The federation
Intranet
Projects of the Sherpas team
The Sherpas team’s activities cover a wide range of physical phenomena: high-energy radiation, particle physics, plasma physics, turbulence and astrocladistics.
Observation of high-energy emission
Active Galaxy Nuclei, Galactic Center, X-ray binaries, Diffuse Emission
Motivations
Observations of the high-energy sky reveal many compact objects, Galactic and extragalactic black holes, neutron stars and white dwarfs, and allow us to study the radiation processes associated with these objects and their extreme environment. Observational diagnostics (morphology, variability, flux and spectrum) reveal the physical phenomena at work in these objects and are essential to test the theoretical predictions developed in our team.
Objectives
Although compact objects have been studied for several decades, many questions remain unanswered:
- What is the geometry of the central regions close to the compact object? What are their dynamics? What do we learn from X-ray polarisation?
- What are the radiative processes at the origin of the observed high-energy emission? What is the origin of its variability?
- What can we deduce from the signatures of winds and jets on the ejection processes ? on the accretion-ejection links ?
- What are the observational differences/ similarities between active galactic nuclei and X-ray binaries?
- What can we learn from the populations of detected X-ray and Gamma-ray sources?
Methods
We access the physical conditions close to the compact object through multi-wavelength observations, covering the whole electromagnetic spectrum from radio to gamma rays. Our group is more specifically involved, as PI and co-I, in high energy observations with high energy satellites like XMM-Newton, Chandra, NuSTAR, NICER, Fermi, INTEGRAL or IXPE.
Contacts
Pierre-Olivier Petrucci, Maïca Clavel, Guillaume Dubus, Gilles Henri, Rafan Muhammad
Kinetic processes and particle acceleration
Pulsars, pulsar winds and pulsar wind nebulae. Stellar-mass black holes and supermassive black holes. Relativistic jets, blazars. Gamma-ray bursts. Nova-like explosions.
Overview
Compact objects (white dwarfs, neutron stars, black holes) are involved in some of the most extreme particle acceleration phenomena in the Universe. Particle spectra escaping from these cosmic accelerators are broad steep power laws extending to tremendous energies, many orders of magnitude above the particles rest mass energies (i.e., they are ultra-relativistic). There are currently two main acceleration mechanisms under scrutiny by the community: diffusive shock acceleration and magnetic reconnection.
Objectives
The work of the team in this field tries to answer the following fundamental questions:
- How are particles accelerated within relativistic magnetic reconnection sites?
- How and where are particle accelerated within neutron star and black hole magnetospheres?
- How are particle accelerated at relativistic, magnetized collisionless shocks?
- What are the observable signatures?
Methods
Studying particle acceleration requires a complete description of plasma phenomena at the kinetic scale, i.e., below the particle Larmor radius scale. The particle-in-cell (PIC) technique is a numerical method well-suited for this purpose: it solves the motion of a large collection of charged particles along with the evolution of the electromagnetic fields and their coupling between them. The team uses and develops the Zeltron code, a state-of-the-art PIC code optimized for studying astrophysical relativistic plasmas.
The ERC-SPAWN project
The ERC-SPAWN project (Simulating Particle Acceleration WithiN black hole magnetospheres, 2020-2026) aims at producing an ab-initio model of black hole magnetospheres. This work is essential to interpret current and future horizon-scale observations by Gravity@VLTI and the Event Horizon Telescope of nearby supermassive black holes.
Contacts
Benoît Cerutti, Guillaume Dubus, John Mehlhaff, Enzo Figueiredo, Valentina Richard Romei, Adrien Soudais
Modelling accretion-ejection phenomena
Young Stellar Objects, Active Galactic Nuclei, Microquasars and X-ray Binaries
Motivations
The team’s work on accretion-ejection phenomena seeks to model and interpret the emission of astrophysical objects with accretion disks. We now know that accretion phenomena are almost systematically associated with ejection, in the form of jets or winds. A specificity of the Sherpas team is to focus on the interaction between the large-scale magnetic field and the accretion disc to explain this link, using both phenomenological models, observations and high-performance numerical simulations.
