Showing 1–50 of 75 results for author: Zayats, V

Robust Preference Optimization through Reward Model Distillation

Authors: Adam Fisch, Jacob Eisenstein, Vicky Zayats, Alekh Agarwal, Ahmad Beirami, Chirag Nagpal, Pete Shaw, Jonathan Berant

Abstract: Language model (LM) post-training (or alignment) involves maximizing a reward function that is derived from preference annotations. Direct Preference Optimization (DPO) is a popular offline alignment method that trains a policy directly on preference data without the need to train a reward model or apply reinforcement learning. However, typical preference datasets have only a single, or at most a… ▽ More Language model (LM) post-training (or alignment) involves maximizing a reward function that is derived from preference annotations. Direct Preference Optimization (DPO) is a popular offline alignment method that trains a policy directly on preference data without the need to train a reward model or apply reinforcement learning. However, typical preference datasets have only a single, or at most a few, annotation per preference pair, which causes DPO to overconfidently assign rewards that trend towards infinite magnitude. This frequently leads to degenerate policies, sometimes causing even the probabilities of the preferred generations to go to zero. In this work, we analyze this phenomenon and propose distillation to get a better proxy for the true preference distribution over generation pairs: we train the LM to produce probabilities that match the distribution induced by a reward model trained on the preference data. Moreover, to account for uncertainty in the reward model we are distilling from, we optimize against a family of reward models that, as a whole, is likely to include at least one reasonable proxy for the preference distribution. Our results show that distilling from such a family of reward models leads to improved robustness to distribution shift in preference annotations, while preserving the simple supervised nature of DPO. △ Less

Submitted 29 May, 2024; originally announced May 2024.

Zipper: A Multi-Tower Decoder Architecture for Fusing Modalities

Authors: Vicky Zayats, Peter Chen, Melissa Ferrari, Dirk Padfield

Abstract: Integrating multiple generative foundation models, especially those trained on different modalities, into something greater than the sum of its parts poses significant challenges. Two key hurdles are the availability of aligned data (concepts that contain similar meaning but is expressed differently in different modalities), and effectively leveraging unimodal representations in cross-domain gener… ▽ More Integrating multiple generative foundation models, especially those trained on different modalities, into something greater than the sum of its parts poses significant challenges. Two key hurdles are the availability of aligned data (concepts that contain similar meaning but is expressed differently in different modalities), and effectively leveraging unimodal representations in cross-domain generative tasks, without compromising their original unimodal capabilities. We propose Zipper, a multi-tower decoder architecture that addresses these concerns by using cross-attention to flexibly compose multimodal generative models from independently pre-trained unimodal decoders. In our experiments fusing speech and text modalities, we show the proposed architecture performs very competitively in scenarios with limited aligned text-speech data. We also showcase the flexibility of our model to selectively maintain unimodal (e.g., text-to-text generation) generation performance by freezing the corresponding modal tower (e.g. text). In cross-modal tasks such as automatic speech recognition (ASR) where the output modality is text, we show that freezing the text backbone results in negligible performance degradation. In cross-modal tasks such as text-to-speech generation (TTS) where the output modality is speech, we show that using a pre-trained speech backbone results in superior performance to the baseline. △ Less

Submitted 31 May, 2024; v1 submitted 28 May, 2024; originally announced May 2024.

Comments: Under review at NeurIPS

Propagation of focused scalar and vector vortex beams in anisotropic media: A semi-analytical approach

Authors: Vittorio Aita, Mykyta Shevchenko, Francisco J. Rodríguez-Fortuño, Anatoly V. Zayats

Abstract: In the field of structured light, the study of optical vortices and their vectorial extension--vectorial vortex beams--has garnered substantial interest due to their unique phase and polarisation properties, which make them appealing for many potential applications. Combining the advantages of vortex beams and anisotropic materials, new possibilities for electromagnetic field tailoring can be achi… ▽ More In the field of structured light, the study of optical vortices and their vectorial extension--vectorial vortex beams--has garnered substantial interest due to their unique phase and polarisation properties, which make them appealing for many potential applications. Combining the advantages of vortex beams and anisotropic materials, new possibilities for electromagnetic field tailoring can be achieved in nonlinear optics, quantum and topological photonics. These applications call for a comprehensive modelling framework that accounts for properties of both anisotropic materials and vector vortex beams. In this paper, we describe a semi-analytical model that extends the vectorial diffraction theory to focused vortex beams propagating through a uniaxial slab, considering the cases of scalar and vectorial vortexes in the common framework of a Laguerre-Gaussian modes. The model aims to provide a comprehensive description of the methodology, enabling the implementation of complex beams transmission through, reflection from and propagation in uniaxial anisotropic materials for specific applications. We apply the developed approach to propagation of high-order vortex beams in uniaxial materials with various dispersion characteristics> elliptic, hyperbolic and epsilon-near-zero regimes. We show how variations of the medium anisotropy modify the beam structure due to the vectorial nature of their interaction, which results from the different permittivities of the medium for transverse and longitudinal field components. The applicability of the approach can be extended to artificially structured media if they can described by effective medium parameters. The developed formalism will be useful for modelling of interaction of complex beams with uniaxial materials, allowing a common framework for a large variety of situations, which can also be extended beyond the electromagnetic waves. △ Less

Submitted 6 February, 2024; originally announced February 2024.

Large-area, freestanding single-crystal gold of single nanometer thickness

Authors: Chenxinyu Pan, Yuanbiao Tong, Haoliang Qian, Alexey V. Krasavin, Jialin Li, Jiajie Zhu, Yiyun Zhang, Bowen Cui, Zhiyong Li, Chenming Wu, Zhenxin Wang, Lufang Liu, Linjun Li, Xin Guo, Anatoly V. Zayats, Limin Tong, Pan Wang

Abstract: Two-dimensional single-crystal metals are highly sought after for next-generation technologies. Here, we report large-area (>10^4 μm2), single-crystal two-dimensional gold with thicknesses down to a single-nanometer level, employing an atomic-level-precision chemical etching approach. The ultrathin thickness and single-crystal quality endow two-dimensional gold with unique properties including sig… ▽ More Two-dimensional single-crystal metals are highly sought after for next-generation technologies. Here, we report large-area (>10^4 μm2), single-crystal two-dimensional gold with thicknesses down to a single-nanometer level, employing an atomic-level-precision chemical etching approach. The ultrathin thickness and single-crystal quality endow two-dimensional gold with unique properties including significantly quantum-confinement-augmented optical nonlinearity, low sheet resistance, high transparency and excellent mechanical flexibility. By patterning the two-dimensional gold into nanoribbon arrays, extremely-confined near-infrared plasmonic resonances are further demonstrated with quality factors up to 5. The freestanding nature of two-dimensional gold allows its straightforward manipulation and transfer-printing for integration with other structures. The developed two-dimensional gold provides an emerging platform for fundamental studies in various disciplines and opens up new opportunities for applications in high-performance ultrathin optoelectronic, photonic and quantum devices. △ Less

Submitted 13 November, 2023; originally announced November 2023.

Journal ref: Nature Commun. 15 (2024) 2840-2849

AudioPaLM: A Large Language Model That Can Speak and Listen

Authors: Paul K. Rubenstein, Chulayuth Asawaroengchai, Duc Dung Nguyen, Ankur Bapna, Zalán Borsos, Félix de Chaumont Quitry, Peter Chen, Dalia El Badawy, Wei Han, Eugene Kharitonov, Hannah Muckenhirn, Dirk Padfield, James Qin, Danny Rozenberg, Tara Sainath, Johan Schalkwyk, Matt Sharifi, Michelle Tadmor Ramanovich, Marco Tagliasacchi, Alexandru Tudor, Mihajlo Velimirović, Damien Vincent, Jiahui Yu, Yongqiang Wang, Vicky Zayats , et al. (5 additional authors not shown)

Abstract: We introduce AudioPaLM, a large language model for speech understanding and generation. AudioPaLM fuses text-based and speech-based language models, PaLM-2 [Anil et al., 2023] and AudioLM [Borsos et al., 2022], into a unified multimodal architecture that can process and generate text and speech with applications including speech recognition and speech-to-speech translation. AudioPaLM inherits the… ▽ More We introduce AudioPaLM, a large language model for speech understanding and generation. AudioPaLM fuses text-based and speech-based language models, PaLM-2 [Anil et al., 2023] and AudioLM [Borsos et al., 2022], into a unified multimodal architecture that can process and generate text and speech with applications including speech recognition and speech-to-speech translation. AudioPaLM inherits the capability to preserve paralinguistic information such as speaker identity and intonation from AudioLM and the linguistic knowledge present only in text large language models such as PaLM-2. We demonstrate that initializing AudioPaLM with the weights of a text-only large language model improves speech processing, successfully leveraging the larger quantity of text training data used in pretraining to assist with the speech tasks. The resulting model significantly outperforms existing systems for speech translation tasks and has the ability to perform zero-shot speech-to-text translation for many languages for which input/target language combinations were not seen in training. AudioPaLM also demonstrates features of audio language models, such as transferring a voice across languages based on a short spoken prompt. We release examples of our method at https://google-research.github.io/seanet/audiopalm/examples △ Less

Submitted 22 June, 2023; originally announced June 2023.

