Search

We study the resistivity of three-dimensional semimetals with linear dispersion in the presence of on-site electron-electron interaction. The well-known quadratic temperature dependence of the resistivity of conventional metals is turned into an unusual $T^6$-behavior. An analogous change affects the thermal transport, preserving the linearity in $T$ of the ratio between thermal and electrical conductivities. These results hold from weak coupling up to the non-perturbative region of the Mott transition. Our findings yield a natural explanation for the hitherto not understood large exponents characterizing the temperature-dependence of transport experiments on various topological semimetals., Comment: 6 pages, 3 Figures, Suppl. Mat

Since the early days of Dirac flux quantization, magnetic monopoles have been sought after as a potential corollary of quantized electric charge. As opposed to magnetic monopoles embedded into the theory of electromagnetism, Weyl crystals exhibit Berry flux monopoles in reciprocal parameter space. As a function of crystal momentum, such monopoles locate at the degeneracy point of the Weyl cone. Here, we report momentum-resolved spectroscopic signatures of Berry flux monopoles in TaAs as a paradigmatic Weyl semimetal. We have probed the orbital and spin angular momentum (OAM and SAM) of the Weyl-fermion states by angle-resolved photoemission spectroscopy at bulk-sensitive soft X-ray energies (SX-ARPES) combined with photoelectron spin detection and circular dichroism. Supported by first-principles calculations, our measurements image characteristics of a topologically non-trivial winding of the OAM at the Weyl nodes and unveil a chirality-dependent SAM of the Weyl bands. Our results experimentally visualize the non-trivial momentum-space topology in a Weyl semimetal, promising to have profound implications for the study of quantum-geometric effects in solids.

Safely interacting with humans is a significant challenge for autonomous driving. The performance of this interaction depends on machine learning-based modules of an autopilot, such as perception, behavior prediction, and planning. These modules require training datasets with high-quality labels and a diverse range of realistic dynamic behaviors. Consequently, training such modules to handle rare scenarios is difficult because they are, by definition, rarely represented in real-world datasets. Hence, there is a practical need to augment datasets with synthetic data covering these rare scenarios. In this paper, we present a platform to model dynamic and interactive scenarios, generate the scenarios in simulation with different modalities of labeled sensor data, and collect this information for data augmentation. To our knowledge, this is the first integrated platform for these tasks specialized to the autonomous driving domain.

Thanks to large-scale labeled training data, deep neural networks (DNNs) have obtained remarkable success in many vision and multimedia tasks. However, because of the presence of domain shift, the learned knowledge of the well-trained DNNs cannot be well generalized to new domains or datasets that have few labels. Unsupervised domain adaptation (UDA) studies the problem of transferring models trained on one labeled source domain to another unlabeled target domain. In this paper, we focus on UDA in visual emotion analysis for both emotion distribution learning and dominant emotion classification. Specifically, we design a novel end-to-end cycle-consistent adversarial model, termed CycleEmotionGAN++. First, we generate an adapted domain to align the source and target domains on the pixel-level by improving CycleGAN with a multi-scale structured cycle-consistency loss. During the image translation, we propose a dynamic emotional semantic consistency loss to preserve the emotion labels of the source images. Second, we train a transferable task classifier on the adapted domain with feature-level alignment between the adapted and target domains. We conduct extensive UDA experiments on the Flickr-LDL & Twitter-LDL datasets for distribution learning and ArtPhoto & FI datasets for emotion classification. The results demonstrate the significant improvements yielded by the proposed CycleEmotionGAN++ as compared to state-of-the-art UDA approaches.

Available information concerning the elastic moduli of refractory carbides at temperatures (T) of relevance for practical applications is sparse and/or inconsistent. We carry out ab initio molecular dynamics (AIMD) simulations at T = 300, 600, 900, and 1200 K to determine the temperature-dependences of the elastic constants of rocksalt-structure (B1) TiC, ZrC, HfC, VC, and TaC compounds as well as multicomponent high-entropy carbides (Ti,Zr,Hf,Ta,W)C and (V,Nb,Ta,Mo,W)C. The second order elastic constants are calculated by least-square fitting of the analytical expressions of stress vs. strain relationships to simulation results obtained from three tensile and three shear deformation modes. Moreover, we employ sound velocity measurements to evaluate the bulk, shear, elastic moduli and Poisson's ratios of single-phase B1 (Ti,Zr,Hf,Ta,W)C and (V,Nb,Ta,Mo,W)C at ambient conditions. Our experimental results are in excellent agreement with the values obtained by AIMD simulations. In comparison with the predictions of previous ab initio calculations - where the extrapolation of finite-temperature elastic properties accounted for thermal expansion while neglecting intrinsic vibrational effects - AIMD simulations produce a softening of elastic moduli with T in closer agreement with experiments. Results of our simulations show that TaC is the system which exhibits the highest elastic resistances to both tensile and shear deformation up to 1200 K, and identify the high-entropy (V,Nb,Ta,Mo,W)C system as candidate for applications that require good ductility and toughness at room as well as elevated temperatures.

