Archives ouvertes Hal IMT Atlantique
Not a member yet
20414 research outputs found
Sort by
Modules Front-End RF en LTCC pour télécommunications et magnétométrie en bande W
No full textThis PhD thesis deals with the design and implementation of RF front-end modules in Low Temperature Co-Fired Ceramic (LTCC) technology for W-band (75–110 GHz) applications in telecommunications and nitrogen–vacancy (NV) centre magnetometry. After a review of high-frequency packaging solutions, an MCM-LTCC approach combined with wire bonding is selected as a trade-off between compactness, low RF losses, mechanical robustness and the ability to integrate passive functions. The first part focuses on the development of a planar LTCC resonator dedicated to the microwave excitation of an NV-based diamond magnetometer operated under a 3 T static magnetic field. The structure, based on a coplanar waveguide with a circularly loaded slot, is optimized to generate at the diamond position a microwave magnetic field of about 0.5 kA/m over a few-GHz bandwidth, with a homogeneous field region compatible with the crystal size. These performances are validated through S-parameter measurements and near-field mapping. Building on this resonator, a W-band LTCC front-end module is designed. It integrates two commercial MMICs (a ×8 frequency multiplier and a power amplifier) assembled on the LTCC substrate and interconnected to the resonator, so as to bring the microwave power as close as possible to the diamond and remove the need for a long, lossy W-band cable. Finally, a cylindrical dielectric resonator antenna (CDRA) integrated in LTCC is investigated for telecommunication applications around 100 GHz. Several slot-fed configurations are simulated and realized. Measured S-parameters and radiation patterns confirm the feasibility of monolithic LTCC CDRAs in the W-band. These results demonstrate the suitability of LTCC technology for compact, high-performance W-band RF front-end modules for both quantum metrology and millimetre-wave communication systems.Cette thèse porte sur la conception et la réalisation de modules frontend RF en technologie LTCC (Low Temperature Co-Fired Ceramic) pour la bande W (75–110 GHz), dédiés à des applications de télécommunications et de magnétométrie à centres azote–vacance (NV) dans le diamant. Après un état de l’art des technologies de packaging millimétriques, la solution MCM-LTCC associée au micro-câblage est retenue comme compromis entre compacité, faibles pertes RF, robustesse et capacité d’intégration de fonctions passives. Une première partie est consacrée au développement d’un résonateur plan en LTCC destiné à l’excitation micro-ondes d’un magnétomètre NV soumis à un champ statique de 3 T. La structure, de type ligne CPW associée à une fente chargée circulaire, est optimisée afin de générer au niveau du diamant un champ magnétique micro-ondes d’environ 0,5 kA/m sur une bande utile de quelques GHz, avec une zone de champ homogène adaptée à la taille du cristal. Ces performances sont validées par des mesures de paramètres S et des cartographies de champ proche. Sur cette base, un module frontend W-band en LTCC est conçu. Il intègre deux MMIC commerciaux (multiplicateur de fréquence ×8 et amplificateur de puissance) montés en surface et interconnectés à la structure résonnante, afin d’amener la puissance micro-ondes au plus près du diamant et de supprimer l’usage d’un câble W-band long et dissipatif. Enfin, une antenne à résonateur diélectrique cylindrique (CDRA) intégrée en LTCC est étudiée pour des applications de télécommunications autour de 100 GHz. Différents schémas d’alimentation par fente sont simulés et réalisés. Les mesures de paramètres S et de rayonnement confirment la faisabilité d’antennes CDRA monolithiques en LTCC en bande W. L’ensemble de ces travaux met en évidence le potentiel de la technologie LTCC pour la réalisation de modules frontends RF compacts et performants en bande W, pour la métrologie quantique et les systèmes de communication millimétriques
Initial Solution Sampling for the Variable Neighborhood Search Method : Fitness Landscapes Analysis
No full textInternational audienceInitial Solution Sampling for the Variable Neighborhood Search Method : Fitness Landscapes Analysi
Measurement of -hadron correlations relative to the event plane in semicentral Pb-Pb collisions at TeV
No full textInternational audiencePer-trigger yields of -hadron correlations were measured in semicentral Pb-Pb collisions at TeV in ALICE at the LHC. The reconstructed , with a transverse momentum of GeV/, no significant event-plane dependence is observed within uncertainties. The results are compared with predictions from the JEWEL model, which implements jet energy loss in an expanding medium. JEWEL predicts no significant modification of either the near- or away-side associated yields, independent of whether medium recoils are included. The observed behavior may indicate the presence of additional energy-loss mechanisms beyond those governed by path-length dependence
