The impact of waveform systematics and Gaussian noise on the interpretation of GW231123

This study demonstrates that the interpretation of the gravitational-wave event GW231123 as a merger of massive, highly-spinning black holes is robust against waveform model systematics and Gaussian detector noise, confirming that its key properties, particularly high mass and spin magnitudes, are reliably inferred using the NRSur7dq4 model.

The Gist

The Big Picture: A Cosmic Mystery

Imagine the universe is a giant, dark room, and a massive event just happened: two incredibly heavy black holes smashed together. This event, named GW231123, sent ripples through space-time called gravitational waves.

Scientists caught a glimpse of these ripples using giant detectors (LIGO). But here's the problem: the signal was very short and faint, like hearing a single, sharp clap in a noisy room. Because the signal was so brief, scientists were using different "translation manuals" (mathematical models) to figure out what the black holes looked like.

The Conflict: When they used different manuals, they got very different answers.

• Manual A said: "These black holes are huge and spinning incredibly fast."
• Manual B said: "Actually, they might be smaller, and maybe not spinning as fast."
• Manual C said: "They are far away and facing sideways."

This disagreement made scientists nervous. Was the universe actually weird, or were the "manuals" broken? This paper investigates whether the disagreement is real or just an illusion caused by noise and bad math.

---

The Investigation: Three Main Tests

The authors ran a series of computer experiments to see if they could reproduce these confusing results. Think of it like a detective trying to figure out if a witness is lying or if the lighting just made them look different.

1. The "Perfect Signal" Test (Waveform Systematics)

The Analogy: Imagine you are trying to identify a song by listening to a very short, distorted clip. You have three different apps that try to guess the song. One app says it's a rock song, another says it's jazz. You wonder: Is the song actually both? Or are the apps just bad at guessing?

What they did:
Instead of using random guesses, the authors took the "best guess" version of the signal (the one that fit the data best) and played it back into a computer with zero noise. It was a perfect, clean signal.

The Result:
Even with a perfect signal and no noise, the different apps (models) still gave different answers.

• The Takeaway: The disagreement isn't because the signal is messy; it's because the mathematical "manuals" themselves have built-in differences. However, the authors found that one specific manual (NRSur) matched the "perfect signal" best. When they used that manual, the results were consistent.

2. The "Static Noise" Test (Gaussian Noise)

The Analogy: Now, imagine you are trying to hear that same song, but you turn on a radio with static. Sometimes the static makes the song sound like a drum solo; other times it sounds like a flute. Does the static change the song? No, but it changes what you think the song is.

What they did:
The authors took that same "perfect signal" and added 20 different types of random static (simulating the real noise in the detectors). They ran the analysis again and again.

The Result:

• Mass and Spin: Even with the static, the "NRSur" manual consistently said: "These black holes are heavy and spinning very fast." The noise made the numbers wiggle a little, but it never changed the main story.
• The "Other" Manuals: The other manuals (XPHM and XO4a) were more confused by the noise. They sometimes guessed the black holes were smaller or spinning slower.
• The Takeaway: The most important conclusion—that these black holes are massive and spinning wildly—is robust. It survives the static. The confusion comes from a mix of the noise and the flaws in the other mathematical models.

3. The "Two Ears" Test (Single-Detector Inference)

The Analogy: You have two ears (LIGO Hanford and LIGO Livingston). Sometimes, if a loud truck drives by near your left ear, your left ear hears a different sound than your right ear. Scientists worried that for GW231123, the two detectors were hearing different things, suggesting one detector might be broken or the data was bad.

What they did:
They simulated the signal and listened to it with "Left Ear" only, then "Right Ear" only, using random static.

The Result:
They found that even with perfect, clean data, random static often makes the two ears hear slightly different things. The differences seen in the real GW231123 event were not unusual. They are exactly what you would expect from normal static noise.

• The Takeaway: There is no evidence that the data is "broken" or that the detectors are malfunctioning. The slight differences between the two detectors are just normal statistical noise.

---

The Verdict: What is Real?

The paper concludes that the "weird" nature of GW231123 is real, not an illusion.

1. The Black Holes are Massive: They likely fall into a "mass gap" (a range of weights where black holes are usually thought not to exist).
2. They are Spinning Fast: They are rotating at extreme speeds.
3. The Confusion is Math, Not Physics: The reason scientists argued about the details was because they were using different mathematical tools. One tool (NRSur) is the most accurate for this specific type of signal.

The Future: Better Ears

The paper ends by looking at the future. Currently, our "ears" (LIGO) are a bit fuzzy. But in the mid-2030s, there is a planned upgrade called LIGO A#.

The Analogy: Imagine upgrading from a cheap, crackly radio to a high-fidelity studio microphone.

• Now: We can guess the song is "fast and loud," but we aren't sure of the exact notes.
• With LIGO A#: We will hear the song perfectly. The uncertainty about the black holes' mass and spin will shrink from a wide guess to a precise measurement.

Summary in One Sentence

This paper proves that the strange, heavy, fast-spinning black holes detected in GW231123 are real and not just a trick of the math or the noise, and that future upgrades to our detectors will let us hear them with crystal-clear precision.

Technical Summary

1. Problem Statement

The gravitational-wave event GW231123 is an exceptional signal consistent with the merger of two massive, highly-spinning black holes (BHs). Its properties challenge current astrophysical formation models:

• Mass Gap: The total mass ($190\text{--}265 M_\odot$) places the primary BH within or beyond the pair-instability mass gap ($\sim60\text{--}130 M_\odot$), while the secondary likely spans the entire gap.
• High Spins: Component spins are inferred to be extremely high ($\sim0.9$), which contradicts standard isolated binary formation models predicting low natal spins.

