Guesswork in the gap: the impact of uncertainty in the compact binary population on source classification

This study analyzes 66 gravitational wave events to demonstrate that the probability of classifying compact objects as neutron stars is highly sensitive to population model assumptions—particularly pairing preferences and spin distributions—rather than just measurement noise or equation of state constraints, leading to significant classification uncertainties for events like GW230529 and GW190425.

The Gist

The Big Picture: The "Missing Link" in the Cosmic Zoo

Imagine the universe as a giant zoo of heavy objects. We have two main types of heavyweights: Neutron Stars (the "small" heavyweights, about the size of a city but as heavy as the Sun) and Black Holes (the "giant" heavyweights, infinitely dense and invisible).

For a long time, astronomers thought there was a clear fence between them. They believed there was a "gap" in mass where no objects existed. It was like a zoo where you have a row of lions, then a big empty space, and then a row of elephants. You never saw a "mini-elephant" or a "super-lion."

But recently, gravitational wave detectors (like LIGO and Virgo) started hearing "thumps" from colliding objects that landed right in that empty space. Events like GW190814 and GW230529 are like finding a mysterious creature in the gap. Is it a giant lion? Or a baby elephant?

The Problem: The "Guesswork" in the Title

The paper asks a simple but tricky question: How do we know what these creatures are?

Usually, you'd just weigh them. If it's under a certain weight, it's a Neutron Star. If it's over, it's a Black Hole. But these cosmic collisions are messy. The data is fuzzy (like trying to weigh a cat while it's running on a shaky scale).

The authors of this paper realized that our guess isn't just based on the scale (the data). It's heavily influenced by what we expect to see (our "population model").

The Analogy: The "Crowded Room" Detective

Imagine you walk into a crowded room and see a person standing alone in a corner. You don't know their height. You have to guess if they are a child or an adult.

• The Data (The Scale): You can see them, but the lighting is bad. They look like they might be 5 feet tall, or maybe 6 feet.
• The Population Model (The Crowd): This is the key.
  • Scenario A: If you know this room is filled only with professional basketball players, and you see someone who looks 5'10", you might guess, "That's a short basketball player."
  • Scenario B: If you know this room is filled only with kindergarteners, and you see someone who looks 5'10", you might guess, "That's a very tall kid, or maybe an adult who snuck in."

The paper argues that our guess about the person's identity changes completely depending on what we think the rest of the crowd looks like.

The Three Main "Guessing Games"

The authors analyzed 66 confirmed cosmic collisions to see how much our "crowd assumptions" change the answer. They found three main things that mess with our classification:

1. The "Date Night" Preference (Pairing)

• The Concept: Do Neutron Stars like to pair up with other Neutron Stars of similar size, or do they like to pair with much heavier Black Holes?
• The Analogy: Imagine a dance floor.
  • If Neutron Stars are "matchy-matchy" lovers (they only dance with partners of the same size), and we see a couple where one partner is heavy and the other is light, we might guess the heavy one is actually a "light" Black Hole, because a "heavy" Neutron Star wouldn't dance with such a light partner.
  • If Neutron Stars are "opposites attract" lovers, the same heavy object might be classified as a Neutron Star because it's perfectly happy dancing with a light partner.
• The Result: Changing this "dating preference" in the math changed the probability of an object being a Neutron Star by up to 67% for some events!

2. The "Spinning Top" Effect (Spin)

• The Concept: Neutron stars spin. Some spin slowly; some spin like a top. Fast spinning can make a heavy object stay stable as a Neutron Star instead of collapsing into a Black Hole.
• The Analogy: Think of a figure skater. If they spin fast, they can balance on one toe (stay a Neutron Star). If they stop spinning, they fall over (become a Black Hole).
• The Result: If we assume these objects spin wildly, we are more likely to call a heavy object a Neutron Star. If we assume they spin slowly, we call it a Black Hole. The authors found that our uncertainty about how fast they spin creates huge swings in our classification.

