Kick matters: The impact of a new recoil model on the retention of hierarchical black-hole remnants in globular clusters

By employing a state-of-the-art recoil-kick model (gwModel_flow_prec) instead of traditional analytic approximations, this study demonstrates that the retention probability of hierarchical black-hole merger remnants in globular clusters is significantly higher, thereby altering the resulting mass and spin distributions and offering new insights into the formation of massive binaries like GW231123.

The Gist

Imagine a crowded dance floor inside a giant, ancient ballroom called a Globular Cluster. This ballroom is packed with thousands of dancers: stars and, more importantly for our story, Black Holes.

For a long time, scientists thought these black holes were mostly solo dancers. But recently, we've realized they sometimes pair up, spin together, and crash into each other. When two black holes merge, they don't just sit there; they create a new, heavier black hole.

Here is the problem: The "Kick."

The Problem: The Unwanted Backflip

When two black holes merge, they release a massive amount of energy in the form of gravitational waves (ripples in space-time). Because of how physics works, this energy doesn't just vanish; it pushes the new, merged black hole in the opposite direction. It's like a cannon firing a shell: the shell flies forward, but the cannon kicks backward.

In the world of black holes, this "kick" can be incredibly violent. If the kick is too hard, the new black hole is launched out of the dance floor (the cluster) entirely. Once it's gone, it can never dance again. It can't merge with another partner to become even bigger.

The Old Map vs. The New GPS

For years, scientists used an old, simplified map (called the HLZ model) to predict how hard this kick would be. This map was based on limited data and assumed the kicks were usually very strong.

Because of this old map, scientists thought: "Oh, the kicks are so strong that almost all merged black holes get kicked out of the cluster. So, it's impossible for them to stay, find a new partner, and merge again to become super-massive monsters."

This led to a dead end in our understanding of how giant black holes form.

Enter the new GPS:
The authors of this paper, Tousif Islam, Digvijay Wadekar, and Konstantinos Kritos, built a brand new, high-tech map called gwModel_flow_prec. They used super-computers and advanced math (like "normalizing flows," which is a fancy way of saying "AI that learns patterns from data") to create a much more accurate picture of these kicks.

The Big Discovery: The Kicks Are Gentler Than We Thought

When they ran their new simulations, they found something surprising: The old map was overestimating the violence.

In many cases, the new model shows that the "kick" is actually much softer than the old model predicted.

• Old Model: "Boom! You're out of the club!"
• New Model: "Whoa, easy there. You stumbled, but you're still on the dance floor."

Why This Matters: The "Ladder" to Giant Black Holes

This changes everything about how we think giant black holes are made.

Think of building a massive black hole like climbing a ladder:

1. Step 1: Two small black holes merge.
2. Step 2: If the kick is too hard, they fall off the ladder. Game over.
3. Step 3: If the kick is soft enough, the new black hole stays in the cluster. It finds a new partner (maybe another small one, or even another "merged" one) and climbs to Step 3.
4. Step 4: They merge again, getting even bigger.

With the old map, scientists thought the ladder was broken after the first step. With the new map, the ladder is much longer and sturdier. It means black holes can keep merging, merging, and merging, growing into the "Intermediate Mass Black Holes" (IMBHs) that we see in events like GW231123 (a recent gravitational wave detection involving a very heavy black hole).

The Results in Plain English

The team ran thousands of computer simulations of these star clusters:

• Retention: They found that with the new model, many more black holes stay inside the cluster after merging.
• Growth: Because they stay, they can merge again. This leads to a population of much heavier black holes than we previously thought possible.
• The "GW231123" Connection: The new model predicts that the specific heavy black holes we saw in the recent GW231123 event are actually quite common in these clusters. The old model struggled to explain how such heavy objects could form without getting kicked out.

The Takeaway

Imagine you were trying to build a tower of blocks, but you were told the wind was so strong it would blow the tower over after every two blocks. You'd stop building.

This paper says, "Wait, the wind isn't that strong! It's actually a gentle breeze."

Because the "wind" (the recoil kick) is gentler than we thought, black holes can stack up on top of each other, merging over and over again to create the cosmic giants we are now starting to detect. It's a new, more hopeful (and more violent) story about how the universe builds its heaviest monsters.

Technical Summary

Here is a detailed technical summary of the paper "Kick matters: The impact of a new recoil model on the retention of hierarchical black-hole remnants in globular clusters" by Islam, Wadekar, and Kritos.

1. Problem Statement

The formation of massive black holes (BHs), particularly those in the intermediate-mass range ($>100 M_\odot$) and the upper mass gap, is a leading hypothesis for the origin of gravitational wave (GW) events like GW190521 and GW231123. A primary mechanism for this growth is hierarchical merging, where merger remnants (2G, 3G, etc.) remain trapped in dense stellar environments (globular clusters) and undergo further mergers.

The critical bottleneck for this process is the gravitational recoil kick imparted to the remnant BH during a merger. If the kick velocity ($v_{kick}$) exceeds the cluster's escape velocity ($v_{esc}$), the remnant is ejected, halting the hierarchical chain.

The Core Issue: Current population synthesis and N-body simulations almost universally rely on the HLZ analytic model (Lousto, Zlochower, et al.) to estimate recoil kicks. This model was developed years ago using a relatively small dataset of numerical relativity (NR) simulations, heavily biased toward equal-mass, aligned-spin binaries. The authors argue that the HLZ model systematically overestimates kick velocities in many regions of the parameter space (particularly for asymmetric mass ratios and low spins), leading to an artificial suppression of predicted hierarchical merger rates and the retention of massive BHs.

