Introduction

There are two apparent types of cosmological fast radio bursts (FRBs): repeaters1 and apparently one-off bursts2 (which could be in principle repeaters as well). On the other hand, two nearby events offer clues to the engine and formation channels of FRBs: The apparently one-off burst FRB 20200428 from the Galactic magnetar SGR 1935+21543,4, points toward a magnetar engine formed from core-collapse supernovae. The FRB 20200120E, identified as a repeating FRB source through the Canadian Hydrogen Intensity Mapping Experiment FRB project (CHIME/FRB)5, and then localized by the European Very Long Baseline Interferometry Network (EVN) to a globular cluster [PR95] 30244 in the M81 galactic system at a distance of 3.63 Mpc6, points toward an unknown engine formed from a certain delayed channel. It is curious to see whether these two channels can be generalized to cosmological FRBs.

High-time resolution analysis of the super-narrow components (≲100 ns) of the bursts from FRB 20200120E shows high spectral luminosity7,8, which falls on the low-end of the luminosity distribution of cosmological FRBs. However, considering the burst energy release, there is still a gap between these nearby sources and cosmological FRBs. FRB 202004283,4 was about 1–2 orders of magnitude less energetic than the weakest known extragalactic FRB9,10,11. Many bursts were detected from FRB 20200120E: CHIME detected three bursts with fluences from 2.0 to 2.4 Jy ms in the frequency band of 400–800 MHz5, Effelsberg recorded 65 bursts with fluences from 0.04 to 0.71 Jy ms within the frequency band of 1200–1600 MHz6,12, and DSS-63 observed a burst with a fluence of 0.75 ± 0.15 Jy ms within the frequency band of 2192.5−2307.5 MHz7. Nonetheless, FRB 20200120E has even smaller energies, with the most intense burst still less energetic than FRB 20200428. All previously detected bursts from FRB 20200120E fall roughly in the middle between the Crab super-giant pulses13 and typical repeating FRBs10,11,14 (Fig. 1). One may speculate that the globular cluster channel may not form cosmological FRBs, and hence, cannot be a contributor to the cosmological FRB population.

Fig. 1: Isotropic-equivalent spectral luminosity versus the product of observing frequency and pulse width for various coherent radio pulses.
figure 1

The brightest burst from FRB 20200120E, detected in this study, is denoted by the red-filled star. Other bursts from FRB 20200120E, observed by CHIME5, DSS-637, and Effelsberg telescopes6,12, are represented by the orange, green, and tomato unfilled stars, respectively. Three well-studied localized repeating FRBs with known distances are also plotted: FRB 20121102A10 in violet crosses, FRB 20201124A11 in thistle crosses, and FRB 20180916B8,14,68 in blue crosses. Radio bursts or pulses from the Galactic magnetar SGR 1935+21543,4,15,69,70 are denoted by black squares, while super-giant pulses from the Crab pulsar13 are shown in tan. The pulsar65 and rotating radio transient (RRAT) populations (RRATalog) are marked in thistle and slateblue, respectively. The gray dashed lines denote constant isotropic-equivalent energy release, with no clear gap existing between various radio pulses. The gray dotted lines represent constant brightness temperature. Source data are provided as a Source Data file.

Full size image

In this work, we presented the detection of a bright radio burst from FRB 20200120E using the 40-meter Haoping radio telescope. The detection of this bright burst bridges the energy gap between nearby and cosmological FRBs. It suggests that FRBs hosted in globular clusters are bright enough to be observed at cosmological distances, providing further evidence that globular clusters could host cosmological FRBs, at least for the faint population.

Results

We observed FRB20200120E using the Haoping radio telescope from September 16 to 17, 2021, lasting 22.7 hours, and from April 13 to 15, 2023, spanning 39.8 hours. We detected a burst on September 16, 2021, with a barycentric arrival time of 14:39:27.2016 Universal Time (UT) at an infinite frequency. The burst was identified by both PRESTO and HEIMDALL, with a signal-to-noise ratio (S/N) of 13.8 and 11.0, respectively, for the full-band data (1100–1700 MHz). In addition, the narrow-band data (1100–1250 MHz) yielded S/N of 29.8 and 28.3 by PRESTO and HEIMDALL, respectively. Figure 2 presents the dynamic spectra and pulse profile of the detected burst, after being de-dispersed using the optimal dispersion measure (DM) value of 87.82 pc cm−3 (see Methods subsection “Observations and burst detection”), in general agreement with previous detections5,7,12,15,16.

