Search

High-repetition-rate sources of bright electron bunches have a wide range of applications. They can directly be employed as probes in electron-scattering setups, or serve as a backbone for the generation of radiation over a broad range of the electromagnetic spectrum. This paper describes the development of a compact sub-Mega-electronvolt (sub-MeV) electron-source setup capable of operating at MHz repetition rates and forming sub-picosecond electron bunches with transverse emittance below 20~nm. The setup relies on a conduction-cooled superconducting single-cell resonator with its geometry altered to enhance the field at the surface of the emitter. The system is designed to accommodate cooling using a model a $2$~W at 4.2 K pulse tube cryogen-free cryocooler. Although we focus on the case of a photoemitted electron bunch, the scheme could be adapted to other emission mechanisms.

The first investigation of the single-particle structure of the bound states of 17C, via the d(16C, p) transfer reaction, has been undertaken. The measured angular distributions confirm the spin-parity assignments of 1/2+ and 5/2+ for the excited states located at 217 and 335 keV, respectively. The spectroscopic factors deduced for these states exhibit a marked single-particle character, in agreement with shell model and particle-core model calculations, and combined with their near degeneracy in energy provide clear evidence for the absence of the N = 14 sub-shell closure. The very small spectroscopic factor found for the 3/2+ ground state is consistent with theoretical predictions and indicates that the {\nu}1d3/2 strength is carried by unbound states. With a dominant l = 0 valence neutron configuration and a very low separation energy, the 1/2+ excited state is a one-neutron halo candidate., Comment: 7 pages, 5 figures, Accepted for publication in Physics Letters B

Bootstrap Your Own Latent (BYOL) is a self-supervised learning approach for image representation. From an augmented view of an image, BYOL trains an online network to predict a target network representation of a different augmented view of the same image. Unlike contrastive methods, BYOL does not explicitly use a repulsion term built from negative pairs in its training objective. Yet, it avoids collapse to a trivial, constant representation. Thus, it has recently been hypothesized that batch normalization (BN) is critical to prevent collapse in BYOL. Indeed, BN flows gradients across batch elements, and could leak information about negative views in the batch, which could act as an implicit negative (contrastive) term. However, we experimentally show that replacing BN with a batch-independent normalization scheme (namely, a combination of group normalization and weight standardization) achieves performance comparable to vanilla BYOL ($73.9\%$ vs. $74.3\%$ top-1 accuracy under the linear evaluation protocol on ImageNet with ResNet-$50$). Our finding disproves the hypothesis that the use of batch statistics is a crucial ingredient for BYOL to learn useful representations.

Moir\'e superlattices in two-dimensional (2D) van der Waals (vdW) heterostructures provide 20 an efficient way to engineer electron band properties. The recent discovery of exotic quantum phases and their interplay in twisted bilayer graphene (tBLG) has built this moir\'e system one of the most renowned condensed matter platforms (1-10). So far the studies of tBLG has been mostly focused on the lowest two flat moir\'e bands at the first magic angle {\theta}m1 ~ 1.1{\deg}, leaving high-order moir\'e bands and magic angles largely unexplored. Here we report 25 an observation of multiple well-isolated flat moir\'e bands in tBLG close to the second magic angle {\theta}m2 ~ 0.5{\deg}, which cannot be explained without considering electron-election interactions. With high magnetic field magneto-transport measurements, we further reveal a qualitatively new, energetically unbound Hofstadter butterfly spectrum in which continuously extended quantized Landau level gaps cross all trivial band-gaps. The 30 connected Hofstadter butterfly strongly evidences the topologically nontrivial textures of the multiple moir\'e bands. Overall, our work provides a new perspective for understanding the quantum phases in tBLG and the fractal Hofstadter spectra of multiple topological bands.

We introduce Bootstrap Your Own Latent (BYOL), a new approach to self-supervised image representation learning. BYOL relies on two neural networks, referred to as online and target networks, that interact and learn from each other. From an augmented view of an image, we train the online network to predict the target network representation of the same image under a different augmented view. At the same time, we update the target network with a slow-moving average of the online network. While state-of-the art methods rely on negative pairs, BYOL achieves a new state of the art without them. BYOL reaches $74.3\%$ top-1 classification accuracy on ImageNet using a linear evaluation with a ResNet-50 architecture and $79.6\%$ with a larger ResNet. We show that BYOL performs on par or better than the current state of the art on both transfer and semi-supervised benchmarks. Our implementation and pretrained models are given on GitHub.

