A Joint Optimal Search for Gravitational Waves from Resolved and Unresolved Supermassive Binary Black Holes with Pulsar Timing Arrays

This paper introduces a joint model for the gravitational wave background and resolvable supermassive black hole binaries to constrain their population properties using simulated NANOGrav 15-year data, revealing that most active galactic nucleus candidates are in tension with observations while projecting a 5% probability of detecting individual sources in the upcoming 20-year dataset.

The Gist

Imagine the universe is filled with a constant, low hum, like the sound of a busy crowd in a stadium. In the world of astronomy, this "hum" is called the Gravitational Wave Background (GWB). For a long time, scientists thought this hum was just a smooth, steady noise created by millions of supermassive black holes dancing in pairs (called binaries) across the cosmos. They treated it like a single, giant ocean wave.

However, this new paper suggests that the ocean isn't actually smooth. It's made of individual raindrops.

Here is a simple breakdown of what the authors, Goncharov and his team, did and found:

1. The New Model: From a "Smooth Ocean" to "Individual Raindrops"

Previously, scientists looked for the "ocean" (the background hum) and the "raindrops" (individual, bright black hole pairs) as two separate problems. They used one set of rules for the background and a different set for the bright spots.

The authors created a single, unified model that treats the background and the bright spots as part of the same family.

• The Analogy: Imagine trying to hear a specific person shouting in a stadium.
  • Old Way: You try to measure the average noise of the crowd, and separately, you try to find the loudest person. You don't assume the loudest person is part of the crowd's noise.
  • New Way: You realize the crowd's noise is actually made up of thousands of people, and the "loudest person" is just the one who happens to be shouting the hardest right now. Your model accounts for both the general roar and the specific shout simultaneously.

2. The "Counting" Statistic ($N_c$)

The authors introduced a new way to measure the data called the "Characteristic Number" ($N_c$).

• The Analogy: Think of $N_c$ as a "crowd density meter."
  • If $N_c$ is high (like 1,000), it means there are so many black hole pairs that their individual signals blend together perfectly into a smooth, predictable hum (a Gaussian process). It's like a dense fog where you can't see individual drops.
  • If $N_c$ is low (like 1 or 2), it means there are very few pairs. The "hum" becomes choppy and bumpy because you can hear the distinct "thump" of individual pairs. It's like a sparse rainstorm where you hear individual drops hitting the roof.

The paper argues that if we can prove $N_c$ is low, we have found the "smoking gun" that the hum is indeed caused by these specific black hole pairs, rather than some other mysterious cosmic source.

3. The Test: Simulating the NANOGrav Data

The team didn't just do math on paper; they tested their model against a simulation of real data from the NANOGrav project (a group of scientists using pulsars—cosmic lighthouses—to detect gravitational waves). They made the simulation look exactly like the real 15-year dataset, including all the known "static" and noise.

What they found:

• No "Loud Shouters" yet: In their simulation of the 15-year data, they found no evidence of a single, individually resolvable black hole pair (a "CW" or Continuous Wave). The data was too quiet to pick out a specific shout from the crowd.
• The "Smooth Ocean" wins: The data was consistent with a high $N_c$ (around 1,000). This means the current data looks like a smooth, Gaussian ocean, not a choppy rainstorm of individual drops. They couldn't prove the "Poisson" nature (the individual raindrops) just yet.

4. Checking the "Guest List" (AGN Candidates)

Astronomers have a list of suspected black hole pairs (candidates) found by looking at active galaxies (AGN). The authors checked if these candidates fit the new model.

• The Result: Using their new, stricter rules, they found that 21 out of 114 of these candidates are "in tension" with the data.
• The Analogy: Imagine a security guard (the new model) checking a guest list. The old guard (previous analysis) let almost everyone in. The new guard, using a more precise checklist, kicked 21 people out because their "shout" would have been too loud for the current background noise. Only one candidate remained on the "maybe" list based on the old, looser rules.

5. Future Predictions: Will We Hear a Shout Soon?

The authors looked ahead to what might happen with 20 years of data (instead of 15).

• The Odds: They calculated the probability of finally hearing a distinct "shout" (detecting a single black hole pair) with a strong signal.
  • 15-year data: Only a 2% chance.
  • 20-year data: A 5% chance.
• The "Outlier" Chance: While a strong shout is unlikely, they found a 40% chance of finding a "whisper" (a weak signal with a low score) that stands out slightly from the noise in the 20-year data.

Summary

This paper builds a better, more physics-based "microphone" to listen to the universe. It combines the search for the general background noise with the search for specific loud sources into one model.

When they tested this new microphone on simulated data, it told them: "We are still hearing a smooth hum, not individual drops. We haven't found a specific black hole pair shouting yet, and our new rules suggest many of the suspected candidates are actually too loud to fit with what we see."

It's a step toward a more complete understanding of the cosmic hum, even if the "individual voices" haven't been clearly heard just yet.

