Semianalytic Sensitivity Estimates for Out-of-Bank Gravitational-Wave Signals

This paper introduces a fast semianalytic method using fitting factors to estimate the sensitivity of gravitational-wave searches to physical effects not explicitly modeled in template banks, such as spin, eccentricity, and deviations from general relativity.

The Gist

Imagine you are a detective trying to find a specific type of whisper in a very noisy room. In the world of gravitational waves, these "whispers" are ripples in space-time caused by massive objects like black holes crashing into each other. To find them, scientists use a giant library of "templates"—pre-recorded, perfect versions of what these whispers should sound like. They scan the noisy data, looking for a match between the real noise and the templates in their library.

However, there's a problem: the real universe is messy. Sometimes, the objects spinning, the orbits are slightly oval-shaped (eccentric), or the laws of physics might be slightly different than we think. If the real signal doesn't perfectly match any of the "perfect" templates in the library, the search might miss it, or it might think the signal is weaker than it really is.

The Problem with Current Methods
Traditionally, to figure out how good their search is, scientists have to run millions of computer simulations. They take a fake signal, hide it in fake noise, and run it through their search engine to see if it gets caught. This is like testing a metal detector by burying thousands of coins in a beach and digging them all up to see how many you missed. It works, but it takes a massive amount of time and computer power.

Furthermore, old methods assumed the library of templates was so huge and dense that every possible signal would have a perfect match. But in reality, the library has gaps. If a signal falls into a gap, the old methods would still say, "We would have found this!" because they ignore the fact that the library is incomplete.

The New Solution: A Fast, Smart Shortcut
The authors of this paper (Vijaykumar and Essick) developed a new, fast way to estimate how well these searches work without running millions of slow simulations.

Think of it like this: Instead of burying a million coins and digging them all up, they created a mathematical "calculator" that instantly tells you how likely you are to find a coin based on two things:

1. How loud the whisper is (the signal's strength).
2. How well the whisper matches the library (a score they call the "Fitting Factor").

If a signal is very loud but doesn't match any template well (maybe because the black holes are spinning in a weird way), the calculator says, "You might miss this one." If it matches perfectly, it says, "You'll catch this one easily."

What They Tested
They tested their new calculator against real-world scenarios to see if it was accurate:

• The "Missing Pages" Test: They looked at a library that was missing pages about spinning objects. They showed that their calculator correctly predicted that the search would miss signals with high spin, whereas the old methods would have falsely claimed they would find them.
• The "Oval Orbit" Test: They tested signals where the objects orbit in an oval shape rather than a perfect circle. Their method correctly estimated that the search would struggle to find these, losing about 20-50% of them depending on how oval the orbit was.
• The "New Physics" Test: They simulated signals that broke the standard rules of physics (General Relativity). Again, their calculator accurately predicted that the search would miss these signals because the library didn't have templates for them.

Why This Matters
This new method is like having a super-fast GPS for gravitational wave searches. Instead of driving every possible route to see which ones are blocked (the slow simulation method), this calculator instantly maps out the "blind spots" in the search.

It allows scientists to quickly answer questions like:

• "If we are looking for black holes with high spin, how many will we miss?"
• "How much does our search sensitivity drop if the orbits are oval?"
• "If gravity works slightly differently than we think, will our current search find it?"

By using this fast, semi-analytic approach, scientists can quickly understand the limitations of their searches and plan better experiments to catch the elusive whispers of the universe, all without waiting days or weeks for computer simulations to finish.

Technical Summary

Technical Summary: Semianalytic Sensitivity Estimates for Out-of-Bank Gravitational-Wave Signals

Problem Statement
Estimating the sensitivity of gravitational-wave (GW) searches is critical for astrophysical and fundamental physics applications, particularly for correcting selection biases in population studies. Currently, this is achieved through computationally intensive "injection campaigns," where millions of simulated signals are embedded in detector noise and processed by full search pipelines. This process is resource-heavy and scales linearly with catalog size. Furthermore, existing semianalytic approaches (e.g., Essick 2023) often assume that the template bank is infinitely dense and that all relevant physics is modeled within the bank. These assumptions fail to account for "out-of-bank" effects, such as signals with orbital eccentricity, spin-precession, or deviations from General Relativity (GR), or signals falling into gaps within discrete template banks. Consequently, traditional estimates provide only an upper bound on sensitivity for such signals, ignoring the signal-to-noise ratio (SNR) loss caused by mismatches between the true signal and the best available template.

