HAL Portal IOGS (nstitut d'Optique Graduate School)
Not a member yet
12589 research outputs found
Sort by
Réalisation d’une expérience d’atomes froids dans une fibre à coeur creux : vers la cristallisation spontanée
No full textCold atoms are versatile quantum systems that enable the study of numerous quantum phenomena through precise tools and a high degree of control. Many experiments use these systems to explore spontaneous organization phenomena : systems that are initially disordered undergoing a phase transition to an ordered state due to interactions between their constituents. Most studies in this field rely on atoms confined in cavities. However, such cavities impose the light modes interacting with the atoms by their boundary conditions,there by predetermining the final geometry the system. Some experiments using ring cavities,which lack standing waves, offer an additional degree of freedom. In our experiment, we aim to go even further: the experiment we developed and describe in this thesis aims to induce the spontaneous crystallization of cold atoms in free space. Since the atoms are not in a cavity, they are considered as in free space, even if they are radially confined within the fiber’s core. The quasi 1D geometry increases the light-atom interaction, needed to implement the long range interatomic forces required to trigger the phase transition towards a crystal structure The cold atom crystal would be formed by long-range interactions induced by the light interacting with the atoms, simultaneously generating a light crystal with an associated superradiant pulse, that did not initially exist. Beyond a certain intensity threshold, these interactions would break the translational symmetry of the atomic cloud, with a spacing determined only by the characteristics of the cloud and the light, rather than by a standing wave.This type of experiment could contribute to the study of long-range interactions, the mechanisms of crystal formation, and could find applications in quantum simulations or magnetic field sensors. Finally, an important aspect is that the size of cold atom experiments generally remains too large to enable measurements outside vacuum chambers. In the case of atomic sensors,the measured forces often depend on the inverse of the distance from the observed fields. It is therefore crucial to work towards the miniaturization of cold atom experiments. Although other studies exploit HCFs, this experiment is unique in that it connects two vacuum chambersvia an HCF, with the goal of using cold atoms as magnetic field sensors between the chambers.This represents a step towards more compact devices for cold atom experiments. This project began with my doctoral research therefore, in this thesis, I present the development and construction of this new experiment, as well as the initial results concerning the cold atom source, the loading of atoms into the HCF, and a brief introduction to spontaneous organization phenomena.Les atomes froids sont des systèmes quantiques versatiles qui permettent l’étude de nombreux phénomènes quantiques grâce à des outils précis et un contrôle avec un haut degré de précision. De nombreuses expériences exploitent ces systèmes pour explorer des phénomènes d’organisation spontanée : des systèmes initialement désordonnés subissant une transition de phase vers un état ordonné sous l’effet des interactions entre leurs constituants. La plupart des études dans ce domaine utilisent des atomes confinés dans des cavités. Cependant, ces cavités imposent des restrictions sur les modes de lumière qui interagissent avec les atomes,prédéterminant ainsi la géométrie du système. Certaines expériences utilisant des cavités annulaires, dépourvues d’onde stationnaire, offrent un degré de liberté supplémentaire.Dans notre expérience, nous visons à aller encore plus loin : L’expérience qu’on a développé et décrite dans cette thèse a pour objectif de provoquer la cristallisation spontanée d’atomes froids en espace libre. En confinant les atomes dans une fibre à cœur creux (HCF), nous nous concentrons sur une seule dimension, considérant que les atomes sont en espace libre dans la direction longitudinale de la fibre. Le cristal d’atomes froids serait formé par des interactions de longue portée induites par la lumière avec laquelle ils, générant simultanément un cristal de lumière qui n’existait pas initialement. Au-delà d’un seuil d’intensité lumineuse, ces interactions biseraient la symétrie de translation du nuage atomique, avec un pas déterminé uniquement par les caractéristiques du nuage et de la lumière et non pas par une onde stationnaire.Ce type d’expérience pourrait contribuer à l’étude des interactions de longue portée, des mécanismes de formation des cristaux et pourrait trouver des applications dans les simulations quantiques ou les capteurs de champs magnétiques.Enfin, un aspect important est que la taille des expériences d’atomes froids reste généralement trop grande pour permettre des mesures en dehors des chambres à vide. Or, dans le cas de capteurs atomiques, les forces mesurées dépendent souvent de l’inverse de la distance par rapport aux champs observés. Il est donc crucial de tendre vers la miniaturisation des expériences d’atomes froids. Bien que d’autres travaux exploitent des HCF, cette expérience est unique en ce qu’elle connecte deux chambres à vide via une HCF, avec pour objectif l’utilisation d’atomes froids comme capteurs de champs magnétiques entre les chambres. Cela constitue une avancée vers des dispositifs plus compacts pour les expériences d’atomes froids.Ce projet a débuté avec mon doctorat. Dans ma thèse, je présente le développement et la construction de cette nouvelle expérience, ainsi que les premiers résultats concernant la source d’atomes froids, le chargement d’atomes dans la fibre à cœur creux, et une brève introduction aux phénomènes d’organisation spontanée
Benchmarking systématique des ordinateurs quantiques: statut et recommandations