2. Methodology

The authors employ a multi-tiered approach to compare the legacy HLZ model against their new state-of-the-art model, gwModel_flow_prec (developed in a companion paper, IW25).

• The New Model (gwModel_flow_prec):

  • Built on a significantly larger dataset combining NR simulations and Black Hole Perturbation Theory (BHPT).
  • Utilizes normalizing flows (a deep learning technique) to capture the complex, non-Gaussian distribution of kick velocities arising from varying spin orientations.
  • Incorporates post-Newtonian (PN) structural insights.
  • Designed to be computationally efficient enough for large-scale cluster simulations.

• Simulation Frameworks:

  1. Semi-Analytical "Back-of-the-Envelope" Calculations: Using the custom gwGenealogy package to rapidly test retention probabilities across varying initial mass functions (IMF), spin distributions, and binary pairing functions.
  2. Detailed Cluster Simulations:
     • Post-processing: Re-analyzing existing catalogs from the CMC (Cluster Monte Carlo) and colossus codes. These simulations self-consistently model cluster evolution, stellar dynamics, and binary hardening.
     • New Simulations: Running $\sim7,500$ new simulations using rapster, a fast semi-analytical cluster evolution code. rapster evolves global cluster properties (density, velocity dispersion) and treats dynamical channels as Poisson processes, allowing for massive statistical sampling of different cluster initial conditions (mass, radius, metallicity).

• Metrics:

  • Jensen-Shannon Divergence (JSD): Used to quantify the statistical difference between the kick velocity distributions predicted by HLZ and gwModel_flow_prec.
  • Retention Probability: The fraction of merger remnants retained ($v_{kick} < v_{esc}$).
  • IMBH Formation Rate: The probability of forming BHs with masses $\ge 100 M_\odot$, $150 M_\odot$, and$250 M_\odot$.

3. Key Contributions

1. Quantification of Model Discrepancy: The paper demonstrates that the HLZ model and gwModel_flow_prec yield significantly different kick distributions, particularly for asymmetric mass ratios ($q \gtrsim 2$) and low-spin binaries. The JSD between the models is substantial ($>0.1$) in these regimes.
2. Systematic Overestimation by HLZ: The authors confirm that HLZ tends to predict higher kick velocities than the more accurate NR-informed model, leading to an underestimation of BH retention in previous studies.
3. Unified Framework: Introduction of gwGenealogy and the integration of gwModel_flow_prec into rapster and existing N-body catalogs, providing a reproducible pipeline for future hierarchical merger studies.

4. Key Results

A. Retention Probabilities

• Low Escape Velocities ($v_{esc} \lesssim 200$ km/s): The difference is most pronounced here. For $v_{esc} = 50$ km/s, HLZ predicts near-zero retention for many configurations, whereas gwModel_flow_prec predicts significant retention (up to 40% for specific mass ratios and low spins).
• High Escape Velocities ($v_{esc} \approx 500$ km/s): Differences diminish as nearly all remnants are retained regardless of the model.
• Cluster Properties: In clusters with $v_{esc}$ between 30–300 km/s (typical of many globular clusters), the new model predicts a significantly larger fraction of retained 2G and 3G BHs.

B. Intermediate-Mass Black Hole (IMBH) Formation

• Increased Efficiency: Using gwModel_flow_prec increases the cumulative probability of forming IMBHs ($\ge 100 M_\odot$) by approximately 5–8% compared to HLZ across various metallicities and escape velocities.
• Massive Remnants: The new model allows for the formation of BHs up to $\sim 280 M_\odot$ in high-escape velocity clusters, whereas HLZ often fails to retain the necessary progenitors to reach these masses.
• GW231123 Connection: The new model populates the mass-spin parameter space consistent with the progenitors of GW231123 (a massive binary event) much more effectively than HLZ, which struggles to produce such massive, retained remnants in standard cluster scenarios.

C. Mass and Spin Distributions

• Generational Evolution: Across 5 generations of mergers, gwModel_flow_prec consistently produces a larger population of high-mass BHs.
• Spin Effects: The retention of low-spin, asymmetric binaries is significantly higher under the new model. Since low-spin binaries are more common in certain formation channels, this drastically alters the final mass and spin distribution of the hierarchical population.
• Pairing Functions: The study confirms that "selective pairing" (mass-weighted) combined with the new kick model yields the most massive BHs, but the magnitude of the improvement is driven by the kick model.

5. Significance and Implications

• Re-evaluation of Hierarchical Channels: The paper suggests that hierarchical mergers are a more efficient pathway for forming massive black holes than previously thought. The "runaway" formation of IMBHs via successive mergers is more viable in globular clusters when using accurate recoil models.
• Interpretation of GW Events: The results provide a stronger theoretical basis for interpreting events like GW190521 and GW231123 as products of hierarchical mergers in dense clusters, rather than requiring exotic formation channels (e.g., primordial black holes or AGN disks).
• Future Modeling: The authors advocate for the immediate adoption of gwModel_flow_prec (or similar NR-informed models) in all major population synthesis codes (e.g., BPOP, fastcluster, CMC, MOCCA). They also propose distilling the normalizing flow model into a closed-form analytic expression to reduce computational overhead in future simulations.

Conclusion: The "kick" is a decisive factor in the survival of merger remnants. By correcting the systematic overestimation of kicks in the widely used HLZ model, this work revitalizes the hierarchical merger scenario as a primary mechanism for building the massive black holes observed by LIGO/Virgo/KAGRA.