Deep reinforcement learning (RL) has led to many recent and groundbreaking advances. However, these advances have often come at the cost of both increased scale in the underlying architectures being trained as well as increased complexity of the RL algorithms used to train them. These increases have in turn made it more difficult for researchers to rapidly prototype new ideas or reproduce published RL algorithms. To address these concerns this work describes Acme, a framework for constructing novel RL algorithms that is specifically designed to enable agents that are built using simple, modular components that can be used at various scales of execution. While the primary goal of Acme is to provide a framework for algorithm development, a secondary goal is to provide simple reference implementations of important or state-of-the-art algorithms. These implementations serve both as a validation of our design decisions as well as an important contribution to reproducibility in RL research. In this work we describe the major design decisions made within Acme and give further details as to how its components can be used to implement various algorithms. Our experiments provide baselines for a number of common and state-of-the-art algorithms as well as showing how these algorithms can be scaled up for much larger and more complex environments. This highlights one of the primary advantages of Acme, namely that it can be used to implement large, distributed RL algorithms that can run at massive scales while still maintaining the inherent readability of that implementation. This work presents a second version of the paper which coincides with an increase in modularity, additional emphasis on offline, imitation and learning from demonstrations algorithms, as well as various new agents implemented as part of Acme., Comment: This work presents a second version of the paper which coincides with an increase in modularity, additional emphasis on offline, imitation and learning from demonstrations algorithms, as well as various new agents implemented as part of Acme

The exact nature of the lowest $K^\pi =2_\gamma ^+$ rotational bands in all deformed nuclei remains obscure. Traditionally they are assumed to be collective vibrations of the nuclear shape in the $\gamma$ degree of freedom perpendicular to the nuclear symmetry axis. Very few such $\gamma$-bands have been traced past the usual back-bending rotational alignments of high-j nucleons. We have investigated the structure of positive-parity bands in the N=90 nucleus 156Dy, using the 148Nd(12C,4n)156Dy reaction at 65 MeV, observing the resulting ${\gamma}$-ray transitions with the Gammasphere array. The even- and odd-spin members of the $K^\pi =2_\gamma^+$ $\gamma$-band are observed to 32+ and 31+ respectively. This rotational band faithfully tracks the ground-state configuration to the highest spins. The members of a possible $\gamma$-vibration built on the aligned yrast S-band are observed to spins 28+ and 27+. An even-spin positive-parity band, observed to spin 24+, is a candidate for an aligned S-band built on the seniority-zero configuration of the $0_2^+$ state at 676 keV. The crossing of this band with the $0_2^+$ band is at $\hbar\omega$= 0.28(1) MeV and is consistent with the configuration of the $0_2^+$ band not producing any blocking of the monopole pairing., Comment: Published in Physics Review C

Rotational structures have been measured using the Jurogam II and GAMMASPHERE arrays at low spin following the 155Gd({\alpha},2n)157Dy and 148Nd(12C, 5n)155Dy reactions at 25 and 65 MeV, respectively. We report high-K bands, which are conjectured to be the first candidates of a K{\pi}= 2+ {\gamma} vibrational band, built on the [505]11/2- neutron orbital, in both odd-A 155, 157Dy isotopes. The coupling of the first excited K=0+ states or the so-called \b{eta} vibrational bands at 661 and 676 keV in 154Dy and 156Dy to the [505]11/2- orbital, to produce a K{\pi}=11/2- band, was not observed in both 155Dy and 157Dy, respectively. The implication of these findings on the interpretation of the first excited 0+ states in the core nuclei 154Dy and 156Dy are also discussed., Comment: Published in Physical Rev C