Objectives
The work in this area seeks to answer the following questions:
- By what mechanism do these disks accrete (hydro/MHD turbulence, ejection...)?
- What are the links between ejection and accretion phenomena?
- How does the magnetic field evolve in these objects and what would be the associated signatures?
- What are the links between these accretion-ejection processes and the radiative processes occurring in the environment of compact objects (X-binaries, AGNs, cataclysmic variables, FuOr)?
Methods
For the modelling part, we use both semi-analytical solutions (self-similar solutions), phenomenological models of disks, and direct numerical simulations in the magnetohydrodynamic approximation (IDEFIX, PLUTO and SNOOPY codes).
Semi-analytical models allow an easier comparison with observations. This is for example the case of the self-similar Jet Emitting Disk (JED) solutions developed in the sherpas team for years. They can be compared with X-ray observations of compact objects such as AGNs or X-ray binaries (e.g., Marcel et al. 2019, Ursini et al. 2020, Marino et al. 2021, Barnier et al. 2022). They can also be used to study the effects of ejecting disk properties, such as magnetization, on the characteristics of the ejected stream and the spectral signatures (e.g. absorption lines) expected (Datta et al. 2024).
Contacts
Jonathan Ferreira, Geoffroy Lesur, Pierre-Olivier Petrucci, Guillaume Dubus, Nicolas Scepi, Nathan Zimniak
Dynamics of protoplanetary disks
Young stellar objects and Planetary systems
Motivations
Protoplanetary disks are cold, low ionisation environments. They can nevertheless be described as a non-ideal plasma, subject to a coupling to the magnetic field depending on the location inside the disc. It is now thought that this coupling is necessary to explain the accretion rates in protoplanetary disks but also the molecular winds that are almost systematically observed.
Objectives
The objective of this research theme is to develop non-ideal plasma models of these disks from first principles of physics, in order to deduce observational signatures of the dynamics of these objects. In particular, we aim to:
- Understand how these objects accrete, despite their very low ionisation fraction
- Understand how the winds interact with the dust in the disc (sedimentation, entrainment and transport)
- Study the interaction between the plasma and the forming planets still buried in the disc, in particular the migration of the planets, depending on their mass.
- Understand how gas turbulence interact with grain growth
Methods
To study these phenomena, we use numerical methods adapted to the solution of the compressible magnetohydrodynamics (MHD) equations. In practice, we use the Idefix code, which is a new generation Godunov finite volume code, adapted to massively parallel accelerated super-computers (e.g. using GPUs) such as Jean Zay at Idris or Adastra at CINES and Lumi in Finland, the largest European supercomputer open to academic research.
The ERC-MHDiscs project
The ERC MHDiscs project aims to perform ab-initio simulations of protoplanetary disks taking into account the interaction with the surrounding magnetic field and the low degree of ionisation of these objects. In particular, we aim to obtain observable quantitative predictions that would indicate the presence of planets buried in these discs, or magnetohydrodynamic phenomena that could be precursors to planetary formation.
Contacts
Geoffroy Lesur, Nicolas Scepi
Astrocladistics and astrostatistics
Motivations
Big data are an inescapable reality in astrophysics today, and the adapted tools are starting to spread and develop: Statistics, Big Data, Data Mining, Machine Learning, Deep Learning, AI (Artificial Intelligence), these terms are now part of the landscape of astrophysicists. Still in its infancy a fifteen years ago, astrostatistics now has its own dedicated journal that Didier Fraix-Burnet created in 2019.
