By Philip Mansfield

Let me tell you a cosmic ghost story.

Act I: A Mysterious Galaxy Killer on the Loose

Every large galaxy in the universe is surrounded by an orbiting swarm of smaller galaxies. These orbits are not the safe, regular circles of a GPS satellite. They are the chaotic last gasps of dying galaxies. These small galaxies are dragged to their doom by the intense gravity of their larger companions. They get shredded apart by enormous gravitational tides, and may even smash directly into the huge galaxy at the center of it all. The problem is clear: intergalactic killers are on the loose, pulling the locals into their haunted houses.

But if you look at the movement of these galaxies closely, you’ll notice that something is… wrong. The doors on these haunted houses slam at odd times, books are floating through the air, and the orbits of all these galaxies just aren’t right.  These galaxies are not in control of their own fates—even the giant central galaxies, initially the prime suspects, are victims. They are being dragged around like marionettes by the gravity of much larger and more mysterious objects. The objects dragging them around are completely invisible no matter what wavelength you look at them in, and their gravity is so strong that they forced galaxies to form out of the swirling mass of featureless Hydrogen that filled the early Universe. Astronomers call them “dark matter halos.” But on dark and windswept nights, I like to think of them as ghosts that can possess entire galaxies.

Act II: The Killer Was Inside the Computer All Along

These ghosts have been the subject of a great hunt in astronomy that has lasted for generations. At first, this hunt centered on answering the question “Do dark matter halos exist?” And now many of us are hoping to continue it by answering the question “What exactly are dark matter halos made of?” I look to these dying galaxies for clues, and so do many people in my field. Dark matter is the main thing ripping satellite galaxies apart and the main thing keeping them together: these ghosts can’t help but leave clues of its existence stamped upon dying galaxies like a muddy footprint under a windowsill.

We run simulations of dark matter halos and their orbiting galaxies, compare those simulations to the real Universe, and use that comparison to infer whether the simulation’s underlying physics model is reasonable. Studying a simulated dark matter halo is supposed to be easier than studying a real, invisible one. We know exactly what all the pieces of our simulations are doing and exactly where they are. We have spent decades developing tools and codes for analyzing these simulations. We were confident that even though dark matter halos might haunt the sky, they certainly wouldn’t haunt our computers.

We were wrong.

Act III: Digital Ghostbusting

In all good horror movies, the killer chases the heroes for the first two acts, but the hero finally chases the killer in the third act. Perhaps our chase has led us deep into the haunted house, and perhaps in this inner sanctum we see a great painting at the center of it all:

This painting is an example of Pointillism, where the painter only allows themselves to build up pictures by placing simple, individual dots on the canvas. This style of painting is a good analogy for how simulations of galaxies work: all the objects inside our simulations are built up of little “dots” of dark matter, called particles. As an artist, what I like about Pointillism is seeing how complex objects can be made out of simple components. As a physicist, what I love about particles is that they’re simple enough that a computer can always simulate them, even if the objects made up of them are too complicated for any human to fully understand.

But Pointillism and simulations encounter similar problems when trying to represent objects made out of very few dots. The pitcher in this image has high contrast and a lot of dots, so it’s easy to make out… but what about those blobs next to it? Is that light color at the base a placemat? Is the lump to the left a discarded shirt? It’s hard to tell. Similarly, what about a dark matter halo made up of only a few particles? Is that actually a halo? What if it’s just random noise or the remnant of an old collision? For simulations, the situation is even more dire than that. Rather than a painting with high-contrast objects like pitchers and beds, imagine that a simulation of dark matter halos is a painting of a horde of gray ghosts fading through and eventually consuming one another. Maybe you can reliably find a ghost that is made up of thousands of dots, but what if it only has ten?

These “problems” become quite fun with the right mindset. In the same way that an artist might fall in love with getting as much expressivity out of their ten dots as possible, hundreds of physicists have fallen in love with the idea of writing software that can find halos made up of only a few particles. Research led by several teams across the world (1, 2, 3), including KIPAC, has focused on improving an old family of techniques called “particle-tracking.” Particle-tracking follows the movement of particles over time to make it easier for the code to find the halos these particles make up. Returning to our painting analogy, imagine that instead of a single painting, you had a movie made up of hundreds of paintings. The ghosts in your painting are always made up of the same dots, even though the dots might shift and fade as the ghosts move and change forms. If you could figure out which dots belonged to which ghost, finding a ghost again later would be much easier.

However, researchers discovered something… bone-chilling with these new techniques. There were many more ghosts than we thought. We expected that improved analysis would only improve our ability to find halos made up of just a few particles. But beyond finding more of these small halos, these new particle-tracking codes were also finding more large ones: ones that all the previous techniques should have been able to find. After years of testing, the conclusion was clear: these new halos were not hallucinations, they were real predictions of the simulation that had been missed by the previous methods. 

To make matters more exciting, these missing halos are the most important for the quest to understand dark matter. The halos that are most easily missed by traditional methods are hanging on by a thread near the ends of their lives. The specific details of dark matter have the biggest impact on these halos, whether or not they exist at all. Although it was shocking to find them there, the ghosts in our machines might be the secret to understanding the ghosts in our skies.

The author would like to thank Sanskriti Das, Jack Dinsmore, Xinnan Du, and Lori White for their assistance with this article.