We propose a new probabilistic programming language for the design and analysis of cyber-physical systems, especially those based on machine learning. Specifically, we consider the problems of training a system to be robust to rare events, testing its performance under different conditions, and debugging failures. We show how a probabilistic programming language can help address these problems by specifying distributions encoding interesting types of inputs, then sampling these to generate specialized training and test data. More generally, such languages can be used to write environment models, an essential prerequisite to any formal analysis. In this paper, we focus on systems like autonomous cars and robots, whose environment at any point in time is a 'scene', a configuration of physical objects and agents. We design a domain-specific language, Scenic, for describing scenarios that are distributions over scenes and the behaviors of their agents over time. As a probabilistic programming language, Scenic allows assigning distributions to features of the scene, as well as declaratively imposing hard and soft constraints over the scene. We develop specialized techniques for sampling from the resulting distribution, taking advantage of the structure provided by Scenic's domain-specific syntax. Finally, we apply Scenic in a case study on a convolutional neural network designed to detect cars in road images, improving its performance beyond that achieved by state-of-the-art synthetic data generation methods., Comment: Supercedes arXiv:1809.09310

Seeking the largest solution to an expression of the form A x <= B is a common task in several domains of engineering and computer science. This largest solution is commonly called quotient. Across domains, the meanings of the binary operation and the preorder are quite different, yet the syntax for computing the largest solution is remarkably similar. This paper is about finding a common framework to reason about quotients. We only assume we operate on a preorder endowed with an abstract monotonic multiplication and an involution. We provide a condition, called admissibility, which guarantees the existence of the quotient, and which yields its closed form. We call preordered heaps those structures satisfying the admissibility condition. We show that many existing theories in computer science are preordered heaps, and we are thus able to derive a quotient for them, subsuming existing solutions when available in the literature. We introduce the concept of sieved heaps to deal with structures which are given over multiple domains of definition. We show that sieved heaps also have well-defined quotients., Comment: In Proceedings GandALF 2020, arXiv:2009.09360

Large-scale labeled training datasets have enabled deep neural networks to excel across a wide range of benchmark vision tasks. However, in many applications, it is prohibitively expensive and time-consuming to obtain large quantities of labeled data. To cope with limited labeled training data, many have attempted to directly apply models trained on a large-scale labeled source domain to another sparsely labeled or unlabeled target domain. Unfortunately, direct transfer across domains often performs poorly due to the presence of domain shift or dataset bias. Domain adaptation is a machine learning paradigm that aims to learn a model from a source domain that can perform well on a different (but related) target domain. In this paper, we review the latest single-source deep unsupervised domain adaptation methods focused on visual tasks and discuss new perspectives for future research. We begin with the definitions of different domain adaptation strategies and the descriptions of existing benchmark datasets. We then summarize and compare different categories of single-source unsupervised domain adaptation methods, including discrepancy-based methods, adversarial discriminative methods, adversarial generative methods, and self-supervision-based methods. Finally, we discuss future research directions with challenges and possible solutions.

Incipient anomalies present milder symptoms compared to severe ones, and are more difficult to detect and diagnose due to their close resemblance to normal operating conditions. The lack of incipient anomaly examples in the training data can pose severe risks to anomaly detection methods that are built upon Machine Learning (ML) techniques, because these anomalies can be easily mistaken as normal operating conditions. To address this challenge, we propose to utilize the uncertainty information available from ensemble learning to identify potential misclassified incipient anomalies. We show in this paper that ensemble learning methods can give improved performance on incipient anomalies and identify common pitfalls in these models through extensive experiments on two real-world datasets. Then, we discuss how to design more effective ensemble models for detecting incipient anomalies., Comment: Submitted to Transactions on Cyber-Physical Systems for Special Issue on AI and Cyber-Physical Systems