No full textArchitectures for quantum computing can only be scaled up when they are accompanied by suitable benchmarking techniques. The document provides a comprehensive overview of the state and recommendations for systematic benchmarking of quantum computers. Benchmarking is crucial for assessing the performance of quantum computers, including the hardware, software, as well as algorithms and applications. The document highlights key aspects such as component-level, system-level, software-level, HPC-level, and application-level benchmarks. Component-level benchmarks focus on the performance of individual qubits and gates, while system-level benchmarks evaluate the entire quantum processor. Software-level benchmarks consider the compiler's efficiency and error mitigation techniques. HPC-level and cloud benchmarks address integration with classical systems and cloud platforms, respectively. Application-level benchmarks measure performance in real-world use cases. The document also discusses the importance of standardization to ensure reproducibility and comparability of benchmarks, and highlights ongoing efforts in the quantum computing community towards establishing these benchmarks. Recommendations for future steps emphasize the need for developing standardized evaluation routines and integrating benchmarks with broader quantum technology activities
Development of an all-fiber spliced laser system using stimulated Brillouin scattering mitigation techniques and achieving 50 ns, 10 kW peak power for lidar applications
No full textInternational audienceThis work compares several Yb-doped large mode area (LMA) fiber amplifiers operating in the ∼50 ns regime at 1030 nm designed for airborne wind lidar application. The sinusoidal phase modulation applied to the signal enhances the SBS threshold by a factor 8.8 compared to a single frequency operation while still complying with lidar receiver spectral requirements. By combining this modulation with short, highly doped LMA fibers, more than 10 kW output peak power has been reached for 50 ns pulses at 40 kHz repetition rate
Atomistic Modeling as a Key to Unraveling Ultra-Short Laser Nano-Synthesis and Nano-Treatment
No full textInternational audienceThe development of short and ultrashort laser pulses has revolutionized the field of micro- and nanostructuring, enabling precise nanoparticle synthesis, assembly, and fragmentation. These laser-induced material modifications have proven invaluable in diverse fields such as catalysis, energy storage, medicine, and microelectronics. Through atomistic simulations, including the two-temperature model and molecular dynamics, this study is focused on the main mechanisms of these interactions. Our findings reveal that depending on the laser parameters and material properties, lasers can induce a spectrum of effects—from surface melting and amorphization to nanostructure formation, as well as nanoparticle synthesis, sintering, and fragmentation. At lower fluences, subtle modifications such as thin-layer melting and amorphization are observed, which are highly sensitive to the material's crystal orientation. In contrast, higher fluences lead to critical phenomena, generating explosive target disintegration, creating nanoparticles, and even fragmenting them. These processes are of significant interest for nano-catalysis, energy storage, and sensor development. We show, furthermore, that ultrashort laser pulses offer a pathway to tailor surface properties such as wettability and electrical conductivity. Furthermore, such interactions, often exhibit unique dynamics and even anisotropic behavior, offering new insights for the fabrication of next-generation nano-catalytic materials and energy storage systems. The study emphasizes how atomistic modeling not only enhances our understanding of these complex phenomena but also guides the development of innovative applications in advanced technologies
Next-Generation Ultrafast Photoluminescence Spectroscopy: Integration of Transient Grating Optical Gate and Advanced Femtosecond Laser Technology
No full textInternational audienceWe demonstrate a high-performance ultrafast broadband timeresolved photoluminescence (TRPL) system based on the transient grating photoluminescence spectroscopy (TGPLS) technique. The core of the system is a Kerr effect-induced transient grating (TG) optical gate driven by high repetition rate ultrashort laser pulses at 1030 nm with micro-Joule pulse energy. Satisfying the demands of spectroscopy applications, the setup achieves high sensitivity, rapid data acquisition, ultrafast time resolution, and a wide spectral window from ultraviolet to near-infrared. The time resolution can be further improved to achieve <80 fs instrument response function by employing the multiple plate compression (MPC) technique to temporally compress the driving pulses. This work presents a new benchmark for ultrafast TRPL.</div
Twelve Years of LabEx PRIMES (2012-2024)
No full textInternational audienceAn overview of the research performed in the Auvergne Rhône Alpes Region within theLaboratoire d’Excellence Physics, Radiobiology, Medical Imaging, and Simulatio
High Precision Astrometry Science in the Context of Space Mission Prospectives
No full textInternational audienceAstrometry is one of the oldest branches of astronomy which measures the position, the proper motion and parallax of celestial objects. Following the Hipparcos and Gaia missions that have measured several billions of them using global astrometry, we propose to increase astrometry precision on pointed objects using differential astrometry in a large field in order to unravel rocky planets in habitable zones of stars in the Sun vicinity and investigate the nature of dark matter in galactic environments as recommended by the ESA Senior Committee in the Voyager 2050 prospective. Substantial technology developments in a number of critical areas is needed in order to reach the highest required precision of sub-micro-arcsecond. One of them is CMOS image sensors using the stitching technique to merge the multiple design structures on the wafer and produce array with very large number of pixels. Another one is to calibrate the pixel positions using projecting modulating interferometric laser fringes on the array. Finally, the distortion of the optical system can be monitored and compensated using reference stars as metrology sources. The final precision depends on the diameter and the field of view of the telescope that is used as well as the time spent on each target. We present here the science goals that can be achieved with such missions either within the framework of an ESA Medium-class mission or even in the NASA most challenging Habitable Worlds Observatory, a large space telescope recommended by the American Astronomy and Astrophysics prospective for the 2020s and designed specifically to search for signs of life on planets orbiting other stars
