Systematic bias in LISA ringdown analysis due to waveform inaccuracy

This paper demonstrates that accurate parameter estimation for massive black hole binaries in the LISA band requires ringdown templates to include a specific minimum number of modes (ranging from 3–6 for typical systems to over 10 for high-SNR events) to prevent systematic biases caused by waveform truncation.

The Gist

Imagine the universe is a giant, silent concert hall. For decades, we've been listening to the "music" of colliding black holes using detectors like LIGO. But soon, a new, much more sensitive instrument called LISA (a space-based detector) will join the orchestra. LISA will be able to hear the "ringing" of massive black holes after they smash together, a sound known as the ringdown.

This paper is essentially a recipe guide for how to listen to that ringdown without getting the recipe wrong.

Here is the breakdown of what the authors are doing, using some everyday analogies:

1. The Problem: The "Broken Piano" Effect

When two black holes merge, they don't just stop; they vibrate like a struck bell. This vibration is made up of many different "notes" (frequencies) fading away at different speeds. In physics, we call these notes Quasi-Normal Modes (QNMs).

• The Analogy: Imagine hitting a piano key. You hear the main note (the fundamental tone), but you also hear a complex mix of higher-pitched harmonics and overtones that fade out quickly.
• The Issue: If you try to describe that sound using only the main note, you get a rough approximation. But if you want to know exactly which piano was hit, how hard it was struck, and where it is in the room, you need to hear the subtle harmonics too.
• The Risk: If your model (the sheet music you use to analyze the sound) ignores too many of these harmonics, you might think you're listening to a Steinway when it's actually a Yamaha. This leads to systematic bias—a consistent error in your measurements that doesn't go away even if you listen longer.

2. The Solution: How Many Notes Do We Need?

The authors asked a crucial question: "How many of these 'notes' (modes) do we need to include in our model to get the answer right?"

• Too few notes: You get a blurry picture. Your measurements of the black hole's mass and spin will be wrong.
• Too many notes: You might start "overfitting." This is like trying to explain a simple melody by adding random, made-up notes just to make the math look complex. It confuses the computer and makes the analysis unstable.

3. The Experiment: The "Mode Hierarchy"

The team created a "perfect" model containing 13 different notes (a mix of the main tones and the higher overtones). Then, they tested what happened if they only used the top 1, top 2, top 3, etc., of the loudest notes.

They found that the answer depends on how "loud" the event is:

• For faint, distant events: You only need to hear the 3 to 6 loudest notes to get a good idea of what happened.
• For very loud, close events: Because LISA will be so sensitive, it will hear the faint whispers of the higher notes. To avoid errors here, you might need to include at least 10 notes in your model.

4. The "Window" Problem (Spectral Leakage)

One of the trickiest parts of this paper is dealing with how we cut the sound off.

• The Analogy: Imagine trying to record a bell ringing, but you have to stop the recording abruptly. If you just hit "stop" (a sharp cut), it creates a weird "click" or "hiss" in the audio file that sounds like static. In physics, this is called spectral leakage. It makes the loud notes bleed into the quiet notes, contaminating the data.
• The Fix: The authors used a clever mathematical trick (called "mirroring") to smooth out the start of the recording, like fading the volume in rather than snapping it on. They also set a "high-frequency cutoff" (a speed limit for the notes they analyze) to ensure that the "hiss" from the loud notes doesn't mess up the quiet ones.

5. The Takeaway: Why This Matters

The authors conclude that as LISA gets ready to launch, we can't just use a "one-size-fits-all" model.

• For typical black hole mergers (which happen far away in the early universe), we need to include 3 to 6 modes to avoid lying to ourselves about the black hole's properties.
• For the loudest, closest mergers, we need to be even more careful and include up to 10 modes.

In simple terms: This paper tells the future LISA team, "Don't just listen to the bass drum of the black hole collision. You need to tune your ears to the cymbals and the violins too, or you'll think the whole orchestra is out of tune."

By figuring out exactly how many "notes" are needed, they ensure that when LISA finally hears the universe ring, we will know exactly what kind of black holes made that sound.

Technical Summary

Here is a detailed technical summary of the paper "Systematic bias in LISA ringdown analysis due to waveform inaccuracy" by Capuano et al.

1. Problem Statement

As gravitational wave (GW) detector sensitivities improve, particularly with the upcoming Laser Interferometer Space Antenna (LISA), the statistical uncertainties in parameter estimation will decrease significantly. However, this increased precision exposes systematic biases introduced by inaccurate waveform modeling.

The specific problem addressed is the truncation of Quasi-Normal Modes (QNMs) in ringdown templates. The ringdown signal of a merging massive black hole binary (MBHB) is theoretically a superposition of exponentially damped sinusoids (QNMs). In practice, templates often include only a few dominant modes to avoid overfitting. The authors investigate:

• How many modes are required to ensure that systematic errors (caused by missing modes) are smaller than statistical errors (caused by detector noise).
• How this requirement varies across the LISA parameter space (mass, redshift, spin, mass ratio).
• The impact of spectral leakage caused by time-domain windowing on frequency-domain analysis.