Fig. 2: The detected burst using the Haoping telescope.
figure 2

a Integrated pulse profile and (b) dynamic spectra. The burst is plotted with time and frequency resolutions of 40.96 μs and 0.78 MHz, respectively, after being de-dispersed using a DM of 87.82 pc cm−3. The burst was detected on September 16, 2021, with a barycentric arrival time of 14:39:27.2016 UT at an infinite frequency. Source data are provided as a Source Data file.

Full size image

Comparison of the burst with other bursts from FRB 20200120E

The burst has a frequency extent of 1130(4)–1220(4) MHz, a peak flux density of 188 ± 38 Jy, and an averaged fluence of 30 ± 6 Jy ms (see Table 1). This fluence value is about 13 times greater than the brightest burst previously detected5. Focusing on the L-band, its fluence is 42–754 times brighter than previous ones12,16. An energy distribution analysis of the burst sample from the Effelsberg telescope revealed a steep power-law function with an index of approximately −2.4 at high fluences, and suggest that unless FRB 20200120E has a bi-modal and/or time-variable energy distribution, observations in L-band would be unlikely to detect bursts much exceeding a fluence of 2 Jy ms, such as the detection we made12. Our only detection with S/N about 30 and the search threshold of S/N = 7 imply a much shallower energy distribution function. This discrepancy might be explained by introducing two distinct energy distributions in different emission states.

Table 1 The properties of the bright burst from FRB 20200120E
Full size table

As a significant number of bursts within the Effelsberg sample were obtained during a 40-minute burst storm, we divided this Effelsberg dataset into two subsets: bursts in the burst storm and bursts excluding the burst storm. Remarkably, the latter subset displays a relatively flat energy function with a power-law index of approximately ‒0.98. The detected Haoping burst aligns with the extrapolation of this energy function (Fig. 3). Despite being detected at different frequency ranges, it is noteworthy that the bursts recorded by the CHIME and DSS-63 telescopes also align well with the second flatter power-law line. Notably, while the first steep power-law line could also provide a fit for these two samples, our detection significantly deviates from its extrapolation. Incidentally, a two-component energy function has been identified from other active repeating sources10, with FRB 20201124A showing a flatter power-law tail at higher fluences17. This suggests that the two-component energy function may be a characteristic of some repeating FRBs.

Fig. 3: Distribution of specific energies for bursts originating from FRB 20200120E.
figure 3

The scatter markers in gray and light blue correspond to the burst storm and other bursts detected by the Effelsberg telescope (about 1400 MHz), respectively. These bursts were fitted with dashed black and blue power law lines with 1 σ uncertainty for higher fluences, respectively. The orange, green, and red crosses represent bursts with associated errorbars detected by the DSS-63 telescopes (about 2250 MHz), CHIME telescopes (about 600 MHz), and this work, respectively (see Methods subsection “Bursts of FRB 20200120E from different telescopes''). Notably, the bursts of the CHIME and DSS-63 telescopes were detected at different frequency ranges. Source data are provided as a Source Data file.

Full size image

Based on Equation (1), high luminosities derived from the super-narrow components typically do not significantly increase the detectability of the bursts (see Methods subsection “Detectability of the bursts from FRB 20200120E”). For previous bursts detected at L-band with the largest fluence8 and peak flux8, the Five-hundred-meter Aperture Spherical Radio Telescope (FAST)18 could obtain 7σ detections at distances of approximate 43 Mpc and 46 Mpc, respectively. Notably, if positioning FRB 20200120E at a cosmological distance, such as the distance of the nearest localized cosmological repeating FRB 20180916B at 149 Mpc14, only this bright “Haoping burst” would be detectable by FAST. Consequently, the repeating FRB 20200120E would appear as a cosmological apparent one-off FRB.