Comments: Technical report

MultiTurnCleanup: A Benchmark for Multi-Turn Spoken Conversational Transcript Cleanup

Authors: Hua Shen, Vicky Zayats, Johann C. Rocholl, Daniel D. Walker, Dirk Padfield

Abstract: Current disfluency detection models focus on individual utterances each from a single speaker. However, numerous discontinuity phenomena in spoken conversational transcripts occur across multiple turns, hampering human readability and the performance of downstream NLP tasks. This study addresses these phenomena by proposing an innovative Multi-Turn Cleanup task for spoken conversational transcript… ▽ More Current disfluency detection models focus on individual utterances each from a single speaker. However, numerous discontinuity phenomena in spoken conversational transcripts occur across multiple turns, hampering human readability and the performance of downstream NLP tasks. This study addresses these phenomena by proposing an innovative Multi-Turn Cleanup task for spoken conversational transcripts and collecting a new dataset, MultiTurnCleanup1. We design a data labeling schema to collect the high-quality dataset and provide extensive data analysis. Furthermore, we leverage two modeling approaches for experimental evaluation as benchmarks for future research. △ Less

Submitted 27 October, 2023; v1 submitted 19 May, 2023; originally announced May 2023.

Comments: EMNLP 2023 main conference. Dataset: https://github.com/huashen218/MultiTurnCleanup

Dynamic dielectric metasurfaces via control of surface lattice resonances in non-homogeneous environment

Authors: Izzatjon Allayarov, Andrey B. Evlyukhin, Diane J. Roth, Boris Chichkov, Anatoly V. Zayats, Antonio Calà Lesina

Abstract: Dynamic control of metamaterials and metasurfaces is crucial for many photonic technologies, such as flat lenses, displays, augmented reality devices, and beam steering, to name a few. The dynamic response is typically achieved by controlling the phase and/or amplitude of individual meta-atom resonances using electro-optic, phase-change or nonlinear effects. Here, we propose and demonstrate a new… ▽ More Dynamic control of metamaterials and metasurfaces is crucial for many photonic technologies, such as flat lenses, displays, augmented reality devices, and beam steering, to name a few. The dynamic response is typically achieved by controlling the phase and/or amplitude of individual meta-atom resonances using electro-optic, phase-change or nonlinear effects. Here, we propose and demonstrate a new practical strategy for the dynamic control of the resonant interaction of light with dielectric metasurfaces, exploiting the dependence of the interaction between meta-atoms in the array on the inhomogeneity of the surrounding medium. The revealed tuning mechanisms are based on the concept of the surface lattice resonance (SLR), the development of which strongly depends on the difference between permittivities of superstrate and substrate materials. We experimentally demonstrate surface lattice resonances in dielectric (Si) metasurfaces, and reveal two tuning mechanisms corresponding to shifting or dam** of the SLR in optofluidic environment. The demonstrated dynamic tuning effect with the observed vivid colour changes may provide a dynamic metasurface approach with high spectral selectivity and enhanced sensitivity for sensors, as well as high-resolution for small pixel size displays. △ Less

Submitted 4 April, 2023; originally announced April 2023.

Comments: Main text: 10 pages, 4 figures. Supplementary information: 18 pages, 14 figures

Hot-electron dynamics in plasmonic nanostructures

Authors: Jacob Khurgin, Anton Yu. Bykov, Anatoly V. Zayats

Abstract: The coherent oscillations of mobile charge carriers near the surface of good conductors-surface plasmons-are been exploited in many applications in information technologies, clean energy, high-density data storage, photovoltaics, chemistry, biology, medicine and security. Light can be coupled to surface plasmons and trapped near the interface between a metal and an adjacent material. This leads to… ▽ More The coherent oscillations of mobile charge carriers near the surface of good conductors-surface plasmons-are been exploited in many applications in information technologies, clean energy, high-density data storage, photovoltaics, chemistry, biology, medicine and security. Light can be coupled to surface plasmons and trapped near the interface between a metal and an adjacent material. This leads to the nanoscale confinement of light, impossible by any other means, and a related electromagnetic field enhancement. Microscopic electron dynamic effects associated with surface plasmons are capable of significantly influencing physical and chemical processes near a conductor surface, not only as a result of the high electric fields, but also via the excitation of energetic charge carriers: holes below Fermi level or electrons above it. When remaining inside plasmonic media, these so-called hot carriers result in nonlinear, Kerr-type, optical effects important for controlling light with light. They can also transfer into the surroundings of the nanostructures, resulting in photocurrent, or they can interact with adjacent molecules and materials, inducing photochemical transformations. Understanding the dynamics of hot carriers and related effects in plasmonic nanostructures is essential for the development of ultrafast detectors and nonlinear optical components, broadband photocatalysis, enhanced nanoscale optoelectronic devices, nanoscale and ultrafast temperature control, and other technologies of tomorrow. This review will discuss the basics of plasmonically-engendered hot electrons, theoretical descriptions and experimental methods to study them, and describe prototypical processes and examples of the most promising applications of hot-electron processes at the metal interfaces. △ Less

Submitted 19 December, 2023; v1 submitted 20 February, 2023; originally announced February 2023.

Roadmap on structured waves

Authors: K. Y. Bliokh, E. Karimi, M. J. Padgett, M. A. Alonso, M. R. Dennis, A. Dudley, A. Forbes, S. Zahedpour, S. W. Hancock, H. M. Milchberg, S. Rotter, F. Nori, Ş. K. Özdemir, N. Bender, H. Cao, P. B. Corkum, C. Hernández-García, H. Ren, Y. Kivshar, M. G. Silveirinha, N. Engheta, A. Rauschenbeutel, P. Schneeweiss, J. Volz, D. Leykam , et al. (25 additional authors not shown)

Abstract: Structured waves are ubiquitous for all areas of wave physics, both classical and quantum, where the wavefields are inhomogeneous and cannot be approximated by a single plane wave. Even the interference of two plane waves, or a single inhomogeneous (evanescent) wave, provides a number of nontrivial phenomena and additional functionalities as compared to a single plane wave. Complex wavefields with… ▽ More Structured waves are ubiquitous for all areas of wave physics, both classical and quantum, where the wavefields are inhomogeneous and cannot be approximated by a single plane wave. Even the interference of two plane waves, or a single inhomogeneous (evanescent) wave, provides a number of nontrivial phenomena and additional functionalities as compared to a single plane wave. Complex wavefields with inhomogeneities in the amplitude, phase, and polarization, including topological structures and singularities, underpin modern nanooptics and photonics, yet they are equally important, e.g., for quantum matter waves, acoustics, water waves, etc. Structured waves are crucial in optical and electron microscopy, wave propagation and scattering, imaging, communications, quantum optics, topological and non-Hermitian wave systems, quantum condensed-matter systems, optomechanics, plasmonics and metamaterials, optical and acoustic manipulation, and so forth. This Roadmap is written collectively by prominent researchers and aims to survey the role of structured waves in various areas of wave physics. Providing background, current research, and anticipating future developments, it will be of interest to a wide cross-disciplinary audience. △ Less

Submitted 12 January, 2023; originally announced January 2023.

Comments: 110 pages, many figures

Journal ref: Journal of Optics, 25(10), 103001 (2023)

Non-Diffractive 3D Polarisation Features of Optical Vortex Beams

Authors: Andrei Afanasev, Jack J. Kingsley-Smith, Francisco J. Rodríguez-Fortuño, Anatoly V. Zayats

Abstract: Vector optical vortices exhibit complex polarisation patterns due to the interplay between spin and orbital angular momenta. Here we demonstrate, both analytically and with simulations, that certain polarisation features of optical vortex beams maintain constant transverse spatial dimensions independently of beam divergence due to diffraction. These polarisation features appear in the vicinity of… ▽ More Vector optical vortices exhibit complex polarisation patterns due to the interplay between spin and orbital angular momenta. Here we demonstrate, both analytically and with simulations, that certain polarisation features of optical vortex beams maintain constant transverse spatial dimensions independently of beam divergence due to diffraction. These polarisation features appear in the vicinity of the phase singularity and are associated with the presence of longitudinal electric fields. The predicted effect may prove important in metrology and high resolution imaging applications. △ Less

Submitted 8 November, 2022; v1 submitted 18 August, 2022; originally announced August 2022.

Comments: 18 pages, 4 figures; revised text and figures

Topological Transformation and Free-Space Transport of Photonic Hopfions

Authors: Yijie Shen, Bingshi Yu, Haijun Wu, Chunyu Li, Zhihan Zhu, Anatoly V. Zayats

Abstract: Structured light fields embody strong spatial variations of polarisation, phase and amplitude. Understanding, characterization and exploitation of such fields can be achieved through their topological properties. Three-dimensional (3D) topological solitons, such as hopfions, are 3D localized continuous field configurations with nontrivial particle-like structures, that exhibit a host of important… ▽ More Structured light fields embody strong spatial variations of polarisation, phase and amplitude. Understanding, characterization and exploitation of such fields can be achieved through their topological properties. Three-dimensional (3D) topological solitons, such as hopfions, are 3D localized continuous field configurations with nontrivial particle-like structures, that exhibit a host of important topologically protected properties. Here, we propose and demonstrate photonic counterparts of hopfions with exact characteristics of Hopf fibration, Hopf index, and Hopf map** from real-space vector beams to homotopic hyperspheres representing polarisation states. We experimentally generate photonic hopfions with on-demand high-order Hopf indices and independently controlled topological textures, including Néel-, Bloch-, and anti-skyrmionic types. We also demonstrate a robust free-space transport of photonic hopfions, thus, showing potential of hopfions for develo** optical topological informatics and communications. △ Less

Submitted 11 July, 2022; originally announced July 2022.

