A Generalized Algorithmic Framework for Detecting Faraday Rotation Measure Flares in Repeating Fast Radio Bursts

1. Problem Statement

Repeating Fast Radio Bursts (FRBs) serve as probes for the magneto-ionic environments surrounding their progenitors. While the temporal evolution of the Faraday Rotation Measure (RM) is a critical diagnostic tool, distinguishing discrete, transient "RM flares" (caused by passing plasma clumps or stellar coronal mass ejections) from intrinsic environmental volatility (secular trends or chaotic stochasticity) is a significant challenge.

Current detection methods suffer from:

• Non-uniform temporal sampling: Observations range from hours to months apart, making fixed smoothing windows ineffective.
• Complex morphologies: RM evolution exhibits diverse behaviors, from smooth long-term trends to chaotic fluctuations.
• Lack of standardization: Previous discoveries (e.g., FRB 20220529A) were largely opportunistic and relied on manual inspection, lacking a statistically robust, automated framework to differentiate genuine physical flares from noise or baseline drift.

2. Methodology

The authors propose a generalized algorithmic framework designed to automate the detection and characterization of RM flares. The pipeline consists of three core components:

A. Adaptive Baseline Estimation (Quiescent Background Modeling)

To handle irregular sampling cadence, the algorithm avoids fixed sample counts in favor of time-duration-based rolling windows.

• Adaptive Window Size ($W$): Calculated based on the median time interval between observations ($\Delta t_{med}$) scaled by a multiplier ($k_w = 25$), with bounds $W_{min}=30$ days and $W_{max}=150$ days. This ensures the baseline is robust against noise in sparse data while remaining responsive to secular trends in dense data.
• Iterative Rejection: To prevent "signal self-subtraction" (where flare data skews the baseline), the algorithm employs a two-stage filtering process:
  1. Global Screening: Masks extreme outliers deviating from the global median by $>10 \times \text{MAD}_{glob} + 200 \text{ rad m}^{-2}$.
  2. Local Filtering: Within each window, points deviating from a local rolling median by $>5 \times \text{MAD}_{local} + 25 \text{ rad m}^{-2}$ are excluded.
• Baseline Construction: A rolling median is applied to the remaining "quiescent" points, followed by linear interpolation to create a continuous, clean reference baseline $B(t)$.

B. Significance Scoring

The framework models the noise floor by combining instrumental errors with intrinsic environmental volatility.

• Noise Components:
  • $\sigma_{glob}$: Global residual noise (Median Absolute Deviation of residuals between raw data and baseline).
  • $\sigma_{loc}$: Local noise (rolling 5-point standard deviation).
  • $\sigma_{err}$: Instrumental error.
• Total Noise ($\sigma_{tot}$): Calculated as $\sqrt{\sigma_{glob}^2 + \sigma_{loc}^2 + \sigma_{err}^2}$.
• Significance Score ($S$): Defined as $S(t) = \frac{|RM(t) - B(t)|}{\max(1, \sigma_{tot})}$. A numerical floor of 1.0 prevents significance inflation in low-noise regions.

C. Flare Detection Logic

• Trigger Threshold: A candidate flare is initiated only if a contiguous segment contains a point where $S(t) > T_{tri}$, with a rigorous threshold of $T_{tri} = 10$. (A reference threshold of $T_{ref} = 5$ is used for visual inspection).
• Temporal Boundary Definition: To achieve sub-cadence precision, the "Flare Phase" duration is determined using a Full Width at Tenth Maximum (FWTM) criterion. The algorithm interpolates the time points where the score curve crosses $\eta \times S_{peak}$ (where $\eta = 0.1$).

3. Key Contributions

• First Uniform Census: The paper presents the first systematic, automated search for RM flares across a population of repeating FRBs, moving away from manual, case-by-case analysis.
• Robust Statistical Framework: The introduction of adaptive windowing and iterative baseline rejection solves the problem of non-uniform sampling and signal self-subtraction.
• Differentiation of Dynamics: The method provides a rigorous statistical basis to distinguish between discrete transient events (flares) and continuous secular evolution or chaotic stochasticity.
• Open Source Tool: The Python implementation is made available to the community to facilitate future analysis of large datasets from facilities like FAST, CHIME, and the SKA.

4. Results

The framework was applied to a sample of 15 repeating FRBs with significant historical RM variability.

• Primary Finding: High-confidence RM flares are remarkably rare.
• FRB 20220529A: This was the only source in the sample to satisfy the rigorous detection criteria ($S > 10$). It exhibited a sudden jump from a median of ~17 rad m$^{-2}$ to ~2000 rad m$^{-2}$, followed by a rapid recovery. The algorithm successfully isolated this as a discrete event with a clear FWTM-defined duration.
• Other Sources:
  • FRB 20201124A: Showed a peak score marginally close to the reference threshold ($S \approx 5$) but failed to trigger the rigorous threshold ($S < 10$), suggesting lower statistical confidence or a different physical mechanism.
  • FRB 20121102A, FRB 20180916B, FRB 20190520B: These sources displayed large absolute RM changes (secular evolution or chaotic fluctuations) but lacked the localized, peak-scaled structure required to be classified as a "flare" under this framework.

5. Significance and Implications

• Progenitor Diagnostics: The rarity of confirmed RM flares suggests that the specific physical conditions required (e.g., dense magneto-ionic clumps or companion star CMEs crossing the line of sight) are either intrinsically infrequent or require specific geometric alignments.
• Environmental Characterization: The framework allows astronomers to categorize FRB environments into distinct modes: those dominated by discrete transient interactions versus those dominated by continuous secular evolution or stochastic turbulence.
• Future Prospects: As polarization data volumes increase with next-generation telescopes (SKA), this standardized tool is essential for managing big data, correlating RM flares with activity cycles, and refining models of FRB progenitor systems (e.g., distinguishing between magnetars in binary systems vs. isolated magnetars).

In conclusion, this work establishes a critical, statistically robust methodology for identifying transient plasma structures around FRBs, transforming RM analysis from an opportunistic endeavor into a systematic census of the high-energy universe.