Comparison of the FRB 20200120E with other radio sources

Using the distance to the M81 globular cluster [PR95] 30244 of 3.63 Mpc6, we derive a specific luminosity of 3.0 ± 0.6  × 1030 erg s−1 Hz−1 and an isotropic-equivalent energy of 5.6 ± 1.1 × 1035 erg for the burst. This energy is about five times greater than the brightest radio burst from SGR 1935+21543,19, and is about 1/3 of the weakest burst of FRB 20121102A10 (see Methods subsection “Estimation and comparison of isotropic-equivalent energy”). With the inclusion of this detected burst, the energy level of FRB 20200120E aligns closely with the main FRB population (Fig. 1). The energy release range of this source now spans up to three orders of magnitude.

Comparison of the FRB 20200120E environment with other FRBs

The environments of transients usually reveal the population of the progenitors. In general, progenitors from young populations reside in the central bright region of the host galaxies, with small offsets from the galaxy centers, while those from old populations are in the outer faint region of the host galaxies and have large offsets20,21. FRB 20200120E is located on the edge of the host galaxy, with a large offset of \(2{0}_{-2}^{+3}\) kpc5, consistent with a globular cluster origin.

In order to examine whether other FRBs are consistent with a globular cluster origin, we examine the normalized offsets of them with respect to the host galaxy centers based on the available data22,23. In addition, for FRBs that lack host galaxy radius or offset information, we searched for the host galaxy information with the FRB positions in the SDSS, DESI/Legacy Survey as well as Pan-STARRS catalog, and estimated the offsets with the FRBs and host galaxy position. For host galaxies detected with Pan-STARRS catalog only, we fit the r-band with GALFIT to obtain the half-light radii as well as the uncertainty. The results are presented in Fig. 4, which show that a fraction of cosmological FRBs indeed have a normalized offset similar to or larger than that of FRB 20200120E. In particular, FRB 20190611B and FRB 20190523A have normalized offsets of 5.4 ± 2.7 and 8.3 ± 6.9, respectively22, which are away from the stellar light of the host galaxies. These positions are more likely associated with older population environments such as globular clusters.

Fig. 4: Cumulative distribution of FRB offsets.
figure 4

The offset of 27 FRBs from the center of the host galaxies, normalized to the half-light radii r50 of the hosts. The blue line and gray region represent median value and 1 σ uncertainty. One can see that two FRBs have normalized offsets comparable to that of FRB 20200120E. Source data are provided as a Source Data file.

Full size image

Discussion

There is evidence that at least some active repeating FRBs track star formation history and therefore originate from prompt formation channels of the FRB engine (such as magnetars)24,25. However, it is still under debate whether all FRBs formed from the prompt formation channels. For instance, analyses of the DM distribution of all FRBs in the first CHIME FRB catalog26,27 indicate the necessity of a delayed population with respect to star formation history together with the prompt population tracking star formation. A subset of the first CHIME FRB catalog, which excludes bursts of low DM or low DM excess with respect to the Galactic contribution, seems to track star formation history28, similar to some nearby localized FRBs29. The detection of this bright event from FRB 20200120E, which is bright enough to be detected at cosmological distances, suggests that the delayed channel can contribute to the FRB population, at least to the population of faint FRBs. Moreover, few localized FRBs at larger distances have similar or even larger offsets from the center of the host galaxy than FRB 20200120E (Fig. 4). These bursts could be candidates that reside in globular clusters.

The central engine that powers FRB 20200120E remains unclear. Theoretically, there are three possibilities for FRBs in globular clusters (see Methods subsection “Analysis of the implications for FRB models” for details): young magnetar bursts30,31,32,33, giant pulses from young pulsars34, and interacting neutron stars35,36,37,38,39,40. The first two models involve young magnetars or pulsars. In globular clusters, this would require a different formation channel from the core-collapse scenario. It has been suggested that the mergers of compact binary stars, such as white dwarfs and neutron stars, and the accretion-induced collapse of white dwarfs could be potential formation channels6,16,41,42. The interacting neutron stars model involves the release of gravitational energy during the inspiral of a neutron star binary or the collision between asteroids and a neutron star.

