Limits of vacuum-template subtraction for LISA massive black hole binary sources in realistic environments

This paper demonstrates that while gravitational wave dephasing caused by gas accretion in massive black hole binaries will likely bias the inference of other LISA signals due to imperfect vacuum-template subtraction, the resulting residual is unlikely to be confidently distinguished from instrumental noise as a stochastic signal.

The Gist

Imagine the LISA (Laser Interferometer Space Antenna) mission as a giant, ultra-sensitive ear floating in space, designed to listen to the "songs" of the universe. Specifically, it wants to hear the deep, rumbling notes of Massive Black Hole Binaries—two giant black holes dancing around each other before they crash and merge.

However, LISA faces a unique challenge: unlike ground-based detectors that hear short, sharp "pops" (like gunshots), LISA hears a continuous, overlapping choir of thousands of these black hole songs happening at the same time. To make sense of this, scientists use a technique called a "Global Fit." Think of this like trying to identify every individual singer in a massive choir by mathematically subtracting their known voices from the total sound until only the silence (or the noise) remains.

The Problem: The "Vacuum" Assumption

For years, scientists have built their "voice templates" (the mathematical models used to identify the singers) based on a perfect, empty universe. They assume the black holes are dancing in a total vacuum, with nothing but gravity pulling them together.

But in reality, the universe isn't empty. These black holes are often surrounded by swirling clouds of gas and dust (accretion discs). As the black holes spin, they drag through this gas, like a swimmer moving through thick honey instead of water. This "honey" (the gas) changes the rhythm of their dance, causing them to spin slightly out of sync with the "vacuum" predictions.

The Experiment: What Happens When We Ignore the Gas?

In this paper, Lorenz Zwick asks a critical question: What happens if we try to subtract the "vacuum" song from the "gas-filled" reality?

1. The Mismatch: Because the gas changes the timing (phase) of the gravitational waves, the "vacuum template" doesn't perfectly match the real signal.
2. The Residual: When you subtract the wrong song from the real one, you don't get silence. You get a residual noise—a messy, leftover static.
3. The Accumulation: If you have just one black hole, this leftover static is tiny and unnoticeable. But LISA will hear thousands of these sources. When you add up the tiny leftovers from thousands of imperfectly subtracted songs, they create a new, collective background noise.

The Analogy: The Choir and the Static

Imagine a choir of 1,000 singers.

• The Reality: Each singer is wearing a slightly different pair of shoes that makes them step a tiny bit out of rhythm.
• The Template: The conductor (LISA) assumes everyone is wearing perfect, silent shoes.
• The Subtraction: The conductor tries to cancel out the sound of the choir by playing a recording of a choir with perfect shoes.
• The Result: Because the shoes in the recording don't match the real shoes, the cancellation isn't perfect. You are left with a faint, rhythmic thumping sound. If you do this for one singer, you hear a tiny thump. If you do it for 1,000 singers, that thumping becomes a loud, confusing rumble that drowns out the quiet parts of the music.

The Findings: Is the Rumble a Problem?

The paper calculates the strength of this "rumble" (the residual noise) based on how many black holes LISA will likely find and how much gas they are eating.

• The Good News: The rumble is not loud enough to be mistaken for a new, mysterious signal from the early universe. It won't trick scientists into thinking they found a new type of cosmic event. It's too weak to be "heard" as a distinct signal above the instrument's own static.
• The Bad News: The rumble is loud enough to be annoying. It acts like a persistent background hiss that makes it harder to hear the details of the individual singers.
  • If scientists try to measure the mass or spin of a specific black hole binary, this leftover static will introduce biases. It's like trying to tune a radio while someone is constantly tapping on the speaker; you might think the station is slightly off-key when it's actually just the tapping interfering.

The Conclusion: A New Kind of Noise

The paper concludes that while this "gas-induced static" won't be a discovery in itself, it is a systematic error that LISA scientists must account for.