One particularly important area is the real-time classification of new observations, as the rate and volume of data acquisition by large modern telescopes and detectors no longer allow for visual or manual analysis. To this end, astronomers have mainly tried to make the algorithms reproduce our knowledge (so-called supervised classification), mainly via neural networks or Deep Learning. A priori, this makes it possible to classify new observations very quickly into classes that are already known... provided that we have a very large learning base and an appropriate classification. In response to these impasses, more and more unsupervised or semi-supervised methods are being developed.
But it was well before that, in 2000, that our team invested in unsupervised and evolutionary approaches.
Objectives
Understand the power of unsupervised classification, i.e. the ability of the algorithm to build its learning base by itself. The objective of unsupervised classification is to search for structures in a data space with n objects and p variables. These structures characterize either similarities (groups or classes) or relationships such as evolutionary paths. The former are rather studied by statistical methods (machine learning, data mining), the latter are based on graph theory and are the prerogative of phylogenetic methods (path minimization).
Methods
Astrocladistics
Initiated in 2001 by Didier Fraix-Burnet, astrocladistics concerns the use of phylogenetic methods in astrophysics, cladistics, also called Maximum Parsimony, being one of them but certainly the most general. Initially thought for the case of galaxy diversification, astrocladistics has been successfully applied on all kinds of astrophysical entities, such as stars, stellar clusters (globular and open), and even Jupiter and Saturn satellites! To read more about astrocladistics, visit D. Fraix-Burnet’s dedicated web-blog.
A great deal of work has been published by our team, but we quickly came up against a fundamental limitation: the absence of classes obtained by multivariate analysis, in the same way as for living species. A little detour into statistics has enabled us to classify almost two million of galaxies and thus obtain objective classifications based on morphometric data (structure, photometry, spectroscopy, etc.).
Astrostatistics
The unsupervised approach that we implement is based on the technique of mixing Gaussian distributions in a latent subspace. In other words, we look for groupings of various Gaussian distributions not in the space of the variables, but in a subspace where these groupings are clearer. The algorithm used is called Fisher-EM and was developed by Charles Bouveyron and is available under R.
We applied it to a sample of 702 248 optical spectra of galaxies (from the SDSS), each spectrum containing 1437 monochromatic fluxes. We find an optimum of 86 classes, of which the 37 largest cover 99% of the sample (Fraix-Burnet et al. 2021). A first quick analysis shows the physical relevance of these classes in classically used diagrams. Une more in-depth interpretation of each class is underway in order to build a true atlas of galaxy spectra.
This work was extended to redshifts of 1.2 by Julien Dubois using VIPERS data (Dubois et al 2024). The redshift effect shifts the optical spectra to such an extent that it is not possible to process the whole sample at once. We therefore divided it into 23 redshift bins, and linked the classes from one bin to another. The resulting tree shows the chronological behaviour of galaxy species. We can see three large branches whose spectra indicate that they are respectively passive galaxies (red branches), star-forming galaxies (green branches), and very blue galaxies (with very numerous emission lines, blue branches).
Hugo Chambon is extending this work to a redshift of around 4, in an attempt to improve this tree to represent a more realistic evolutionary scheme (in the form of a network) and thus get closer to a true representation of the diversification of galaxies.
Unsupervised classification of hyperspectral data
Hugo Chambon has, for the first time, carried out an unsupervised classification of hyperspectral images of galaxies. After the classification of spectra (above) and the classification of galaxy images (Fraix-Burnet 2023), it became quite natural to use statistical classification methods to identify spectrally similar regions within galaxies (Chambon & Fraix-Burnet 2024).
Perspectives
We have amply demonstrated the value of unsupervised approaches and should be able to produce an evolutionary diagram of galaxies in the course of cosmic evolution. It is important to bear in mind that this is pioneering work, which requires and will continue to require profound reflection on the physics of galaxies for a long time to come. The results obtained do not immediately fit into our usual schemes, but isn’t it the aim of research to look at data from ever different and new angles?
Contacts
Didier Fraix-Burnet, Hugo Chambon