Fast radio bursts (FRBs) are millisecond-duration radio pulses of unknown origin, but at least some are linked to magnetars. Whether burst sequences are random or exhibit memory remains an open question: some studies point to clustering and power-law tails, others to nearly Poissonian behavior. Likely, the statistics depend on the observational timescale. FRB 20240114A is the richest data source: 11,553 bursts over 214 days, allowing memory to be probed over an unprecedentedly wide range. This work joins recent studies searching for order in the apparent chaos of repeating FRBs, but stands out by tracing, for the first time, the transition from randomness to non-stationarity within a single source.
We used the burst catalog of FRB 20240114A obtained with the 500-meter spherical radio telescope FAST. For the most active day (MJD 60381, 3,197 bursts in 4.38 hours), we analyzed the coherence of the event rate, searching for consecutive intervals with increasing burst frequency. Waiting-time distributions between bursts were fitted with sums of exponentials and a power law with a threshold, selecting models by the BIC information criterion. To reveal long-term memory, we applied the rescaled range (R/S) method, which estimates the Hurst exponent: 0.5 corresponds to a random walk, >0.5 to persistence, and >1 to non-stationarity. Energy distributions of bursts with different waiting times were compared using a model with a power-law segment and an exponential cutoff.
We found four independent pieces of evidence for scale-dependent memory. First, on the most active day, we found coherent rate-increasing structures lasting up to 3,600 seconds—the longest such interval among all repeating FRBs. Second, the waiting-time distribution for a single day is described by three exponentials, while the full 214-day sample requires a power-law tail—a signature of self-organized criticality. Third, R/S analysis of the full sample shows a break in the power law: at small lags, the Hurst exponent is about 0.63 (weak memory), and at large lags, 1.04 (non-stationary drift). The break corresponds to roughly one hour, matching the coherence limit. For energies, the picture is similar: H1 = 0.60, H2 = 1.10, but the break occurs at a different scale. Fourth, the slope of the energy distribution depends on waiting time: for short pauses it is shallower (α ≈ 1.39), for longer ones steeper (α ≈ 1.56), while the high-energy cutoff remains constant. This points to stress-dependent behavior akin to the seismic Gutenberg–Richter law and suggests that the maximum energy is an intrinsic property of the source.
This work demonstrates for the first time that the same FRB source can appear as a random process on short timescales and as a system with long memory and drift on long timescales. This resolves the contradiction between previous studies that saw either Poissonian statistics or power-law tails: it all depends on the duration of observations. Unlike neighboring works that either limited themselves to global measures of chaos or found stationary memory, here we identify a transition timescale of about an hour and non-stationarity with a Hurst exponent greater than one. This sets new benchmarks for magnetar burst models and emphasizes that long data series are needed to understand the dynamics.
The topic could develop toward comparative analysis of other hyperactive repeaters with rich statistics. It is important to determine whether the hourly timescale is universal for all FRBs or depends on source properties such as rotation period or magnetic field. Also promising is joint modeling of waiting times and energies within a unified physical picture, for example, accounting for slow magnetosphere evolution. Future observations with higher sensitivity and time resolution will help separate intrinsic effects from observational biases.
The results will influence the interpretation of data from radio telescopes such as FAST, CHIME, and future sky surveys, and the development of magnetar activity models. They may also be useful in analyzing other astrophysical transients, such as X-ray bursts from magnetars.
Immediate next steps include applying similar multi-level analysis to other repeating FRBs with sufficient burst counts, and testing the influence of observational effects on measured parameters through simulations. Additionally, monitoring of FRB 20240114A should be extended to refine the upper limit on memory timescale.
For radio astronomy, the key unsolved problem is the physics of FRB generation: what processes in the neutron star's magnetosphere produce such short and powerful pulses. This work links the statistical properties of bursts to possible self-organized criticality and slow magnetic field evolution, but does not answer the question of the emission mechanism. Another problem is the universality of behavior across different repeaters: it remains unclear how transferable conclusions from one source are to the entire population.
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
We analyzed 11,553 bursts from FRB 20240114A, recorded by FAST over 214 days. We studied waiting-time distributions, rate coherence, R/S analysis, and the relationship between energy and waiting time. We found a transition from Poisson statistics on short timescales to power-law tails and non-stationarity on long timescales, with a characteristic scale of about one hour.
Where its strength is
For the first time, we show that a single source exhibits both randomness on short timescales and long memory with drift on long timescales, resolving a contradiction between previous observations. The unprecedentedly large sample allowed us to robustly identify the transition scale using several independent methods.
Where it could go next
Worth trying
Why this may matter
If the transition from randomness to non-stationarity is a general property of repeating FRBs, it imposes constraints on the source physics: short bursts may be independent events, while long-term trends reflect the evolution of the magnetar or its environment. Understanding such memory is important for interpreting future observations and for model selection.
Nearby works in our archive
The archive is crowded around this work: 5 reviews nearby, closest at 0.72. This ground is well walked — and there is plenty to compare against.
The most active observing day yielded 3,197 bursts in 4.38 hours — on average one burst every 5 seconds. This is the highest rate among all known repeating FRBs. For comparison, a typical repeater emits only a few bursts per hour.