• Current Strategy: We can't just ignore the gas. If we use "vacuum" templates, we will get the details of the black holes slightly wrong.
• Future Strategy: Scientists need to develop "gas-aware" templates. Instead of assuming the black holes are dancing in a vacuum, the models need to include the "honey" of the gas.
• The Silver Lining: Interestingly, if we can learn to distinguish this static, it might actually tell us something cool about the universe: how much gas is swirling around these black holes and how fast they are eating it.

In short: The universe is messier than our perfect models. If we don't account for the "gas" around black holes, LISA's data will be slightly "out of tune," making it harder to measure the stars accurately, even if the noise itself isn't loud enough to be a headline-grabbing discovery.

Technical Summary

Here is a detailed technical summary of the paper "Limits of vacuum-template subtraction for LISA massive black hole binary sources in realistic environments" by Lorenz Zwick.

1. Problem Statement

The Laser Interferometer Space Antenna (LISA) will observe a dense population of gravitational wave (GW) sources in the millihertz band, necessitating a "global-fit" data analysis approach where thousands of overlapping signals are modeled and subtracted simultaneously. Current global-fit pipelines primarily rely on vacuum General Relativity (GR) waveform templates.

However, the astrophysical formation of Massive Black Hole (MBH) binaries likely involves sustained interactions with circumbinary gas discs and stellar environments prior to and during their GW-driven inspiral. These Environmental Effects (EEs), particularly gas accretion, induce a cumulative phase shift (dephasing) in the GW signal relative to vacuum predictions.

• The Core Issue: If the true signals contain unmodeled environmental dephasing but the analysis uses vacuum templates, the subtraction will be imperfect.
• The Consequence: While the dephasing for any single MBH binary might be too small to detect individually, the collective residual from thousands of imperfectly subtracted sources could accumulate into a stochastic background. This background could either mimic a stochastic GW signal or, more critically, bias the parameter estimation of other signals in the global fit.

2. Methodology

The author employs a statistical approach to quantify the impact of unmodeled gas effects on LISA data over a 4-year observation period.

• Population Modeling:

  • The study utilizes state-of-the-art predictions for the MBH merger population (based on ASTRID simulations, L-Galaxies, and semi-analytical models).
  • The differential merger rate is parameterized as a product of independent log-normal distributions for primary mass ($M_1$), mass ratio ($q$), and redshift ($z$).
  • Uncertainty envelopes are defined to bracket variations between different population models.
  • The total number of sources is drawn from a Poisson distribution with a mean merger rate $\langle \dot{n} \rangle$.

• Environmental Physics (Gas Accretion):

  • The study assumes MBH binaries are embedded in circumbinary accretion discs.
  • The gas-induced orbital evolution is modeled using a phenomenological prescription where the orbital frequency derivative $\dot{\Omega}_{gas}$ scales with the accretion rate.
  • The accretion rate is scaled by a fiducial Eddington ratio ($f_{Edd}$), assumed to be typical for the population.
  • The resulting dephasing ($\delta\Psi$) relative to vacuum evolution is derived in the Stationary Phase Approximation (SPA). Crucially, this dephasing scales as $f^{-13/3}$ (equivalent to a "negative 4PN" term), making it highly frequency-dependent and difficult to mimic with standard vacuum parameters.

• Residual Calculation:

  • Two subtraction methods are tested:
    1. Simple Subtraction: Subtracting a vacuum waveform with identical parameters from the true (gas-affected) waveform.
    2. Best-Match Subtraction: Finding the vacuum waveform that maximizes the inner product with the true signal (mimicking a maximum likelihood fit) and subtracting that.
  • The study finds that due to the steep frequency scaling of the gas dephasing, the "best-match" vacuum template cannot effectively absorb the environmental phase shift. Consequently, the residuals from both methods are similar in power distribution.