Optical skyrmions and other topological quasiparticles of light

Authors: Yijie Shen, Qiang Zhang, Peng Shi, Lu** Du, Xiaocong Yuan, Anatoly V. Zayats

Abstract: Skyrmions are topologically stable quasiparticles that have been predicted and demonstrated in quantum fields, solid-state physics, and magnetic materials, but only recently observed in electromagnetic fields, triggering fast expanding research across different spectral ranges and applications. Here we review the recent advances in optical skyrmions within a unified framework. Starting from fundam… ▽ More Skyrmions are topologically stable quasiparticles that have been predicted and demonstrated in quantum fields, solid-state physics, and magnetic materials, but only recently observed in electromagnetic fields, triggering fast expanding research across different spectral ranges and applications. Here we review the recent advances in optical skyrmions within a unified framework. Starting from fundamental theories, including classification of skyrmionic states, we describe generation and topological control of different kinds of optical skyrmions in structured and time-dependent optical fields. We further highlight generalized classes of optical topological quasiparticles beyond skyrmions and outline the emerging applications, future trends, and open challenges. A complex vectorial field structure of optical quasiparticles with versatile topological characteristics emerges as an important feature in modern spin-optics, imaging and metrology, optical forces, structured light and topological and quantum technologies. △ Less

Submitted 3 January, 2024; v1 submitted 20 May, 2022; originally announced May 2022.

Teaching BERT to Wait: Balancing Accuracy and Latency for Streaming Disfluency Detection

Authors: Angelica Chen, Vicky Zayats, Daniel D. Walker, Dirk Padfield

Abstract: In modern interactive speech-based systems, speech is consumed and transcribed incrementally prior to having disfluencies removed. This post-processing step is crucial for producing clean transcripts and high performance on downstream tasks (e.g. machine translation). However, most current state-of-the-art NLP models such as the Transformer operate non-incrementally, potentially causing unacceptab… ▽ More In modern interactive speech-based systems, speech is consumed and transcribed incrementally prior to having disfluencies removed. This post-processing step is crucial for producing clean transcripts and high performance on downstream tasks (e.g. machine translation). However, most current state-of-the-art NLP models such as the Transformer operate non-incrementally, potentially causing unacceptable delays. We propose a streaming BERT-based sequence tagging model that, combined with a novel training objective, is capable of detecting disfluencies in real-time while balancing accuracy and latency. This is accomplished by training the model to decide whether to immediately output a prediction for the current input or to wait for further context. Essentially, the model learns to dynamically size its lookahead window. Our results demonstrate that our model produces comparably accurate predictions and does so sooner than our baselines, with lower flicker. Furthermore, the model attains state-of-the-art latency and stability scores when compared with recent work on incremental disfluency detection. △ Less

Submitted 1 May, 2022; originally announced May 2022.

Comments: To be published at NAACL 2022

Atomically smooth single-crystalline platform for low-loss plasmonic nanocavities

Authors: Lufang Liu, Alexey V. Krasavin, Junsheng Zheng, Yuanbiao Tong, Pan Wang, Xiaofei Wu, Bert Hecht, Chenxinyu Pan, Jialin Li, Linjun Li, Xin Guo, Anatoly V. Zayats, Limin Tong

Abstract: Nanoparticle-on-mirror plasmonic nanocavities, capable of extreme optical confinement and enhancement, have triggered state-of-the-art progress in nanophotonics and development of applications in enhanced spectroscopies and molecular detection. However, the optical quality factor and thus performance of these nanoconstructs are undermined by the granular polycrystalline metal films used as a mirro… ▽ More Nanoparticle-on-mirror plasmonic nanocavities, capable of extreme optical confinement and enhancement, have triggered state-of-the-art progress in nanophotonics and development of applications in enhanced spectroscopies and molecular detection. However, the optical quality factor and thus performance of these nanoconstructs are undermined by the granular polycrystalline metal films used as a mirror. Here, we report an atomically smooth single-crystalline platform for low-loss nanocavities using chemically-synthesized gold microflakes as a mirror. Nanocavities constructed using gold nanorods on such microflakes exhibit a rich structure of plasmonic modes, which are highly sensitive to the thickness of optically-thin (down to ~15 nm) microflakes. The atomically smooth single-crystalline microflakes endow nanocavities with significantly improved quality factor (~2 times) and scattering intensity (~3 times) compared with their counterparts based on deposited films. The developed low-loss nanocavities further allow for the integration with a mature platform of fiber optics, opening opportunities for realizing nanocavity-based miniaturized photonic devices with high performance. △ Less

Submitted 22 January, 2022; originally announced January 2022.

Comments: 38 pages, 5 figures

Residual Adapters for Parameter-Efficient ASR Adaptation to Atypical and Accented Speech

Authors: Katrin Tomanek, Vicky Zayats, Dirk Padfield, Kara Vaillancourt, Fadi Biadsy

Abstract: Automatic Speech Recognition (ASR) systems are often optimized to work best for speakers with canonical speech patterns. Unfortunately, these systems perform poorly when tested on atypical speech and heavily accented speech. It has previously been shown that personalization through model fine-tuning substantially improves performance. However, maintaining such large models per speaker is costly an… ▽ More Automatic Speech Recognition (ASR) systems are often optimized to work best for speakers with canonical speech patterns. Unfortunately, these systems perform poorly when tested on atypical speech and heavily accented speech. It has previously been shown that personalization through model fine-tuning substantially improves performance. However, maintaining such large models per speaker is costly and difficult to scale. We show that by adding a relatively small number of extra parameters to the encoder layers via so-called residual adapter, we can achieve similar adaptation gains compared to model fine-tuning, while only updating a tiny fraction (less than 0.5%) of the model parameters. We demonstrate this on two speech adaptation tasks (atypical and accented speech) and for two state-of-the-art ASR architectures. △ Less

Submitted 14 September, 2021; originally announced September 2021.

Comments: Accepted to EMNLP 2021

Disfluency Detection with Unlabeled Data and Small BERT Models

Authors: Johann C. Rocholl, Vicky Zayats, Daniel D. Walker, Noah B. Murad, Aaron Schneider, Daniel J. Liebling

Abstract: Disfluency detection models now approach high accuracy on English text. However, little exploration has been done in improving the size and inference time of the model. At the same time, automatic speech recognition (ASR) models are moving from server-side inference to local, on-device inference. Supporting models in the transcription pipeline (like disfluency detection) must follow suit. In this… ▽ More Disfluency detection models now approach high accuracy on English text. However, little exploration has been done in improving the size and inference time of the model. At the same time, automatic speech recognition (ASR) models are moving from server-side inference to local, on-device inference. Supporting models in the transcription pipeline (like disfluency detection) must follow suit. In this work we concentrate on the disfluency detection task, focusing on small, fast, on-device models based on the BERT architecture. We demonstrate it is possible to train disfluency detection models as small as 1.3 MiB, while retaining high performance. We build on previous work that showed the benefit of data augmentation approaches such as self-training. Then, we evaluate the effect of domain mismatch between conversational and written text on model performance. We find that domain adaptation and data augmentation strategies have a more pronounced effect on these smaller models, as compared to conventional BERT models. △ Less

Submitted 27 July, 2021; v1 submitted 21 April, 2021; originally announced April 2021.

Comments: INTERSPEECH 2021

Photonic spin lattices: symmetry constraints for skyrmion and meron topologies

Authors: Xinrui Lei, Ai** Yang, Peng Shi, Zhenwei Xie, Lu** Du, Anatoly V. Zayats, Xiaocong Yuan

Abstract: Symmetry governs many electronic and photonic phenomena in optics and condensed matter physics. Skyrmions and merons are prominent topological structures in magnetic materials, with the topological features determined by the interplay between anisotropy of a material and its magnetization. Here we theoretically show and experimentally demonstrate that the symmetry of the electromagnetic field dete… ▽ More Symmetry governs many electronic and photonic phenomena in optics and condensed matter physics. Skyrmions and merons are prominent topological structures in magnetic materials, with the topological features determined by the interplay between anisotropy of a material and its magnetization. Here we theoretically show and experimentally demonstrate that the symmetry of the electromagnetic field determines the spin topological properties of the guided modes via spin-orbit coupling and may only result in either hexagonal spin-skyrmion or square spin-meron lattices. We also show that in the absence of spin-orbit coupling these spin topologies are degenerated in dynamic field-skyrmions, unifying description of electromagnetic field topologies. The results provide new understanding of electromagnetic field topology and its transformations as well as new opportunities for applications in quantum optics, spin-optics and topological photonics. △ Less

Submitted 29 March, 2021; originally announced March 2021.

Representations for Question Answering from Documents with Tables and Text

Authors: Vicky Zayats, Kristina Toutanova, Mari Ostendorf

Abstract: Tables in Web documents are pervasive and can be directly used to answer many of the queries searched on the Web, motivating their integration in question answering. Very often information presented in tables is succinct and hard to interpret with standard language representations. On the other hand, tables often appear within textual context, such as an article describing the table. Using the inf… ▽ More Tables in Web documents are pervasive and can be directly used to answer many of the queries searched on the Web, motivating their integration in question answering. Very often information presented in tables is succinct and hard to interpret with standard language representations. On the other hand, tables often appear within textual context, such as an article describing the table. Using the information from an article as additional context can potentially enrich table representations. In this work we aim to improve question answering from tables by refining table representations based on information from surrounding text. We also present an effective method to combine text and table-based predictions for question answering from full documents, obtaining significant improvements on the Natural Questions dataset. △ Less

Submitted 26 January, 2021; originally announced January 2021.

