WorksPhysics and Astronomy
A high-frequency turnover in the persistent radio source of FRB20201124A
Abstract
Persistent radio sources (PRSs) associated with repeating fast radio bursts probe the immediate magneto-ionic environment of their engine and thereby constrain the nature and evolutionary stage of the progenitor. The persistent radio source of FRB20201124A is the most extreme in the current sample, with a spectrum that remains flat or inverted up to 22 GHz. We aim to constrain the centimetre-to-millimetre spectral energy distribution of this source and to test whether its properties are consistent with a young, rotation-powered magnetar wind nebula. We analysed archival ALMA Band 3 observations at GHz and combined them with existing VLA measurements at 6, 15 and 22 GHz. The resulting spectrum was fitted with the Granot–Sari broken power law in the slow-cooling regime using Bayesian nested sampling. The two-parameter model yields a peak frequency GHz and a peak flux Jy. This solution is consistent with theoretical expectations for a young magnetar wind nebula, which predict a peak frequency in the same range and a self-absorption frequency below GHz. With its inverted spectrum extending to tens of gigahertz and the highest peak frequency among known PRSs, the source associated with FRB20201124A is the most extreme in the current sample. Its properties match those expected for a rotation-powered magnetar wind nebula, making it a strong candidate for the youngest ( yr) magnetar associated with a repeating FRB to date.
Introduction
Fast radio bursts (FRBs) are millisecond-duration radio transients of extragalactic origin, with dispersion measures (column density of free electrons) that place the majority of them at cosmological distances [19, 6]. A growing fraction of FRBs are observed to repeat, and the repeating population has enabled precise localisations and detailed studies of their host galaxies and local environments [24, 5]. A subset of repeating FRBs is associated with compact PRSs, whose luminosities and spectra are difficult to reconcile with star formation alone and have been interpreted as emission from a magnetar wind nebula or a young supernova remnant surrounding the central engine [13, 14, 3, 2, 1, 18] (Moroianu et al. 2026).
PRSs are the immediate magneto-ionic environment of the FRB central engine. Their luminosities and rotation measures therefore provide direct constraints on the progenitor and on the evolutionary stage of the surrounding nebula. A simple relation between PRS radio luminosity and FRB Faraday rotation measure (RM) was predicted on theoretical grounds by Yang et al. [26, 27], under the assumption that both quantities arise from the same magnetised plasma. This relation was observationally established and extended into the low-luminosity, low-RM regime by the detection of the PRS associated with FRB 20201124A [3]. More recently, Yang [25] showed that the intrinsic scatter of the – relation encodes information on the temporal evolution of the nebula radius, offering a diagnostic of the expansion history of the FRB environment.
The repeating fast radio burst FRB 20201124A was localised to a barred spiral galaxy at and found to be embedded in a region of elevated star formation (Piro et al. [20]; Fong et al. [7]). Multi-frequency VLA observations revealed a compact persistent radio source coincident with the milliarcsecond FRB position, with a spectrum that remains flat or inverted up to 22 GHz (Bruni et al. [3]). Although high-resolution VLBI observations have not yet yielded a detection (see Sect. 2.2), the source is considered a bona fide PRS on the basis of several independent indicators: (i) its precise positional coincidence with the FRB, (ii) its inverted spectrum, which is inconsistent with the steep or flat spectra typical of star-formation-related emission, and (iii) the fact that the star-formation rate required to account for its radio luminosity ( yr) exceeds both the H-derived SFR measured with GTC/MEGARA at the FRB location ( yr) and the upper limit from NOEMA millimetre observations ( yr) (Bruni et al. [3]). This inverted spectrum already set FRB 20201124A apart from the other known PRSs, whose emission typically peaks at lower frequencies and declines above a few gigahertz, and made it the most extreme object in the current sample in terms of spectral energy distribution.
First indications of the spectral shape of other PRSs have recently become available. Bruno et al. [4] reported a flat spectrum () between 1 and 6 GHz for the PRS associated with FRB 20190417A, and placed a lower limit on the turnover frequency ( MHz at 95% confidence) from the non-detection at 144 MHz. Pelliciari et al. [18] presented a similarly tentative SED for another PRS, constructed from a single VLBI measurement and two upper limits. While these data already hint at a possible turnover at relatively low frequencies, the limited sampling and mixed resolution make the constraints preliminary.
Subsequent modelling of PRS spectra has interpreted the radio emission as synchrotron radiation from a magnetar wind nebula or a young supernova remnant, with the peak frequency marking the transition between optically thick and thin regimes (Rahaman et al. [22]). In the Granot–Sari formalism adopted by Rahaman et al. [22], the location of the peak and the spectral indices on either side constrain the magnetic field, the electron energy distribution, and the size of the emitting region. For FRB 20201124A the existing data do not go beyond 22 GHz, leaving the peak frequency and the optically thin slope essentially unconstrained.
In this work we present the first radio-to-millimetre spectral energy distribution of the PRS associated with FRB 20201124A, extending from 6 to GHz with new ALMA Band 3 observations. We fit the SED with the Granot–Sari model as formulated by Rahaman et al. [22] and observe the first quantitative constraints on the physical parameters of the emitting region. Throughout the work we adopt a standard CDM cosmology (Planck Collaboration et al. [21]) with km s Mpc, and .