However, FRB 20200120E is the only source so far localized to a globular cluster6, and only one detected burst from this source could be observed at a cosmological distance (see Methods subsection “Detectability of the bursts from FRB 20200120E”). The absence of such bursts could be attributed to several factors: (1) The low event rate of bright bursts from these sources. Even with a 1σ confidence level, about 360 hours of observation is necessary to detect a burst similar to the “Haoping burst”. (2) The requirement for very sensitive telescopes, which typically have a small field of view (FOV), to detect such bursts at cosmological distances. (3) The current most efficient FRB hunter, CHIME, is only sensitive to such bursts at a distance of ≲ 10 Mpc (see Methods subsection “Detectability of the bursts from FRB 20200120E”), limiting the number of potential host galaxies for generating this phenomenon. Next-generation telescopes that combine both a large FOV and high sensitivity (e.g., SKA) could provide a more complete FRB sample. Until then, searching for more bright bursts from FRB 20200120E with sufficient but not very sensitive observations, and localizing more FRBs in nearby galaxies or globular clusters through specified designed observational campaigns43, would be valuable for further understanding FRB engines and their formation channels.

Methods

Observations and burst detection

Observations for this study were carried out using the 40 m − diameter Haoping radio telescope, which consists of two epochs: Epoch 1 from September 16 to 17, 2021, lasting 22.7 h, and Epoch 2 from April 13 to 15, 2023, spanning 39.8 hours. The receiver was within the L-band, covering a frequency range of 1100 to 1700 MHz. The single-polarization signals were 8-bit sampled and channelized using the Reconfigurable Open Architecture Computing Hardware generation 2 (ROACH 2)44. Subsequently, the data were stored in the PSRFITS search mode format45. The specific sample time and channel width for Epoch 1’s observation were set at 40.96 μs and 0.195 MHz, respectively. Due to the site’s observation constraints during Epoch 2, the sample time was 81.92 μs, and the channel width was 0.390 MHz. To verify the observation setup and search pipelines, a bright pulsar PSR J0332+5434 has been observed prior to each FRB 20200120E observation.

The data collected from the Haoping radio telescope were processed to create three different kinds of band datasets for search: the full-band (1100–1700 MHz), half-band (1100–1400 MHz and 1400–1700 MHz) and narrow-band (1100–1250 MHz, 1250–1400 MHz, 1400–1550 MHz and 1550–1700 MHz). These datasets were analyzed via two individual search pipelines, based on the pulsar/FRB single pulse searching packages presto46 and heimdall47. The datasets were dedispersed in a range of DM values from 78 to 98 cm−3 pc, with a step size of 0.01 cm−3.

Any candidate with an S/N greater than seven was recorded and subject to visual inspection. The DM for any detected burst will be derived by maximizing the S/N of its integrated pulse profile. Within the 62.5 hours of observation, a burst was detected in the data from Epoch 1, occurring on September 16, 2021, with a barycentric arrival time of 14:39:27.2016 Universal Time (UT) at an infinite frequency. Both PRESTO and HEIMDALL successfully identified this burst, yielding S/N of 13.8 and 11.0 for the full-band data (1100–1700 MHz), S/N of 22.5 and 18.8 for the half-band data (1100–1400 MHz), and S/N of 29.8 and 28.3 for the narrow-band data (1100–1250 MHz), respectively.

Estimation of flux, fluence, and energy of the burst

To estimate the burst’s flux densities, we employed two methods:

(1) Radiometer equation approach: The expected root-mean-square (RMS) of the off-pulse data in Jy could be derived using the radiometer equation48:

$$\Delta {S}_{{{{\rm{sys}}}}}=\frac{{T}_{{{{\rm{sys}}}}}}{G\sqrt{\Delta \nu {N}_{p}{t}_{{{{\rm{obs}}}}}}},$$
(1)

where we ignore the loss factor owing to the 8-bit sampling, Δν = 90 MHz and tobs = 40.96 μs are the specified frequency range (1130–1220 MHz) and time resolution, and Np = 1 as only the right-handed single-polarization was employed. Tsys is the system temperature and G is the telescope antenna gain. Notably, the Haoping telescope is equipped with a room-temperature receiver, and the latest measurements of Tsys and G for the telescope at L band is about 120 K and 0.26 K/Jy, respectively, which are consistent with the previously reported values for Haoping49 and another similar 40-m telescope from Kunming, China50. By calibrating the data using ΔSsys, we obtained a peak flux of 188 ± 38 Jy, accounting for a 20% uncertainty in system temperature fluctuation. The fluence was computed by integrating the burst flux above the baseline, while the effective width was determined by dividing the fluence by the burst peak flux.