Comments: To appear at EACL 2021

Machine learning -- based diffractive imaging with subwavelength resolution

Authors: Abantika Ghosh, Diane J. Roth, Luke H. Nicholls, William P. Wardley, Anatoly V. Zayats, Viktor A. Podolskiy

Abstract: Far-field characterization of small objects is severely constrained by the diffraction limit. Existing tools achieving sub-diffraction resolution often utilize point-by-point image reconstruction via scanning or labelling. Here, we present a new imaging technique capable of fast and accurate characterization of two-dimensional structures with at least wavelength/25 resolution, based on a single fa… ▽ More Far-field characterization of small objects is severely constrained by the diffraction limit. Existing tools achieving sub-diffraction resolution often utilize point-by-point image reconstruction via scanning or labelling. Here, we present a new imaging technique capable of fast and accurate characterization of two-dimensional structures with at least wavelength/25 resolution, based on a single far-field intensity measurement. Experimentally, we realized this technique resolving the smallest-available to us 180-nm-scale features with 532-nm laser light. A comprehensive analysis of machine learning algorithms was performed to gain insight into the learning process and to understand the flow of subwavelength information through the system. Image parameterization, suitable for diffractive configurations and highly tolerant to random noise was developed. The proposed technique can be applied to new characterization tools with high spatial resolution, fast data acquisition, and artificial intelligence, such as high-speed nanoscale metrology and quality control, and can be further developed to high-resolution spectroscopy △ Less

Submitted 7 May, 2020; originally announced May 2020.

Transverse spinning of unpolarized light

Authors: J. S. Eismann, L. H. Nicholls, D. J. Roth, M. A. Alonso, P. Banzer, F. J. Rodríguez-Fortuño, A. V. Zayats, F. Nori, K. Y. Bliokh

Abstract: It is well known that spin angular momentum of light, and therefore that of photons, is directly related to their circular polarization. Naturally, for totally unpolarized light, polarization is undefined and the spin vanishes. However, for nonparaxial light, the recently discovered transverse spin component, orthogonal to the main propagation direction, is largely independent of the polarization… ▽ More It is well known that spin angular momentum of light, and therefore that of photons, is directly related to their circular polarization. Naturally, for totally unpolarized light, polarization is undefined and the spin vanishes. However, for nonparaxial light, the recently discovered transverse spin component, orthogonal to the main propagation direction, is largely independent of the polarization state of the wave. Here we demonstrate, both theoretically and experimentally, that this transverse spin survives even in nonparaxial fields (e.g., tightly focused or evanescent) generated from a totally unpolarized light source. This counterintuitive phenomenon is closely related to the fundamental difference between the degrees of polarization for 2D paraxial and 3D nonparaxial fields. Our results open an avenue for studies of spin-related phenomena and optical manipulation using unpolarized light. △ Less

Submitted 6 April, 2020; originally announced April 2020.

Comments: 11 pages, 3 figures

Journal ref: Nature Photonics 15, 156 (2021)

Transverse spin dynamics of light: the generalized spin-momentum locking for structured guided modes

Authors: Peng Shi, Lu** Du, Congcong Li, Anatoly V. Zayats, Xiaocong Yuan

Abstract: Quantum spin-Hall effect, a manifestation of topological properties that govern the behavior of surface states, was studied intensively in condensed matter physics resulting in the discovery of topological insulators. The quantum spin-Hall effect of light was introduced for surface plane-waves which intrinsically carry transverse optical spin, leading to many intriguing phenomena and applications… ▽ More Quantum spin-Hall effect, a manifestation of topological properties that govern the behavior of surface states, was studied intensively in condensed matter physics resulting in the discovery of topological insulators. The quantum spin-Hall effect of light was introduced for surface plane-waves which intrinsically carry transverse optical spin, leading to many intriguing phenomena and applications in unidirectional waveguiding, metrology and quantum technologies. In addition to spin, optical waves can exhibit complex topological properties of vectorial electromagnetic fields, associated with orbital angular momentum or nonuniform intensity variations. Here, by considering both spin and angular momentum, we demonstrate a generalized spin-momentum relationship that governs vectorial properties of guided electromagnetic waves, extending optical quantum spin-Hall effect to a two-dimensional vector field of structured guided wave. The effect results in the appearance of the out-of-plane transverse optical spins, which vary progressively from the 'up' state to the 'down' state around the energy flow, and their variation is uniquely locked to the energy propagation direction. The related spin-momentum locking in a chiral spin swirl is demonstrated with four kinds of surface structured waves and proven both theoretically and experimentally. The results provide understanding of the spin dynamics in electromagnetic guided waves and show great importance in spin optics, topological photonics and optical spin-based devices and techniques. △ Less

Submitted 5 August, 2020; v1 submitted 9 October, 2019; originally announced October 2019.

Comments: 30 pages; 13 figures; 44 equations; 1 table; 41 references for main text and 45 references for Supplementary Materials

Amplitude and phase control of guided modes excitation from a single dipole source:engineering far- and near-field directionality

Authors: Michela F. Picardi, Anatoly V. Zayats, Francisco J. Rodríguez-Fortuño

Abstract: The design of far-field radiation diagrams from combined electric and magnetic dipolar sources has recently found applications in nanophotonic metasurfaces that realize tailored reflection and refraction. Such dipolar sources also exhibit important near-field evanescent coupling properties with applications in polarimetry and quantum optics. Here we introduce a rigorous theoretical framework for e… ▽ More The design of far-field radiation diagrams from combined electric and magnetic dipolar sources has recently found applications in nanophotonic metasurfaces that realize tailored reflection and refraction. Such dipolar sources also exhibit important near-field evanescent coupling properties with applications in polarimetry and quantum optics. Here we introduce a rigorous theoretical framework for engineering the angular spectra encompassing both far- and near-fields of electric and magnetic sources and develop a unified description of both free space and guided mode directional radiation. The approach uses the full parametric space of six complex-valued components of magnetic and electric dipoles in order to engineer constructive or destructive near-field interference. Such dipolar sources can be realized with dielectric or plasmonic nanoparticles. We show how a single dipolar source can be designed to achieve the selective coupling to multiple waveguide modes and far-field simultaneously with a desired amplitude, phase, and direction. △ Less

Submitted 15 July, 2019; originally announced July 2019.

Disfluencies and Human Speech Transcription Errors

Authors: Vicky Zayats, Trang Tran, Richard Wright, Courtney Mansfield, Mari Ostendorf

Abstract: This paper explores contexts associated with errors in transcrip-tion of spontaneous speech, shedding light on human perceptionof disfluencies and other conversational speech phenomena. Anew version of the Switchboard corpus is provided with disfluency annotations for careful speech transcripts, together with results showing the impact of transcription errors on evaluation of automatic disfluency… ▽ More This paper explores contexts associated with errors in transcrip-tion of spontaneous speech, shedding light on human perceptionof disfluencies and other conversational speech phenomena. Anew version of the Switchboard corpus is provided with disfluency annotations for careful speech transcripts, together with results showing the impact of transcription errors on evaluation of automatic disfluency detection. △ Less

Submitted 8 April, 2019; originally announced April 2019.

Comments: Submitted to INTERSPEECH 2019

Giving Attention to the Unexpected: Using Prosody Innovations in Disfluency Detection

Authors: Vicky Zayats, Mari Ostendorf

Abstract: Disfluencies in spontaneous speech are known to be associated with prosodic disruptions. However, most algorithms for disfluency detection use only word transcripts. Integrating prosodic cues has proved difficult because of the many sources of variability affecting the acoustic correlates. This paper introduces a new approach to extracting acoustic-prosodic cues using text-based distributional pre… ▽ More Disfluencies in spontaneous speech are known to be associated with prosodic disruptions. However, most algorithms for disfluency detection use only word transcripts. Integrating prosodic cues has proved difficult because of the many sources of variability affecting the acoustic correlates. This paper introduces a new approach to extracting acoustic-prosodic cues using text-based distributional prediction of acoustic cues to derive vector z-score features (innovations). We explore both early and late fusion techniques for integrating text and prosody, showing gains over a high-accuracy text-only model. △ Less

Submitted 8 April, 2019; originally announced April 2019.