• Signal-to-Noise Ratio (SNR) & Distinguishability:

  • The total residual power spectral density (PSD) is computed by summing the residuals of all sources in quadrature.
  • The SNR of the residual background is calculated against the LISA noise PSD.
  • Distinguishability is assessed by comparing the residual energy density ($\Omega_{GW}$) against Bayesian Power-Law Sensitivity (BPLS) curves, which account for uncertainties in the noise covariance matrix (a stricter criterion than simple SNR thresholds).

3. Key Contributions

• Quantification of the "Residual Background": The paper provides the first quantitative estimate of the GW residual power generated by the collective failure of vacuum templates to model gas-induced dephasing in a realistic MBH population.
• Scaling Laws: The authors derive a simple fitting function for the median residual amplitude, showing it scales with the square root of the merger rate ($\sqrt{\langle \dot{n} \rangle}$) and the square root of the Eddington ratio ($\sqrt{f_{Edd}}$).
• Dependence on Hyper-parameters: The study characterizes how the residual SNR depends on population properties, finding a strong dependence on the merger rate and redshift, but an exponential suppression with increasing primary mass (lighter binaries are more susceptible to gas effects).
• Distinguishability Analysis: The paper rigorously tests whether this residual can be distinguished from instrumental noise using Bayesian model selection, moving beyond simple SNR calculations.

4. Key Results

• Residual SNR: For a fiducial merger rate of $\langle \dot{n} \rangle = 20 \text{ yr}^{-1}$ and $f_{Edd} = 1$, the accumulated residual has an SNR of:
  $$\text{SNR} \approx 3.2^{+5.4}_{-1.9} \times \sqrt{f_{Edd} \frac{\langle \dot{n} \rangle}{20 \text{ yr}^{-1}}}$$
  This indicates a significant power excess over the LISA noise floor.
• Spectral Shape: The residual spectrum resembles a stochastic background with a peak around $10^{-4}$ Hz (for high merger rates), decaying exponentially at higher frequencies.
• Bias vs. Detection:
  • Bias: The residual is large enough to bias the inference of other signals (e.g., individual MBH binaries or the stochastic background from white dwarfs) because its amplitude exceeds the noise floor and the SNR threshold for parameter estimation errors.
  • Distinguishability: However, the residual is unlikely to be confidently distinguished from instrumental noise as a standalone stochastic signal. Even under optimistic assumptions (10% noise uncertainty), the residual spectrum lies well below the Bayesian detection threshold (log-Bayes factor of 1) for the median population realization. Only the loudest ~5% of realizations would be detectable as a distinct stochastic signal.
• Mass Dependence: The residual is dominated by lighter MBH binaries ($M_1 \sim 10^5 M_\odot$) because gas effects scale inversely with chirp mass. These lighter systems are also the most uncertain in current population models.

5. Significance and Implications

• Global Fit Challenges: The study highlights a critical limitation in current LISA data analysis strategies. Even if environmental effects are undetectable in individual sources, their cumulative effect in a global fit creates a systematic error floor.
• Parameter Estimation Risks: The primary risk is not the detection of the residual itself, but the bias it introduces into the measurement of other astrophysical parameters (masses, spins, distances) if the environmental physics is ignored.
• Mitigation Strategies: The authors suggest that targeted mitigation is necessary. Strategies could include:
  • Identifying and re-analyzing subsets of lighter MBH sources with waveform models that explicitly incorporate environmental effects (e.g., negative-PN terms).
  • Moving from frequency-domain stochastic treatments to time-domain analyses to correlate residuals with specific merger events.
• Future Work: The paper emphasizes the need for more realistic modeling of the correlation between binary mass and accretion rate (Eddington ratio), as lighter binaries may require higher accretion rates to merge, potentially increasing the residual amplitude further.

In conclusion, the paper demonstrates that vacuum-template subtraction is insufficient for the full LISA data stream in realistic astrophysical environments. While the resulting "noise" may not be a detectable stochastic signal on its own, it poses a significant threat to the precision of the global fit, potentially compromising the scientific return of the LISA mission if not addressed.