Learning a good representation is an essential component for deep reinforcement learning (RL). Representation learning is especially important in multitask and partially observable settings where building a representation of the unknown environment is crucial to solve the tasks. Here we introduce Prediction of Bootstrap Latents (PBL), a simple and flexible self-supervised representation learning algorithm for multitask deep RL. PBL builds on multistep predictive representations of future observations, and focuses on capturing structured information about environment dynamics. Specifically, PBL trains its representation by predicting latent embeddings of future observations. These latent embeddings are themselves trained to be predictive of the aforementioned representations. These predictions form a bootstrapping effect, allowing the agent to learn more about the key aspects of the environment dynamics. In addition, by defining prediction tasks completely in latent space, PBL provides the flexibility of using multimodal observations involving pixel images, language instructions, rewards and more. We show in our experiments that PBL delivers across-the-board improved performance over state of the art deep RL agents in the DMLab-30 and Atari-57 multitask setting.

Here we report on Landau level spectroscopy in magnetic fields up to 34 T performed on a thin film of topological insulator Bi$_2$Te$_3$ epitaxially grown on a BaF$_2$ substrate. The observed response is consistent with the picture of a direct-gap semiconductor in which charge carriers closely resemble massive Dirac particles. The fundamental band gap reaches $E_g=(175\pm 5)$~meV at low temperatures and it is not located on the trigonal axis, thus displaying either six or twelvefold valley degeneracy. Notably, our magneto-optical data do not indicate any band inversion. This suggests that the fundamental band gap is relatively distant from the $\Gamma$ point where profound inversion exists andgives rise to relativistic-like surface states of Bi$_2$Te$_3$., Comment: 12 pages, 11 figures, to be published in Phys. Rev. B

An amplifier based on a highly-doped Chromium Zinc-Selenide (Cr:ZnSe) crystal is proposed to increase the pulse energy emitted by an electron bunch after it passes through an undulator magnet. The primary motivation is a possible use of the amplified undulator radiation emitted by a beam circulating in a particle accelerator storage ring to increase the particle beam's phase-space density---a technique dubbed Optical Stochastic Cooling (OSC). This paper uses a simple four energy level model to estimate the single-pass gain of Cr:ZnSe and presents numerical calculations combined with wave-optics simulations of undulator radiation to estimate the expected properties of the amplified undulator wave-packet.

Understanding and tuning correlated states is of great interest and significance to modern condensed matter physics. The recent discovery of unconventional superconductivity and Mott-like insulating states in magic-angle twisted bilayer graphene (tBLG) presents a unique platform to study correlation phenomena, in which the Coulomb energy dominates over the quenched kinetic energy as a result of hybridized flat bands. Extending this approach to the case of twisted multilayer graphene would allow even higher control over the band structure because of the reduced symmetry of the system. Here, we study electronic transport properties in twisted trilayer graphene (tTLG, bilayer on top of monolayer graphene heterostructure). We observed the formation of van Hove singularities which are highly tunable by twist angle and displacement field and can cause strong correlation effects under optimum conditions, including superconducting states. We provide basic theoretical interpretation of the observed electronic structure., Comment: 16 pages, 9 figures

The simultaneous measurement of the isotopic fission-fragment yields and fission-fragment velocities of $^{239}$U has been performed for the first time. The $^{239}$U fissioning system was produced in one-neutron transfer reactions between a $^{238}$U beam at 5.88 MeV/nucleon and a $^{9}$Be target. The combination of inverse kinematics at low energy and the use of the VAMOS++ spectrometer at the GANIL facility allows the isotopic identification of the full fission-fragment distribution and their velocity in the reference frame of the fissioning system. The proton and neutron content of the fragments at scission, their total kinetic and total excitation energy, as well as the neutron multiplicity were determined. Information from the scission point configuration is obtained from these observables and the correlation between them. The role of the octupole-deformed proton and neutron shells in the fission-fragment production is discussed., Comment: 8 pages, 7 figures

Atari games have been a long-standing benchmark in the reinforcement learning (RL) community for the past decade. This benchmark was proposed to test general competency of RL algorithms. Previous work has achieved good average performance by doing outstandingly well on many games of the set, but very poorly in several of the most challenging games. We propose Agent57, the first deep RL agent that outperforms the standard human benchmark on all 57 Atari games. To achieve this result, we train a neural network which parameterizes a family of policies ranging from very exploratory to purely exploitative. We propose an adaptive mechanism to choose which policy to prioritize throughout the training process. Additionally, we utilize a novel parameterization of the architecture that allows for more consistent and stable learning.