Comments: Accepted at NAACL-HLT 2019

Spontaneous photon-pair generation at the nanoscale

Authors: Giuseppe Marino, Alexander S. Solntsev, Lei Xu, Valerio F. Gili, Luca Carletti, Alexander N. Poddubny, Mohsen Rahmani, Daria A. Smirnova, Haitao Chen, Aristide Lemaître, Guoquan Zhang, Anatoly V. Zayats, Costantino De Angelis, Giuseppe Leo, Andrey A. Sukhorukov, Dragomir N. Neshev

Abstract: Optical nanoantennas have shown a great capacity for efficient extraction of photons from the near to the far-field, enabling directional emission from nanoscale single-photon sources. However, their potential for the generation and extraction of multi-photon quantum states remains unexplored. Here we demonstrate experimentally the nanoscale generation of two-photon quantum states at telecommunica… ▽ More Optical nanoantennas have shown a great capacity for efficient extraction of photons from the near to the far-field, enabling directional emission from nanoscale single-photon sources. However, their potential for the generation and extraction of multi-photon quantum states remains unexplored. Here we demonstrate experimentally the nanoscale generation of two-photon quantum states at telecommunication wavelengths based on spontaneous parametric down-conversion in an optical nanoantenna. The antenna is a crystalline AlGaAs nanocylinder, possessing Mie-type resonances at both the pump and the bi-photon wavelengths and when excited by a pump beam generates photonpairs with a rate of 35 Hz. Normalized to the pump energy stored by the nanoantenna, this rate corresponds to 1.4 GHz/Wm, being one order of magnitude higher than conventional on-chip or bulk photon-pair sources. Our experiments open the way for multiplexing several antennas for coherent generation of multi-photon quantum states with complex spatial-mode entanglement and applications in free-space quantum communications and sensing. △ Less

Submitted 9 April, 2019; v1 submitted 16 March, 2019; originally announced March 2019.

Comments: A new author, Aristide Lemaître, has been added to the author list. No change has been made to the main text

Journal ref: Optica 6, 1416-1422 (2019)

Experimental demonstration of linear and spinning Janus dipoles for polarisation and wavelength selective near-field coupling

Authors: Michela F. Picardi, Martin Neugebauer, Joerg S. Eismann, Gerd Leuchs, Peter Banzer, Francisco J. Rodríguez-Fortuño, Anatoly V. Zayats

Abstract: The electromagnetic field scattered by nano-objects contains a broad range of wave vectors and can be efficiently coupled to waveguided modes. The dominant contribution to scattering from subwavelength dielectric and plasmonic nanoparticles is determined by electric and magnetic dipolar responses. Here, we experimentally demonstrate spectral and phase selective excitation of Janus dipoles, sources… ▽ More The electromagnetic field scattered by nano-objects contains a broad range of wave vectors and can be efficiently coupled to waveguided modes. The dominant contribution to scattering from subwavelength dielectric and plasmonic nanoparticles is determined by electric and magnetic dipolar responses. Here, we experimentally demonstrate spectral and phase selective excitation of Janus dipoles, sources with electric and magnetic dipoles oscillating out of phase, in order to control near-field interference and directional coupling to waveguides. We show that by controlling the polarisation state of the dipolar excitations and the excitation wavelength to adjust their relative contributions, directionality and coupling strength can be fully tuned. Furthermore, we introduce a novel spinning Janus dipole featuring cylindrical symmetry in the near and far field, which results in either omnidirectional coupling or noncoupling. Controlling the propagation of guided light waves via fast and robust near-field interference between polarisation components of a source is required in many applications in nanophotonics and quantum optics. △ Less

Submitted 22 January, 2019; originally announced January 2019.

Robust cross-domain disfluency detection with pattern match networks

Authors: Vicky Zayats, Mari Ostendorf

Abstract: In this paper we introduce a novel pattern match neural network architecture that uses neighbor similarity scores as features, eliminating the need for feature engineering in a disfluency detection task. We evaluate the approach in disfluency detection for four different speech genres, showing that the approach is as effective as hand-engineered pattern match features when used on in-domain data a… ▽ More In this paper we introduce a novel pattern match neural network architecture that uses neighbor similarity scores as features, eliminating the need for feature engineering in a disfluency detection task. We evaluate the approach in disfluency detection for four different speech genres, showing that the approach is as effective as hand-engineered pattern match features when used on in-domain data and achieves superior performance in cross-domain scenarios. △ Less

Submitted 17 November, 2018; originally announced November 2018.

Comments: This paper was submitted to EMNLP 2018 and was rejected. Our EMNLP submission is posted here to establish concurrency with "Disfluency Detection using Auto-Correlational Neural Networks" by P. Lou, P. Anderson, M. Johnson which was submitted to EMNLP at the same time

Optical forces from near-field directionalities in planar structures

Authors: Jack J. Kingsley-Smith, Michela F. Picardi, Lei Wei, Anatoly V. Zayats, Francisco J. Rodríguez-Fortuño

Abstract: Matter manipulation with optical forces has become commonplace in a wide range of research fields and is epitomized by the optical trap. Calculations of optical forces on small illuminated particles typically neglect multiple scattering on nearby structures. However, this scattering can result in large recoil forces, particularly when the scattering includes directional near-field excitations. Nea… ▽ More Matter manipulation with optical forces has become commonplace in a wide range of research fields and is epitomized by the optical trap. Calculations of optical forces on small illuminated particles typically neglect multiple scattering on nearby structures. However, this scattering can result in large recoil forces, particularly when the scattering includes directional near-field excitations. Near-field recoil forces have been studied in the case of electric, magnetic and circularly polarized dipoles, but they exist for any type of directional near-field excitation. We use the force angular spectrum as a concise and intuitive analytical expression for the force on any dipole near planar surfaces, which allows us to clearly distinguish the effect due to the dipole, and due to the surface. We relate this directly to the coupling efficiency of surface or guided modes via Fermi's golden rule. To exemplify this, a near-field force transverse to the illumination is computationally calculated for a Huygens dipole near a metallic waveguide. We believe this formalism will prove insightful for various nanomanipulation systems within areas such as nanofluidics, sensing, biotechnology and nano-assembly of nanostructures. △ Less

Submitted 24 May, 2019; v1 submitted 13 November, 2018; originally announced November 2018.

Comments: 9 pages, 3 figures

Journal ref: Phys. Rev. B 99, 235410 (2019)

Multilevel nonvolatile optoelectronic memory based on memristive plasmonic tunnel junctions

Authors: Pan Wang, Mazhar E. Nasir, Alexey V. Krasavin, Wayne Dickson, Anatoly V. Zayats

Abstract: Highly efficient information processing in brain is based on processing and memory components called synapses, whose output is dependent on the history of the signals passed through them. Here we have developed an artificial synapse with both electrical and optical memory effects using reactive tunnel junctions based on plasmonic nanorods. In an electronic realization, the electrons tunneled into… ▽ More Highly efficient information processing in brain is based on processing and memory components called synapses, whose output is dependent on the history of the signals passed through them. Here we have developed an artificial synapse with both electrical and optical memory effects using reactive tunnel junctions based on plasmonic nanorods. In an electronic realization, the electrons tunneled into plasmonic nanorods under low bias voltage are harvested to write information into the tunnel junctions via hot-electron-mediated chemical reactions with the environment. In an optical realization, the information can also be written optically by external light illumination to excite hot electrons in plasmonic nanorods. The stored information is non-volatile and can be read in both realizations either electrically or optically by measuring the resistance or inelastic-tunnelling-induced light emission, respectively. These memristive light-emitting plasmonic tunnel junctions can be used as memory, logic units or artificial synapses in future optoelectronic or neuromorphic information systems. △ Less

Submitted 8 November, 2018; originally announced November 2018.

Comments: 4 figures

Deep-subwavelength features of photonic skyrmions in a confined electromagnetic field with orbital angular momentum

Authors: Lu** Du, Ai** Yang, Anatoly V. Zayats, Xiaocong Yuan

Abstract: In magnetic materials, skyrmions are nanoscale regions where the orientation of electron spin changes in a vortex-type manner. Here we show that spin-orbit coupling in a focused vector beam results in a skyrmion-like photonic spin distribution of the excited waveguided fields. While diffraction limits the spatial size of intensity distributions, the direction of the field, defining photonic spin,… ▽ More In magnetic materials, skyrmions are nanoscale regions where the orientation of electron spin changes in a vortex-type manner. Here we show that spin-orbit coupling in a focused vector beam results in a skyrmion-like photonic spin distribution of the excited waveguided fields. While diffraction limits the spatial size of intensity distributions, the direction of the field, defining photonic spin, is not subject to this limitation. We demonstrate that the skyrmion spin structure varies on the deep-subwavelength scales down to 1/60 of light wavelength, which corresponds to about 10 nanometre lengthscale. The application of photonic skyrmions may range from high-resolution imaging and precision metrology to quantum technologies and data storage where the spin structure of the field, not its intensity, can be applied to achieve deep-subwavelength optical patterns. △ Less

Submitted 12 June, 2018; originally announced June 2018.

Interferometric evanescent wave excitation of nano-antenna for ultra-sensitive displacement and phase metrology

Authors: Lei Wei, Anatoly V. Zayats, Francisco J. Rodríguez-Fortuño

Abstract: We propose a method for ultra-sensitive displacement and phase metrology based on the interferometric evanescent wave excitation of nano-antennas. We show that with a proper choice of nano-antenna, tiny displacements or relative phase variations can be converted into sensitive scattering direction changes in the Fourier $k$-space. These changes stem from the strong position dependence of the imagi… ▽ More We propose a method for ultra-sensitive displacement and phase metrology based on the interferometric evanescent wave excitation of nano-antennas. We show that with a proper choice of nano-antenna, tiny displacements or relative phase variations can be converted into sensitive scattering direction changes in the Fourier $k$-space. These changes stem from the strong position dependence of the imaginary Poynting vector orientation within interfering evanescent waves. Using strongly-evanescent standing waves, high sensitivity is achieved in the nano-antenna's zero scattering direction, which varies linearly with displacement over a long range. With weakly-evanescent wave interference, even higher sensitivity to tiny displacement or phase changes can be reached around chosen location. The high sensitivity of the proposed method can form the basis for many applications. △ Less

Submitted 12 June, 2018; originally announced June 2018.

