieve state-of-the-art performance. In this paper, we propose a framework, called SeqFace, for learning discriminative face features. Besides a traditional identity training dataset, the designed SeqFace can train CNNs by using an additional dataset which includes a large number of face sequences collected from videos. Moreover, the label smoothing regularization (LSR) and a new proposed discriminative sequence agent (DSA) loss are employed to enhance discrimination power of deep face features via making full use of the sequence data. Our method achieves excellent performance on Labeled Faces in the Wild (LFW), YouTube Faces (YTF), only with a single ResNet. The code and models are publicly available on-line (https://github.com/huangyangyu/SeqFace).

cs.CV↗

arXiv2018

Probing Landau levels of strongly interacting massive Dirac electrons in layer-polarized MoS$_2$

Monolayer transition metal dichalcogenides are recently emerged 2D electronic systems with various novel properties, such as spin-valley locking, circular dichroism, valley Hall effects, Ising superconductivity. The reduced dimensionality and large effective masses further produce unconventional many-body interaction effects. Although recent hole transport measurements in WSe$_2$ indicate strong interactions in the valence bands, many-body interaction effects, particularly in the conduction bands, remain elusive to date. Here, for the first time, we perform transport measurements up to a magnetic field of $29$T to study the massive Dirac electron Landau levels (LL) in layer-polarized MoS$_2$ samples with mobilities of $22000$cm$^2$/(V$\cdot$s) at $1.5$K and densities of $\sim10^{12}$cm$^{-2}$. With decreasing the density, we observe LL crossing induced valley ferrimagnet-to-ferromagnet transitions, as a result of the interaction enhancement of the g-factor from $5.64$ to $21.82$. Near integer ratios of Zeeman-to-cyclotron energies, we discover LL anticrossings due to the formation of quantum Hall Ising ferromagnets, the valley polarizations of which appear to be reversible by tuning the density or an in-plane magnetic field. Our results provide compelling evidence for many-body interaction effects in the conduction bands of monolayer MoS$_2$ and establish a fertile ground for exploring strongly correlated phenomena of massive Dirac electrons.

cond-mat.mes-hall↗

arXiv2018

Neutrinos propagating in curved spacetimes

In the Dirac-Weyl equation that describes massless neutrino propagation in the minimal Standard Model, the ($2\times 2$ equivalence of the) gamma matrices convert Weyl spinors into spacetime tensors, and vice versa. They can thus be regarded as amalgamations of three different types of mappings, one that connects particle spinors directly forming representations to internal gauge symmetries to their spacetime counterparts that are embodied by null flags, another that translates the spacetime spinors into their corresponding tensors expressed in an orthonormal tetrad, and finally a purely tensorial transformation into the coordinate tetrad. The splitting of spinors into particle and spacetime varieties is not usually practised, but we advocate its adoption for better physical clarity, in terms of distinguishing internal and spacetime transformations, and also for understanding the scattering of neutrinos by spacetime curvature. We construct the basic infrastructure required for this task, and provide a worked example for the Schwarzschild spacetime. Our investigation also uncovers a possible under-determinacy in the flavoured Dirac-Weyl equation, which could serve as a new incision point for introducing flavour oscillation mechanisms.

physics.gen-ph↗