HAL Portal IOGS (nstitut d'Optique Graduate School)
Not a member yet
12589 research outputs found
Sort by
When to Learn: Conformal Scores as Online Update Criteria
No full textInternational audienceWe present Conformal Online Learning (COL), a novel framework designed for adaptive model updates in streaming environments. Inspired by conformal prediction but repurposed for learning, COL evaluates the consistency of models over time, rather than constructing prediction sets. At each step, the model's performance is compared against a calibrated conformity threshold derived from recent history. An update is triggered only if the model's error exceeds this threshold, ensuring that learning occurs only when necessary. This selective adaptation leads to a significant reduction in redundant updates while preserving long-term accuracy. Unlike classical online learning methods that continuously retrain or assume stationarity, COL offers a principled, data-driven mechanism for dynamic model refinement with minimal computational overhead. The framework is general and model-agnostic, making it suitable for a wide range of applications involving evolving data streams and non-stationary environments. We demonstrate the effectiveness of COL through its application to online learning of Koopman linear embeddings for non-linear dynamical systems.</div
Dose Rate Effects in Ag-Doped Metaphosphate Glass Radiophotoluminescent Dosimeters Up to MGy Range
No full textInternational audienceThis work presents a comprehensive study of dose rate effects on Ag-doped metaphosphate glass radiophotoluminescence (RPL) dosimeters for doses ranging between 1 Gy and 100 MGy. Data from multiple irradiation campaigns with diverse radiation fields, such as 60Co gamma radiation, X-rays with energy spectrum up to 100 keV, monoenergetic electron beams, and mixed fields offer valuable insights for accurate dosimetry in high-radiation environments. Dose rates ranging over eight decades, from 0.1 kGy/h to 18 MGy/h, are experimentally tested. Low dose rate effects can be observed at low doses; however, for doses higher than 100 kGy, both overestimation and underestimation can occur depending on the dose rate
Revealing nonequilibrium pathways in ultrafast laser-silica interaction via first-principles modeling
No full textInternational audienceThe advancement of micro-and nanoscale technologies increasingly depends on precise control over material structuring processes. Ultrafast laser irradiation has proven effective in inducing nanoscale modifications, provided that key irradiation parameters are carefully matched to the nonlinear response characteristics of the material [1]. A detailed understanding of the ultrafast processes underlying photoexcitation-induced band renormalization and energy deposition is essential for controlling laser-driven modifications. These electronic dynamics initiate rapid thermal and hydrodynamic relaxation pathways that mediate structural changes, including amorphization, polymorphic transitions, and the emergence of nanoscale features such as voids and gratings.This study focuses on the interaction of femtosecond laser pulses with wide-bandgap materials, specifically SiO2 in its amorphous (fused silica) and crystalline (α-quartz) forms. Using ab initio methods and multiphysics simulations, we investigate how sub-diffraction-limited energy localization and fast nonequilibrium dynamics can guide structural transformations.Under intense excitation, conduction-band electron densities rise sharply, inducing transient modifications of the electronic structure. These changes include notable bandgap renormalization and band distortion, which we analyze using Density Functional Theory (DFT) with GW corrections, Time-Dependent DFT, and molecular dynamics [2]. Simulations show that within tens of femtoseconds, the bandgap can shift by several electronvolts, accompanied by a reduction in bond strength and changes in thermal and optical properties [3,4].At later stages of the interaction, energy absorption mechanisms become more complex, contributing to the formation of nanoscale features. By modeling the coupled evolution of electronic excitation and lattice response, we explore pathways that lead to the confinement of absorbed energy, modification of bonding configurations, and rapid energy redistribution. Combined electromagnetic and atomistic simulations reveal the emergence of structural reorganizations driven by localized heating, bond destabilization, and plasma formation. These effects lead to the nucleation of nanopores, which can self-organize into polarization-aligned