Incipient anomalies present milder symptoms compared to severe ones, and are more difficult to detect and diagnose due to their close resemblance to normal operating conditions. The lack of incipient anomaly examples in the training data can pose severe risks to anomaly detection methods that are built upon Machine Learning (ML) techniques, because these anomalies can be easily mistaken as normal operating conditions. To address this challenge, we propose to utilize the uncertainty information available from ensemble learning to identify potential misclassified incipient anomalies. We show in this paper that ensemble learning methods can give improved performance on incipient anomalies and identify common pitfalls in these models through extensive experiments on two real-world datasets. Then, we discuss how to design more effective ensemble models for detecting incipient anomalies., Comment: Submitted to Transactions on Cyber-Physical Systems for Special Issue on AI and Cyber-Physical Systems

Ensemble learning is widely applied in Machine Learning (ML) to improve model performance and to mitigate decision risks. In this approach, predictions from a diverse set of learners are combined to obtain a joint decision. Recently, various methods have been explored in literature for estimating decision uncertainties using ensemble learning; however, determining which metrics are a better fit for certain decision-making applications remains a challenging task. In this paper, we study the following key research question in the selection of uncertainty metrics: when does an uncertainty metric outperforms another? We answer this question via a rigorous analysis of two commonly used uncertainty metrics in ensemble learning, namely ensemble mean and ensemble variance. We show that, under mild assumptions on the ensemble learners, ensemble mean is preferable with respect to ensemble variance as an uncertainty metric for decision making. We empirically validate our assumptions and theoretical results via an extensive case study: the diagnosis of referable diabetic retinopathy., Comment: Preprint of submission to NeurIPS 2020

Intermediate-Severity (IS) faults present milder symptoms compared to severe faults, and are more difficult to detect and diagnose due to their close resemblance to normal operating conditions. The lack of IS fault examples in the training data can pose severe risks to Fault Detection and Diagnosis (FDD) methods that are built upon Machine Learning (ML) techniques, because these faults can be easily mistaken as normal operating conditions. Ensemble models are widely applied in ML and are considered promising methods for detecting out-of-distribution (OOD) data. We identify common pitfalls in these models through extensive experiments with several popular ensemble models on two real-world datasets. Then, we discuss how to design more effective ensemble models for detecting and diagnosing IS faults.

The coupling of a spin to an underlying substrate is the basis for a plethora of phenomena. In the case of a metallic substrate, Kondo screening of the adatom magnetic moment can occur. As the substrate turns superconducting, an intriguing situation emerges where the pair breaking due to the adatom spins leads to Yu-Shiba-Rusinov bound states, but also intertwines with Kondo phenomena. Through scanning tunneling spectroscopy, we analyze the interdependence of Kondo screening and superconductivity. Our data obtained on single Fe adatoms on Nb(110) show that the coupling and the resulting YSR states are strongly adsorption site-dependent and reveal a quantum phase transition at a Kondo temperature comparable to the superconducting gap. The experimental signatures are rationalized by combined density functional theory and continuous-time quantum Monte-Carlo calculations to rigorously treat magnetic and hybridization effects on equal footing., Comment: 6 pages, 4 figures

Industry 4.0 is changing fundamentally data collection, its storage and analysis in industrial processes, enabling novel application such as flexible manufacturing of highly customized products. Real-time control of these processes, however, has not yet realized its full potential in using the collected data to drive further development. Indeed, typical industrial control systems are tailored to the plant they need to control, making reuse and adaptation a challenge. In the past, the need to solve plant specific problems overshadowed the benefits of physically isolating a control system from its plant. We believe that modern virtualization techniques, specifically application containers, present a unique opportunity to decouple control from plants. This separation permits us to fully realize the potential for highly distributed, and transferable industrial processes even with real-time constraints arising from time-critical sub-processes. In this paper, we explore the challenges and opportunities of shifting industrial control software from dedicated hardware to bare-metal servers or (edge) cloud computing platforms using off-the-shelf technology. We present a migration architecture and show, using a specifically developed orchestration tool, that containerized applications can run on shared resources without compromising scheduled execution within given time constraints. Through latency and computational performance experiments we explore limits of three system setups and summarize lessons learned., Comment: 24 pages, 5 figures, 5 tables, submitted do Journal, under Review

Using a numerically exact first-principles many-body approach, we revisit the "prototypical" Kondo case of a cobalt impurity on copper. Even though this is considered a well understood example of the Kondo effect, we reveal an unexpectedly strong dependence of the screening properties on the parametrization of the local Coulomb tensor. As a consequence, the Kondo temperature can vary by orders of magnitude depending on the complexity of the parametrization of the electron-electron interaction. Further, we challenge the established picture of a spin-$1$ moment involving two cobalt $d$-orbitals only, as orbital-mixing interaction terms boost the contribution of the remainder of the $d$-shell., Comment: 15 pages, 14 figures; final version with extended bibliography and appendix

