Unveiling the Spectral Morphological Division of Fast Radio Bursts with CHIME/FRB Catalog 2
Wan-Peng Sun, Yin-Long Cao, Yong-Kun Zhang, Ji-Guo Zhang, Xiaohui Liu, Yichao Li, Fu-Wen Zhang, Wan-Ting Hou, Jing-Fei Zhang, Xin Zhang · 2026
Read Unveiling the Spectral Morphological Division of Fast Radio Bursts with CHIME/FRB Catalog 2 by Wan-Peng Sun, Yin-Long Cao, Yong-Kun Zhang, Ji-Guo Zhang, Xiaohui Liu, Yichao Li, Fu-Wen Zhang, Wan-Ting Hou, Jing-Fei Zhang, Xin Zhang (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2601-16048.txt.
Introduction
Wan-Peng Sun and nine coauthors sort fast radio bursts by their light shape. Their 2026 preprint uses the CHIME burst list, its second full list. Bursts are oft split into those that repeat and those seen once. It is not clear if the two groups are truly two kinds. The team lets the data sort bursts with no fixed labels.
They join a map tool with a clump search to group the bursts. The groups rest on the form of each burst in light and tone. Clear groups would hint at more than one birth path for bursts. Weak groups would hint at one wide kind. The work gives a fresh, data led view of this split.
Catalog record
| CREATOR | Wan-Peng Sun, Yin-Long Cao, Yong-Kun Zhang, Ji-Guo Zhang, Xiaohui Liu, Yichao Li, Fu-Wen Zhang, Wan-Ting Hou, Jing-Fei Zhang, Xin Zhang |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 49 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2601-16048
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Unveiling the Spectral Morphological Division of Fast Radio Bursts with CHIME/FRB Catalog 2 Wan-Peng Sun, Yin-Long Cao, Yong-Kun Zhang, Ji-Guo Zhang, Xiaohui Liu, Yichao Li, Fu-Wen Zhang, Wan-Ting Hou, Jing-Fei Zhang, Xin Zhang Abstract Fast radio bursts (FRBs) are commonly divided into repeating and apparently non-repeating sources, but whether these represent distinct physical populations remains uncertain. In this work, we apply an unsupervised machine learning methods combining Uniform Manifold Approximation and Projection (UMAP) with density-based clustering to analyze CHIME/FRB Catalog 2. We find that FRBs remain primarily separated into two clusters in the multi-dimensional parameter space, with a recall of 0.94 for known repeaters, indicating strong robustness. Consistent with Catalog 1 analyses, we confirm that the spectral morphology parameter, specifically spectral running remains the key discriminator between the two populations, indicating that narrowband emission is an intrinsic and persistent property of repeating FRBs. With the enlarged Catalog 2 sample, we further identify a stable subclass of atypical repeaters (about 6% of repeating bursts) that are broadband, shorter in duration, and more luminous, resembling non-repeating bursts. The Nonrepeater-like cluster also shows higher inferred energies and dispersion measures, consistent with a scenario in which apparently non-repeating FRBs may result from observational incompleteness, with low-energy repeating bursts remaining undetected. Our results provide new statistical evidence for a physical connection between repeating and non-repeating FRBs. I Introduction Fast Radio Bursts (FRBs) are millisecond-duration, extremely energetic radio transients. Since their discovery in 2007 ( Lorimer et al. 2007 ) , their extreme energetics and possible cosmological origins have made them one of the most active frontiers in astrophysics, although their physical origin has not yet been fully understood. Observationally, most FRBs exhibit dispersion measures (DMs) well in excess of the maximum Galactic contribution, indicating that they predominantly arise from compact sources at cosmological distances ( Thornton et al. 2013 ; Spitler et al. 2014 ; Tendulkar et al. 2017 ; Marcote et al. 2017 ; Petroff et al. 2022 ; Zhang 2023 ; Zhang 2024 ) . The burst from the Galactic magnetar SGR 1935+2154 ( Bochenek et al. 2020 ; CHIME/FRB Collaboration et al. 2020 ) , however, has demonstrated that magnetars can account for at least a fraction of the population. Owing to their large DMs, high brightness, and cosmological distances, FRBs have gradually emerged as powerful probes of the intergalactic medium and the cosmic matter distribution ( Macquart et al. 2020 ; Zhao et al. 2020 ; Zhang et al. 2023 ; Zhang et al. 2025a ; Zhang et al. 2025b ) . Over the past decade, FRB observations have revealed diverse activity: some sources produce multiple bursts (repeaters), whereas others have only been detected once (non-repeaters). Whether this distinction reflects intrinsically different progenitors or is instead driven by observational incompleteness and selection biases remains one of the most debated questions in current FRB research ( Zhang 2023 ) . Analyzing the observational characteristics of repeating and non-repeating FRBs is a key approach to understanding their physical origins. Early studies noted that repeating FRBs tend to exhibit more complex temporal structures, longer pulse widths, and relatively narrower spectra ( Pleunis et al. 2021 ; Kumar et al. 2021 ; CHIME/FRB Collaboration et al. 2023 ) , whereas non-repeating sources are typically short-duration, broadband bursts ( CHIME/FRB Collaboration et al. 2021 ; Pleunis et al. 2021 ; Cui et al. 2021 ; Zhang et al. 2022 ) . Sun’s analysis based on the first CHIME/FRB sample further revealed that the most significant differences between the two classes lie in their spectral shape parameters, particularly the spectral running ( Sun et al. 2025a ) , indicating that repeaters generally have steeper, narrower spectra, while non-repeaters tend to show flatter and broader spectra. These systematic differences were robustly confirmed in the analysis of the larger CHIME/FRB Catalog 2 sample (Catalog 2) ( FRB Collaboration et al. 2026 ) . Nevertheless, these distinctions are not absolute, as a small fraction of repeating bursts display spectral and pulse-duration properties highly similar to those of typical broadband, short-duration non-repeating bursts ( FRB Collaboration et al. 2026 ) . Notably, statistical studies of the burst rates for the two classes show that, after correcting for observational duration and detection efficiency, their distributions do not exhibit a bimodal structure CHIME/FRB Collaboration et al. 2023 . This suggests that repeaters and non-repeaters may not constitute entirely distinct populations, and that non-repeating FRBs could potentially repeat on longer timescales. In particular, Sun et al. 2025b found that, in the \gamma - r parameter space, the burst rate correlates positively with pulse width and negatively with spectral bandwidth, lending strong support to the single-progenitor scenario proposed by CHIME/FRB Collaboration et al. 2023 . Furthermore, no significant differences have yet been found in the host galaxy properties of the two classes ( Ma et al. 2025 ; CHIME/FRB Collaboration et al. 2025 ; Pastor-Marazuela et al. 2026 ) , and analyses of their energy distributions indicate that non-repeating FRBs may simply represent the rare, high-energy tail of the repeating population ( Kirsten et al. 2024 ; Ould-Boukattine et al. 2024 ) . Collectively, these findings highlight the complexity of the FRB dichotomy and underscore the need to reassess classification in a higher-dimensional, multi-parameter observational space. In recent years, machine learning has demonstrated unique advantages in the classification of FRBs. Unsupervised methods can perform dimensionality reduction and clustering on multi-dimensional observational parameters without relying on prior labels, thereby revealing the natural clustering and structural differences between repeating and non-repeating FRBs in parameter space ( Chen et al. 2022 ; Yang et al. 2023 ; Zhu-Ge et al. 2023 ; García et al. 2024 ; Sun et al. 2025a ; Qiang et al. 2025 ; Liu et al. 2025 ; Madheshwaran et al. 2025 ; Júnior et al. 2026 ) . On the other hand, supervised methods leverage information from known repeating FRBs to establish nonlinear mappings between input features and classes in high-dimensional feature spaces ( Luo et al. 2023 ; Sharma & Rajpaul 2024 ; Kharel et al. 2025 ; Kharel et al. 2025 ; Arni et al. 2025 ) , enabling the prediction of potential repeating candidates. Compared to supervised approaches, unsupervised methods are more model-independent and do not require prior training, making them particularly well suited for uncovering statistical correlations among multi-dimensional FRB observables. Our previous work ( Sun et al. 2025a ) demonstrated that the spectral shape parameters r and \gamma , which are closely related to FRB spectral morphology, serve as the key features for distinguishing repeaters from non-repeaters, achieving maximal separation between the two classes. Similar conclusions have been independently confirmed in recent studies ( Kharel et al. 2025 ; Arni et al. 2025 ) . Although unsupervised methods have demonstrated distinct advantages, their effectiveness largely depends on sample size and the completeness of the feature space. Previous studies, often based on early samples such as CHIME/FRB Catalog 1, were limited by the number of sources and statistical coverage, making it difficult to fully characterize rare but potentially physically meaningful events in parameter space, particularly the uncommon repeaters exhibiting broadband emission. With the release of Catalog 2, the FRB sample size has increased by an order of magnitude. This