Research

My research

A journey through the faint Universe

Galaxies

Galaxies are formed of billions of stars, gas and dust, embedded in a dark matter halo and held together by gravity. They are found in a huge variety of shapes, sizes, colours, luminosity, from the smallest dwarf galaxie to the large elliptical and beautiful spiral galaxies such as our own Milky Way.

The Hubble fork, showing the variety of galaxy shapes and colours. © Karen Masters

Galaxies are not static objects, they evolve over time. According to the current models of galaxy formation, galaxies are formed from the gravitational collapse of gas in the early Universe, which then cools and forms stars. Over time, galaxies merge and interact with each other, they accrete gas, leading to the formation of larger galaxies and the evolution of their structure and properties. This is known as hierarchical galaxy formation.
Check out this video of simulated galaxy mergers, overlaid with images of real galaxies, to help visualise what mergers do to galaxies !

Tidal features

During these mergers, stars and gas are pulled out of their host galaxy due to the gravitational forces at play. These stars form extended, faint structures around galaxies, known as tidal features. The tidal features are crucial to study, as they can provide important clues about the later stages of galactic assembly. Indeed, numerical simulations show that their properties (morphology, luminosity, colour) are a function of the parameters of the merger, such as the mass ratio of the merging galaxies, their relative velocity and the impact parameters.

For instance, a merger between two silimar-mass galaxies (called major merger) will produce wide, elongated tidal tails such as the famous Antennae galaxies. A merger with a low-mass companion galaxy (minor merger) will strip the small galaxy of its stars, producing stellar streams (thinner than tidal tails). When the two galaxies collide 'head-on' (radial merger), they produce shells, which are sharp, concentric arc-shaped features around the host galaxy.

Look out the beautiful examples below !

Examples of tidal tails in Euclid Q1 data. Tails are formed during major mergers between similar-mass galaxies. © E. Sola

Examples of streams in Euclid Q1 data. Streams are formed during minor mergers between a massive galaxy and a low-mass companion. © E. Sola

Examples of shells in Euclid Q1 data. Shells are formed during radial mergers (i.e. 'head-on' collisions). © E. Sola

Therefore, if you can identify the type of tidal feature around a galaxy, you can get clues on the late merger history of the galaxy, and thus on its formation and evolution. However, a big challenge is that these tidal features are very faint and difficult to detect. Careful data reduction and image processing are required to reveal these structures.

The Low Surface Brightness (LSB) Universe

In our own Milky Way, stars can be individually resolved and we can have access to their positions and velocities. Streams can then be identified as groups of stars sharing similar properties (e.g. colour, distance, velocity).

Atlas of the known streams around the Milky Way (as of May 2024), © Bonaca & Price-Whelan 2024

However, in more distant galaxies, the stars are too far away and too faint to be resolved individually. Instead, we can only see the diffuse light emitted by the stars in the tidal features. As this light is very faint, careful image processing is required to reveal these faint, Low Surface Brightness (LSB) structures. Images complying with these requirements are called deep images.

Left: illustration of a 'standard' data reduction processing, perfect for small, point-like sources. However, for large objects, the light from the object's outskirts is erased during a step called 'sky subtraction'. The image is from the Pan-STARRS survey. Right: illustration of the same image but with a 'deep' data reduction processing, which preserves the light from the object's outskirts, revealing the LSB structures. The image is from the MATLAS survey. The galaxy here is NGC0474, a giant elliptical galaxy with prominent streams and shells. The Aladin sky atlas was used to create this illustration.

Tidal features are not the only stellar structures that inhabit the low surface brightness regime. Diffuse dwarf galaxies, the extended outskirts of massive systems such as Malin 1, and the faint glow of intracluster light (i.e. the diffuse stars bound to galaxy clusters) also belong to this hidden domain.

CFHT and Euclid

A wide variety of telescopes and instruments can be used to capture deep images, as long as the data are carefully processed. These range from small amateur telescopes to intermediate and large professional facilities, and even space-based observatories.

For my research, I use deep images from the Canada France Hawaii Telescope (CFHT) and the Euclid space telescope. But these are only two examples among many others, such as (to name only a few) the Dragonfly Telephoto Array, the DGSAT network, the Milanković Telescope, the Hyper Suprime-Cam @ Subaru telescope, the Burrell Schmidt telescope, the VST telescope, the Gran Telescopio Canarias, the Large Binocular Telescope, as well as the brand-new Vera C. Rubin Observatory and the future ESA's mission ARRAKIHS.

The Canada France Hawaii Telescope (CFHT) is a 3.6m telescope located on Mauna Kea, Hawaii. It is equipped with the MegaCam camera, which has a large field of view and which is associated to several optical filters. A LSB-compliant processing pipeline (Elixir-LSB, © J.-C. Cuillandre) has been developed to process the data from this camera. It was used for large, LSB surveys such as NGVS, MATLAS, VESTIGE and CFIS/UNIONS.

