HAL-CEA
Not a member yet
145450 research outputs found
Sort by
Wall-crossing phenomenon for the liquid bin model
No full textWe introduce the liquid bin model as a continuous-time deterministic dynamics, arising as the hydrodynamic limit of a discrete-time stochastic interacting particle system called the infinite bin model. For the liquid bin model, we prove the existence and uniqueness of a stationary evolution, to which the dynamics converges exponentially fast. The speed of the front of the system is explicitly computed as a continuous piecewise rational function of the parameters of the model, revealing an underlying wall-crossing phenomenon. We show that the regions on which the speed is rational are of non-empty interior and are naturally indexed by Dyck paths. We provide a complete description of the adjacency structure of these regions, which generalizes the Stanley lattice for Dyck paths. Finally we point out an intriguing connection to the topic of extensions of partial cyclic orders to total cyclic orders
Key role of short-lived halogens on global atmospheric oxidation during historical periods
No full textInternational audienceAtmospheric oxidation largely determines the abundance and lifetime of short-lived climate forcers like methane, ozone and aerosols, as well as the removal of pollutants from the atmosphere. Hydroxyl, nitrate and chlorine radicals (OH, NO 3 and Cl), together with ozone (O 3 ), are the main atmospheric oxidants. Short-lived halogens (SLH) affect the concentrations of these oxidants, either through direct chemical reactions or indirectly by perturbing their main sources and sinks. However, the effect of SLH on the combined abundance of global oxidants during historical periods remains unquantified and is not accounted for in air quality and climate models. Here, we employ a state-of-the-art chemistry-climate model to comprehensively assess the role of SLH on atmospheric oxidation under both pre-industrial (PI) and present-day (PD) conditions. Our results show a substantial reduction in present-day atmospheric oxidation caused by the SLH-driven combined reduction in the global boundary layer levels of OH (16%), NO 3 (38%) and ozone (26%), which is not compensated by the pronounced increase in Cl (2632%). These global differences in atmospheric oxidants show large spatial heterogeneity due to the variability in SLH emissions and their nonlinear chemical interactions with anthropogenic pollution. Remarkably, we find that the effect of SLH was more pronounced in the pristine PI atmosphere, where a quarter (OH: -25%) and half (NO 3 : -49%) of the boundary layer concentration of the main daytime and nighttime atmospheric oxidants, respectively, were controlled by SLH chemistry. The lack of inclusion of the substantial SLH-mediated reduction in global atmospheric oxidation in models may lead to significant errors in calculations of atmospheric oxidation capacity, and the concentrations and trends of short-lived climate forcers and pollutants, both historically and at present. Environmental signicanceIn this work, we used a state-of-the-art chemistry-climate model to evaluate the effect of short-lived halogens (SLH) on the combined abundance of global atmospheric oxidants in the pre-industrial (PI) and present-day (PD) atmospheres. The results show a signicant global reduction in the PD levels of hydroxyl (OH), nitrate (NO 3 ), and ozone (O 3 ) concentrations. The simulations also show that the role of halogens on atmospheric oxidation was more prominent in the pristine pre-industrial atmosphere where a quarter of OH and half of NO 3 concentrations in the boundary layer are accounted for by SLH. We conclude that atmospheric oxidation, particularly in the preindustrial atmosphere, cannot be fully understood without consideration of SLH emissions and chemistry.</div
A Multi-Agent Deep Reinforcement Learning Approach for Traffic Management in Complex Communication Networks
No full textInternational audienceModern communication networks like 5G and 6G are increasingly integrating Distributed Artificial Intelligence (DAI) to provide fast decision-making services like traffic management despite both the unpredictable patterns of network traffic and the intrinsic dynamism of the underlying communication network. In particular, Distributed Artificial Intelligence will enable optimal network resource usage and prevent network congestion, addressing the challenges posed by the dynamic patterns characterizing complex communication networks like 5G and 6G networks. This paper focuses on designing and assessing a new traffic management solution based on a Multi-Agent Deep Reinforcement Learning (MA-DRL). Our solution aims at adapting to network conditions to prevent network traffic congestion, while improving throughput, latency, and loss compared to existing traffic management methods