(2) Comparison to know pulsar: Due to the relatively low sensitivity and severe RFI environment of the Haoping telescope, we only used the bright pulsar PSR J0332+5434 as the comparison. We examined data (MJD 59473) of PSR J0332+5434 prior to Epoch 1 and opted for a 30-min observation minimally affected by radio interference (RFI). To reduce the scintillation impacts, we also examined the Haoping archive of J0332+5434 and obtained an additional three observations (MJD 58831,59579,59588) with sufficient tracking (≥30 min) and minimal RFI impact. The pulsar data were folded using the dspsr51 and psrchive52 packages, based on its time ephemeris53. Utilizing the mean flux density at 1400 MHz and the period of PSR J0332+543454, we scaled the fluence of our burst data to Jy ms units based on the four observations of PSR J0332+5434. This approach resulted in a fluence of 25 ± 8 Jy ms (corresponding to peak flux of 154 ± 52 Jy), accounting for a 30% uncertainty related to the variation of pulsar data.

Notably, the flux estimations attained from both methods exhibit congruence. For the main text, we adopted the value derived from the first approach based on the latest measurement for gain and system temperature for the Haoping telescope. The estimation based on the pulsar could be a useful comparison and reference point.

Detectability of the bursts from FRB 20200120E

If positioning this “Haoping burst” at the distance of the nearest other extragalactic repeating FRB, FRB 20180916B at 149 Mpc, it would exhibit a peak flux of approximately 111 mJy. The FAST telescope has a system temperature of about 20 K and a gain of 16 K/Jy18. Based on Equation (1), taking into account Np = 2 and tobs using the burst width, the sensitivity of this high-sensitive instrument at L-band could yield a 16σ detection. For the “Haoping burst”, previous bursts detected at L-band with the largest fluence8 and peak flux8, if sufficient time resolution is available to identify the bursts, FAST could obtain 7σ detections at distances of approximate 225 Mpc, 43 Mpc and 46 Mpc, respectively.

Additionally, considering the gain and system temperature for CHIME and the Parkes cryoPAF, which are 1.16 K/Jy and 50 K55, and 0.735 K/Jy and 20 K56, respectively, these two instruments could obtain 7σ detection for “Haoping burst” at distances of approximate 37 Mpc and 47 Mpc, respectively. The CHIME FRBs with baseband data could significantly increase source localizations57, enabling the determination of whether bursts with low DM or low DM excesses with respect to the Galactic contribution are extragalactic sources. However, the current CHIME system would only record the baseband data containing real-time detections with high S/N (usually set between 10 and 12) at a time resolution of approximately 1 ms. Such a system would only be sensitive to a burst similar to the “Haoping burst” at distances of ≲10 Mpc.

Effective time resolution

The effective time resolution of our observations could be estimated by58

$${w}_{{{{\rm{eff}}}}}=\sqrt{{w}_{{{{\rm{DM}}}}}^{2}+{w}_{\Delta {{{\rm{D}}}}M}^{2}+{w}_{{{{\rm{sample}}}}}^{2}},$$
(2)

where wDM is the dispersion smearing about 73–107 μs at a frequency of 1100–1250 MHz with channel width of 0.195 MHz, wΔDM is caused by the dedispersion error which can be ignored for our observation, wsample = 40.96 μs is the sample time of Epoch 1. Our observation has an effective time resolution of approximately from 83 to 114 μs for frequency from 1100 to 1250 MHz. This resolution is similar to the width of the burst we detected of 161 μs. Due to the effective time resolution limit, we can not explore its possible narrower components, which are well studied by DSS-637 and Effelsberg16. However, it is notable that even derived from such a large time resolution, the peak flux density of our burst of approximately 188 Jy is slightly smaller than their peak flux derived from time resolutions of  <100 ns.

Bursts of FRB 20200120E from different telescopes

In previous studies, burst detections from FRB 20200120E have been reported by three telescopes. According to the CHIME real-time FRB report, eight candidate bursts from this source were identified within the frequency range of 400–800 MHz, yet the properties of only three of them are available5. These three bursts were detected from January 20 to November 29, 2020, corresponding to an expected on-source time of about 41 h59. However, the on-source exposure would increase to about 111 h if one uses the starting date of the CHIME FRB project on July 25, 2018. To avoid included observation of CHIME with different sensitivity at different stages, we only employed the properties and expected exposure time of these three bursts to estimate the event rate of CHIME detections60. The Effelsberg telescope recorded a total of 65 bursts, with fluences ranging from 0.04–0.71 Jy ms within the frequency band of 1200–1600 MHz6,12. The DSS-63 telescope identified a burst with a fluence of 0.75 ± 0.15 Jy ms within the frequency band of 2192.5–2307.5 MHz7.