Journal ref: Phys. Rev. Lett. 121, 193901 (2018)

Directional scattering from particles under evanescent wave illumination: the role of reactive power

Authors: Lei Wei, Michela F. Picardi, Jack J. Kingsley-Smith, Anatoly V. Zayats, Francisco J. Rodríguez-Fortuño

Abstract: Study of photonic spin-orbital interactions, which involves control of the propagation and spatial distributions of light with the polarization of electromagnetic fields, is not only important at the fundamental level but also has significant implications for functional photonic applications that require active tuning of directional light propagation. Many of the experimental demonstrations have b… ▽ More Study of photonic spin-orbital interactions, which involves control of the propagation and spatial distributions of light with the polarization of electromagnetic fields, is not only important at the fundamental level but also has significant implications for functional photonic applications that require active tuning of directional light propagation. Many of the experimental demonstrations have been attributed to the spin-momentum locking characteristic of evanescent waves. In this letter, we show another property of evanescent waves: the polarization dependent direction of the imaginary part of the Poynting vector, i.e. reactive power. Based on this property, we propose a simple and robust way to tune the directional far-field scattering from nanoparticles near a surface under evanescent wave illumination by controlling linear polarization and direction of the incident light. △ Less

Submitted 13 March, 2018; originally announced March 2018.

Journal ref: Optics Letters 43, 3393-3396 (2018)

Circular Dichroism Enhancement in Plasmonic Nanorod Metamaterials

Authors: D. Vestler, I. Shishkin, E. A. Gurvitz, M. E. Nasir, A. Ben-Moshe, A. P. Slobozhanyuk, A. V. Krasavin, T. Levi-Belenkova, A. S. Shalin, P. Ginzburg, G. Markovich, A. V. Zayats

Abstract: Optical activity is a fundamental phenomenon originating from the chiral nature of crystals and molecules. While intrinsic chiroptical responses of ordinary chiral materials to circularly polarized light are relatively weak, they can be enhanced by specially tailored nanostructures. Here, nanorod metamaterials, comprising a dense array of vertically aligned gold nanorods, is shown to provide signi… ▽ More Optical activity is a fundamental phenomenon originating from the chiral nature of crystals and molecules. While intrinsic chiroptical responses of ordinary chiral materials to circularly polarized light are relatively weak, they can be enhanced by specially tailored nanostructures. Here, nanorod metamaterials, comprising a dense array of vertically aligned gold nanorods, is shown to provide significant enhancement of the circular dichroism response of an embedded material. A nanorod composite, acting as an artificial uniaxial crystal, is filled with chiral mercury sulfide nanocrystals embedded in a transparent polymer. The nanorod based metamaterial, being inherently achiral, enables optical activity enhancement or suppression. Unique properties of inherently achiral structures to tailor optical activities pave a way for flexible characterization of optical activity of molecules and nanocrystal-based compounds. △ Less

Submitted 11 January, 2018; originally announced January 2018.

Structural second-order nonlinearity in metamaterials

Authors: B. Wells, A. Yu. Bykov, G. Marino, M. E. Nasir, A. V. Zayats, V. A. Podolskiy

Abstract: Nonlinear processes are at the core of many optical technologies including lasers, information processing, sensing, and security, and require optimised materials suitable for nanoscale integration. Here we demonstrate the emergence of a strong bulk second-order nonlinear response in a composite plasmonic nanorod material comprised of centrosymmetric materials. The metamaterial provides equally str… ▽ More Nonlinear processes are at the core of many optical technologies including lasers, information processing, sensing, and security, and require optimised materials suitable for nanoscale integration. Here we demonstrate the emergence of a strong bulk second-order nonlinear response in a composite plasmonic nanorod material comprised of centrosymmetric materials. The metamaterial provides equally strong generation of the p-polarized second harmonic light in response to both s- and p-polarized excitation. We develop an effective-medium description of the underlying physics, compare its predictions to the experimental results and analyze the limits of its applicability. We show that while the effective medium theory adequately describes the nonlinear polarization, the process of emission of second harmonic light cannot be described in the same framework. The work provides an understanding of the emergent nonlinear optical response in composites and opens a doorway to new nonlinear optical platform designs for integrated nonlinear photonics. △ Less

Submitted 29 December, 2017; originally announced January 2018.

Efficient energy propagation through self-assembled gold nanoparticle chain waveguides

Authors: Fatih N. Gür, Cillian P. T. McPolin, Søren Raza, Martin Mayer, Diane J. Roth, Anja Maria Steiner, Markus Löffler, Andreas Fery, Mark L. Brongersma, Anatoly V. Zayats, Tobias A. F. König, Thorsten L. Schmidt

Abstract: The strong interaction of light with metallic nanoparticles enables field confinement well below the diffraction limit. Plasmonic waveguides consisting of metal nanoparticle chains could be used for the propagation of energy or information on the nanoscale, but high losses have thus far impeded practical applications. Here we demonstrate that efficient waveguiding is possible through gold nanopart… ▽ More The strong interaction of light with metallic nanoparticles enables field confinement well below the diffraction limit. Plasmonic waveguides consisting of metal nanoparticle chains could be used for the propagation of energy or information on the nanoscale, but high losses have thus far impeded practical applications. Here we demonstrate that efficient waveguiding is possible through gold nanoparticle chains despite the high dissipative losses of gold. A DNA origami directed self-assembly of monocrystalline, spherical nanoparticles allows the interparticle spacing to be decreased to 2 nm or below, which gives rise to lower-energy plasmon resonance modes. Our simulations imply that these lower energy modes allow efficient waveguiding but collapse if interparticle gap sizes are increased. Individual waveguides are characterized with nanometer-resolution by electron energy loss spectroscopy, and directed propagation of energy towards a fluorescent nanodiamond and nanoscale energy conversion is shown by cathodoluminescence imaging spectroscopy on a single-device level. With this approach, micrometer-long propagation lengths might be achieved, enabling applications in information technology, sensing and quantum optics. △ Less

Submitted 10 July, 2018; v1 submitted 25 December, 2017; originally announced December 2017.

Stability of Thin Film Refractory Plasmonic Materials Taken to High Temperatures in Air

Authors: Matthew P. Wells, Gomathi Gobalakrichenane, Ryan Bower, Bin Zou, Rebecca Kilmurray, Andrei P. Mihai, Neil McN. Alford, Rupert F. M. Oulton, Lesley F. Cohen, Stefan A. Maier, Anatoly V. Zayats, Peter K. Petrov

Abstract: Materials such as W, TiN, and SrRuO3 (SRO) have been suggested as promising alternatives to Au and Ag in plasmonic applications owing to their refractory properties. However, investigation of the reproducibility of the optical properties after thermal cycling at high operational temperatures is so far lacking. Here, thin films of W, Mo, Ti, TiN, TiON, Ag, Au, and SrRuO3 are investigated to assess… ▽ More Materials such as W, TiN, and SrRuO3 (SRO) have been suggested as promising alternatives to Au and Ag in plasmonic applications owing to their refractory properties. However, investigation of the reproducibility of the optical properties after thermal cycling at high operational temperatures is so far lacking. Here, thin films of W, Mo, Ti, TiN, TiON, Ag, Au, and SrRuO3 are investigated to assess their viability for robust refractory plasmonic applications. Films ranging in thickness from 50 - 180 nm are deposited on MgO and Si substrates by RF magnetron sputtering and, in the case of SrRuO3, pulsed laser deposition, prior to characterisation by means of AFM, XRD, spectroscopic ellipsometry, and DC resistivity. Measurements are conducted before and after annealing in air at temperatures ranging from 300 - 1000° C for one hour, to establish the maximum cycling temperature and potential longevity at temperature for each material. It is found that SrRuO3 retains metallic behaviour after annealing at 800° C, however, importantly, the optical properties of TiN and TiON are degraded as a result of oxidation. Nevertheless, both TiN and TiON may be better suited than Au or SRO for high temperature applications operating under vacuum conditions. △ Less

Submitted 24 November, 2017; originally announced November 2017.

Comments: 15 pages, 4 Figures, 3 Tables

Janus and Huygens' dipolar sources for near-field directionality

Authors: Michela F. Picardi, Anatoly V. Zayats, Francisco J. Rodríguez-Fortuño

Abstract: Controlling directionality of emission, scattering and waveguiding is an important requirement in quantum optical technology, integrated photonics and new metasurface designs, as well as radio and microwave engineering. Recently, several approaches have been developed to achieve unidirectional scattering in the far-field relying on Huygens' dipolar sources, and in waveguided optics based on spin-H… ▽ More Controlling directionality of emission, scattering and waveguiding is an important requirement in quantum optical technology, integrated photonics and new metasurface designs, as well as radio and microwave engineering. Recently, several approaches have been developed to achieve unidirectional scattering in the far-field relying on Huygens' dipolar sources, and in waveguided optics based on spin-Hall effects involving circularly polarised electric or magnetic dipoles, all of which can be realised with plasmonic or dielectric nanoparticles. Here we show that there exists a dipolar source complimentary to Huygens' dipole, termed Janus dipole, which is not directional in the far-field, but its coupling to waveguided modes is topologically protected so that it is allowed on one side of the dipole but not on the opposite side. The near field directionality of the Huygens' dipole is also revealed and a generalised Kerker's condition for far- and near-field directionality is introduced. Circular electric and magnetic dipoles, together with Huygens' and Janus dipolar sources, form a complete set of directional dipolar sources in far- and near-field, paving the way for promising applications. △ Less

Submitted 14 August, 2017; v1 submitted 8 August, 2017; originally announced August 2017.