We propose a reinforcement learning agent to solve hard exploration games by learning a range of directed exploratory policies. We construct an episodic memory-based intrinsic reward using k-nearest neighbors over the agent's recent experience to train the directed exploratory policies, thereby encouraging the agent to repeatedly revisit all states in its environment. A self-supervised inverse dynamics model is used to train the embeddings of the nearest neighbour lookup, biasing the novelty signal towards what the agent can control. We employ the framework of Universal Value Function Approximators (UVFA) to simultaneously learn many directed exploration policies with the same neural network, with different trade-offs between exploration and exploitation. By using the same neural network for different degrees of exploration/exploitation, transfer is demonstrated from predominantly exploratory policies yielding effective exploitative policies. The proposed method can be incorporated to run with modern distributed RL agents that collect large amounts of experience from many actors running in parallel on separate environment instances. Our method doubles the performance of the base agent in all hard exploration in the Atari-57 suite while maintaining a very high score across the remaining games, obtaining a median human normalised score of 1344.0%. Notably, the proposed method is the first algorithm to achieve non-zero rewards (with a mean score of 8,400) in the game of Pitfall! without using demonstrations or hand-crafted features., Comment: Published as a conference paper in ICLR 2020

Landau-level spectroscopy, the optical analysis of electrons in materials subject to a strong magnetic field, is a versatile probe of the electronic band structure and has been successfully used in the identification of novel states of matter such as Dirac electrons, topological materials or Weyl semimetals. The latter arise from a complex interplay between crystal symmetry, spin-orbit interaction and inverse ordering of electronic bands. Here, we report on unusual Landau-level transitions in the monopnictide TaP that decrease in energy with increasing magnetic field. We show that these transitions arise naturally at intermediate energies in time-reversal-invariant Weyl semimetals where the Weyl nodes are formed by a partially gapped nodal-loop in the band structure. We propose a simple theoretical model for electronic bands in these Weyl materials that captures the collected magneto-optical data to great extent., Comment: 6 pages, 3 figures + supplementary materials, accepted in Phys. Rev. Lett

We consider the problem of efficient credit assignment in reinforcement learning. In order to efficiently and meaningfully utilize new data, we propose to explicitly assign credit to past decisions based on the likelihood of them having led to the observed outcome. This approach uses new information in hindsight, rather than employing foresight. Somewhat surprisingly, we show that value functions can be rewritten through this lens, yielding a new family of algorithms. We study the properties of these algorithms, and empirically show that they successfully address important credit assignment challenges, through a set of illustrative tasks., Comment: NeurIPS 2019

Of the two stable forms of graphite, hexagonal (HG) and rhombohedral (RG), the former is more common and has been studied extensively. RG is less stable, which so far precluded its detailed investigation, despite many theoretical predictions about the abundance of exotic interaction-induced physics. Advances in van der Waals heterostructure technology have now allowed us to make high-quality RG films up to 50 graphene layers thick and study their transport properties. We find that the bulk electronic states in such RG are gapped and, at low temperatures, electron transport is dominated by surface states. Because of topological protection, the surface states are robust and of high quality, allowing the observation of the quantum Hall effect, where RG exhibits phase transitions between gapless semimetallic phase and gapped quantum spin Hall phase with giant Berry curvature. An energy gap can also be opened in the surface states by breaking their inversion symmetry via applying a perpendicular electric field. Moreover, in RG films thinner than 4 nm, a gap is present even without an external electric field. This spontaneous gap opening shows pronounced hysteresis and other signatures characteristic of electronic phase separation, which we attribute to emergence of strongly-correlated electronic surface states., Comment: 18 pages, 12 figures

A direct and complete measurement of isotopic fission-fragment yields of $^{239}$U has been performed for the first time. The $^{239}$U fissioning system was produced with an average excitation energy of 8.3 MeV in one-neutron transfer reactions between a $^{238}$U beam and a $^{9}$Be target at Coulomb barrier energies. The fission fragments were detected and isotopically identified using the VAMOS++ spectrometer at the GANIL facility. This measurement allows to directly evaluate the fission models at excitation energies of fast neutrons, relevant for next-generation nuclear reactors. The present data, in agreement with model calculations, do not support the recently reported anomaly in the fission-fragment yields of $^{239}$U and confirm the persistence of spherical shell effects in the Sn region at excitation energies exceeding the fission barrier by few MeV.

