Discovery of 27 new Rotating Radio Transients by MeerTRAP

The paper reports the discovery of 27 new Galactic radio transients by the MeerTRAP single pulse search programme at the MeerKAT telescope. These transients are classified as rotating radio transients (RRATs), a subclass of pulsars that exhibit occasional single radio pulses. The discoveries bring the total number of Galactic transients discovered by MeerTRAP to 95.

Key findings and results:

• Discovery details: Of the 27 new transients, 15 were discovered at L-band, 9 at UHF band, and 3 at S-band (the first MeerTRAP Galactic transients found at S-band). Six sources were found in the incoherent beam (IB) and 21 in coherent beams (CB). Discovery signal-to-noise ratios ranged from 8.4 to 47.0, and dispersion measures (DMs) ranged from 21.4 to 398.0 pc cm−3.

• Localization: Image-domain localization was achieved for 14 of the 27 discoveries using transient buffer data, providing arcsecond accuracy. For the previously published MeerTRAP source MTP0040 (now PSR J1357−6507), the position was determined to be RA = 13:57:55.8±1.3 s and Dec = -65:07:18.8±1.6″, and its period was measured as 1.8744(4) s.

• Period measurements: Preliminary periods were estimated for 8 sources using arrival times of multiple pulses, ranging between 0.78 s and 4 s. Coherent timing solutions were obtained for 4 sources: PSR J1726−6456 (P = 1.6543954814(2) s, Ṗ = 2.94(2)×10−15 ss−1), PSR J1811−1740 (P = 1.8065860258(4) s, Ṗ = 159.89(3)×10−15 ss−1), PSR J1454−3338 (P = 2.9409767754(6) s, Ṗ = 2.46(2)×10−15 ss−1), and PSR J1750−4450 (P = 2.195311081(6) s, with only an upper limit on Ṗ).

• Burst rates and fluences: Burst rates ranged from 0.05 to 88 hr−1. The highest burst rates were observed from PSR J1726−6456 (63 hr−1), MTP0087 (88 hr−1), PSR J2225+1243 (77 hr−1), and PSR J1750−4450 (49 hr−1). Fluences ranged from 0.13 to 11 Jy ms. Most CB discoveries had fluences below 1 Jy ms, while four CB sources (MTP0087, PSR J1410−4804, PSR J2225+1243, PSR J1750−4450) showed bright pulses >1 Jy ms. Three IB discoveries that were localized (PSR J0937−6344, PSR J1449−5953, PSR J0929−1855) showed high fluences between 3 and 11 Jy ms.

• Follow-up observations: Follow-up observations were conducted with the 100-m Effelsberg telescope and the 64-m Murriyang telescope at Parkes. Four sources (PSR J1726−6456, PSR J1811−1740, PSR J1155−5021, PSR J2225+1243) showed regular pulsar-like periodic emission in the Fast Folding Algorithm (FFA) search. PSR J1726−6456 and PSR J2225+1243 showed signs of nulling, while PSR J1811−1740 and PSR J1155−5021 showed persistent signals. PSR J1155−5021 showed two components in its folded profile. No periodic signal was found for PSR J1410−4804, but two bright single pulses (S/N = 38 and 55) were detected with peak fluxes of 685±45 mJy and 925±45 mJy, visible across the full 0.7–4 GHz band, indicating a low burst rate of 2 pulses per hour.

• Independent discoveries: Some sources were independently discovered by other surveys. MTP0084 was also found by CRAFTS at DM 25.7 pc cm−3. MTP0106 was independently discovered by the MPIfR-MeerKAT UHF Galactic Plane Survey with period 668 ms and DM 133.3 pc cm−3. MTP0100 was considered different from pulsar J1832−0901t based on pulse width and position differences.

• Discussion: The results illustrate the blurred separation between RRATs and canonical pulsars, as many RRAT-like sources show regular periodic emission in long follow-up observations. Some sources (e.g., MTP0093, PSR J1449−5953, MTP0103, PSR J0929−1855) were observed for tens of hours but only a few pulses were detected, indicating extremely sporadic emission. PSR J2225+1243, located at Galactic latitude b = −36.6°, has a minimum DM distance of 12.9 kpc (NE2025 model) and a z-distance of 7.7 kpc, suggesting an approximate age of 30.9 million years, placing it close to the pulsar death valley on the P−Ṗ diagram. The paper notes that the current RRAT population is a mixture of weak pulsars, extremely nulling pulsars, and prototype RRATs, and a dedicated effort is needed to separate these populations.