Journal ref: Phys. Rev. Lett. 120, 117402 (2018)

Repulsion of polarized particles from two-dimensional materials

Authors: Francisco J. Rodríguez-Fortuño, Michela F. Picardi, Anatoly V. Zayats

Abstract: Repulsion of nanoparticles, molecules and atoms from surfaces can have important applications in nanomechanical devices, microfluidics, optical manipulation and atom optics. Here, through the solution of a classical scattering problem, we show that a dipole source can experience a robust and strong repulsive force when its near-field interacts with a two-dimensional material that has a metallic ch… ▽ More Repulsion of nanoparticles, molecules and atoms from surfaces can have important applications in nanomechanical devices, microfluidics, optical manipulation and atom optics. Here, through the solution of a classical scattering problem, we show that a dipole source can experience a robust and strong repulsive force when its near-field interacts with a two-dimensional material that has a metallic character. As an example, the case of graphene is considered, showing that a broad bandwidth of repulsion can be obtained spanning the frequency range $0<\hbarω<(5/3)μ_c$, where $μ_c$ is the chemical potential of graphene, tuneable electrically or by chemical do**. △ Less

Submitted 11 July, 2017; originally announced July 2017.

Journal ref: Phys. Rev. B 97, 205401 (2018)

Conversation Modeling on Reddit using a Graph-Structured LSTM

Authors: Vicky Zayats, Mari Ostendorf

Abstract: This paper presents a novel approach for modeling threaded discussions on social media using a graph-structured bidirectional LSTM which represents both hierarchical and temporal conversation structure. In experiments with a task of predicting popularity of comments in Reddit discussions, the proposed model outperforms a node-independent architecture for different sets of input features. Analyses… ▽ More This paper presents a novel approach for modeling threaded discussions on social media using a graph-structured bidirectional LSTM which represents both hierarchical and temporal conversation structure. In experiments with a task of predicting popularity of comments in Reddit discussions, the proposed model outperforms a node-independent architecture for different sets of input features. Analyses show a benefit to the model over the full course of the discussion, improving detection in both early and late stages. Further, the use of language cues with the bidirectional tree state updates helps with identifying controversial comments. △ Less

Submitted 6 April, 2017; originally announced April 2017.

Comments: Submitted to TACL

Titanium oxynitride thin films with tunable double epsilon-near-zero behaviour

Authors: Laurentiu Braic, Nikolaos Vasilantonakis, Andrei P. Mihai, Ignacio J. Villar Garcia, Sarah Fearn, Bin Zou, Brock Doiron, Rupert F. Oulton, Lesley Cohen, Stefan A. Maier, Neil McN. Alford, Anatoly V. Zayats, Peter K. Petrov

Abstract: Titanium Oxynitride (TiOxNy) thin films are fabricated using reactive magnetron sputtering. The mechanism of their growth formation is explained and their optical properties are presented. The films grown when the level of residual Oxygen in the background vacuum was between 5E-9Torr to 20E-9Torr exhibit double Epsilon-Near-Zero (2-ENZ) behaviour with ENZ1 and ENZ2 wavelengths tunable in the 700-8… ▽ More Titanium Oxynitride (TiOxNy) thin films are fabricated using reactive magnetron sputtering. The mechanism of their growth formation is explained and their optical properties are presented. The films grown when the level of residual Oxygen in the background vacuum was between 5E-9Torr to 20E-9Torr exhibit double Epsilon-Near-Zero (2-ENZ) behaviour with ENZ1 and ENZ2 wavelengths tunable in the 700-850 nm and in the 1100-1350 nm spectral ranges, respectively. Samples fabricated when the level of residual Oxygen in the background vacuum was above 2E-8Torr exhibit non-metallic behaviour, while the layers deposited when the level of residual Oxygen in the background vacuum was below 5E-9Torr, show metallic behaviour with a single ENZ value. The double ENZ phenomenon is related to the level of residual Oxygen in the background vacuum and is attributed to the mixture of TiN and TiOxNy/TiOx phases in the films. Varying the partial pressure of nitrogen during the deposition can further control the amount of TiN, TiOx and TiOxNy compounds in the films and, therefore, tune the screened plasma wavelength. A good approximation of the ellipsometric behaviour is achieved with Maxwell-Garnett theory for a composite film formed by a mixture of TiO2 and TiN phases suggesting that double ENZ TiOxNy films are formed by inclusions of TiN within a TiO2 matrix. These oxynitride compounds could be considered as new materials exhibiting double ENZ in the visible and near-IR spectral ranges. △ Less

Submitted 28 March, 2017; originally announced March 2017.

Two-dimensional wave propagation without anomalous dispersion

Authors: Carl M. Bender, Francisco J. Rodriguez, Sarben Sarkar, Anatoly V. Zayats

Abstract: In two space dimensions and one time dimension a wave changes its shape even in the absence of a dispersive medium. However, this anomalous dispersive behavior in empty two-dimensional space does not occur if the wave dynamics is described by a linear homogeneous wave equation in two space dimensions and {\it two} time dimensions. Wave propagation in such a space can be realized in a three-dimensi… ▽ More In two space dimensions and one time dimension a wave changes its shape even in the absence of a dispersive medium. However, this anomalous dispersive behavior in empty two-dimensional space does not occur if the wave dynamics is described by a linear homogeneous wave equation in two space dimensions and {\it two} time dimensions. Wave propagation in such a space can be realized in a three-dimensional anisotropic metamaterial in which one of the space dimensions has a negative permittivity and thus serves as an effective second time dimension. These results lead to a fundamental understanding and new approaches to ultrashort pulse sha** in nanostructures and metamaterials. △ Less

Submitted 27 December, 2016; originally announced December 2016.

Journal ref: Phys. Rev. Lett. 119, 114301 (2017)

Lateral Casimir force on a rotating particle near a planar surface

Authors: Alejandro Manjavacas, Francisco J. Rodríguez-Fortuño, F. Javier García de Abajo, Anatoly V. Zayats

Abstract: We study the lateral Casimir force experienced by a particle that rotates near a planar surface. The origin of this force lies in the symmetry breaking induced by the particle rotation in the vacuum and thermal fluctuations of its dipole moment, and, therefore, in contrast to lateral Casimir forces previously described in the literature for corrugated surfaces, it exists despite the translational… ▽ More We study the lateral Casimir force experienced by a particle that rotates near a planar surface. The origin of this force lies in the symmetry breaking induced by the particle rotation in the vacuum and thermal fluctuations of its dipole moment, and, therefore, in contrast to lateral Casimir forces previously described in the literature for corrugated surfaces, it exists despite the translational invariance of the planar surface. Working within the framework of fluctuational electrodynamics, we derive analytical expressions for the lateral force and analyze its dependence on the geometrical and material properties of the system. In particular, we show that the direction of the force can be controlled by adjusting the particle-surface distance, which may be exploited as a new mechanism to manipulate nanoscale objects. △ Less

Submitted 12 December, 2016; originally announced December 2016.

Comments: 8 pages, 5 figures

Journal ref: Phys. Rev. Lett. 118, 133605 (2017)

Spontaneous Emission in Nonlocal Materials

Authors: Pavel Ginzburg, Diane Roth, Mazhar E. Nasir, Paulina Segovia Olvera, Alexey V. Krasavin, James Levitt, Liisa M. Hirvonen, Brian Wells, Klaus Suhling, David Richards, Viktor A. Podolskiy, Anatoly V. Zayats

Abstract: Light-matter interactions can be dramatically modified by the surrounding environment. Here we report on the first experimental observation of molecular spontaneous emission inside a highly nonlocal metamaterial based on a plasmonic nanorod assembly. We show that the emission process is dominated not only by the topology of its local effective medium dispersion, but also by the nonlocal response o… ▽ More Light-matter interactions can be dramatically modified by the surrounding environment. Here we report on the first experimental observation of molecular spontaneous emission inside a highly nonlocal metamaterial based on a plasmonic nanorod assembly. We show that the emission process is dominated not only by the topology of its local effective medium dispersion, but also by the nonlocal response of the composite, so that metamaterials with different geometric parameters but the same local effective medium properties exhibit different Purcell factors. A record-high enhancement of a decay rate is observed, in agreement with the developed quantitative description of the Purcell effect in a nonlocal medium. An engineered material nonlocality introduces an additional degree of freedom into quantum electrodynamics, enabling new applications in quantum information processing, photo-chemistry, imaging, and sensing. △ Less

Submitted 16 May, 2016; originally announced May 2016.