We investigate thin film resistive thermometry based on metal-to-insulator-transition (niobium nitride) materials down to very low temperature. The variation of the NbN thermometer resistance have been calibrated versus temperature and magnetic field. High sensitivity in tempertaure variation detection is demonstrated through efficient temperature coefficient of resistance. The nitrogen content of the niobium nitride thin films can be tuned to adjust the optimal working temperature range. In the present experiment, we show the versatility of the NbN thin film technology through applications in very different low temperature use-cases. We demonstrate that thin film resistive thermometry can be extended to temperatures below 30 mK with low electrical impedance., Comment: Article accepted for publication in JLTP

The ground state of charge neutral graphene under perpendicular magnetic field was predicted to be a quantum Hall topological insulator with a ferromagnetic order and spin-filtered, helical edge channels. In most experiments, however, an otherwise insulating state is observed and is accounted for by lattice-scale interactions that promote a broken-symmetry state with gapped bulk and edge excitations. We tuned the ground state of the graphene zeroth Landau level to the topological phase via a suitable screening of the Coulomb interaction with a SrTiO$_3$ high-$k$ dielectric substrate. We observed robust helical edge transport emerging at a magnetic field as low as 1 tesla and withstanding temperatures up to 110 kelvins over micron-long distances. This new and versatile graphene platform opens new avenues for spintronics and topological quantum computation., Comment: Main text + SI

Temporally-shaped electron bunches at ultrafast time scales are foreseen to support an array of applications including the development of small-footprint accelerator-based coherent light sources or as probes for, e.g., ultrafast electron-diffraction. We demonstrate a method where a transversely-segmented electron bunch produced via photoemission from a transversely-patterned laser distribution is transformed into an electron bunch with modulated temporal distribution. In essence, the presented transformation enables the mapping of the transverse laser distribution on a photocathode surface to the temporal coordinate and provides a proof-of-principle experiment of the method proposed in W. S. Graves, et al. as a path toward the realization of compact coherent X-ray sources, albeit at a larger timescale. The presented experiment is validated against numerical simulations and the versatility of the concept, e.g. to tune the current-distribution parameters, is showcased. Although our work focuses on the generation of electron bunches arranged as a temporal comb it is applicable to other temporal shapes.

As humans we are driven by a strong desire for seeking novelty in our world. Also upon observing a novel pattern we are capable of refining our understanding of the world based on the new information---humans can discover their world. The outstanding ability of the human mind for discovery has led to many breakthroughs in science, art and technology. Here we investigate the possibility of building an agent capable of discovering its world using the modern AI technology. In particular we introduce NDIGO, Neural Differential Information Gain Optimisation, a self-supervised discovery model that aims at seeking new information to construct a global view of its world from partial and noisy observations. Our experiments on some controlled 2-D navigation tasks show that NDIGO outperforms state-of-the-art information-seeking methods in terms of the quality of the learned representation. The improvement in performance is particularly significant in the presence of white or structured noise where other information-seeking methods follow the noise instead of discovering their world.

