Modeling gravitational wave sources in the MillenniumTNG simulations

This paper introduces a flexible, parallel framework called Arepo-GW that integrates binary population synthesis with hydrodynamic simulations to generate gravitational wave event catalogs, demonstrating its application on the MillenniumTNG suite to reveal merger rates and progenitor properties consistent with current observations while highlighting specific discrepancies in binary black hole merger rates and redshift evolution.

The Gist

Imagine the universe as a giant, bustling city that has been growing for 13.8 billion years. In this city, stars are the people, and galaxies are the neighborhoods. Sometimes, two stars that are "best friends" (binary stars) get so close they eventually crash into each other, creating a massive ripple in the fabric of space-time called a gravitational wave.

For a long time, scientists could only listen to these ripples after they happened, like hearing a thunderclap after the storm. But they wanted to know where the storm started, why it happened, and how often it occurs across the whole city.

This paper introduces a new, super-smart tool called Arepo-GW that acts like a "crystal ball" for the universe. Here is how it works, broken down into simple concepts:

1. The Problem: The Universe is Too Big to Watch

Scientists have built massive computer simulations of the universe (like the MillenniumTNG project). These are like incredibly detailed video games where they simulate the formation of billions of galaxies. However, these simulations track gas, stars, and dark matter, but they don't naturally "see" the tiny, invisible collisions between dead stars (black holes and neutron stars) that create gravitational waves.

It's like having a simulation of a whole city, but the software doesn't know how to count the car accidents because the cars are too small and the crashes happen too fast to track in real-time.

2. The Solution: The "Stochastic Lottery"

The authors created a new module (a piece of software add-on) called Arepo-GW. Instead of trying to simulate every single star collision from scratch (which would take forever), they use a clever trick:

• The Library of Possibilities: They first used a super-advanced calculator (called sevn) to simulate millions of star pairs in a lab setting. They figured out the odds: "If you have a star of this age and this metal content, there is a 1-in-a-million chance it will merge in 5 billion years."
• The Lottery: Now, when they run their giant universe simulation, every time a star is born, Arepo-GW checks its "ID card" (age, metal content, mass). It then pulls a number from a hat. If the number matches the odds from the library, the simulation says, "Bingo! This star is destined to become a gravitational wave source."

This allows them to build a massive catalog of gravitational wave events without slowing down the main simulation.

3. What They Found: The "Ghost" Map

Using this tool on their giant simulation (the MTNG740 box, which is one of the largest virtual universes ever made), they generated a map of where these collisions happen.

• The Connection: They found that gravitational waves mostly happen where new stars are being born. It's like saying car accidents happen most often in busy city centers, not in empty fields.
• The "Too Many" Problem: Their simulation predicted that Black Hole collisions happen about 4.5 times more often than what our current telescopes (like LIGO) are actually seeing.
  • Analogy: Imagine your weather app predicts it will rain 4.5 times more than it actually does. This suggests our "weather models" for how stars die might be slightly off, or perhaps our telescopes are missing a lot of the "rain" because it's too far away or too faint.
• The Time Traveler: They found that while the universe is getting quieter (fewer stars are forming now than in the past), the rate of these black hole crashes isn't dropping as fast as we thought. It's like a party that should be winding down, but the DJ (the black holes) keeps the music going longer than expected.

4. Why This Matters

This framework is a game-changer because it connects two worlds that usually don't talk to each other:

1. Cosmology: How the universe grows and changes over time.
2. Stellar Physics: How individual stars live and die.

By linking them, scientists can now say, "If the universe formed stars this way, then we should see gravitational waves that way."

The Bottom Line

The authors have built a bridge between the "big picture" of the universe and the "small picture" of dying stars. They created a tool that can predict where and when the universe will "ring" with gravitational waves.

Even though their prediction for black hole collisions is a bit too high compared to what we see today, this isn't a failure—it's a clue! It tells scientists, "Hey, our model of how stars die needs a little tuning." As we get better at building these virtual universes and better at detecting the real waves, this tool will help us understand the history of the cosmos, one ripple at a time.

Technical Summary

Here is a detailed technical summary of the paper "Modeling gravitational wave sources in the MillenniumTNG simulations."

1. Problem Statement

The detection of gravitational waves (GWs) by LIGO/Virgo/KAGRA has opened a new window into the universe, but interpreting the population demographics of merging compact objects (Binary Black Holes [BBH], Binary Neutron Stars [BNS], and Black Hole-Neutron Star [BHNS] systems) requires a robust theoretical framework. Existing methods face significant limitations:

• Semi-analytic models: Often rely on observationally calibrated star formation histories and parameterized binary evolution, failing to capture complex dynamical environments and feedback-regulated star formation.
• Hybrid approaches: Combine merger trees with simplified binary evolution but lack spatial resolution and rely on empirical scaling relations.
• Post-processing limitations: Previous attempts to map GW sources to cosmological hydrodynamical simulations often struggled with small simulation volumes (failing to sample rare high-density environments) or lacked self-consistency between local physical conditions (metallicity, gas density) and binary evolution.

There is a need for a framework that couples binary population synthesis (BPS) directly into large-volume, full-physics cosmological hydrodynamical simulations to predict GW event catalogs with high fidelity, accounting for the cosmic history of star formation and chemical enrichment.

2. Methodology

The authors introduce Arepo-GW, a flexible, fully parallel module integrated into the moving-mesh code Arepo. The framework couples the state-of-the-art BPS code sevn (Stellar EVolution for N-body) with the MillenniumTNG (MTNG) cosmological simulation suite.

