Statistical inference of fast radio burst environments using galaxy number density

This study investigates the galactic environments of fast radio bursts (FRBs) using galaxy number density, with the aim of constraining their progenitor types without requiring precise localization. The authors analyze 19 repeaters and 253 non-repeaters from CHIME Catalog 1, using galaxies from the WISE × PS1 photometric redshift catalog. FRB redshifts are estimated using the Macquart relation from dispersion measures, and galaxy number densities are calculated within a 6 Mpc aperture radius, normalized against 10,000 random apertures in a 100 × 100 Mpc2 region around each FRB.

The key results are as follows:

1. Comparison between repeaters and non-repeaters: A Kolmogorov–Smirnov (KS) test shows no statistically significant difference between the galactic environments of repeaters and non-repeaters, with a p-value of pKS = 0.673. Gaussian fits to the density increment histograms give peak positions of µ = −0.24 ± 0.15 for repeaters and µ = 0.0088 ± 0.050 for non-repeaters, which are consistent within errors. The authors note this could be due to small-number statistics from the limited repeater sample, or it could indicate that both populations share similar environments and may not be intrinsically distinct.

2. Comparison of FRBs with random galaxy apertures:

• For non-repeaters, the median KS p-value is pKS = 4.47 × 10−2, with 54.1% of Monte Carlo trials satisfying pKS ≤ 0.05, indicating a statistically significant deviation from random apertures. The non-repeater CDF is shifted leftward, suggesting they preferentially reside in underdense environments.

• For repeaters, the median KS p-value is pKS = 5.38 × 10−1, with only 1.2% of trials satisfying pKS ≤ 0.05, showing no significant difference from random, though this is likely limited by the small sample size.

• For the combined FRB sample (repeaters + non-repeaters), the median KS p-value is pKS = 2.84 × 10−2, with 60.6% of trials satisfying pKS ≤ 0.05, indicating a statistically significant preference for underdense environments, mainly driven by non-repeaters.

3. Redshift-matching analysis: After matching the redshift distributions of repeaters and non-repeaters (resulting in 149 matched non-repeaters), the conclusions remain unchanged: no significant difference between repeaters and non-repeaters (pKS = 5.7 × 10−1), but significant differences remain when comparing non-repeaters (pKS = 1.6 × 10−2) and the combined sample (pKS = 1.5 × 10−2) against random apertures.

4. Robustness tests:

• Varying the assumed host DM from 50 to 100 pc cm−3 makes the non-repeater vs. random comparison marginally consistent with the null hypothesis (pKS = 5.18 × 10−2), but the combined sample remains significant (pKS = 1.5 × 10−2).

• Using a 150 × 150 Mpc2 background region yields significant differences for non-repeaters (pKS = 7.36 × 10−4) and combined FRBs (pKS = 2.60 × 10−4), but not for repeaters alone (pKS = 2.07 × 10−1).

• Testing aperture radii of 2 Mpc and 4 Mpc gives consistent results: significant differences for non-repeaters and combined samples against random apertures, but not for repeaters alone.

5. Projection effects: The large DM-derived redshift uncertainties can dilute or enhance density contrasts by about 0.7 ∆δ, but since the same redshift slices are used for both FRB and random apertures, the relative conclusion that FRBs preferentially occur in lower-density environments is not significantly affected.

The authors interpret the preference for underdense environments as suggesting that FRBs are statistically more likely to be associated with star-forming (late-type) galaxies or relatively low-mass quiescent galaxies, rather than massive quiescent galaxies, based on the well-established morphology-density relation. However, they emphasize this is a statistical interpretation and does not preclude FRB occurrence in massive quiescent hosts.

The paper concludes that the statistical difference between repeater and non-repeater environments depends on sample size and can be better constrained with future samples from CHIME Catalog 2 and upcoming facilities like BURSTT. The study provides an important observational constraint on FRB host galaxy types and progenitor models, while acknowledging caveats regarding detection biases in dense environments and systematic uncertainties in the host DM assumption.

Improvements to AI Systems Based on This Paper:

1. Uncertainty-Aware Statistical Comparison Module

• Improvement: Integrate a Monte Carlo–based KS-test framework that explicitly propagates redshift and density uncertainties (e.g., host DM assumptions, aperture size variations) into the significance assessment.