3. The "Hard Limit" (Equation of State)

• The Concept: There is a theoretical maximum weight a Neutron Star can hold before it collapses. This depends on the "Equation of State" (EOS)—basically, how squishy or hard the matter inside is.
• The Analogy: Imagine a bridge. If the bridge is made of steel, it can hold a truck. If it's made of jelly, it collapses under a bicycle. We don't know exactly what the "bridge" inside a Neutron Star is made of.
• The Result: If we assume the bridge is super strong (hard matter), we can classify heavier objects as Neutron Stars. If we assume it's weak, we have to call them Black Holes.

The Good News and the Bad News

The Bad News:
For the "fuzzy" events (like GW230529), the answer is not just "it's a Black Hole." The answer is "It depends on what you think the rest of the universe looks like."

• Under one set of assumptions, GW230529's heavy partner is a 1% chance of being a Neutron Star.
• Under a different set of assumptions, it's a 67% chance.
  This means we are currently "guessing in the gap."

The Good News:
For the "loud and clear" events (like GW190814), the data is so strong that the assumptions don't matter as much.

• GW190814's small partner is almost certainly a Neutron Star (or a very light Black Hole) regardless of the "crowd" assumptions. The signal was so loud (high Signal-to-Noise Ratio) that the "scale" overpowered the "guesswork."

The Conclusion: What's Next?

The paper concludes that we cannot trust our classifications of these "gap" objects until we have more data.

• Right now: We are like detectives trying to solve a crime with a blurry photo and a hunch about the suspect's friends.
• In the future: As we detect more collisions, we will learn the "rules of the crowd" (the population model). We will know exactly how Neutron Stars pair up and how they spin. Once we know the rules, the "guesswork" will disappear, and we will finally know exactly what these mysterious gap creatures are.

In short: The paper warns us not to be too confident in our current labels for these mysterious objects. Our labels are currently more a reflection of our assumptions than the reality of the objects themselves. We need more data to stop guessing and start knowing.

Technical Summary

1. Problem Statement

The nature of compact objects within the "lower mass gap" (approximately $2.5 - 5 M_\odot$) remains uncertain. Gravitational wave (GW) events such as GW190814, GW230529, and GW190425 contain components with masses consistent with both neutron stars (NS) and black holes (BH).

Classifying these objects is challenging because the probability that a component is a neutron star, denoted as $P(\text{NS})$, does not depend solely on the measurement noise of the specific event. Instead, it is heavily influenced by:

1. Population Priors: Assumptions about the underlying distribution of masses, spins, and pairing preferences of compact binaries.
2. Equation of State (EOS) Constraints: Theoretical limits on the maximum mass ($m_{\text{max}}$) and spin ($\chi_{\text{max}}$) a neutron star can support.
3. Degeneracies: Correlations between mass ratio ($q$), component masses ($m_1, m_2$), and effective spin ($\chi_{\text{eff}}$) in the GW waveform, which are particularly pronounced in low Signal-to-Noise Ratio (SNR) events.

The paper investigates how varying assumptions about the astrophysical population affects the classification of ambiguous events, quantifying the reliability of $P(\text{NS})$ as a metric.

2. Methodology

The authors employ a Hierarchical Bayesian Inference framework to analyze 66 confident compact binary coalescence (CBC) events from the GWTC-3 catalog.

• Population Model: They utilize the FullPop-4.0 model (and extensions), which describes the joint mass, spin, and redshift distributions. Key features include:
  • A broken power-law mass distribution with a "notch" (dip) representing the mass gap.
  • Gaussian peaks for NS and BH populations.
  • Mass-dependent spin magnitude and tilt distributions (separating NS and BH regimes at $3 M_\odot$).
  • A pairing function ($\beta_{LL}$) governing the preference for equal-mass binaries (BNS) versus unequal-mass systems.
• Classification Framework:
  • Population-Only: $P(m < \gamma_{\text{low}})$, where $\gamma_{\text{low}}$ is the inferred lower edge of the mass gap.
  • EOS-Informed: $P(m < m_{\text{max}}^{\text{EOS}}(\chi))$, where the boundary is determined by the maximum mass supported by a specific Equation of State, accounting for rotational support.
• Analysis Strategy:
  • The authors perform a sensitivity analysis by fixing specific hyperparameters (e.g., spin tilt mean, pairing preference) to extreme values while sampling the remaining parameters from their posterior distributions.
  • They reweight the posterior samples of individual events (like GW230529 and GW190425) to calculate how $P(\text{NS})$ shifts under different population assumptions.
  • They explicitly model the coupling between the mass ratio ($q$) and effective spin ($\chi_{\text{eff}}$) to understand how spin priors drive mass inferences.