Weak particle presence
No full text30 pages, 5 figures. To be published in Foundations of PhysicsInternational audienceThe concept of presence has been extensively explored in philosophy, yet the notion of particle presence within quantum theory remains under-examined. In this article, we explore particle presence through an analysis of a paradox arising from weak measurements. We show that the classical intuition about particle presence involves an erroneous logical combination of propositions from single-time weak values, leading to inconsistencies that result in the deduction of discontinuous trajectories. Instead, we argue that by treating presence as a property defined across time by measuring sequential weak values, the discontinuity paradox is resolved, providing a coherent, non-classical account of particle presence. We discuss some advantages and drawbacks of this account, and consider applications to other cases of trajectory discontinuity
BEP: Hafnium Oxide Ferroelectric Films: from fundamental understanding to optimized low power device Integration
No full textNational audienceNeuromorphic systems are artificial neural systems inspired by the biological human brain, whose connections can be mimicked by artificial synapses between memory devices, forming neural networks (NN) for deep learning. Training a NN demands plasticity, which requires the possibility to analyze a large number of values stored in memory and allow changes in synaptic strength. Inference, in turn, demands long-term stability. Currently, no non-volatile, low-power memory technology offers both characteristics simultaneously.In the pursuit of a memory solution for simultaneous on-chip learning and inference, a hybrid FeRAM/OxRAM synapse circuit is proposed. Hafnium oxide (HfO2)-based ferroelectric memories (FeRAM) are used for learning, exploiting their high endurance, ultra-fast, low-power (10-50 fJ/bit), and non-volatile character. However, the destructive read operation of FeRAMs makes them unsuitable for inference. This is why HfO2-based resistive memories (OxRAM) are used for inference, as they offer long-term stability and non-destructive reading. A successful development and implementation of an OxRAM/FeRAM synapse on a single CMOS substrate requires a comprehensive characterization of the physico-chemical properties of the HfO2 layer to understand their impact on the performance of the device [1]. Manifest interest has been given to oxygen vacancies (VO) point defects, which appear to have a preeminent influence on the endurance. Here we illustrate the capabilities of hard x-ray photoelectron spectroscopy (HAXPES) for VO quantification and profiling in HfO2. By inserting an ultra-thin, metallic, oxygen scavenging layer it may be possible to engineer the required vacancy concentration and hence optimize the ferroelectric/resistive character
Paving the way for improved representation of coupled Physical and biogeochemical processes in Arctic River Plumes—A case study of the Mackenzie shelf
No full textInternational audienceProcesses affecting the transformation of riverine dissolved organic carbon (DOC) across the land-to-ocean aquatic continuum are still poorly constrained in Arctic models, leading to large uncertainties in simulated air-sea CO fluxes of the coastal periphery. Here we use the ECCO-Darwin regional configuration of the Southeastern Beaufort Sea to analyze the sensitivity of simulated carbon cycling to (1) the model vertical discretization and (2) different parameterizations of Mackenzie River carbon discharge. We show that riverine DOC lifetime rather than its volume largely modulates Mackenzie River plume air-sea CO 2 fluxes, leading to the Southeastern Beaufort Sea (SBS) being either a source (0.03 Tg C year ) or sink (-0.20 Tg C year ) of atmospheric carbon. We show that estuarine processes, such as flocculation, also play an important role and can dampen CO outgassing by up to 0.07 Tg C year . In terms of model physics, by increasing the vertical grid resolution, we better fit observed plume structure, without altering the simulated concentrations of DOC. However, the decrease in river forcing cell volume increases local pCO and promotes elevated outgassing in the vicinity of the Delta. Our work demonstrates that future Arctic land-ocean models must consider the intricate details of river plume systems to realistically simulate coastal-ocean physics and biogeochemistry
3D MC. I. X-Ray Tomography Begins to Unravel the 3D Structure of a Molecular Cloud in our Galaxy’s Center
