HAL-INSU
Not a member yet
203706 research outputs found
Sort by
Geometric deep learning for galaxy-halo connection: a case study for galaxy intrinsic alignments
No full textInternational audienceForthcoming cosmological imaging surveys, such as the Rubin Observatory LSST, require large-scale simulations encompassing realistic galaxy populations for a variety of scientific applications. Of particular concern is the phenomenon of intrinsic alignments (IA), whereby galaxies orient themselves towards overdensities, potentially introducing significant systematic biases in weak gravitational lensing analyses if they are not properly modelled. Due to computational constraints, simulating the intricate details of galaxy formation and evolution relevant to IA across vast volumes is impractical. As an alternative, we propose a Deep Generative Model trained on the IllustrisTNG-100 simulation to sample 3D galaxy shapes and orientations along with correlated scalar features, conditioned on the tidal fields and halo mass. The architecture consists of a SO(3) diffusion generative model, implemented with E(3) equivariant Graph Neural Networks that explicitly respect the Euclidean symmetries of our Universe. The generated and the true values for geometric quantities using two-point statistics are statistically consistent; e.g. Wasserstein-1 distances indicate per cent-level (or better) agreement in the 1D distributions of scalar quantities. Notably, our model demonstrates the ability to jointly model Euclidean-valued scalars (galaxy sizes, shapes, and colours) along with non-Euclidean valued SO(3) quantities (galaxy orientations) that are governed by highly complex galactic physics at non-linear scales
Soil and tree stem xylem water isotope data from two pan-European sampling campaigns
No full textInternational audienceThe stable isotope ratios of hydrogen (δ2H) and oxygen (δ18O) are useful for studying ecohydrological dynamics in forests. However, most isotope-based eco-hydrological studies are limited to single sites, resulting in a lack of large-scale isotope data for understanding tree water uptake. Here, we provide a first systematic isotope dataset for soil and stem xylem water collected during two pan-European sampling campaigns at 40 beech (Fagus sylvatica), spruce (Picea abies), or mixed beech-spruce forest sites in spring and summer 2023 (https://doi.org/10.16904/envidat.542, Lehmann et al., 2024). The dataset is complemented by additional site-, soil-, and tree-specific metadata. The samples and metadata were collected by different researchers across Europe following a standardized protocol. Soil samples were taken at up to 5 depths (ranging from 0 to 90 cm) and stem xylem samples from the trunks of three beech and/or spruce trees per site. All samples were sent to a single laboratory, where all analytical work was conducted. Water was extracted using cryogenic vacuum distillation and analyzed with an isotope laser spectrometer. Additionally, a subset of the samples was analyzed with an isotope ratio mass spectrometer. Data quality checks revealed a high mean total extraction efficiency, mean water amount (>1 mL), accuracy, and precision. The isotopic signature of soil and stem xylem water varied as a function of the geographic origin and changed from spring to summer across all sites. While δ2H and δ18O were strongly correlated, the soil water data plotted closer to the Global Meteoric Water Line (GMWL) than the stem xylem water. Specifically, the δ2H values of the xylem water were more enriched than those of the soil water, leading to a systematic deviation from the GMWL. Isotopic enrichment of the stem xylem water at mixed forest sites was larger for spruce trees than for beech trees. This dataset is particularly useful for large-scale studies on plant water use, ecohydrological model testing, and isotope mapping across Europe
JWST-MIRI Observations of the Irradiated Chemistry in the Inner Disk Cavity of GM Aur
No full textInternational audienceAbstract We present a spectroscopic analysis of the GM Aur disk using JWST MIRI-MRS, as part of the JWST Disk Infrared Spectral Chemistry Survey (JDISCS). The 1D spectrum exhibits faint dust continuum emission and is relatively poor in molecular gas compared to protoplanetary disks with no inner dust cavities. We identify fundamental CO emission, bright rotationally excited OH lines longward of 9 μ m, weak rotational H 2 O emission, the methyl cation CH 3 + , HCO + , and tentatively identify CO 2 . Otherwise, the spectrum is dominated by H 2 and atomic emission. We model the molecular spectra using slab models and retrieve excitation temperatures, column densities, and emitting areas in a Bayesian framework. The OH line fluxes are used to estimate the photodissociation rate of water. We find that 5.4 × 10 40 H 2 O molecules are photodestroyed every second, which suggests that the highest-energy OH lines trace a region near the 0.2 au dust cavity wall with an incident far-ultraviolet flux of G 0 ∼ 10 7 . The OH emission is compared to DALI models, which point toward a high incident flux of Ly α photons in the OH emitting region. We further use the H I lines to provide an independent estimate of the accretion luminosity and mass accretion rate, in agreement with prior measurements. A comparison with other cavity disks indicates that more evolved systems such as GM Aur may be less volatile rich and are characterized by brighter OH prompt emission and abundant molecular ions
The role of turbulence in setting the phase of the ISM and implications for the star formation rate
No full textInternational audienceWhat regulates star formation in different regions of the Galaxy is still debated and especially the role of turbulence is not fully understood. In this work, we explore the link between star formation, turbulence and the thermal state of the multi-phase interstellar medium (ISM). We analyse a suite of stratified box simulations modelling a realistic ISM that aims to probe environments similar to those found in the Milky Way. Turbulence is injected through stellar feedback and an external large-scale driving force. We find that star formation can be either boosted or reduced when increasing the external driving strength, depending on the environment. When the density is sufficiently high or the initial UV background weak, warm neutral gas naturally transitions to the cold phase, leading to high cold neutral medium (CNM) fractions of around 30 -40%. Under these conditions, excessive large-scale driving leads to a slight reduction of the CNM fraction and an increase in the amount of gas that is thermally unstable. What limits the star formation in this regime is a reduced fraction of dense gas due to additional turbulent support against collapse. For low density regions subject to significant external UV background, overdensities in which cooling is efficient are much rarer and we find that star formation is regulated by the formation of cold gas. In such cases, turbulence can significantly boost star formation by compressing gas in shocks and increasing the CNM fraction dramatically. In our simulations we see an increase from almost no CNM to up to a fraction of 15 % when including external turbulence driving; leading to an associated increase in the star formation rate. We provide a model to quantify this behaviour and predict the CNM fraction by combining the standard ISM cooling/heating model with the density PDF generated by turbulence. The change in the dominant limiting process for star formation between low-density /externally heated and intermediate-density /feedback heated environments could provides a natural explanation for the observed break in the Kennicutt-Schmidt relation around column densities of 9 M ⊙ pc -2
