Search

The superconducting version of a diode effect has been the subject of extensive research in the last few years. So far, the focus has almost exclusively been on charge transport, but a natural question is if it is possible to obtain a dissipationless diode effect for spin transport. Here, we demonstrate that it is possible to generate an electrically tunable spin-diode effect with ideal efficiency by using superconductor/ferromagnet multilayers: the system completely forbids spin transport in one direction whereas it transport spins with zero energy loss in the other direction. We identify two mechanisms generating this phenomenon. This result provides a way to integrate the superconducting diode effect with spin-polarized currents., Comment: 6 pages, 6 figures

We use strong-coupling Eliashberg theory to study the competition of separate superconducting orders at low temperatures. Specifically, we study magnon-mediated superconductivity in a trilayer heterostructure with a thin normal metal between two antiferromagnetic insulators. Spin-triplet $p$-wave, spin-triplet $f$-wave, and spin-singlet $d$-wave superconducting gaps have been predicted to occur close to the critical temperature for the superconducting instability. The gap symmetry with the largest critical temperature depends on parameters in the model. We confirm that the same gap symmetries appear at any temperature below the critical temperature. Furthermore, we show that the temperature can affect the competition between the different superconducting orders. In addition, we consider time-reversal-symmetry-breaking, complex linear combinations of candidate pairings, such as chiral $p$-, $f$-, and $d$-wave gaps, as well as $p_x+if_y$-wave gaps. We find indications that some of these time-reversal-symmetry-breaking, nodeless gaps offer a greater condensation energy than the time-reversal symmetric gaps. This indicates that superconducting states with spontaneously broken time-reversal symmetry and nontrivial topology may be preferred in this system., Comment: 12 pages, 7 figures, accepted in Physical Review B

Convolutional Neural Networks (CNNs) and Transformers have achieved remarkable success in computer vision tasks. However, their deep architectures often lead to high computational redundancy, making them less suitable for resource-constrained environments, such as edge devices. This paper introduces ParFormer, a novel vision transformer that addresses this challenge by incorporating a Parallel Mixer and a Sparse Channel Attention Patch Embedding (SCAPE). By combining convolutional and attention mechanisms, ParFormer improves feature extraction. This makes spatial feature extraction more efficient and cuts down on unnecessary computation. The SCAPE module further reduces computational redundancy while preserving essential feature information during down-sampling. Experimental results on the ImageNet-1K dataset show that ParFormer-T achieves 78.9\% Top-1 accuracy with a high throughput on a GPU that outperforms other small models with 2.56$\times$ higher throughput than MobileViT-S, 0.24\% faster than FasterNet-T2, and 1.79$\times$ higher than EdgeNeXt-S. For edge device deployment, ParFormer-T excels with a throughput of 278.1 images/sec, which is 1.38 $\times$ higher than EdgeNeXt-S and 2.36$\times$ higher than MobileViT-S, making it highly suitable for real-time applications in resource-constrained settings. The larger variant, ParFormer-L, reaches 83.5\% Top-1 accuracy, offering a balanced trade-off between accuracy and efficiency, surpassing many state-of-the-art models. In COCO object detection, ParFormer-M achieves 40.7 AP for object detection and 37.6 AP for instance segmentation, surpassing models like ResNet-50, PVT-S and PoolFormer-S24 with significantly higher efficiency. These results validate ParFormer as a highly efficient and scalable model for both high-performance and resource-constrained scenarios, making it an ideal solution for edge-based AI applications., Comment: Under Review in IEEE Transactions on Cognitive and Developmental System

Localized spins of single atoms adsorbed on surfaces have been proposed as building blocks for spintronics and quantum computation devices. However, identifying a way to achieve current-induced switching of spins with very low dissipation is an outstanding challenge with regard to practical applications. Here, we show that the indirect exchange interaction between spin impurities can be controlled by a dissipationless supercurrent. All that is required is a conventional superconductor and two spin impurities placed on its surface. No triplet Cooper pairs or exotic material choices are needed. This finding provides a new and accessible way to achieve the long-standing goal of supercurrent-induced spin switching., Comment: 5 pages, 4 figures

To further improve the speaking styles of synthesized speeches, current text-to-speech (TTS) synthesis systems commonly employ reference speeches to stylize their outputs instead of just the input texts. These reference speeches are obtained by manual selection which is resource-consuming, or selected by semantic features. However, semantic features contain not only style-related information, but also style irrelevant information. The information irrelevant to speaking style in the text could interfere the reference audio selection and result in improper speaking styles. To improve the reference selection, we propose Contrastive Acoustic-Linguistic Module (CALM) to extract the Style-related Text Feature (STF) from the text. CALM optimizes the correlation between the speaking style embedding and the extracted STF with contrastive learning. Thus, a certain number of the most appropriate reference speeches for the input text are selected by retrieving the speeches with the top STF similarities. Then the style embeddings are weighted summarized according to their STF similarities and used to stylize the synthesized speech of TTS. Experiment results demonstrate the effectiveness of our proposed approach, with both objective evaluations and subjective evaluations on the speaking styles of the synthesized speeches outperform a baseline approach with semantic-feature-based reference selection., Comment: Accepted by InterSpeech 2022

