The nanohertz gravitational-wave background, discovered by pulsar timing arrays, is most likely produced by merging supermassive black holes. Measuring the spatial anisotropies of this background is the next step in extracting information about the black hole population. Shot noise from rare, bright sources is expected to be strong at high frequencies, but its statistics have so far been described only by mean values. This work joins a series of studies of background anisotropies: neighboring papers examine the large-scale structure, Hellings–Downs correlations, and mapping methods, but the realization-to-realization scatter of shot noise had remained uncharacterized.
The authors built a population model of merging supermassive black holes based on empirical calibrations (LM24) and generated 700,000 realizations of the sky using Monte Carlo methods. For each realization, they computed the second and fourth moments of the strain field and the shot-noise amplitude as their ratio. They compared the distributions of these quantities with analytical predictions and investigated their dependence on frequency, population model, and the time a source spends in a frequency bin.
The distribution of the shot-noise amplitude turns out to be broad: the 95% interval spans a factor of ~50 (1.7 decades) at a fixed frequency. Median values lie 2–3 times below the mean, and the moment-based estimate ⟨h⁴⟩/⟨h²⟩² overestimates the mean by a factor of ~3 at 0.1 yr⁻¹ and by more than two orders of magnitude at 1 yr⁻¹. The reason is the inequality ⟨X/Y⟩ ≠ ⟨X⟩/⟨Y⟩ and the fact that shot noise is bounded above by unity. The median shot noise grows from 3×10⁻³ at 3 nHz to 0.05 at 30 nHz, following a power law $f^1$.1–1.2 instead of the expected $f^8$/3. The distribution of the fourth moment is even wider: the 95% interval spans ~3 decades, and the mean falls outside this interval above 15 nHz. Conditioning on a measurement of the characteristic strain narrows the scatter to 0.7–0.8 decades at an accuracy of 0.05 dex.
This work shows that interpreting PTA anisotropy measurements requires modeling the full distribution function of shot noise, not just relying on averages. This sets it apart from previous studies that limited themselves to analytical estimates of the mean. Accounting for realization scatter is critical for testing black hole population models and for planning future observations.
The method needs to be extended to finite frequency bins with PTA weighting schemes, the covariances between multipoles must be investigated, and realistic sky maps with non-Gaussian angular structure should be constructed. Conditioning on the detection of individual bright sources will narrow the distributions and improve model discrimination.
The results will affect data analysis for NANOGrav, EPTA, and other PTAs, as well as modeling of supermassive black hole populations and their connection to galaxies.
The next step is to incorporate full shot-noise PDFs into Bayesian inference pipelines for PTAs and to develop methods for joint analysis of anisotropies with individual source detections.
This work is directly related to the unresolved problem of determining the contribution of different sources to the nanohertz gravitational-wave background: astrophysical black hole mergers, cosmological processes, or exotic physics. Correct shot-noise statistics are essential for separating these components.
This section is addressed to the author of the paper and their colleagues. Neighbouring works were found by meaning-based search across our archive, not by word overlap. We do not review: we show the field around the work.
What the work does
The authors built a population model of merging supermassive black holes based on empirical calibrations (LM24) and generated 700,000 realizations of the sky using Monte Carlo methods. For each realization, they computed the second and fourth moments of the strain field and the shot noise amplitude as their ratio. They compared the distributions of these quantities with analytical predictions, investigating their dependence on frequency, population model, and the time a source spends in a frequency bin.
Where its strength is
This work provides, for the first time, the full distribution of shot noise, showing that typical realizations lie 2–3 times below the mean, and that moment-based estimates overestimate the signal by orders of magnitude at high frequencies. This offers a concrete tool for correctly interpreting PTA data.
Where it could go next
Worth trying
Why this may matter
Accounting for the realization scatter of shot noise is critical for testing black hole population models and planning observations, because a typical realization can differ significantly from the average, affecting the interpretation of background anisotropy measurements.
Nearby works in our archive
The closest work in the archive sits at 0.70 by meaning — the usual level for our corpus.
The median shot noise at 1 yr⁻¹ reaches 0.05, which is only four times below the current NANOGrav upper limit of 0.20—future observations may detect background anisotropy for the first time.