Typeset using
No full textCold gaseous systems play important roles in galaxy evolution by possibly providing fuel to ignite active galactic nuclei (AGN) activity and star-formation. In this work, we analyze [C ii]158µm and continuum observations from ALMA for a sample of four radio AGN at z ≈ 3.5, focusing on eight associated companion cloud systems discovered within projected distances of tens of kiloparsecs or less. The spatial distribution of these companions indicates that the majority of cold gas is not located at the AGN position, i.e., not in their host galaxies. With the assistance of [C ii] at 0.2 ′′ resolution, we further confirm the gas-poor nature of the hosts by re-analyzing archival, is consistent with sources discussed in the literature. Our findings show the gaspoor radio AGN hosts have nearby gas-rich companions. We propose that these companions may be stripped clouds resulting from merger processes, which could be a trigger of radio-loud AGN. They may also be a signature of negative AGN feedback (e.g., shock heating) on these infalling companions and on the host galaxy. In general, our analysis shows that powerful AGN at and before Cosmic Noon are impacting and being impacted by cold gaseous clouds in their circumgalactic or protointracluster media.</div
Learning to generate physical ocean states: Towards hybrid climate modeling
No full textOcean General Circulation Models require extensive computational resources to reach equilibrium states, while deep learning emulators, despite offering fast predictions, lack the physical interpretability and long-term stability necessary for climate scientists to understand climate sensitivity (to greenhouse gas emissions) and mechanisms of abrupt % variability such as tipping points. We propose to take the best from both worlds by leveraging deep generative models to produce physically consistent oceanic states that can serve as initial conditions for climate projections. We assess the viability of this hybrid approach through both physical metrics and numerical experiments, and highlight the benefits of enforcing physical constraints during generation. Although we train here on ocean variables from idealized numerical simulations, we claim that this hybrid approach, combining the computational efficiency of deep learning with the physical accuracy of numerical models, can effectively reduce the computational burden of running climate models to equilibrium, and reduce uncertainties in climate projections by minimizing drifts in baseline simulations
What environmental and human factors influence the decision of a beachgoer to enter the water at a high-energy beach? Application to South Western France
No full textInternational audienceBackgroundCoastal areas are among the most attractive destinations worldwide, but engaging in water-based recreational activities is not without risk. The overall bathing risk ultimately results from the combination of natural physical hazards (e.g. rip currents, shore break waves) and the individuals who expose themselves to them. Among the growing body of beach safety studies, many have identified the lack of exposure data as a severe limitation (1). A first attempt to address this was made by considering the beachgoer population rather than the total population to assess incident rates (2). We believe our research takes a step further by estimating the proportion of beachgoers who enter the water on a given day.MethodsWe built a unique multidisciplinary database combining data collected by an on-site beachgoers survey, weather stations, marine buoys and tidal reconstruction. We employed a logistic regression analysis to predict beachgoer’s decision to enter the water on any given day at a high-energy recreational beach.ResultsWe demonstrated that both environmental and human factors influence a beachgoer’s decision to enter the water. Daily mean wave height and daily mean insolation duration were significant predictors at the p<0.001 level, while age, place of residence and self-confidence in swimming out of a rip current were significant at the p<0.05 level or higher. Our model has an accuracy, F-Score, precision and recall of 71%, 73%, 86%, 79% respectively.ConclusionBeachgoer exposure on any given day can ultimately be predicted by coupling our model with beach attendance models (3). This would allow for the design of rescue and preventive operations on days with high expected exposure. While models based solely on environmental factors can be used to forecast beach risks, incorporating human factors into the model provides valuable insight for crafting prevention messages. To this end, forecasting tools must be based on behavioural analytical framework as much as possibl
Cosmological simulations of the same spiral galaxy: satellite properties, the role of baryonic physics and star formation history in shaping dark matter cores/cusps
No full textWe investigate the role of baryonic physics in shaping the population, structure, and internal dynamics of galactic subhalos using the Mochima suite of cosmological zoom-in simulations. A refined method is developed to identify bound subhalo material by isolating the local gravitational potential and applying multi-criteria phase-space selection. This approach enables a robust characterisation of subhalo properties across five baryonic runs with varying prescriptions for star formation, and supernova and protostellar feedback, as well as a dark matter-only baseline. At the population level, we find that host halo concentration, modulated by baryonic feedback, is a key predictor of subhalo survival. Subhalos with more massive stellar components exhibit deeper internal potentials and enhanced resilience to tidal disruption. At the structural level, we identify a broad diversity in inner dark matter profiles, consistent with observations of dwarf galaxies. We show that this diversity correlates with both star formation history and environmental interaction. In particular, galaxies that form most of their stars early tend to retain steep cusps, while those with extended or recent star formation exhibit oscillating inner slopes shaped by bursty feedback and tidal perturbations. These findings suggest that the so-called "diversity problem" may reflect the complex interplay between feedback history and gravitational environment, rather than a breakdown of cold dark matter predictions.</div