We report on transport and capacitance spectroscopy study of two kinds of quantum wells, namely Cd$_{0.02}$Hg$_{0.98}$Te and Cd$_{0.06}$Hg$_{0.94}$Te with the thicknesses of 7.4 and 11.5 nm, accordingly. The fraction of Cd was chosen in a way that the both quantum wells are expected to have gapless band structure typical for a Dirac fermions system. We have established that the first quantum well exhibits a massless Dirac fermions system with a quality slightly better then in conventional HgTe quantum wells of critical thickness. Second quantum well exhibits a high-quality two-dimensional topological insulator state with the energy gap of around 10 meV and well-defined edge transport making it as a good candidate for further study and applications of topological insulators., Comment: 5 pages, 5 figures

The quantum Hall effect (QHE) originates from discrete Landau levels forming in a two-dimensional (2D) electron system in a magnetic field. In three dimensions (3D), the QHE is forbidden because the third dimension spreads Landau levels into multiple overlapping bands, destroying the quantisation. Here we report the QHE in graphite crystals that are up to hundreds of atomic layers thick - thickness at which graphite was believed to behave as a 3D bulk semimetal. We attribute the observation to a dimensional reduction of electron dynamics in high magnetic fields, such that the electron spectrum remains continuous only in the direction of the magnetic field, and only the last two quasi-one-dimensional (1D) Landau bands cross the Fermi level. In sufficiently thin graphite films, the formation of standing waves breaks these 1D bands into a discrete spectrum, giving rise to a multitude of quantum Hall plateaux. Despite a large number of layers, we observe a profound difference between films with even and odd numbers of graphene layers. For odd numbers, the absence of inversion symmetry causes valley polarisation of the standing-wave states within 1D Landau bands. This reduces QHE gaps, as compared to films of similar thicknesses but with even layer numbers because the latter retain the inversion symmetry characteristic of bilayer graphene. High-quality graphite films present a novel QHE system with a parity-controlled valley polarisation and intricate interplay between orbital, spin and valley states, and clear signatures of electron-electron interactions including the fractional QHE below 0.5 K.

Unsupervised representation learning has succeeded with excellent results in many applications. It is an especially powerful tool to learn a good representation of environments with partial or noisy observations. In partially observable domains it is important for the representation to encode a belief state, a sufficient statistic of the observations seen so far. In this paper, we investigate whether it is possible to learn such a belief representation using modern neural architectures. Specifically, we focus on one-step frame prediction and two variants of contrastive predictive coding (CPC) as the objective functions to learn the representations. To evaluate these learned representations, we test how well they can predict various pieces of information about the underlying state of the environment, e.g., position of the agent in a 3D maze. We show that all three methods are able to learn belief representations of the environment, they encode not only the state information, but also its uncertainty, a crucial aspect of belief states. We also find that for CPC multi-step predictions and action-conditioning are critical for accurate belief representations in visually complex environments. The ability of neural representations to capture the belief information has the potential to spur new advances for learning and planning in partially observable domains, where leveraging uncertainty is essential for optimal decision making.

We study the problem of learning classifiers robust to universal adversarial perturbations. While prior work approaches this problem via robust optimization, adversarial training, or input transformation, we instead phrase it as a two-player zero-sum game. In this new formulation, both players simultaneously play the same game, where one player chooses a classifier that minimizes a classification loss whilst the other player creates an adversarial perturbation that increases the same loss when applied to every sample in the training set. By observing that performing a classification (respectively creating adversarial samples) is the best response to the other player, we propose a novel extension of a game-theoretic algorithm, namely fictitious play, to the domain of training robust classifiers. Finally, we empirically show the robustness and versatility of our approach in two defence scenarios where universal attacks are performed on several image classification datasets -- CIFAR10, CIFAR100 and ImageNet.

This paper presents expected THz radiation spectra emitted by micro-bunched electron beams produced using a slit-mask placed within a magnetic chicane in the FAST (Fermilab Accelerator Science and Technology) electron injector at Fermilab. Our purpose is to generate tunable narrow-band THz radiation with a simple scheme in a conventional photo-injector. Using the slit-mask in the chicane, we create a longitudinally micro-bunched beam after the chicane by transversely slicing an energy chirped electron bunch at a location with horizontal dispersion. In this paper, we discuss the theory related to the micro-bunched beam structure, the beam optics, the simulation results of the micro-bunched beam and the bunching factors. Energy radiated at THz frequencies from two sources: coherent transition radiation and from a wiggler is calculated and compared. We also discuss the results of a simple method to observe the micro-bunching on a transverse screen monitor using a skew quadrupole placed in the chicane., Comment: 25 pages, 8 figures