Research
Ancient Echoes of Cosmic Collisions
Galaxies grow over time through collisions, or mergers, with other galaxies. When a smaller galaxy hits a larger galaxy in a head-on collision, it is completely torn apart, but its stars keep traveling through the host galaxy. As these stars turn around in their orbits, like a bungee jumper reaching the end of their rope, they pile up and make large concentric rings, which we call shells. These shells are large, spherical, concentric structures that surround the central galaxy to give them their unique, ethereal appearance.
(Image credit: DESI & NOIRLab)
The shape and number of shells in a galaxy contains a wealth of information about how long ago the collision happened, the properties of the colliding galaxies, and the angle that the galaxies collided at. Below is a very simple simulation of one of these collisions; each column shows a specific time in the simulation, the top column shows both galaxies, and the bottom panel shows only the smaller dwarf galaxy that shreds apart to become shells. See that over time, the number of shells increases, but each shell becomes fainter as the overall structure mixes away. Also notice that the shells on either side of the galaxy are offset in distance from the galactic center, because the shells are oscillating back and forth from one side of the galaxy to the other. 
(Image credit: George Williams/NASA)
In reality, shell galaxies don't always look this clean. The host galaxy might move in the merger, or have some existing structure that makes the shells take on more messy shapes. Alternatively, the galaxy can interact with other neighboring galaxies while shells are forming or after shells have formed, which can further complicate the shapes of shells. For example, see the rich shapes of shells in NGC 474 and its neighboring galaxies. 
(Image credit: George Williams)
Shell galaxies can take on a wide variety of shapes, sizes, colors, and environments, making them an extremely rich dynamical playground to study the life cycle of galaxies throughout the Universe. 
(Image credit: Horizon Europe Legacy Surveys/D. Lang, Perimeter Institute)
Dark Matter & Dynamical Insights
As stars orbit a galaxy, they move according to their energy and the gravitational field of that galaxy. This gravity is created by the combination of the luminous material in a galaxy (stars, gas) plus the "dark" component of the galaxy, made up of unknown material that doesn't emit any light. If we're able to measure the gravity inside a galaxy somehow, then we're able to figure out how much matter is inside that galaxy, and roughly where it's located.
Shells provide a wonderful opportunity for this. As the dwarf galaxy shreds apart in the center of the larger galaxy during collision, all of its stars have roughly the same amount of energy, setting them on nearly identical orbits. As the stars move outwards to become shells, the location that they turn around is set only by the strength and shape of the gravitational field. So, understanding the shape and locations of the shells inside a shell galaxy tells you the properties of the gravitational field for that galaxy.
It's relatively straightforward to understand where the stars and gas are located in a galaxy (because you can see them), and as a result we can determine how much gravity we expect there to be from that luminous matter. This can then be subtracted out of the total gravity that we observe from our analysis of shells, and we are only left with the amount and location of the dark matter within that galaxy. Different theories for dark matter predict it having different shapes, and there may even be clumps of dark matter in some of these galaxies that could be identified from measurements of shells! Knowing more about the distribution of dark matter in shell galaxies will help us narrow down different possibilities for our best guesses about what dark matter actually is.
(Image credit: Wayne Ngan)
Our Goals
Determining where shells are located in a galaxy is hard to do automatically. There are some algorithms that can sometimes tell you where shells are located, but these usually only work for the nice-looking shells that are made from ideal conditions. Only a small fraction of observed shell galaxies fit in this category. Additionally, sometimes it can be hard to tell if a galaxy even contains shells or not from a glance! It would take researchers an extremely long amount of time to properly characterize and model each shell galaxy.
Therefore, we are asking for help from volunteers to help classify shells in galaxies. We have a substantial number of galaxy images from DESI and Euclid, where we have used Galaxy Zoo's models to select likely shell galaxies. As you help us determine which galaxies do and don't contain shells, we will be able to refine our search to better select only shell galaxies. Additionally, we ask that volunteers draw the locations of each shell they see on these images; this will provide us with an extensive catalog of shell galaxies plus the locations and shapes of their shells, at a scale that has never been done before.
With this new catalog of shell galaxy data, we will be able to characterize the amount and shape of the dark matter in each galaxy using the process described above. We hope to use the data for this large population of shell galaxies to learn about how spherical dark matter halos are, put a limit on how "clumpy" the dark matter is in each galaxy, and also learn about the events that lead to shell galaxies and galactic collisions in the first place.
Machine Learning & AI
This project uses data from Zoobot, a publicly available machine learning model trained on Galaxy Zoo classifications. This tool is used to select likely shell galaxies, which cannot be done by hand because the pool of all DESI and Euclid galaxy images is enormous. We plan on using the human classification results from this project to train an updated version of Zoobot that is better at identifying shell galaxies. This will let us show volunteers more images of shell galaxies and fewer images of non-shell galaxies while classifying. It will also let us find as many shell galaxies as possible in the massive DESI and Euclid datasets. We will not use the results from this project to train generative AI models or LLMs.