Quantitative analysis of radium-226 at the microscale by NanoSIMS
No full textInternational audienceThis study presents an analytical protocol for the quantitative imaging of 226Ra by Nanoscale Secondary Ion Mass Spectrometry (NanoSIMS), based on the synthesis and characterization of barite reference materials enriched in 226Ra (2.89 and 9.65 µg g−1), 232Th (2120 µg g−1), and Pb (85, 755 and 8080 µg g−1). This methodological development addresses key analytical challenges related to the low abundance of 226Ra and spectral interferences. We identified 88Sr138Ba+ and 208Pb18O+ as potential polyatomic interferences for 226Ra+. However, we demonstrated that these species do not significantly impact the 226Ra+ signal when appropriate mass resolution and precise peak alignment are applied. Thanks to the use of homemade 226Ra–barite reference materials, we demonstrated that NanoSIMS measurements can achieve reproducibility ranging from 6% to 12% (2σ) across analytical sessions, depending on the Ra content and the size of the analyzed area. Comparison with known concentrations confirmed the accuracy of the 226Ra+/138Ba+ ratio as a quantitative proxy for radium barite content. A conservative quantification limit achieved for 226Ra is 0.135 µg g−1, making this approach applicable to samples from the natural environment. These results demonstrate that NanoSIMS provides accurate and reproducible quantitative imaging of 226Ra at the micrometric scale using matrix-matched internal reference materials. This method opens up new opportunities for studying 226Ra distribution in samples from the natural environment at the mineral scale
Tight MDD Integration for Global Constraints: the Multi-Cumulatives Case
No full textIn problems with objects that have multiple feasible configurations, also known as multi-mode problems, a lack of communication often arises between global constraints and side constraints. To address this issue, we propose a novel approach where object configurations are represented using multi-valued decision diagrams (MDDs). Within this framework, global constraint filtering algorithms can directly query the MDDs to obtain a more accurate estimate of the bounds of a variable. We formally introduced this new method and a direct application is proposed through the Cumulative constraint. We propose two new constraints, named MDD-Cumulative and MDD-Multi-Cumulatives. The former is based on the state-of-the-art timetabling algorithm, while the latter takes advantage of the method for filtering mode assignments. We provide numerical evidence that demonstrates the effectiveness of these methods over the state-of-the-art multi-mode resource-constrained project scheduling problem (MM-RCPSP) benchmarks
Is your AI model secure? A study of data poisoning attacks on AI models
No full textInternational audienceAs Artificial Intelligence (AI) systems have become increasingly integral to applications spanning healthcare, finance, autonomous vehicles, and critical infrastructure, securing models against adversarial threats has become essential. Among these, data poisoning—where adversaries tamper with training data to corrupt or subvert model behavior—poses a distinctive risk by targeting the learning phase rather than inference, degrading accuracy, manipulating predictions, or implanting future backdoors. Given the safety?critical nature of many deployments, understanding and mitigating poisoning risks is paramount for reliable Machine Learning (ML). This paper surveys data poisoning attacks across supervised, unsupervised, and semi?supervised learning, alongside corresponding defenses, and consolidates the literature into a taxonomy to aid practitioners. In addition, we propose a model?agnostic audit methodology for assessing robustness to data poisoning that operationalizes threat modeling, clean baselining, a representative attack suite, robustness metrics with confidence intervals, detectability screening, defense evaluation, and governance?oriented reporting. Together, the taxonomy and audit methodology provide a practical foundation for auditing AI models and selecting defenses tailored to specific use cases, while highlighting open gaps, particularly the limited effectiveness of current defenses and the field’s focus on neural classifiers at the expense of other model families, that motivate the development of more robust, broadly applicable protections
All-sky search for continuous gravitational-wave signals from unknown neutron stars in binary systems in the first part of the fourth LIGO-Virgo-KAGRA observing run