Technical Summary

Technical Summary: A Joint Optimal Search for Gravitational Waves from Resolved and Unresolved Supermassive Binary Black Holes with Pulsar Timing Arrays

Problem Statement
Current Pulsar Timing Array (PTA) analyses treat the Gravitational Wave Background (GWB) and individually resolvable Continuous Wave (CW) signals from Supermassive Black Hole Binaries (SMBHBs) as separate, disconnected problems. The GWB is typically modeled as a stochastic, isotropic Gaussian process, while searches for individual sources use uninformative priors on strain amplitude. This fragmentation prevents a unified inference of the astrophysical population properties (such as number density and mass scale) that seed both the background and the brightest individual sources. Furthermore, standard models often fail to account for the transition from Gaussian large-number statistics to Poissonian small-number statistics, which is fundamental to the discrete nature of SMBHB sources.

Methodology
The authors introduce a joint hierarchical model derived from first principles that simultaneously describes the unresolved GWB and the brightest resolvable SMBHB sources.

• Theoretical Framework: Building on the work of Sato-Polito and Zaldarriaga [15], the model utilizes the characteristic number of sources, $N_c$, and the characteristic strain amplitude, $h_c$, as astrophysical hyperparameters. The model defines the probability density function (PDF) for the total characteristic strain ($h_t$) as a superposition of the brightest source ($h_{cw}$) and the remainder of the population ($h_{t-1}$).
• Statistical Approach: The brightest source is treated as a deterministic directional signal (CW), while the remaining sources form an isotropic stochastic background. The model incorporates the PDF of the brightest source to marginalize over gravitational wave interference and deviations from isotropy.
• Data Analysis: Unlike previous approaches that refit the GWB power spectrum inferred from data, this method performs a global fit of the PTA likelihood to all signal and noise parameters. The authors replace disconnected priors with their hierarchical model $\pi(h_t, h_{cw} | N_c, h_c)$, allowing information to flow between the inference of the background and the inference of individual continuous waves.
• Validation: The methodology is tested on simulated NANOGrav 15-year data that replicates known noise, observations, and the inferred GWB power spectrum. The simulation includes 20 pulsars with 100 ns timing precision over 15.06 years.

Key Contributions

1. Unified Model: The paper presents the first definitive, testable model that jointly models fully resolved and completely unresolved SMBHB sources within a single likelihood framework.
2. Detection Statistic: The authors propose the characteristic number of sources, $N_c$, as a primary detection statistic for the astrophysical origin of the GWB. A value of $N_c \ll 1000$ indicates a Poissonian, non-Gaussian regime, whereas $N_c \geq 1000$ corresponds to the standard Gaussian approximation.
3. Improved Constraints: By using an astrophysically motivated hierarchical prior rather than an uninformative one, the model provides more stringent constraints on the strain amplitude of individually resolvable SMBHBs.
4. Population Inference: The framework allows for the direct recasting of GWB parameters ($N_c, h_c$) into physical population properties: the SMBHB mass scale ($M_{peak}$) and abundance ($\rho_{BH}$).

Results
Applied to the simulated NANOGrav 15-year data, the analysis yields the following findings:

• GWB Nature: There is insufficient evidence in the simulated data to reject the null hypothesis of a Gaussian GWB ($N_c \approx 1000$). The data is insensitive to the SMBHB-specific Poisson regime.
• Individual Sources: No individually resolvable CW signal was detected. The posterior support for the frequency of the brightest source ($f_{cw}$) spans the entire prior range, and the posterior-predictive samples for $h_{cw}$ match the data without requiring an exceptionally bright outlier.
• Tension with AGN Candidates: The new, tighter upper limits on CW strain amplitude place 21 out of 114 SMBHB candidates from Active Galactic Nuclei (AGN) observations in tension with the NANOGrav data. In contrast, the original analysis (using uninformative priors) found only one candidate in tension.
• Detection Probabilities:
  • The probability of detecting a CW with a Signal-to-Noise Ratio (SNR) $\geq 5$ in the 15-year data is estimated at 2%.
  • For the projected 20-year data, this probability increases to 5%.
  • However, the probability of finding a low-significance outlier (SNR $\geq 2$) in the 20-year data is estimated at 40%.

Significance and Claims
The paper claims that this joint model represents a "smoking gun" approach for identifying the astrophysical nature of the GWB. By moving beyond the assumption of a purely Gaussian background, the model enables the detection of source-count fluctuations (non-Gaussianity) that would confirm the discrete SMBHB origin of the signal.

The authors emphasize that their method is more complete and correct than previous factorized likelihood approaches, as it avoids parameter estimation degeneracies by performing a global fit. While the current simulated data does not show a resolvable CW, the framework establishes a rigorous checklist for future detections: consistency between posterior and posterior-predictive samples, broad support for the frequency of the brightest source, and low probability of exceeding detection thresholds.

Ultimately, the paper argues that future PTA data will not only measure the GWB amplitude but also test the SMBHB interpretation by narrowing the posterior toward the astrophysically expected region in the ($\rho_{BH}, M_{peak}$) plane and toward smaller $N_c$ values, thereby confirming the transition from Gaussian to Poissonian statistics expected from a finite population of supermassive black holes.