The Canada France Hawaii Telescope (CFHT) on Mauna Kea, Hawaii. © CFHT

The Euclid space telescope is a mission of the European Space Agency (ESA) to study the dark Universe. Launched in July 2023, it is equipped with a visible and near-infrared camera, which will produce deep images of the sky in several filters. It's unprecedented combination of large field of view, high resolution, multiple filters and high sensitivity will allow it to detect the faintest structures in the Universe. Euclid will observe the sky in two main surveys: the Wide Survey and the Deep Survey.

Euclid's first images were showcased as part as the Early Release Observations , were its performance in the LSB regime was demonstrated (with a dedicated pipeline). While we wait for the Data Release 1 in 2026, the Euclid Consortium has released a Quick Data Release (Q1) in March 2025, which contains 63 square degrees of the sky.

Euclid Early Release Observations. © ESA/Euclid/Euclid Consortium/NASA, image processing by J.-C. Cuillandre (CEA Paris-Saclay), G. Anselmi

My work

My goal is to detect and characterise these faint, diffuse tidal features in the local Universe in order to unravel their late mass assembly history. My focus is on focus on massive nearby galaxies. I study their tidal features not only by looking at them (i.e. a census by visual inspection), but also by using quantitative methods to measure their properties.

Jafar

To that end, I contributed to develop and maintain an online annotation tool called Jafar, that allows contributors to precisely delineate the shapes of tidal features in deep images. The annotations (labels and coordinates of the contours) are stored in a database that can then be queried to extract the properties of the tidal features, such as their geometry, photometry and colour.

Annotation interface with its main facilities: drawing buttons (label 1), classification menu (label 2), examples of already drawn annotations(label 3) and summary table (label 4). © Sola et al 2022

Why don't you try it out yourself ?

Jafar annotation sandbox

Recent work

In Sola et al 2022, we present the Jafar tool and its first results on the annotation of tidal features around 352 galaxies in the footprints of the CFIS/UNIONS and MATLAS surveys. We extended this work to 475 galaxies in CFIS/UNIONS, MATLAS, NGVS and VESTIGE in Sola et al 2025. We investigated the effects of the host galaxy's properties and environment on the tidal feature properties. Overall, 36% of the galaxies in our sample have tidal features, and we found a strong correlation with the host galaxy's stellar mass.

Our annotation database can be used to train machine learning algorithms to automatically segment tidal features in deep images. Indeed, manual annotation is a time-consuming task, and automated method are required to scale up the analysis to larger samples, such as the deluge of data from Euclid or Rubin. A first attempt was carried out by Richards et al 2021 who developed a segmentation network trained on our database. This promising work is still ongoing.

I had the opportunity to work with the very first Euclid Early Release Observations (ERO) data. I was involved in the search for dwarf galaxies in the Perseus cluster using Jafar in Marleau et al 2024, where we found 1100 dwarf galaxy candidates, including 600 new discoveries. I contributed to the analysis of tidal features in the ERO Dorado group field in Urbano et al 2025, to the study of the intracluster light in the ERO Perseus cluster in Kluge et al 2024, and of some ram-pressure stripped galaxies in George et al 2025, as well as some minor contributions to other Euclid papers.
With the release of the Euclid Q1 data, I was involved in the semi-automated detection of dwarfs galaxies in Marleau et al 2025. I am leading a project to search for tidal features in the Euclid Q1 data in Sola et al 2025 (in prep.), in preparation for the Euclid Data Release 1.

Additionnally, stellar streams also offer insights on the dark matter distribution in galaxies. In the Milky Way, where stars are individually resolved, streams can be used to put strong constraints the dark matter halo properties. However, recent work (e.g. Nibauer et al 2023) have shown that streams in other galaxies, where the diffuse light is the only observable, can also be used to some extent.
We followed up in Chemaly et al 2025 (in prep.) and showed with simulations that we can actually recover the underlying distribution dark matter halo properties ('flattening'). The next step is to apply this method to a sample of galaxies with long, curved streams suitable for modelling with this method.

Although we already had our annotation database, and some tidal features catalogues from the literature, we were missing a large sample of suitable streams. Therefore, in Sola et al 2025, we classified about 20,000 galaxies from the DESI Legacy Imaging Survey (DESI-LS). We used a combination of original images, models of all sources and residuals images (i.e. images with all sources subtracted) to identify the streams. We ended up with the 35 streams in our first release of the STRRINGS, standing for STReams in Residuals of Images of Nearby GalaxieS that we characterised using Jafar.

Find out more in my publication list