Luminosities and burst durations for various coherent radio pulses

To ensure consistency with previous investigations of various coherent radio pulses16,61, we presented diverse coherent radio pulses in a two-dimensional plane in Fig. 1. The horizontal axis represents the transient width multiplied by the central observing frequency, while the vertical axis depicts the isotropic-equivalent spectral luminosity. The burst detected in this study was present using the central burst frequency. The primary focus of this study is energy release, we omitted the results from the detailed burst temporal structures and applied the effective width or the width of the pulse at 50% of its peak (W50) of the whole event as the burst duration. For pulsars and RRATs, we presented the average properties of pulses from these sources. However, individual pulses could vary in terms of brightness and duration, typically by approximately 1 order of magnitude, with some deviations even reaching up to 3 orders of magnitude62. The luminosities were estimated through peak flux density. It’s noteworthy that the luminosities of pulsars were calculated using the mean flux density averaged over the pulsar period (P0)48. Throughout this study, we modified the pulsar luminosities to isotropic-equivalent spectral luminosity by multiplying them by the ratio of P0 to W50.

Estimation and comparison of isotropic-equivalent energy

Three methods could be used to compare the energies of different radio pulses63: (1) the specific isotropic-equivalent energy:

$${E}_{S}=\frac{4\pi {D}_{L}^{2}}{1+z}\,F,$$
(3)

where DL is the luminosity distance, z is the redshift, and F is the pulse fluence. (2) the isotropic-equivalent energy by scaling ES by central frequency (Ecf), and (3) isotropic-equivalent energy by scaling ES by burst frequency extent (Efe).

As our primary comparisons involve our burst, FRB 202004283 and FRB 20121102A10, in the main text, we used (2) to be consistent with their calculations. The estimated energy of our burst, FRB 20200428 and the weakest burst of FRB 20121102A are about 5.6  × 1035, 1.1  × 1035, and 1.7  × 1036 erg, respectively. Applying (1), the estimated specific energy of these three events are about 4.7  × 1026, 8.0  × 1025, and 1.3  × 1027 erg Hz−1, respectively. Applying (3) the estimated energy of these three events are about 4.2  × 1034, 1.5  × 1034, and 1.4  × 1035 erg, respectively.

Analysis of the implications for FRB models

The observational properties of FRBs (short duration, high brightness temperature, strong linear polarization, etc) have propelled theories involving compact stars, especially neutron stars, as potential sources64. These theories encompass scenarios involving highly magnetized neutron stars or magnetars30,31,32,33, young Crab-like pulsars34, or various interacting neutron stars in binary systems36,38,39,40 or with small bodies35,37. The fact that the Haoping burst carries an energy level exceeding FRB 20200428 from SGR 1935+2154 and is comparable with the faintest cosmological FRBs raised the perspective of interpreting all FRBs with a unified magnetar engine. Nonetheless, in terms of energetics, several other scenarios are still allowed besides the magnetar model. In the following, we discuss various models in turn.

1. Young magnetar burst model. For the magnetar models, the energy of the bursts mainly comes from the dissipation of internal magnetic energy. For a typical magnetar with an internal field B* = 1014B14 G an a canonical radius RNS = 106 cm, the internal magnetic energy is \({E}_{B}=1.67\times {B}_{14}^{2}1{0}^{45}\) erg. The average energy dissipation rate is

$$\dot{{E}_{B}}={E}_{B}/{t}_{{{{\rm{amb}}}}}=5.6\times 1{0}^{32}{B}_{14}^{3.2}\,{{{\rm{erg/s}}}},$$
(4)