Interscale Mixing Microscopy: far field imaging beyond the diffraction limit

Authors: Christopher M. Roberts, Nicolas Olivier, William P. Wardley, Sandeep Inampudi, Wayne Dickson, Anatoly V. Zayats, Viktor A. Podolskiy

Abstract: We present an analytical description and an experimental realization of interscale mixing microscopy, a diffraction-based imaging technique that is capable of detecting wavelength/10 objects in far-field measurements with both coherent and incoherent broadband light. This method aims at recovering the spatial spectrum of light diffracted by a subwavelength object based on far-field measurements of… ▽ More We present an analytical description and an experimental realization of interscale mixing microscopy, a diffraction-based imaging technique that is capable of detecting wavelength/10 objects in far-field measurements with both coherent and incoherent broadband light. This method aims at recovering the spatial spectrum of light diffracted by a subwavelength object based on far-field measurements of the interference created by the object and a finite diffraction grating. A single measurement, analyzing the multiple diffraction orders, is often sufficient to determine the parameters of the object. The presented formalism opens the door for spectroscopy of nanoscale objects in the far-field. △ Less

Submitted 18 April, 2016; originally announced April 2016.

Comments: 5 pages, 5 figures

Disfluency Detection using a Bidirectional LSTM

Authors: Vicky Zayats, Mari Ostendorf, Hannaneh Hajishirzi

Abstract: We introduce a new approach for disfluency detection using a Bidirectional Long-Short Term Memory neural network (BLSTM). In addition to the word sequence, the model takes as input pattern match features that were developed to reduce sensitivity to vocabulary size in training, which lead to improved performance over the word sequence alone. The BLSTM takes advantage of explicit repair states in ad… ▽ More We introduce a new approach for disfluency detection using a Bidirectional Long-Short Term Memory neural network (BLSTM). In addition to the word sequence, the model takes as input pattern match features that were developed to reduce sensitivity to vocabulary size in training, which lead to improved performance over the word sequence alone. The BLSTM takes advantage of explicit repair states in addition to the standard reparandum states. The final output leverages integer linear programming to incorporate constraints of disfluency structure. In experiments on the Switchboard corpus, the model achieves state-of-the-art performance for both the standard disfluency detection task and the correction detection task. Analysis shows that the model has better detection of non-repetition disfluencies, which tend to be much harder to detect. △ Less

Submitted 11 April, 2016; originally announced April 2016.

Comments: Submitted to INTERSPEECH 2016

Benchmarking system-level performance of passive and active plasmonic components: integrated circuits approach

Authors: Alexey V. Krasavin, Anatoly V. Zayats

Abstract: Using criteria of bandwidth and energy consumption for signal guiding and processing, system-level figures of merit for both passive and active plasmonic circuit components are introduced, benchmarking their performance for the realisation of high-bandwidth optical data communication on a chip. The figure of merit for passive plasmonic interconnects has been derived in terms of the system level pe… ▽ More Using criteria of bandwidth and energy consumption for signal guiding and processing, system-level figures of merit for both passive and active plasmonic circuit components are introduced, benchmarking their performance for the realisation of high-bandwidth optical data communication on a chip. The figure of merit for passive plasmonic interconnects has been derived in terms of the system level performance of the plasmonic circuitry, emphasising the bandwidth and power consumption densities. These parameters are linked to the local waveguide characteristics, such as the mode propagation length, bend radius and mode size. The figure of merit enables a comparison of the main types of plasmonic waveguides and can serve as a benchmark for future designs of photonic integrated circuits. A figure of merit for active photonic- or plasmonic-based electro-optical, thermo-optical and all-optical modulators is also derived to reflect the same benchmarking principles. A particular emphasis is made on establishing a practically oriented benchmark where the integral performance of the circuit, not the size or energy consumption of individual components, plays the defining role. △ Less

Submitted 23 August, 2016; v1 submitted 20 October, 2015; originally announced October 2015.

Second-harmonic generation from hyperbolic plasmonic nanorod metamaterial slab

Authors: Giuseppe Marino, Paulina Segovia, Alexey V. Krasavin, Pavel Ginzburg, Nicolas Olivier, Gregory A. Wurtz, Anatoly V. Zayats

Abstract: Hyperbolic plasmonic metamaterials provide numerous opportunities for designing unusual linear and nonlinear optical properties. We show that the modal overlap of fundamental and second-harmonic light in an anisotropic plasmonic metamaterial slab results in the broadband enhancement of radiated second-harmonic intensity by up to 2 orders of magnitudes for TM- and TE-polarized fundamental light, co… ▽ More Hyperbolic plasmonic metamaterials provide numerous opportunities for designing unusual linear and nonlinear optical properties. We show that the modal overlap of fundamental and second-harmonic light in an anisotropic plasmonic metamaterial slab results in the broadband enhancement of radiated second-harmonic intensity by up to 2 orders of magnitudes for TM- and TE-polarized fundamental light, compared to a smooth Au film under TM-polarised illumination. The results open up possibilities to design tuneable frequency-doubling metamaterial with the goal to overcome limitations associated with classical phase matching conditions in thick nonlinear crystals. △ Less

Submitted 7 January, 2016; v1 submitted 30 August, 2015; originally announced August 2015.

Comments: A misprint in the code has been found that affected relative intensities of the SHG reported in the previous version, but not its spectral dependencies

Talking to the crowd: What do people react to in online discussions?

Authors: Aaron Jaech, Victoria Zayats, Hao Fang, Mari Ostendorf, Hannaneh Hajishirzi

Abstract: This paper addresses the question of how language use affects community reaction to comments in online discussion forums, and the relative importance of the message vs. the messenger. A new comment ranking task is proposed based on community annotated karma in Reddit discussions, which controls for topic and timing of comments. Experimental work with discussion threads from six subreddits shows th… ▽ More This paper addresses the question of how language use affects community reaction to comments in online discussion forums, and the relative importance of the message vs. the messenger. A new comment ranking task is proposed based on community annotated karma in Reddit discussions, which controls for topic and timing of comments. Experimental work with discussion threads from six subreddits shows that the importance of different types of language features varies with the community of interest. △ Less

Submitted 16 August, 2015; v1 submitted 8 July, 2015; originally announced July 2015.

Spin-orbit interactions of light

Authors: K. Y. Bliokh, F. J. Rodriguez-Fortuno, F. Nori, A. V. Zayats

Abstract: Light carries spin and orbital angular momentum. These dynamical properties are determined by the polarization and spatial degrees of freedom of light. Modern nano-optics, photonics, and plasmonics, tend to explore subwavelength scales and additional degrees of freedom of structured, i.e., spatially-inhomogeneous, optical fields. In such fields, spin and orbital properties become strongly coupled… ▽ More Light carries spin and orbital angular momentum. These dynamical properties are determined by the polarization and spatial degrees of freedom of light. Modern nano-optics, photonics, and plasmonics, tend to explore subwavelength scales and additional degrees of freedom of structured, i.e., spatially-inhomogeneous, optical fields. In such fields, spin and orbital properties become strongly coupled with each other. We overview the fundamental origins and important applications of the main spin-orbit interaction phenomena in optics. These include: spin-Hall effects in inhomogeneous media and at optical interfaces, spin-dependent effects in nonparaxial (focused or scattered) fields, spin-controlled sha** of light using anisotropic structured interfaces (metasurfaces), as well as robust spin-directional coupling via evanescent near fields. We show that spin-orbit interactions are inherent in all basic optical processes, and they play a crucial role at subwavelength scales and structures in modern optics. △ Less

Submitted 20 May, 2015; v1 submitted 11 May, 2015; originally announced May 2015.

Comments: 23 pages, 6 figures. Comments welcome

Journal ref: Nature Photon. 9, 796 (2015)

Purcell effect in Hyperbolic Metamaterial Resonators

Authors: Alexey P. Slobozhanyuk, Pavel Ginzburg, David A. Powell, Ivan Iorsh, Alexander S. Shalin, Paulina Segovia, Alexey V. Krasavin, Gregory A. Wurtz, Viktor A. Podolskiy, Pavel A. Belov, Anatoly V. Zayats

Abstract: The radiation dynamics of optical emitters can be manipulated by properly designed material structures providing high local density of photonic states, a phenomenon often referred to as the Purcell effect. Plasmonic nanorod metamaterials with hyperbolic dispersion of electromagnetic modes are believed to deliver a significant Purcell enhancement with both broadband and non-resonant nature. Here, w… ▽ More The radiation dynamics of optical emitters can be manipulated by properly designed material structures providing high local density of photonic states, a phenomenon often referred to as the Purcell effect. Plasmonic nanorod metamaterials with hyperbolic dispersion of electromagnetic modes are believed to deliver a significant Purcell enhancement with both broadband and non-resonant nature. Here, we have investigated finite-size cavities formed by nanorod metamaterials and shown that the main mechanism of the Purcell effect in these hyperbolic resonators originates from the cavity hyperbolic modes, which in a microscopic description stem from the interacting cylindrical surface plasmon modes of the finite number of nanorods forming the cavity. It is found that emitters polarized perpendicular to the nanorods exhibit strong decay rate enhancement, which is predominantly influenced by the rod length. We demonstrate that this enhancement originates from Fabry-Perot modes of the metamaterial cavity. The Purcell factors, delivered by those cavity modes, reach several hundred, which is 4-5 times larger than those emerging at the epsilon near zero transition frequencies. The effect of enhancement is less pronounced for dipoles, polarized along the rods. Furthermore, it was shown that the Purcell factor delivered by Fabry-Perot modes follows the dimension parameters of the array, while the decay rate in the epsilon near-zero regime is almost insensitive to geometry. The presented analysis shows a possibility to engineer emitter properties in the structured metamaterials, addressing their microscopic structure. △ Less

Submitted 27 April, 2015; originally announced April 2015.