The Core Challenge:
Reliable inference of these source properties is complicated by two main factors:

1. Waveform Systematics: Different waveform models (e.g., NRSur7dq4, IMRPhenomXPHM, IMRPhenomXO4a) yield systematically different posterior distributions for the same event. For GW231123, these differences are significant, with some models suggesting edge-on systems at close distances and others face-on systems at far distances.
2. Data Limitations: The signal is merger-dominated with very few waveform cycles in the sensitive frequency band, meaning spin inference may rely on as little as a single cycle.
3. Noise Sensitivity: It is unclear whether the observed discrepancies are due to intrinsic model limitations, specific realizations of Gaussian detector noise, or unmodeled non-Gaussian noise (glitches).

The authors investigate whether the astrophysical conclusions (high mass, high spin) are robust against model systematics and Gaussian noise, or if the event's interpretation is an artifact of statistical fluctuation or modeling errors.

2. Methodology

The authors employed a comprehensive simulation campaign to isolate the effects of noise and waveform modeling:

• Signal Simulation:

  • Instead of using generic numerical relativity (NR) simulations (which may not match the specific morphology of GW231123), they generated simulated signals using the maximum-likelihood waveforms derived from the actual GW231123 data analysis.
  • They utilized the maximum-likelihood waveforms from two distinct models: NRSur7dq4 (a time-domain surrogate model) and IMRPhenomXO4a (a frequency-domain phenomenological model).
  • Despite the models inferring different source parameters, their maximum-likelihood waveforms were found to be morphologically very similar.

• Noise Realizations:

  • Zero-Noise Case: Simulated the signal with no noise to isolate pure waveform systematics.
  • Gaussian Noise: Injected the signal into 20 different realizations of Gaussian noise based on the Power Spectral Density (PSD) observed during the GW231123 event.
  • Single-Detector Analysis: Analyzed the simulated data using only LIGO Hanford (H) or LIGO Livingston (L) data to compare against the single-detector results observed for the real event.

• Inference Framework:

  • Recovered parameters using three waveform models: NRSur, XPHM, and XO4a.
  • Used the Jensen-Shannon (JS) divergence to quantitatively measure the similarity between posterior distributions across different models and noise realizations.
  • Analyzed the Bayes factors to understand why different models are preferred despite similar likelihoods.

• Future Sensitivity:

  • Simulated the signal with the projected sensitivity of LIGO A# (mid-2030s upgrade) to assess future measurement precision.

3. Key Contributions

1. Reproduction of Systematics: The study successfully reproduced the large waveform systematics observed in GW231123 using simulated signals in a zero-noise environment. This was achieved by using the maximum-likelihood waveform from the NRSur analysis, contrasting with previous studies that failed to reproduce systematics using generic NR simulations.
2. Decoupling Noise and Systematics: The authors demonstrated that Gaussian noise can either amplify or suppress waveform systematics depending on the specific noise realization.
3. Bayes Factor Interpretation: They clarified that the preference for the XO4a model over NRSur in the original GW231123 analysis (Bayes factor $\sim140:1$) is driven by the prior volume term (less constrained posteriors) rather than a significantly better fit to the data (likelihoods were nearly identical).
4. Single-Detector Consistency: They showed that the differences observed between LIGO Hanford and Livingston inferences for GW231123 are statistically consistent with expected fluctuations from Gaussian noise alone, ruling out data quality issues as the primary cause.

4. Key Results

• Robustness of High Mass and Spin (NRSur):

  • When using the NRSur model, the inference of high component masses (within or above the mass gap) and high spin magnitudes ($\chi_1 \ge 0.7$, $\chi_2 \ge 0.8$) is robust across all 20 Gaussian noise realizations.
  • The effective aligned spin ($\chi_{\text{eff}}$) fluctuates significantly, but the effective precession spin ($\chi_p$) consistently supports high precession ($\chi_p \ge 0.68$).

• Model Dependence:

  • XPHM and XO4a models tend to underestimate the secondary mass and show larger fluctuations in spin parameters when combined with Gaussian noise.
  • However, even with these models, the primary spin magnitude remains high.

• Systematics vs. Noise:

  • In $\sim30\%$ of noise realizations, the difference between single-detector posteriors (H vs. L) was larger than what was observed in the actual GW231123 data. This confirms that the observed detector differences are not anomalous.
  • The degree of systematics observed in the real event is not exceptional; similar or larger discrepancies can be generated purely from Gaussian noise acting on a signal described by existing waveforms.

• Future Prospects (LIGO A#):

  • With the proposed LIGO A# sensitivity, the Signal-to-Noise Ratio (SNR) for a GW231123-like event would increase from $\sim20$ to $\sim100$.
  • This would reduce mass uncertainty from $\sim8\%$ to $\sim2\%$ and primary spin uncertainty from $\sim37\%$ to $\sim6\%$, allowing for definitive characterization of such extreme systems.

5. Significance

This paper provides a critical validation of the astrophysical interpretation of GW231123. By demonstrating that the extreme properties (massive, high-spin black holes) are robust against Gaussian noise and that the observed model discrepancies are consistent with known waveform systematics, the authors conclude that GW231123 is likely a genuine astrophysical object rather than a statistical fluke or an artifact of noise.

The findings support the hypothesis that GW231123 originated from hierarchical mergers (mergers of black holes formed from previous mergers) in dense environments, as this is the only scenario capable of producing such high masses and spins. The study also highlights the limitations of current waveform models and the necessity of future detector upgrades (LIGO A#) to resolve remaining ambiguities in the parameter space of extreme mass ratio and high-spin binaries.