A class of antiferromagnets with spin-polarized electron bands, yet zero net magnetization, called altermagnets is attracting increasing attention due to their potential use in spintronics. Here, we study spin injection into an altermagnet via spin pumping from a ferromagnetic insulator. We find that the spin pumping behaves qualitatively different depending on how the altermagnet is crystallographically oriented relative the interface to the ferromagnetic insulator. The altermagnetic state can enhance or suppress spin pumping, which we explain in terms of spin-split altermagnetic band structure and the spin-flip probability for the incident modes. Including the effect of interfacial Rashba spin-orbit coupling, we find that the spin-pumping effect is in general magnified, but that it can display a non-monotonic behavior as a function of the spin-orbit coupling strength. We show that there exists an optimal value of the spin-orbit coupling strength which causes an order of magnitude increase in the pumped spin current, even for the crystallographic orientation of the altermagnet which suppresses the spin pumping., Comment: 5 pages, 3 figures + supplementary information

We consider a planar heterostructure consisting of a normal metal in proximity to an antiferromagnetic insulator, with an interlayer exchange coupling between the metal and the insulator. The coupling to the two sublattices of the antiferromagnetic insulator is allowed to be asymmetric. An effective electron-electron interaction in the normal metal, mediated by antiferromagnetic magnons in the insulator, is derived to second order in the interlayer exchange coupling. Particular emphasis is placed on including analytically derived expressions for the effective interactions including Umklapp processes in the solutions to the superconducting gap equation. The gap equation is first solved at the critical temperature as an eigenvalue problem by linearizing the gap equation. The eigenvectors yield information on the symmetry of the superconducting gap at the onset of superconductivity, and we derive a phase diagram for the order parameter in this case. In the various regimes of the phase diagram, we find $p$-wave, $f$-wave, and $d$-wave superconductivity, with $p$-wave superconductivity in the dominant part of the phase diagram. Umklapp processes, that come into play with increasing size of the Fermi surface, yield $f$- and $d$-wave symmetries as the preferred symmetries when band filling approaches half filling. To investigate the stability of this order parameter symmetry as the temperature is lowered, we also consider the nonlinear gap equation at zero temperature. We conclude that the phase diagram and the symmetries of the superconducting order parameter essentially are left intact as the temperature is lowered to zero temperature., Comment: 9 pages, 3 figures, accepted in Physical Review B

Recent works have predicted materials featuring bands with a large spin-splitting distinct from ferromagnetic and relativistically spin-orbit coupled systems. Materials displaying this property are known as altermagnets and feature a spin-polarized band structure reminiscent of a $d$-wave superconducting order parameter. We here consider the contact between an altermagnet and a superconductor and determine how the altermagnetism affects the fundamental process of Andreev reflection. We show that the resulting charge conductance depends strongly on the interfacial orientation of the altermagnet relative to the superconductor, displaying features similar to normal metals or ferromagnets. The zero-bias conductance peak present at the interface in the $d$-wave case are robust toward the presence of an altermagnetic interaction. Moreover, the spin conductance strongly depends on the orientation of the altermagnet relative the interface. These results show how the anisotropic altermagnetic state can be probed by conductance spectroscopy and how it offers voltage control over charge and spin currents that are modulated due to superconductivity., Comment: 18 pages, 6 figures

Spin-pumping from a ferromagnetic insulator into a high-$T_c$ superconductor with a $d$-wave superconducting order parameter has recently been experimentally observed. Such unconventional superconducting order is known to produce interfacial bound-states for certain crystallographic orientations. Here, we present a methodology which can be used to study spin-pumping into unconventional superconductors, including the role of interfacial bound-states. As an example, we determine how the crystallographic orientation of the $d$-wave order parameter relative the interface changes the spin-pumping effect. We find that the spin-pumping effect is slightly enhanced at low temperatures for orientations hosting interfacial bound-states compared to other superconducting states. However, the spin-pumping effect does not show a coherence peak close to $T_c$ for such orientations, and instead remains smaller than the normal state value for all $T$. For orientations not hosting interfacial bound-states, we find that the pumped spin current can be increased to several times the normal-state spin current at frequencies that are small compared to the superconducting gap. Our results show that the spin-pumping dependency on frequency and temperature changes qualitatively depending on the crystallographic orientation of unconventional superconducting order parameters relative the interface., Comment: 12 pages, 4 figures

Word Sense Disambiguation (WSD) aims to find the exact sense of an ambiguous word in a particular context. Traditional supervised methods rarely take into consideration the lexical resources like WordNet, which are widely utilized in knowledge-based methods. Recent studies have shown the effectiveness of incorporating gloss (sense definition) into neural networks for WSD. However, compared with traditional word expert supervised methods, they have not achieved much improvement. In this paper, we focus on how to better leverage gloss knowledge in a supervised neural WSD system. We construct context-gloss pairs and propose three BERT-based models for WSD. We fine-tune the pre-trained BERT model on SemCor3.0 training corpus and the experimental results on several English all-words WSD benchmark datasets show that our approach outperforms the state-of-the-art systems., Comment: EMNLP-IJCNLP 2019

Language model pre-training has proven to be useful in learning universal language representations. As a state-of-the-art language model pre-training model, BERT (Bidirectional Encoder Representations from Transformers) has achieved amazing results in many language understanding tasks. In this paper, we conduct exhaustive experiments to investigate different fine-tuning methods of BERT on text classification task and provide a general solution for BERT fine-tuning. Finally, the proposed solution obtains new state-of-the-art results on eight widely-studied text classification datasets.