Methodology
The authors develop a fast semianalytic framework that explicitly incorporates the finite structure of template banks and the effects of unmodeled physics. The core of the method relies on the fitting factor (FF), defined as the maximum overlap between a candidate signal and any template in the bank.

The procedure proceeds as follows:

1. Fitting Factor Calculation: For a given signal characterized by parameters $\theta$, the method calculates $FF(\theta) = \max_{\theta_T \in \text{bank}} M(\theta_T; \theta)$, where $M$ is the match between the signal and a template. This step identifies the best-matching template and the associated SNR loss ($1 - FF$).
2. Noise Modeling: The observed network phase-maximized SNR, $\hat{\rho}^{\text{net}}_{\text{obs}, \phi}$, is modeled not as the optimal SNR, but as a non-central $\chi$ distribution with $2N_{\text{det}}$ degrees of freedom. The non-centrality parameter is scaled by the fitting factor: $\lambda = FF(\theta) \rho^{\text{net}}_{\text{opt}}(\theta)$.
3. Computational Acceleration: To overcome the computational bottleneck of calculating vast numbers of inner products (typically $\gtrsim 10^{12}$), the authors utilize GPU-accelerated waveforms implemented via the ripple software package and the JAX array backend. This allows the fitting factor maximization to be performed efficiently.
4. Thresholding: A detection threshold $\rho_{\text{thr}} \approx 10$ is applied to the modeled SNR distribution to estimate the fraction of detectable signals.

Key Contributions and Results
The paper validates this framework across three distinct application scenarios, demonstrating that it reproduces the behavior of full search pipelines while accounting for template bank limitations:

• Discrete and Incomplete Template Banks: The method successfully reproduces the detection distributions of the GstLAL search for Binary Black Holes (BBHs) and Binary Neutron Stars (BNSs). Crucially, it captures the sensitivity degradation in "gaps" where the template bank lacks coverage. For example, in the O3 BNS search, the bank lacks templates with effective spin $|\chi_{\text{eff}}| > 0.05$ for low masses. The semianalytic estimate correctly predicts a steep drop in detection fraction for signals with high positive $\chi_{\text{eff}}$ (due to the orbital hangup effect and lack of compensating templates), matching empirical GstLAL results.
• Known Missing Physics (Eccentricity): The framework estimates sensitivity to non-spinning, eccentric BNS and BBH systems using template banks designed for circular signals. It quantifies the significant SNR loss: sensitivity drops by $\sim 20\%$ at eccentricity $e \approx 0.1$ and $\sim 30-50\%$ at $e \approx 0.2$ for BNSs. These results align with previous injection studies (e.g., Gadre et al. 2024), confirming that current pipelines miss a substantial fraction of eccentric signals.
• Unknown Missing Physics (Beyond GR): The method evaluates sensitivity to signals with deviations from GR, parameterized by fractional deviations in post-Newtonian coefficients ($\delta \phi_k$). The estimates reproduce the asymmetry observed in real search pipelines where negative deviations (longer signals) are recovered more efficiently than positive deviations. The results show that sensitivity loss increases with total mass and the magnitude of the deviation.

Significance and Claims
The authors claim that this work provides a fast, semianalytic alternative to full injection campaigns, capable of generating realistic sensitivity estimates within hours rather than weeks. By explicitly maximizing the overlap over the finite template bank before applying noise statistics, the method addresses the limitations of previous semianalytic approaches that assumed infinite template density.

The paper positions these estimates as upper limits on achievable search sensitivity. The authors acknowledge that their model approximates the matched-filtering operation but does not replicate the full suite of signal consistency tests (e.g., $\chi^2$ tests) used in pipelines like PyCBC or GstLAL to reject instrumental artifacts. However, they argue that for genuine compact binary signals, sensitivity is primarily driven by SNR loss from template mismatch rather than consistency tests.

The framework is presented as a tool to rapidly identify sensitivity gaps for specific signal classes (e.g., eccentricity, precession, lensing, or GR deviations) and to generate injection sets for training neural network emulators. The authors emphasize that while the current implementation assumes aligned-spin, quadrupolar banks, the methodology is extensible to more complex searches (e.g., higher harmonics or precession) with increased computational cost.