Key Components:

• Binary Population Synthesis (sevn):
  • Utilizes the SEVNbenchmark2401 dataset, containing the evolution of $3 \times 10^8$binary systems across 15 metallicities ($0.0001 \le Z \le 0.034$).
  • Models include core-collapse supernovae (DelayedGauNS), pair-instability supernovae (M20 model), natal kicks, mass transfer (Roche-lobe overflow and winds), common envelope (CE) phases, and gravitational wave decay.
  • Generates merger tables for BBH, BHNS, and BNS channels, sorted by merger time.
• Arepo-GW Integration:
  • Stochastic Sampling: For every star particle in the simulation (treated as a Simple Stellar Population with a Chabrier IMF), the module calculates the probability ($p_*$) of a merger event occurring based on the particle's birth mass, age, and metallicity.
  • Equation: $p_* = M_* f_{bin} f_{IMF} / M_{sim}$, where $M_*$ is the star particle mass, $f_{bin}$ is the binary fraction, and $f_{IMF}$ corrects for the limited mass range sampled in the BPS tables.
  • Operational Modes:
    1. On-the-fly: Executes during the simulation runtime, recording precise spatial positions at the moment of merger.
    2. Post-processing: Analyzes existing snapshots (used in this study for the MTNG740 run). It evolves star particles from birth to the snapshot time to generate a complete catalog.
• Simulation Setup:
  • Applied to the MTNG740 flagship box (comoving volume $\approx 740^3$ Mpc$^3$, side length $\approx 740$ Mpc) and three smaller variants (MTNG46, MTNG92, MTNG185) to test volume convergence.
  • Physics includes full hydrodynamics, metal-dependent cooling, stellar/AGN feedback, and supermassive black hole seeding/growth.

3. Key Contributions

• Novel Framework: Development of Arepo-GW, the first module to seamlessly integrate high-fidelity BPS (sevn) into large-scale cosmological hydrodynamical simulations (Arepo) for GW source modeling.
• Scalability: The algorithm is fully parallel (MPI) and capable of handling massive datasets (e.g., processing a $\approx 7$ TB snapshot to produce a $\approx 30$ TB GW catalog).
• Flexibility: The method is adaptable to other simulation codes and can utilize different BPS tables, allowing for uncertainty quantification in binary evolution models.
• Comprehensive Catalog: Generation of the first GW event catalog for the MTNG suite, covering BBH, BHNS, and BNS channels across cosmic time.

4. Key Results

The analysis of the MTNG740 simulation yields the following insights:

• Merger Rates and Evolution:

  • GW progenitor rates closely track the cosmic Star Formation Rate Density (SFRD).
  • Peaks: BBH and BHNS rates peak at $z \approx 3$, slightly earlier than the SFRD peak ($z \approx 2-3$), while BNS peaks at $z \approx 2$.
  • Redshift Decline: The decline in merger rates at low redshift ($z < 1$) is shallower than the SFRD decline. The BBH rate declines with a slope $\kappa \approx 1.44$, compared to $\kappa \approx 3.2$ inferred from LVK observational constraints.
  • Local Excess: The model overpredicts the local BBH merger rate by a factor of $\sim 4.5$ compared to GWTC-4 constraints. This is consistent with trends in other BPS models, likely due to uncertainties in massive binary evolution and CE efficiency.

• Delay Time Distributions (DTD):

  • BNS: Dominated by short delay times ($t_{delay} \lesssim 0.1$ Gyr), tightly coupling their rates to recent star formation.
  • BBH/BHNS: Exhibit a significant long-delay tail ($t_{delay} \gtrsim 0.5$ Gyr). This allows progenitors to migrate from birth sites, leading to more diffuse spatial distributions compared to BNS.

• Metallicity Dependence:

  • Efficiency: Merger efficiency per unit stellar mass is non-monotonic for BBH/BHNS, peaking at $Z \approx 4 \times 10^{-3}$ and dropping at higher metallicities. BNS efficiency increases monotonically with metallicity.
  • Impact: The rapid chemical enrichment in MTNG (reaching solar metallicity by $z \sim 3$) suppresses BBH formation at late times, but the long-delay tail of BBHs allows low-metallicity progenitors formed earlier to merge later, contributing to the local rate excess.

• Mass Functions:

  • BBH: Primary mass distribution shows a peak around $10 M_\odot$and a secondary feature near$36 M_\odot$, with a sharp drop at$m_1 \approx 44 M_\odot$due to pair-instability supernovae. The distribution extends up to$\sim 70 M_\odot$.
  • Mass Ratios: BBH mergers show a preference for near-equal mass components ($q \to 1$), likely driven by CE evolution.
  • Evolution: Mass functions show only mild evolution with redshift, consistent with previous studies, driven by the rapid metal enrichment of the stellar populations in the simulation.

• Volume Convergence:

  • Results are consistent across all MTNG boxes (46 to 740 Mpc), confirming that the SFRD and GW progenitor properties are well-converged even in smaller volumes, provided mass resolution is constant.

5. Significance

• Cosmological Context: This work provides the first detailed predictions for GW sources within a full cosmological context, linking merger rates directly to galaxy formation physics, feedback, and chemical enrichment.
• Observational Comparison: The framework enables direct cross-correlation with electromagnetic surveys and cosmological tracers, allowing for the study of GW host galaxy properties and spatial clustering.
• Future Surveys: The model offers predictions for next-generation GW detectors (e.g., Einstein Telescope, Cosmic Explorer) that will probe high redshifts ($z \sim 100$).
• Uncertainty Quantification: By isolating the impact of galaxy formation physics from binary evolution uncertainties, the framework helps identify where theoretical models (e.g., CE efficiency, natal kicks) need refinement to match observational data, particularly regarding the BBH rate excess.
• Mock Observations: The generated catalogs serve as a foundation for creating mock GW observations that include selection effects, crucial for interpreting future data from the LVK and future detectors.