No full textInternational audienceWe present the results of a blind all-sky search for continuous gravitational-wave signals from neutron stars in binary systems using data from the first part of the fourth observing run (O4a) using LIGO detectors data. Rapidly rotating, non-axisymmetric neutron stars are expected to emit continuous gravitational waves, whose detection would significantly improve our understanding of the galactic neutron star population and matter under extreme conditions, while also providing valuable tests of general relativity. Neutron stars in binary systems likely constitute a substantial fraction of the unobserved galactic population and, due to potential mass accretion, may emit stronger gravitational-wave signals than their isolated counterparts. This search targets signals from neutron stars with frequencies in the 100-350 Hz range, with orbital periods between 7 and 15 days and projected semi-major axes between 5 and 15 light-seconds. The analysis employs the GPU-accelerated fasttracks pipeline. No credible astrophysical signals were identified, and, in the absence of a detection, we report search sensitivity estimates on the population of neutron stars in binary systems in the Milky Way
One- and three-dimensional identical charged-kaon femtoscopic correlations in Pb--Pb collisions at TeV
No full textInternational audienceThe identical charged-kaon correlations induced by quantum-statistics effects and final-state interactions are measured in PbPb collisions at TeV. The results of one- (1D) and three-dimensional (3D) analyses show that the obtained system-size parameters (radii) are smaller for more peripheral collisions and decrease with increasing pair transverse momentum . The 1D parameters agree within uncertainties with those obtained in PbPb collisions at TeV. The observed power-law dependence of the extracted 3D radii as a function of the pair transverse momentum is a signature of the collective flow in the particle-emitting system created in PbPb collisions. This dependence is well reproduced by the integrated hydrokinetic model calculations except for the outward projection of the radius (measured in the longitudinally co-moving system) for the most central collisions. The time of maximal emission for kaons is extracted from the 3D analysis in a wide collision centrality range from 0 to 90%. Its reduction with decreasing charged-particle multiplicity is well reproduced by the hydrokinetic model predictions, and means that kaons are emitted earlier in more peripheral events
Discovering Data Manifold Geometry via Non-Contracting Flows
No full textWe introduce an unsupervised approach for constructing a global reference system by learning, in the ambient space, vector fields that span the tangent spaces of an unknown data manifold. In contrast to isometric objectives, which implicitly assume manifold flatness, our method learns tangent vector fields whose flows transport all samples to a common, learnable reference point. The resulting arc-lengths along these flows define interpretable intrinsic coordinates tied to a shared global frame. To prevent degenerate collapse, we enforce a non-shrinking constraint and derive a scalable, integration-free objective inspired by flow matching. Within our theoretical framework, we prove that minimizing the proposed objective recovers a global coordinate chart when one exists. Empirically, we obtain correct tangent alignment and coherent global coordinate structure on synthetic manifolds. We also demonstrate the scalability of our method on CIFAR-10, where the learned coordinates achieve competitive downstream classification performance
Light Dark Matter Search with 7.8 Tonne-Year of Ionization-Only Data in XENONnT
No full textInternational audienceWe report on a blinded search for dark matter (DM) using ionization-only (S2-only) signals in XENONnT with a total exposure of over 579 days in three science runs. Dedicated background suppression techniques and the first complete S2-only background model in XENONnT provide sensitivity to nuclear recoils of [0.5, 5.0] and electronic recoils of [0.04, 0.7] . No significant excess over the expected background is observed, and we set 90% confidence level upper limits on spin-independent DM--nucleon and spin-dependent DM--neutron scattering for DM masses between 3 and 8 , as well as on DM--electron scattering, axion-like particles, and dark photons, improving on previous constraints. For spin-independent DM--nucleon scattering, we exclude cross sections above cm at a DM mass of 5 , pushing the XENONnT sensitivity closer to the region where coherent elastic neutrino-nucleus scattering () becomes an irreducible background