2. Methodology

A. Waveform Modeling

• Reference Signal ($h_{GR}$): The authors construct a "true" fiducial waveform using a frequency-domain model containing 13 modes:
  • 11 Linear QNMs (fundamental modes $n=0$ and first overtones $n=1$).
  • 2 Quadratic QNMs (non-linear interactions of linear modes).
  • Modes are selected based on perturbation theory order, overtone number, and rotation (prograde/retrograde).
• Approximate Templates ($h_{AP}^{(N)}$): They generate approximate templates by including only the top $N$ modes, ranked by their single-mode Signal-to-Noise Ratio (SNR).
• Time/Frequency Domain Handling:
  • To mitigate spectral leakage (artifacts from sharp time-domain cutoffs), they employ a mirroring technique at the start time $t_0$ (set to $20M$ after the luminosity peak) rather than a standard Heaviside window.
  • They apply a high-frequency cutoff ($f_{cut}$) based on the PhenomA model. This ensures the analysis is insensitive to the specific choice of time window, effectively providing a conservative lower bound on the required modes.

B. Error Assessment Framework

The authors use the Linear Signal Approximation combined with the Fisher Information Matrix (FIM) to estimate errors:

1. Systematic Error ($\Delta^{(Sys)}\theta$): Calculated as the bias in best-fit parameters when the approximate template is matched against the true signal.
2. Statistical Error ($\Delta^{(St)}\theta$): Derived from the inverse Fisher matrix, representing the uncertainty due to detector noise.
3. Accuracy Criteria: A template is deemed accurate if:
   • Criterion 1: Systematic error is smaller than statistical error for all parameters ($|\Delta^{(Sys)}\theta_i| < \Delta^{(St)}\theta_i$).
   • Criterion 2: The mismatch ($1-M$) is below an SNR-dependent threshold ($1-M < D/2\rho^2$, where$D$is the number of parameters and$\rho$ is the SNR).

C. Parameter Space Exploration

The study scans a wide range of MBHB parameters:

• Masses: $10^5$to$10^8 M_\odot$.
• Redshifts: $z \sim 0$ to $z \sim 10$.
• Spins and Mass Ratios: Various configurations of progenitor spins ($a_{1,2}$) and mass ratios ($q$).
• Orientation: Averaged over sky localization, inclination, and polarization angles.

3. Key Contributions

1. Quantification of Mode Hierarchy: The paper establishes a rigorous hierarchy of QNMs based on SNR, demonstrating that the "loudest" modes depend heavily on the mass ratio and spin. Notably, overtone modes ($n=1$) and retrograde modes become significantly more important for high-spin and mass-asymmetric systems.
2. Window-Insensitive Lower Bound: By utilizing a high-frequency cutoff and mirroring technique, the authors derive a window-insensitive lower bound for the number of modes ($N_{min}$). This means their results represent the minimum modes required regardless of specific time-windowing choices, avoiding overestimation due to spectral leakage artifacts.
3. Validation of Approximations: The authors validate the linear signal approximation by checking the dephasing between the true and approximate waveforms. They confirm that for $N \ge 3$, the dephasing is small enough for the linear approximation to hold, ensuring the robustness of their error estimates.
4. Comparison of Methods: They cross-verify results using two independent criteria (error comparison vs. mismatch threshold) and compare frequency-domain cutoffs with phenomenological tapering and time-domain analyses, finding consistent trends.

4. Key Results

• Minimum Modes Required ($N_{min}$):
  • Typical Systems ($M \sim 10^6 - 10^7 M_\odot$, $z \sim 2-6$): At least 3 to 6 modes are required to avoid systematic biases.
  • High-SNR / Low-Redshift Systems ($z < 1$): For the loudest events (e.g., nearby massive binaries), the requirement increases to 8 to 10 modes.
  • High-Mass / Low-SNR Systems: For very massive ($M > 10^8 M_\odot$) or distant sources where the signal is fainter, fewer modes (often 3–5) suffice because the statistical errors dominate, masking the systematic bias of missing modes.
• Dependence on System Parameters:
  • Mass Ratio: Asymmetric systems ($q < 1$) excite odd-$\ell$ modes and overtones more strongly, often requiring more modes than symmetric systems.
  • Spin: High spins enhance the contribution of retrograde modes and overtones, increasing the complexity of the required template.
  • Redshift: Higher redshift leads to lower SNR, reducing the number of modes needed to keep systematic errors sub-dominant to statistical errors.
• Mode Ordering: The ranking of modes by SNR is not static; it shifts based on the specific binary configuration. For example, the $331$overtone can become louder than the fundamental$330$ mode in highly spinning systems.

5. Significance

• LISA Science Readiness: The results provide a concrete guideline for the LISA data analysis pipeline. To perform Black Hole Spectroscopy (testing the no-hair theorem) and General Relativity tests with high-precision MBHB events, templates must include at least 3–6 modes, and potentially up to 10 for the most favorable events.
• Avoiding False Discoveries: Using truncated templates for high-SNR events could lead to biased estimates of the remnant mass and spin, potentially mimicking deviations from General Relativity or incorrect astrophysical interpretations.
• Methodological Rigor: The paper highlights the critical importance of handling spectral leakage and windowing artifacts in frequency-domain analyses. The proposed "window-insensitive" approach offers a robust framework for future GW data analysis where systematic errors are the limiting factor.
• Future Directions: The authors note that while their frequency-domain approach is robust, future work should incorporate full Time Delay Interferometry (TDI) responses and time-domain Bayesian analyses to fully capture the complexities of the LISA detector response, especially for low-mass sources with short ringdown durations.

In summary, this work establishes that incomplete ringdown modeling is a critical source of systematic error for LISA, and provides a data-driven prescription for the minimum complexity of waveform templates required to ensure unbiased astrophysical and fundamental physics conclusions.