🔭 astrophysics

Bloodhound Unleashed: Particle-based Substructure Tracking for Cosmological Simulations

The paper introduces Bloodhound, a new particle-based algorithm that significantly outperforms standard methods like ROCKSTAR by extending subhalo tracking lifetimes by 3–4 Gyr and providing continuous merger trees, thereby revealing a substantially larger population of surviving subhaloes in the inner regions of dark matter haloes with critical implications for constraining dark matter properties.

Original authors: Hyunsu Kong, Michael Boylan-Kolchin, James S. Bullock

Published 2026-03-27

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Hyunsu Kong, Michael Boylan-Kolchin, James S. Bullock

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Big Picture: The Cosmic "Lost and Found"

Imagine the universe as a giant, bustling city. In this city, the Dark Matter Halos are the massive, invisible skyscrapers. Inside these skyscrapers, there are smaller, dimmer apartments called Subhalos. Some of these apartments are so small and dark that they might not even have any people (stars) living in them, but they are still there.

Scientists use super-computer simulations to watch how these cosmic cities grow over billions of years. They want to know: How many of these tiny apartments survive when a big skyscraper swallows a smaller one?

The problem is that the current tools scientists use to track these apartments are like a bad GPS system. When an apartment gets too close to the center of the big skyscraper (where the gravity is intense), the GPS loses the signal. It thinks the apartment has been destroyed or vanished, even though it's actually still there, just hiding in the crowd.

This paper introduces a new tool called Bloodhound. Instead of trying to find the apartment in every single snapshot of the city, Bloodhound grabs the "tenant list" (the specific particles) the moment the apartment moves in, and then follows those specific people forever, no matter how crowded the building gets.

The Problem: Why the Old GPS Failed

The standard method (called ROCKSTAR + consistent-trees) works like this:

Take a photo of the universe.
Find all the apartments.
Take another photo a million years later.
Try to match the apartments from photo A to photo B.

The Glitch:
When a small apartment (subhalo) falls deep into a massive skyscraper (the host halo), it gets squeezed by gravity. It starts to lose its outer walls (tidal stripping).

The Old GPS gets confused: It sees the apartment losing its walls and thinks, "Oh, it's gone!" It deletes the apartment from the map.
The "Broken Link" Mystery: Sometimes, the GPS loses the apartment for a few years, then sees a tiny, new-looking apartment pop up nearby. It thinks, "Oh, a new apartment was just built here!" In reality, it's the same old apartment that just got really skinny and was invisible for a while. This creates fake data about how many apartments are being built and destroyed.

The Solution: Enter "Bloodhound"

The authors created Bloodhound, a new tracking algorithm. Think of it not as a GPS that looks at the whole city, but as a loyal dog that has a specific scent.

The Scent (The Particles): When a subhalo first falls into the big host halo, Bloodhound takes a "scent sample" of its most loyal, tightly-bound particles (the core members).
The Chase: Instead of looking for the whole building in the next photo, Bloodhound just follows those specific particles. It doesn't care if the building looks messy or if the walls are falling off. As long as the core group of particles is still together, the apartment is still "alive."
The Verdict: Bloodhound only declares an apartment "dead" when the core group of particles has completely scattered and can no longer hold hands.

What Did They Find?

Using simulations of Milky Way-sized galaxies, the team compared the old GPS (ROCKSTAR) with the new Bloodhound. Here is what they discovered:

The "Zombie" Apartments: Bloodhound found that many apartments the old GPS said were dead were actually still alive! On average, Bloodhound kept tracking subhalos for 3 to 4 billion years longer than the old method.
The Inner City is Full: The old GPS thought the center of the galaxy was empty of small apartments. Bloodhound found that the center is actually teeming with them. In the inner 50,000 light-years, Bloodhound found twice as many surviving subhalos as the old method.
Fixing the "Broken Links": Bloodhound fixed the "ghost apartment" problem. It realized that many "newly formed" apartments were actually just old ones that had been lost and then found again. By fixing this, they corrected the history of these galaxies, showing that many small galaxies are actually the descendants of much larger, older ones.
The "Orphan" Mystery: Because Bloodhound tracks things so well, it found that many tiny, dark subhalos survive even when they get very close to the center of the galaxy. This suggests that the "missing" satellite galaxies we see in the real universe might be hiding in these dense inner regions, waiting to be found.