Tukey-Median of Means-Gradients for Langevin Dynamics: a Robust Approach to Fitting Machine Learning Models
No full textMedian-of-Means (MoM) estimators have emerged as powerful tools for robust learning when data are corrupted by outliers or exhibit heavy-tailed distributions. However, incorporating MoM principles to design robust loss functions typically renders the underlying optimization problem highly non-convex, posing significant challenges for the design of efficient, polynomial-time algorithms. In this work, we apply MoM principle at the gradient level using Tukey median to robustly estimate the uncorrupted gradient. More precisely, we propose a novel Tukey MoM-based variant of Stochastic Gradient Langevin Dynamics (TMoM-SGLD) which enjoys strong theoretical guarantees, including hitting-time bounds showing convergence to approximate stationary points of the underlying uncorrupted loss. Contrary to prior MoM estimators, our results hold in broad nonconvex settings, making TMoM-SGLD particularly relevant to modern machine learning tasks. Experiments on corrupted linear regression tasks demonstrate that, unlike prior approaches, TMoM-SGLD remains effective even under high levels of outlier contamination. Additionally, we further validate our approach on a deep learning task, showing its capabilities to robustly train neural networks
Traitement automatique de la langue dans le contexte de la réponse aux appels d’offres complexes
No full textThe tender analysis process, characterized by complex document evaluationsunder tight deadlines, is often resource-intensive and prone to errors. Keyphases of this process rely on Hierarchical Multi-label Text Classification(HMTC) tasks. The goal is to classify a given input document from the tendercorpora into labels derived from a taxonomic hierarchy.While discriminative methods have traditionally dominated HMTC literature,generative models have shown significant promise as alternatives. This thesisdelves deeper into generative methods to analyze their ability to learn both,the hierarchical information and the semantic relationships among labels.Notably, most generative methods perform well at preserving the class hier-archy when making predictions which diminishes concerns over hierarchicalintegrity. However, by design, these solutions set aside what made thestrength of discriminative methods, that is to make predictions based on theentire hierarchical information and to capture correlation from input textand labels for better classifications.We then, introduce a hybrid approach that leverages the strengths of bothmethodologies through a non-autoregressive decoder enhanced with a novelHierarchical Self-Attention mechanism. The decoder input is expanded withlabel embedding. Harnessing the power of the Cross-Attention and Hier-archical Self-Attention mechanisms, we achieve a label representation thatbenefits from instance and global label-wise information. This mechanismfacilitates simultaneous decoding of the entire hierarchy, utilizing global andbidirectional attention alongside labels. Through a set of experimental stud-ies on several large scale datasets, we demonstrate that this design deliverssuperior performance compared to 16 state-of-the-art HMTC models.L'analyse des appels d'offres, qui implique l'évaluation de documents complexes, constitue un processus exigeant en ressources et susceptible de générer des erreurs en raison de sa nature intrinsèquement laborieuse et des contraintes temporelles imposées. Deux phases fondamentales de ce processus reposent sur des tâches de classification hiérarchique multi-label du texte (CHMT), dont l'objectif est d'attribuer à chaque document du corpus d'appels d'offres des catégories spécifiques, organisées au sein d'une taxonomie hiérarchique prédéfinie.Historiquement, les méthodes discriminatives ont dominé la littérature. Toutefois, les modèles génératifs ont récemment émergé comme des alternatives prometteuses. Cette thèse explore en profondeur les méthodes génératives afin d'analyser leur capacité à apprendre à la fois les informations hiérarchiques et les relations sémantiques entre les classes.En particulier, la plupart des méthodes génératives se distinguent par leur aptitude à préserver l'intégrité de la hiérarchie lors des prédictions. Cependant, en raison de leur conception, ces approches ne parviennent pas à exploiter pleinement les informations hiérarchiques ni à capturer efficacement les corrélations fines entre le texte d'entrée et les catégories, limitant ainsi leur capacité à atteindre de meilleures performances en classification.Nous proposons ainsi une approche hybride qui combine les atouts des deux méthodologies, en intégrant un décodeur non auto-régressif, renforcé par un nouveau mécanisme d'attention hiérarchique. L'entrée du décodeur est pré-peuplée par des représentations vectorielles de tous les nœuds de la taxonomie, fournissant ainsi une base complète pour guider le processus de décodage. En exploitant les mécanismes d'attention croisée et d'attention hiérarchique, nous obtenons une représentation des classes qui fusionne à la fois des informations spécifiques aux instances de texte et des informations provenant de la structure globale de la taxonomie. Ces mécanismes permettent donc un décodage simultané de l'ensemble de la hiérarchie.À travers une série d'expérimentations sur plusieurs jeux de données, nous montrons que cette approche surpasse 16 modèles de l'état de l'art en CHMT