No full textInternational audienceAstronomers have used observations of the Galactic gas and dust via infrared, microwave, and radio to study molecular clouds in extreme environments such as the Galactic center. More recently, X-ray telescopes have opened up a new wavelength range in which to study these molecular clouds. Previous flaring events from Sgr A* propagate X-rays outwards in all directions, and these X-rays interact with the surrounding molecular gas, illuminating different parts of the clouds over time. We use a combination of X-ray observations from Chandra and molecular gas tracers (line data from Herschel and the Submillimeter Array) to analyze specific features in the Sticks cloud, one of three clouds in the Three Little Pigs system in the Central Molecular Zone (Galactic longitude and latitude of 0.°106 and −0.°082 respectively). We also present a novel X-ray tomography method we used to create 3D map of the Sticks cloud. By combining X-ray and molecular tracer observations, we are able to learn more about the environment inside the Sticks cloud
Ultrasonic scattering in polycrystalline materials with elongated grains: A comparative 3D and 2D theoretical and numerical analysis
No full textInternational audienceIn this paper, a previously developed theoretical model for the ultrasonic elastic wave scattering, based on the Stanke and Kino model and applicable to both 2D and 3D single-phase untextured polycrystals, is extended to microstructures with elongated grains. The effect of elongated grains on wave attenuation and phase velocity induced by scattering is investigated, highlighting similarities and discrepancies between the 2D and 3D cases. Additionally, 2D and 3D finite element (FE) models are developed to compare and validate the theoretical predictions under fixed assumptions. The morphology of the numerical polycrystalline samples is characterized using a multi-exponential two-point correlation (TPC) function which, when incorporated with the theoretical model, enables a more direct and accurate comparison. The FE models demonstrate excellent quantitative agreement with the theoretical predictions and, moreover, support the wave propagation's directional dependency in the stochastic scattering region and the 2D-3D dimensionality dissimilarities in the Rayleigh region. It is shown that 2D attenuation can predict 3D behavior in the stochastic limit and provide insights into the estimation of 3D grain morphology in the Rayleigh limit
Nonreciprocal resonant surface acoustic wave absorption in YFeO
No full textInternational audienceFerrimagnet insulator yttrium iron garnet (YFeO , YIG) attracts much interest in magnonics and microwave applications due to its relatively low magnetic damping and long magnon relaxation time. The magnetoelastic interaction between YIG magnetization and surface acoustic waves (SAWs) has been the basis in the development of acoustic-surface-wave isolators. However, the investigation of SAW-driven magnetic resonance in YIG is limited by the low compatibility of YIG and SAW devices. In this work, we demonstrate nonreciprocal resonant SAW absorption in an on-chip YIG-SAW device integrated by the focused-ion-beam technique. We observed two distinct SAW absorption signals attributed to the perpendicular standing spin waves (PSSW). Furthermore, the large observed nonreciprocity is compared to theoretical predictions incorporating the influence of asymmetric surface conditions on PSSW mode profiles and nonreciprocal magnon-SAW coupling. The strongly nonreciprocal SAW attenuation in YIG offers a promising opportunity for highly efficient rf signal processing technologies
Bias dependent band alignment in Ga2O3 ferroelectric interface by operando HAXPES
No full textInternational audienceWe present a fundamental study of the band alignment at the interface of HfZrO4 (HZO) with Ge-doped Ga2O3. Ge is an alternative n-type dopant for the wide band gap Ga2O3 due to its shallow donor level and favorable MBE growth conditions. In the perspective of using the ferroelectric polarization of hafnia based oxides, we have used a stack of HZO on highly Ge doped Ga2O3, the latter providing high carrier density. Electrical contacts were ensured by a TiN top electrode deposited on the HZO and an Au pad on the Ge:Ga2O3. The band alignment was measured by carrying out hard X-ray photoelectron spectroscopy (HAXPES) with in situ bias application across the HZO and following the evolution of both HZO and Ga2O3 energy level. Complementary high-resolution transmission electron microscopy (HRTEM) provided structural confirmation of the polar orthorhombic phase however electrical characterization showed that charge injection and trapping at the interface prevents stabilizing the ferroelectric polarization in HZO. The band alignment in the presence of a leaky HZO layer is therefore dominated by the bias induced band skewing