Why Does This Matter?

This isn't just about counting dots in a computer game. It changes how we understand the nature of Dark Matter.

Testing the Rules: If we think Dark Matter is "Cold" (slow-moving), we expect a certain number of tiny apartments. If the old GPS says there are none, but Bloodhound says there are thousands, it changes the rules of the game.
Finding the Invisible: Astronomers are currently looking for "orphan" galaxies—galaxies that have lost their dark matter homes but are still shining. Bloodhound helps us predict where these orphans might be hiding.
The Future: With new telescopes like the Vera Rubin Observatory coming online soon, we will be able to see thousands of these tiny galaxies. We need a reliable map (like Bloodhound) to know if what we see matches our theories about the universe.

The Bottom Line

The old way of tracking cosmic structures was like trying to follow a specific person in a crowded stadium by taking a photo every hour and guessing who they are. You'd lose them the moment the crowd got too thick.

Bloodhound is like giving that person a radio transmitter. You don't need to see them; you just follow the signal. This new tool reveals that the universe is much more crowded with small, dark structures in the centers of galaxies than we previously thought, and it gives us a much clearer picture of how our own galaxy, the Milky Way, grew up.

1. Problem Statement

Modern cosmological simulations, particularly those studying the formation of Milky Way (MW) mass haloes, rely on identifying and tracking dark matter subhaloes to understand the nature of dark matter and the "missing satellites" problem. However, standard subhalo tracking pipelines (specifically the widely used ROCKSTAR halo finder combined with consistent-trees merger tree construction) suffer from significant limitations in dense environments:

Premature Disruption: Standard methods often fail to track subhaloes that orbit deep within the gravitational potential of a massive host halo. As subhaloes lose mass and their particles become mixed with the host's background, snapshot-by-snapshot halo finders lose the subhalo's identity, marking it as "disrupted" even when a self-bound core remains.
Broken-Link Trees: When a subhalo is temporarily lost and then re-identified later, standard merger trees often fail to link the two phases. This results in "broken-link" subhaloes: objects that appear to form de novo inside the host halo (violating $\Lambda$ CDM structure formation theory) with artificially low infall masses and peak velocities.
Biased Statistics: These tracking failures lead to a severe underestimation of the surviving subhalo population, particularly in the inner regions ( $R < 50$ kpc) of the host halo and for subhaloes with small pericentres. This biases predictions for ultra-faint satellite galaxy counts and dark matter substructure constraints.

2. Methodology: The Bloodhound Algorithm

Bloodhound is a post-processing tool designed to augment existing pipelines (specifically ROCKSTAR + consistent-trees) by shifting the tracking paradigm from "snapshot identification" to "particle tracing."

Core Concept: Instead of trying to identify subhaloes at every snapshot and stitch them together, Bloodhound identifies robust, independent haloes immediately before they fall into a larger host. It then tracks the specific particles of these progenitors forward in time until the simulation ends ( $z=0$ ).
Particle Selection:
- It extracts particles from the raw simulation data at the snapshot immediately preceding the subhalo's first infall ( $R_{vir}$ crossing).
- It assumes subhaloes do not accrete significant mass from the host post-infall.
- It tracks these specific particle IDs through all subsequent snapshots.
Center and Property Calculation:
- Center of Mass (COM): Calculated iteratively using the 2% most bound particles (at infall) to ensure the center remains anchored to the dense core, even as the halo is stripped.
- Boundary Definition ( $R_{trunc}$ ): To distinguish bound subhalo particles from stripped tidal debris, Bloodhound analyzes the differential mass profile ( $\rho(r) \times r^2$ ). It identifies the radius where the slope of this profile turns positive, indicating the transition from the self-bound halo to the unbound stripped material.
Disruption Criteria: Rather than relying on arbitrary particle counts, Bloodhound uses the evolution of the physical concentration parameter ( $c_V$ ), defined as the mean density within $R_{max}$ $R_{ma x}$ relative to the critical density. A subhalo is deemed disrupted only if:
1. $c_V$ drops abruptly (to <30% of the previous snapshot).
2. $c_V$ drops significantly relative to infall (factor of 5).
3. The low $c_V$ state is sustained for ~500 Myr (5 snapshots), ruling out transient tidal shocks.