• What the improved AI can do: Automatically flag when a claimed environmental difference (e.g., repeaters vs. non-repeaters) is robust or merely an artifact of small sample sizes or systematic assumptions, by outputting p-value distributions and sensitivity ranges rather than single-point statistics.

2. Redshift-Matching and Bias Correction Algorithm

• Improvement: Implement an automated redshift-distribution matching routine (e.g., propensity score matching or iterative subsampling) to compare populations with unequal selection functions.

• What the improved AI can do: For any astrophysical or cosmological dataset, it can generate matched control samples, quantify residual selection biases, and re-run comparative analyses (e.g., environment density tests) to ensure conclusions are not driven by redshift-dependent detection biases.

3. Multi-Scale Environmental Density Estimator

• Improvement: Build a tool that computes galaxy number density contrasts at multiple aperture radii (e.g., 2, 4, 6 Mpc) and background scales (e.g., 100 vs. 150 Mpc) simultaneously, with built-in normalization against random apertures.

• What the improved AI can do: Automatically identify the spatial scale at which an astronomical transient’s environment deviates most strongly from random, and output a scale-dependent significance map—useful for classifying progenitor environments without requiring precise localization.

4. Progenitor-Type Inference from Statistical Environment Preferences

• Improvement: Develop a Bayesian classifier that maps observed density contrasts (e.g., underdense preferences) to likely host galaxy types (e.g., star-forming late-type vs. massive quiescent) using prior morphology-density relations.

• What the improved AI can do: Given a catalog of transients (FRBs, supernovae, etc.) with only dispersion measures or photometric redshifts, it can probabilistically infer the dominant host galaxy population and flag outliers that may require exotic progenitors—even without individual host identifications.

5. Robustness and Caveat Reporting System

• Improvement: Add an automated robustness-check layer that systematically varies key assumptions (e.g., host DM from 50–100 pc cm−3, background region size, aperture radius) and reports which conclusions survive all perturbations.

• What the improved AI can do: For any scientific claim derived from noisy or indirect measurements, it will generate a caveat matrix showing which results are stable and which are fragile, helping researchers avoid overinterpreting marginal detections.

6. Sample-Size-Aware Significance Correction

• Improvement: Implement a power-analysis module that estimates the minimum sample size needed to detect a given environmental difference (e.g., repeaters vs. non-repeaters) at a specified confidence level, based on observed effect sizes and scatter.

• What the improved AI can do: Before launching new surveys (e.g., CHIME Catalog 2, BURSTT), it can predict how many repeaters are required to confirm or refute environmental distinctions, guiding observational resource allocation.

7. Projection-Effect Deconvolution

• Improvement: Create a deconvolution algorithm that corrects density contrast measurements for redshift uncertainty–induced projection effects (e.g., the 0.7 ∆δ dilution/enhancement factor).

• What the improved AI can do: Given a transient with large distance uncertainties, it can reconstruct the most likely intrinsic density contrast and its confidence interval, enabling more accurate environment classification for poorly localized events.

Fast radio bursts (FRBs) are bright, millisecond-duration radio transients of unknown origin. They are categorized as repeaters and non-repeaters, possibly indicating distinct progenitor types. However, validating this distinction is difficult because of the limited number of localized FRBs. Large-scale galactic environments can provide insight into the nature of the host galaxies of FRBs and their progenitors. High-number-density regions are typically associated with old galaxies, whereas low-number-density regions are linked to young star-forming or less massive quiescent galaxies. In this study, we use galaxy number density to statistically assess the environments of 19 repeaters and 253 non-repeaters from CHIME Catalog 1, using galaxies from the WISE x PS1 catalog. A Kolmogorov-Smirnov (KS) test showed no significant difference between the two populations (p KS = 0.673). This result indicates that the statistical significance of the difference could depend on small-number statistics, highlighting the necessity of future FRB samples. Intriguingly, a comparison of FRBs with random galaxy fields suggests that FRBs may preferentially occur in underdense galactic environments, with a median p-value (p KS) of 2.84 times 10-2 compared to random galaxy apertures.