nanogratings [5]. These results contribute to a quantitative understanding of how ultrafast laser excitation governs structural transformations in dielectrics. The coupling between electronic excitation, structural relaxation, and material reconfiguration is essential for advancing laser processing techniques aimed at controlled nanoscale engineering.1. R. Stoian, and J.P. Colombier, « Advances in ultrafast laser structuring of materials at the nanoscale », Nanophotonics 9(16), 4665-4688 (2020).2. A. Tsaturyan, E. Kachan, R. Stoian, & J.P. Colombier, « Excited‐state dynamics and optical properties of silica Under ultrafast laser irradiation», Advanced Physics Research, 2400106 (2024).3. A. Tsaturyan, E. Kachan, R. Stoian, and J.P. Colombier, « Ultrafast bandgap narrowing and cohesion loss of photoexcited fused silica», The Journal of Chemical Physics 156 (22), 224301 (2022).4. A. Tsaturyan, E. Kachan, R. Stoian, and J.P. Colombier, Unraveling the electronic properties in SiO2 under ultrafast laser irradiation, NPJ Computational Materials, 10 (1), 1-10 (2024).5. A. Rudenko, J.P. Colombier, T. Itina, & R. Stoian, "Genesis of nanogratings in silica bulk via multipulse interplay of ultrafast photo-excitation and hydrodynamics", Advanced Optical Materials, 2100973 (2021).</p
Nonmonotonic Radiative Heat Transfer in the Transition from Far Field to Near Field
No full textInternational audienc
CXCL10-dependent epithelial-vascular cross-talk for endothelial activation following SARS-CoV-2 infection.
No full textInternational audienceThe blood vessel network is heavily impacted by SARS-CoV-2 infection. How SARS-CoV-2 contributes to vascular inflammation and whether epithelio-endothelial cross-talk is involved remain unclear. We investigated in detail the interaction between SARS-CoV-2 and the vasculature using 2D and 3D vesseloid in vitro models. We first assessed whether SARS-CoV-2 is able to directly infect endothelial cells. In the absence of ACE2 in endothelial cells, no productive infection was detected. Low uptake of viral particles by ACE2-overexpressing endothelial cells was observed without efficient viral production. Thus, the indirect effect of SARS-CoV-2 infection may involve epithelio-endothelial cell cross-talk. After infection of the epithelial cells, a significant inflammatory response was detected in the endothelial cells. CXCL10 was the most highly expressed proinflammatory cytokine involved in this intercellular communication, and its function was subsequently explored. Finally, the clinical relevance of our findings was confirmed in two patient cohorts
High-resolution, Time-resolved Terahertz Imaging based on the Fourier Synthetic Aperture technique
No full textInternational audienc
Unveiling how mitotic spindle orientation in 3D human colon organoids affects matrix displacements through a 4D study using DVC
No full textInternational audienceCell division is a major event in tissue homeostasis, enabling renewal and regeneration. In human colon, vertical division is mainly observed in the stem cell compartment while horizontal division is more frequent in the progenitor transit amplifying zone. To study cell division, the human colon epithelium represents a relevant model due to its rapid renewal and high number of mitoses. Studying live mechanical interactions between the epithelium and its matrix in vivo is challenging due to the lack of suitable methods. Colon organoids seeded in Matrigel are good models because they recapitulate the organization and properties of tissue architecture. This culture set-up allows to study the displacements of the matrix around the organoid. We studied the impact of cell division within the human colonic epithelium on the extracellular matrix. We validated an original experimental and analytical process with 3D time-lapse confocal microscopy to follow cell division and matrix displacements, on which we performed a 4D Digital Volume Correlation. Depending on the orientation of the mitotic spindle, cell division affects the matrix differently. Vertical division causes a predominantly uniaxial displacement of the matrix, while horizontal division involves a multiaxial and wider displacement
Electroluminescence and energy transfer mediated by hyperbolic polaritons