3. Key Contributions

Continuous Tracking: Bloodhound eliminates the "lost and found" problem, providing a continuous, unbroken history for subhaloes from infall to $z=0$ .
Correction of Broken Links: It naturally resolves broken-link trees by tracing particles back to their true, massive infalling progenitors, correcting the artificial "in-situ" formation and low-mass bias of standard methods.
Robust Disruption Definition: By using $c_V$ and density profile slopes, it offers a physically motivated definition of disruption that distinguishes between a dispersed, unbound particle cloud and a surviving, albeit stripped, subhalo.
Post-Processing Compatibility: It is designed to work with existing standard pipelines (using their outputs as input for selection), making it an accessible upgrade for current simulation analyses rather than a complete replacement of halo finders.

4. Key Results

The authors tested Bloodhound using the Phat ELVIS suite of Milky Way-mass simulations (both Dark Matter Only and Disc runs).

Extended Survival Time: Bloodhound extends the average tracking time of subhaloes by 3–4 Gyr compared to consistent-trees. For subhaloes with small pericentres, the extension is even longer.
Recovery of "Disrupted" Subhaloes:
- In the Disc runs, 25% of subhaloes deemed disrupted by consistent-trees are found to survive to $z=0$ by Bloodhound.
- In the DMO runs, this fraction rises to nearly 50%.
- For subhaloes infalling early ( $t_{infall} > 10$ Gyr), the survival rate increases dramatically (from ~10% to ~35-80% depending on the run).
Inner Region Abundance: Within the inner 50 kpc of the host:
- Bloodhound finds 30–70% more surviving subhaloes with $V_{peak} \ge 10$ km/s compared to the standard method.
- For $V_{peak} > 13$ km/s, Bloodhound finds more surviving subhaloes than the standard method finds total (surviving + disrupted).
- Standard methods find zero subhaloes with $V_{peak} > 20$ km/s in the inner 50 kpc of Disc runs; Bloodhound recovers ~3 per host.
Correction of Broken Links: Bloodhound identifies that "broken-link" subhaloes (which appear to form inside the host) are actually remnants of massive progenitors ( $V_{infall} \sim 15-30$ km/s) that were lost early. This corrects the infall mass distribution, shifting it to higher values and altering the inferred subhalo mass function.
Disruption Mass: Bloodhound tracks subhaloes to significantly lower masses before declaring them disrupted. The median disruption velocity ( $V_{disrupt}$ ) drops from 10 km/s (standard) to 6 km/s (Bloodhound), representing a mass reduction of 6–65 times.

5. Significance

Milky Way Satellite Census: The results imply that the number of surviving subhaloes capable of hosting ultra-faint dwarf galaxies in the inner Milky Way is significantly higher than previously thought. This helps alleviate the tension between the observed number of satellites and the predicted number of subhaloes in $\Lambda$ CDM.
Dark Matter Constraints: The improved tracking of subhaloes on radial orbits with small pericentres is crucial for interpreting gravitational lensing perturbations and stellar stream gaps, which are sensitive to the inner subhalo population.
Methodological Shift: The paper demonstrates that snapshot-based halo finding is insufficient for dense environments. A particle-tracking approach is necessary to faithfully model the tidal evolution of substructure, particularly for future high-resolution simulations and surveys like LSST.

In conclusion, Bloodhound provides a robust, physically motivated framework for subhalo tracking that resolves critical systematic errors in current cosmological analyses, leading to a more accurate understanding of galaxy formation and dark matter properties.