where the dissipation time due to ambipolar diffusion is \({t}_{{{{\rm{amb}}}}}=2\times 1{0}^{5}{B}_{14}^{-1.2}\) yr30. The average isotropic equivalent luminosity of this event over our observation time is \(\bar{L}=5.79\times 1{0}^{35}\,{{{\rm{erg}}}}/62.5\,{{{\rm{h}}}}=2.6\times 1{0}^{30}\) erg/s41,42. Assuming a radiation efficiency fr = 10−5fr,−5 of the FRB event3, the magnetar producing FRB 20200120E should have an internal magnetic field \({B}_{14}\approx 6.7{f}_{r,-5}^{-0.31}\) and a corresponding dissipation time \({t}_{{{{\rm{amb}}}}}=9.4\times 1{0}^{3}{f}_{r,-5}^{-0.375}\) yr. This suggests that a young magnetar is enough to power this source.

2. Young pulsar giant pulse model. The giant pulse like model assumes that FRBs are powered by the spin-down of NSs34. For a canonical NS with moment of inertia INS ≈ 1045g cm2, the spin-down luminosity due to magnetic dipole radiation is

$${L}_{{{{\rm{sd}}}}}=4.8\times 1{0}^{42}{B}_{12}^{2}{P}_{-3}^{-4}{(1+t/{\tau }_{{{{\rm{sd}}}}})}^{-2}\,{{{\rm{erg/s}}}},$$
(5)

where \({\tau }_{{{{\rm{sd}}}}}=130{P}_{-3}^{2}{B}_{12}^{-2}\) yr is the spin-down time scale of an NS with surface magnetic field B = 1012B12 G and initial period P = 10−3P−3 s. It can be seen that for t < τsd, the luminosity remains almost constant, otherwise it will decrease as Lsd ∝ t−2. As this FRB is still active, we mainly consider the case of for t < τsd. The peak luminosity of this event is Lp = 3.6 × 1039 erg/s. Taking a radiation efficiency of giant pulses at fr = 10−1fr,−134, we found that the NS needs to have an initial period \({P}_{-3} \, \lesssim \, 2.3{B}_{12}^{1/2}{f}_{r,-2}^{1/4}\) and a corresponding spin down time of \({\tau }_{{{{\rm{sd}}}}} \, \lesssim \, 686{B}_{12}^{-1}{f}_{r,-2}^{1/2}\) yr. Recycled pulsars in globular clusters can have millisecond periods, but with a weak surface magnetic field (≲1010 G)65. Therefore, recycled pulsars can not power such an event, and a young NS is required. However, the possibility of associating with pulsar wind nebulae has been disfavored from X-ray observation66.

3. Interacting neutron star model. There are in general two types of magnetosphere interaction models that form NS binaries: NS-NS inspiral model36,38,39,40 and NS-asteroid interaction model35,37. For the NS-asteroid interaction model, the NS may have a main-sequence star companion, which hosts an asteroid belt. The NS may interact with the asteroids and produce FRBs. The FRB energy mainly comes from the gravitational potential of the planet, which is

$$E=1.9\times 1{0}^{38}{m}_{18}\,{{{\rm{erg}}}},$$
(6)

for a planet with mass m = 1018m18 g37. For the observed event with energy 8.70 × 1032 − 5.79 × 1035 erg16, this requires asteroids to be 4.7 × 1012 − 3.1 × 1015 g.

The NS-NS spiral model is usually supposed to be responsible for non-repeating FRBs, as it can produce a luminous event during the last orbits. But it also works when the NSs are widely separated, although the luminosity will become much lower. For two NSs with equal surface magnetic field and anti-parallel magnetic axes, the electromagnetic luminosity from magnetosphere interaction is

$${L}_{{{{\rm{inspiral}}}}}=3.4\times 1{0}^{39}{B}_{12}^{2}{a}_{9}^{-2}\,{{{\rm{erg/s}}}},$$
(7)

at a separation of a = 109a9 cm38. While the merger time for NSs with mass M = 1.4M⊙ is \({t}_{{{{\rm{merger}}}}}=3.2{a}_{9}^{4}\) yr for binaries with separation a ≫ RNS assuming a circular orbit. Again assuming the radiation efficiency of giant pulses, the requirement on the surface magnetic field of both NSs is \({B}_{12}=3.2{a}_{9}{f}_{r,-1}^{-1/2}\). Therefore, in principle, NS-NS inspiral cannot be excluded from this event, which could potentially serve as a new type of electromagnetic counterpart of gravitational wave events.