No full textData are publicly available on Zenodo at 10.5281/zenodo.10625437https://zenodo.org/records/14382617International audienceUnder high electrical current, some materials can emit electromagnetic radiation beyond incandescence. This phenomenon, referred to as electroluminescence, leads to the efficient emission of visible photons and is the basis of domestic lighting devices (for example, light-emitting diodes)1,2. In principle, electroluminescence can lead to mid-infrared emission of confined light–matter excitations called phonon polaritons3,4, resulting from the coupling of photons with crystal lattice vibrations (optical phonons). In particular, phonon polaritons arising in the van der Waals crystal hexagonal boron nitride (hBN) present hyperbolic dispersion, which enhances light–matter coupling5,6. For this reason, electroluminescence of hyperbolic phonon polaritons (HPhPs) has been proposed as an explanation for the peculiar radiative energy transfer within hBN-encapsulated graphene transistors7,8. However, as HPhPs are locally confined, they are inaccessible in the far field, and as such, any hint of electroluminescence has been based on indirect electronic signatures and has yet to be confirmed by direct observation. Here we demonstrate far-field mid-infrared (wavelength approximately 6.5 μm) electroluminescence of HPhPs excited by strongly biased high-mobility graphene within a van der Waals heterostructure, and we quantify the associated radiative energy transfer through the material. The presence of HPhPs is revealed by far-field mid-infrared spectroscopy owing to their elastic scattering at discontinuities in the heterostructure. The resulting radiative flux is quantified by mid-infrared pyrometry of the substrate receiving the energy. This radiative energy transfer is also shown to be reduced in hBN with nanoscale inhomogeneities, demonstrating the central role of the electromagnetic environment in this process
Knowledge Amalgamation for Single-Shot Context-Aware Emotion Recognition
No full textInternational audienceFine-grained emotion recognition using the whole context inside images is a challenging task. Usually, the approaches to solve this problem analyze the scene from different aspects, for example people, place, object or interactions, and make a final prediction that takes all this information into account. Despite giving promising results, this requires specialized pre-trained models, and multiple pre-processing steps, which inevitably results in long and complex frameworks. To obtain a more practicable solution that would work in real time scenario with limited resources, we propose a method inspired by the amalgamation process to incorporate specialized knowledge from multiple teachers inside a student composed of a single architecture. Moreover, the student is not only capable of treating all subjects simultaneously by creating emotion maps, but also to detect the subjects in a bottom-up manner. We also compare our approach with the traditional method of fine-tuning pre-trained models, and show its superiority on two databases used in the context-aware emotion recognition field
Dynamic spatial beam shaping for ultrafast laser processing: a review
No full textInternational audienceThis review examines the state-of-the-art in spatial manipulation of ultrafast laser processing using dynamic light modulators, with a particular focus on liquid crystal-based systems. We discuss phase modulation strategies and highlight the current limitations and challenges in surface and bulk processing. Specifically, we emphasize the delicate balance between high-fidelity beam shaping and energy efficiency, both critical for surface and bulk processing applications. Given the inherent physical limitations of spatial light modulators such as spatial resolution, fill factor, and phase modulation range. We explore techniques developed to bridge the gap between desired intensity distributions and actual experimental beam profiles. We present various laser light modulation technologies and the main algorithmic strategies for obtaining modulation patterns. The paper includes application examples across a wide range of fields, from surgery to surface structuring, cutting, bulk photo-inscription of optical functions, and additive manufacturing, highlighting the significant enhancements in processing speed and precision due to spatial beam shaping. The diverse applications and the technological limitations underscore the need for adapted modulation pattern calculation methods. We discuss several advancements addressing these challenges, involving both experimental and algorithmic developments, including the recent incorporation of artificial intelligence. Additionally, we cover recent progress in phase and pulse front control based on spatial modulators, which introduces an extra control parameter for light excitation with high potential for achieving more controlled processing outcomes.</div