Improvements to AI Systems:

1. Automated Transient Classification and Population Separation

• Improvement: Train a machine learning model on the multi-wavelength, multi-band (L, UHF, S) pulse profiles, burst rates, fluences, and timing solutions (P, Ṗ) to automatically classify new radio transients into subclasses (weak pulsars, extremely nulling pulsars, prototype RRATs).

• What it can do: Given a new single-pulse detection, the AI can predict its likely subclass and flag it for targeted follow-up, reducing manual analysis time and improving survey efficiency.

2. Adaptive Follow-up Scheduling and Resource Allocation

• Improvement: Use reinforcement learning to optimize follow-up observation strategies based on detected burst rate, fluence, and periodicity. The AI can decide which telescope (MeerKAT, Effelsberg, Parkes) and which search algorithm (single-pulse, FFA) to prioritize for each source.

• What it can do: Automatically allocate limited telescope time to maximize the chance of detecting periodic emission or rare bright pulses, as demonstrated by the varied follow-up outcomes (e.g., PSR J1811−1740 persistent vs. PSR J1410−4804 sporadic).

3. Real-time Localization and Cross-matching

• Improvement: Integrate image-domain localization (arcsecond accuracy) with automated cross-matching against existing pulsar catalogs (e.g., CRAFTS, MPIfR-MeerKAT surveys) using a Bayesian framework to assess source uniqueness (e.g., MTP0100 vs. J1832−0901t).

• What it can do: Instantly flag potential duplicate discoveries or new sources, and provide precise positions for multi-wavelength follow-up, reducing false positives and improving catalog completeness.

4. Predictive Modeling of Burst Emission and Nulling Behavior

• Improvement: Develop a generative model (e.g., a transformer or state-space model) trained on the time-series of pulse arrival times and amplitudes to predict nulling patterns, burst rate variability, and the probability of detecting a pulse in a given observation window.

• What it can do: For sources like PSR J1726−6456 (nulling) or MTP0093 (extremely sporadic), the AI can predict optimal observation durations and cadence to maximize pulse detection, and estimate the likelihood of missing rare events.

5. Automated Timing Solution Derivation

• Improvement: Implement a neural network or Gaussian process regressor to fit pulse arrival times and derive coherent timing solutions (P, Ṗ) with uncertainty quantification, even from sparse data (e.g., 4 sources with full solutions).

• What it can do: Automatically produce precise ephemerides for new transients with few pulses, enabling phase-connected timing and accurate P−Ṗ placement (e.g., near the death valley) without manual intervention.

6. Multi-band Fluence and DM Estimation

• Improvement: Use a deep learning model to jointly estimate dispersion measure (DM) and fluence from single-pulse data across different frequency bands, accounting for scattering and band-dependent sensitivity.

• What it can do: Provide real-time DM and fluence estimates for new detections (e.g., DM range 21.4–398.0 pc cm−3, fluence 0.13–11 Jy ms), improving distance and energetics calculations, and enabling immediate comparison with known source populations.

7. Sporadic Source Detection and Prioritization

• Improvement: Train an anomaly detection algorithm on the burst rate distribution (0.05–88 hr−1) to identify ultra-sporadic sources (e.g., MTP0093, PSR J1449−5953) that require long-duration monitoring.

• What it can do: Automatically flag sources with extremely low detection probability and prioritize them for dedicated long-term campaigns, preventing them from being missed in standard survey pipelines.

8. Galactic Population Synthesis and Evolutionary Modeling

• Improvement: Use the new sample (27 sources, plus 95 total) to train a simulation-based inference model that maps observed properties (DM, P, Ṗ, burst rate, z-distance) to underlying neutron star birth rates, magnetic fields, and ages.

• What it can do: Predict the true fraction of RRATs vs. nulling pulsars in the Galaxy, and estimate the detection completeness of current surveys, guiding future telescope designs and survey strategies.