While the breakdown of the perturbation expansion for the many-electron problem has several formal consequences, here we unveil its physical effect: Flipping the sign of the effective electronic interaction in specific scattering channels. By decomposing local and uniform susceptibilities of the Hubbard model via their spectral representations, we prove how entering the non-perturbative regime causes an enhancement of the charge response, ultimately responsible for the phase-separation instabilities close to the Mott MIT. Our analysis opens a new route for understanding phase-transitions in the non-perturbative regime and clarifies why attractive effects emerging from a strong repulsion can induce phase-separations, but not s-wave pairing or charge-density wave instabilities., Comment: 6 pages, 4 Figures, Suppl. Mat.; accepted for publication in Physical Review Letters

High hardness and toughness are generally considered mutually exclusive properties for single-crystal ceramics. Combining experiments and ab initio molecular dynamics (AIMD) atomistic simulations at room temperature, we demonstrate that both the hardness and toughness of single-crystal NaCl-structure VNx/MgO(001) thin films are simultaneously enhanced through the incorporation of anion vacancies. Nanoindentation results show that VN0.8, here considered as representative understoichiometric VNx system, is ~20% harder, as well as more resistant to fracture than stoichiometric VN samples. AIMD modeling of VN and VN0.8 supercells subjected to [001] and [110] elongation reveal that the tensile strengths of the two materials are similar. Nevertheless, while the stoichiometric VN phase systematically cleaves in a brittle manner at tensile yield points, the understoichiometric compound activates transformation-toughening mechanisms that dissipate accumulated stresses. AIMD simulations also show that VN0.8 exhibits an initially greater resistance to both {110}<1-10> and {111}<1-10> shear deformation than VN. However, for progressively increasing shear strains, the VN0.8 mechanical behavior gradually evolves from harder to more ductile than VN. The transition is mediated by anion vacancies, which facilitate {110}<1-10> and {111}<1-10> lattice slip by reducing activation shear stresses by as much as 35%. Electronic-structure analyses show that the two-regime hard/tough mechanical response of VN0.8 primarily stems from its intrinsic ability to transfer d electrons between 2nd-neighbor and 4th-neighbor (i.e., across vacancy sites) V-V metallic states. Our work offers a route for electronic-structure design of hard materials in which a plastic mechanical response is triggered with loading.

The interplay of electronic correlations, spin-orbit coupling and topology holds promise for the realization of exotic states of quantum matter. Models of strongly interacting electrons on honeycomb lattices have revealed rich phase diagrams featuring unconventional quantum states including chiral superconductivity and correlated quantum spin Hall insulators intertwining with complex magnetic order. Material realizations of these electronic states are however scarce or inexistent. In this work, we propose and show that stacking 1T-TaSe$_{2}$ into bilayers can deconfine electrons from a deep Mott insulating state in the monolayer to a system of correlated Dirac fermions subject to sizable spin-orbit coupling in the bilayer. 1T-TaSe$_{2}$ develops a Star-of-David (SoD) charge density wave pattern in each layer. When the SoD centers belonging to two adyacent layers are stacked in a honeycomb pattern, the system realizes a generalized Kane-Mele-Hubbard model in a regime where Dirac semimetallic states are subject to significant Mott-Hubbard interactions and spin-orbit coupling. At charge neutrality, the system is close to a quantum phase transition between a quantum spin Hall and an antiferromagnetic insulator. We identify a perpendicular electric field and the twisting angle as two knobs to control topology and spin-orbit coupling in the system. Their combination can drive it across hitherto unexplored grounds of correlated electron physics including a quantum tricritical point and an exotic first-order topological phase transition., Comment: 8 pages, 3 figures, Supplemental Material with 6 pages and 4 figures, hopping data files included

We study the characteristic time scales of the fluctuating local moments in Hund's metal systems for different degrees of correlation. By analyzing the dynamical spin susceptibility in the real-time domain via the fluctuation-dissipation theorem, we determine the time scales controlling oscillation and damping of on-site fluctuations - a crucial factor for the detection of local moments with different experimental probes. We apply this procedure to realistic many-body calculations of different families of iron-pnictides and chalcogenides, explaining the material-specific trend in the discrepancies reported between experimental and theoretical estimates of the magnetic moments., Comment: 6 pages, 3 figures + supplemental material (7 pages, 2 figures) accepted for publication in Physical Review Letters

Two-dimensional materials can be strongly influenced by their surroundings. A dielectric environment screens and reduces the Coulomb interaction between electrons in the two-dimensional material. Since in Mott materials the Coulomb interaction is responsible for the insulating state, manipulating the dielectric screening provides direct control over Mottness. Our many-body calculations reveal the spectroscopic fingerprints of such Coulomb engineering: we demonstrate eV-scale changes to the position of the Hubbard bands and show a Coulomb engineered insulator-to-metal transition. Based on our proof-of-principle calculations, we discuss the (feasible) conditions under which our scenario of Coulomb engineering of Mott materials can be realized experimentally.