3. Key Contributions

• Quantification of Systematic Uncertainty: The paper provides the first comprehensive mapping of how specific population hyperparameters (pairing, spin distributions, mass peak width) propagate to event-level classification probabilities.
• Identification of Dominant Drivers: It identifies that neutron star pairing preferences ($\beta_{LL}$) and spin tilt distributions ($\mu_{\text{low}}^{\cos\theta}$) are the dominant drivers of classification uncertainty, often outweighing the impact of the EOS itself.
• Re-evaluation of GW230529: The authors correct a discrepancy in the original GW230529 discovery paper (Abac et al.) regarding the implementation of population marginalization, showing that the primary's NS probability is even more sensitive to population assumptions than previously reported.
• Robustness of High-SNR Events: They demonstrate that high-SNR, asymmetric systems (like GW190814) are robust to population assumptions, whereas low-SNR, ambiguous systems are highly sensitive.

4. Key Results

The study reveals substantial variations in $P(\text{NS})$ depending on the chosen population model:

• GW230529 (Primary):
  • Under Population-Only analysis, $P(\text{NS})$ varies drastically between 1% and 67%.
  • Under EOS-Informed analysis, it varies between 0% and 21%.
  • Driver: The variation is primarily driven by the pairing preference ($\beta_{LL}$) and spin tilt ($\mu_{\text{low}}^{\cos\theta}$). Strong pairing (favoring equal masses) pulls the primary mass down, increasing $P(\text{NS})$.
• GW190425 (Primary):
  • Under EOS-Informed analysis, $P(\text{NS})$ spans 51% to 100%.
  • This event is highly sensitive to the width of the NS mass peak and spin magnitude priors.
• GW190814 (Secondary):
  • $P(\text{NS})$ varies by $\le 10\%$ across all models.
  • Reason: The high SNR and extreme mass ratio break the $q-m$ degeneracies, making the classification robust against population priors.
• Spin vs. Mass:
  • Variations in the spin tilt distribution ($\mu_{\text{low}}^{\cos\theta}$) have a comparable or stronger effect on $P(\text{NS})$ than variations in spin magnitude.
  • In EOS-informed analyses, the classification is driven more by the inferred maximum mass ($m_{\text{max}}$) than by the maximum spin support.
• Non-Parametric Validation: The authors tested a non-parametric edge-detection algorithm and found it yielded results consistent with their parametric model, validating the robustness of their findings regarding the mass gap structure.

5. Significance and Implications

• Ambiguity in Current Classifications: The paper concludes that for low-SNR events near the mass gap, $P(\text{NS})$ is not a definitive physical truth but a model-dependent statistic. Current data cannot tightly constrain the population parameters (like pairing and spin tilts) necessary to resolve these ambiguities.
• Need for Uncertainty Budgets: Future GW catalogs must report classification metrics (like $P(\text{NS})$) with an explicit uncertainty budget that surveys a wide range of population priors, rather than relying on a single "best-fit" model.
• Future Observing Runs: As the number of detections increases, particularly high-SNR and asymmetric mergers, the population distributions will be better constrained. This will naturally reduce the variability in $P(\text{NS})$ and allow for more confident identification of the lower mass gap's true nature.
• Methodological Shift: The work advocates for a shift from treating population models as fixed backgrounds to treating them as active, uncertain variables that must be marginalized over when interpreting individual events.

In summary, the paper argues that "guesswork" regarding the astrophysical population currently dominates the classification of ambiguous GW events, and resolving the nature of the lower mass gap requires both better population constraints and higher SNR detections.