Ab initio-calculated ideal strength and toughness describe the upper limits for mechanical properties attainable in real systems and can, therefore, be used in selection criteria for materials design. We employ density-functional ab initio molecular dynamics (AIMD) to investigate the mechanical properties of defect-free rocksalt-structure (B1) TiN and B1 Ti1-xAlxN (x = 0.25, 0.5, 0.75) solid solutions subject to [001], [110], and [111] tensile deformation at room temperature. We determine the alloys' ideal strength and toughness, elastic responses, and ability to plastically deform up to fracture as a function of the Al content. Overall, TiN exhibits greater ideal moduli of resilience and tensile strengths than TiAlN solid solutions. Nevertheless, AIMD modelingshows that, irrespective of the strain direction, the binary compound systematically fractures by brittle cleavage at its yield point. The simulations also indicate that Ti0.5Al0.5N and Ti0.25Al0.75N solid solutions are inherently more resistant to fracture and possess much greater toughness than TiN, due to the activation of local structural transformations (primarily of B1 -> wurtzite type) beyond the elastic-response regime. In sharp contrast, TiAlN alloys with 25% Al exhibit similar brittleness as TiN. The results of this work are examples of the limitations of elasticity-based criteria for prediction of strength, brittleness, ductility, and toughness in materials able to undergo phase transitions with loading. Furthermore, comparing present and previous findings, we suggest a general principle for design of hard ceramic solid solutions that are thermodynamically inclined to dissipate extreme mechanical stresses via transformation toughening mechanisms.

Linear atomic chains containing a single Kondo atom, Co, and several nonmagnetic atoms, Cu, were assembled atom by atom on Cu(111) with the tip of a scanning tunneling microscope. The resulting one-dimensional wires, Cu$_m$CoCu$_n$ ($0\leq m, n\leq 5$), exhibit a rich evolution of the single-Co Kondo effect with the variation of $m$ and $n$, as inferred from changes in the line shape of the Abrikosov-Suhl-Kondo resonance. The most striking result is the quenching of the resonance in CuCoCu$_2$ and Cu$_2$CoCu$_2$ clusters. State-of-the-art first-principles calculations were performed to unravel possible microscopic origins of the remarkable experimental observations., Comment: 10 pages, 9 figures

The Rashba effect is fundamental to the physics of two-dimensional electron systems and underlies a variety of spintronic phenomena. It has been proposed that the formation of Rashba-type spin splittings originates microscopically from the existence of orbital angular momentum (OAM) in the Bloch wave functions. Here, we present detailed experimental evidence for this OAM-based origin of the Rashba effect by angle-resolved photoemission (ARPES) and two-photon photoemission (2PPE) experiments for a monolayer AgTe on Ag(111). Using quantitative low-energy electron diffraction (LEED) analysis we determine the structural parameters and the stacking of the honeycomb overlayer with picometer precision. Based on an orbital-symmetry analysis in ARPES and supported by first-principles calculations, we unequivocally relate the presence and absence of Rashba-type spin splittings in different bands of AgTe to the existence of OAM.

We investigate the divergences appearing in the two-particle irreducible vertex functions of many-fermion systems with attractive on-site interactions. By means of dynamical mean-field theory calculations, we determine the location of singularity lines in the phase diagram of the attractive Hubbard model at half-filling, where the local Bethe-Salpeter equations are non invertible. We find that divergences appear both in the magnetic and in the density scattering channels. The former affect a sector of suppressed fluctuations and comply with the mapping of the physical susceptibilities of the repulsive case. At the same time, the appearance of singularities in the density channel of the attractive model demonstrates that vertex divergences can also plague the dominant scattering sectors associated with enhanced local susceptibilities. This constitutes a counterexample to previously proposed interpretations. Eventually, by exploiting the underlying physical symmetries and a spectral representation of the susceptibilities, we clarify the relation between vertex divergences and the local response of the system in different channels.