Observations and archival data
ALMA
We used archival ALMA Band 3 observations from project 2023.1.01090.S (PI: Hatsukade), taken on 2023 December 31. The data are centred at GHz, with a synthesised beam of 1.69 1.33 at a position angle of in the 12-m array configuration. The epoch is 2 yr after the VLA observations reported by Bruni et al. [3].
The measurement set was calibrated with the ALMA pipeline and imaged in CASA (McMullin et al. [15]) with natural weighting, chosen to minimise the noise. The rms in the final image is 11 Jy beam. The host is detected and appears elongated north–south, tracing the bar seen in the HST imaging of Bruni et al. [3], with a brightness peak near the optical centre of the galaxy.
Because the synthesised beam is larger than the projected separation between the FRB and the host-galaxy centre, the Gaussian wings of the point-spread function of the brighter central component contaminate the measurement at the FRB position. To separate the contribution of the central component from that at the FRB position, we fitted the elongated structure with two Gaussians using the CASA task imfit, with initial estimates taken from a preliminary two-component fit and providing the image rms so that the fit properly accounts for the noise.
The fit converged to a northern component at 05h08m03.5016s, +26°03′38.595″ and a southern component at 05h08m03.483s, +26°03′37.363″ (see Figure 1, left panel). The northern position is consistent with the milliarcsecond FRB localisation of Nimmo et al. [17]; the southern position coincides with the host centre. The integrated flux densities returned by the two-component fit are 20 11 Jy for the northern component and 46 19 Jy for the southern one. The uncertainties combine the image noise with the formal error of the Gaussian fit. We note that the emission at the FRB position is significant at in the image before deblending; the lower flux density obtained for the northern component is the expected result of subtracting the contribution from the Gaussian wings of the brighter southern component. At this signal-to-noise ratio both components remain unresolved, so the reported values should be interpreted as point-source flux densities.

Figure 1. Left: ALMA Band 3 image of the host of FRB 20201124A at GHz (colour scale, in Jy beam). The emission was fitted with two Gaussian components (dashed ellipses indicate the major and minor axes FWHM) centred on the positions marked by the red squares. Contours are 2, 3, 4 the local rms of 11 Jy beam. The green square indicates the milliarcsecond FRB position reported by Nimmo et al. [17]. Right: the same ALMA image with superimposed VLA 15 GHz contours from Bruni et al. [3] (red; 3–9 the local rms of 2.3 Jy beam). The northern ALMA peak and the compact VLA persistent source coincide with the FRB within the positional uncertainties; the southern ALMA peak traces the host centre. The synthesised ALMA beam (1.69) is shown in the lower left corner of both panels, together with the VLA beam in the right panel (0.38, small circle).
Figure 1 (right panel) overlays the ALMA contours (black) on the VLA 15 GHz contours (red) from Bruni et al. [3]. The
FRB position, the compact persistent source identified with the VLA, and the northern ALMA component agree within the positional uncertainties.
HSA
We additionally observed the PRS of FRB 20201124A with the High Sensitivity Array (HSA) in two epochs, both centered to the milliarcsecond FRB position of Nimmo et al. [17]. The first epoch (BB464B) was taken on 2025 April 20–21, from 21:31 to 02:48 UTC ( 5.3 h on source), with eleven stations: the VLBA antennas BR, FD, HN, KP, LA, MK, NL, OV, PT and SC, plus the phased VLA (Y). The second epoch (BB464A) was taken on 2025 May 16, from 13:45 to 19:03 UTC ( 5.3 h), with ten stations: BR, HN, KP, LA, MK, NL, OV, PT, plus Y and Effelsberg (EB). Both epochs used four 128 MHz spectral windows centred at 15.176, 15.304, 15.432 and 15.560 GHz (total bandwidth 512 MHz), recording full polarisation. The data were calibrated with the rPICARD pipeline (Janssen et al. [11]), and imaged in CASA. With natural weighting the image rms is 8 Jy beam. No source is detected at the FRB position, corresponding to a 5 upper limit of 40 Jy beam. This limit lies above the VLA flux density of 20.0 3.5 Jy measured at the same frequency by Bruni et al. [3], so the observation is not deep enough to test whether the persistent emission is compact on parsec scales. A detection, or a limit below the VLA flux, would require a significantly lower rms. The non-detection is therefore not in tension with the VLA flux density and does not challenge the association of the persistent source with the FRB.
Modelling the magnetar wind nebula spectrum
The centimetre-to-millimetre spectrum of the persistent source associated with FRB 20201124A is not a single power law. VLA measurements at 6, 15 and 22 GHz give 8.2 3.8, 20.0 3.5 and 30.0 9.7 Jy (Bruni et al. [3]), while our ALMA detection at 100 GHz, 20 11 Jy, lies below a continuation of the rising VLA spectrum. The spectral turnover suggested by these four measurements is the key observable that we compare with the predictions of a magnetar-wind nebula (MWN) model.
We adopt the slow-cooling synchrotron spectrum of Rahaman et al. [22], which is the Granot–Sari broken power law (Granot & Sari [9]) in the slow-cooling phase (). Synchrotron self-absorption is predicted below GHz for the nebular parameters of this source, so it does not shape the VLA–ALMA window. Between and the injection break the optically thin spectrum rises as
while above and below the cooling break it falls as
We fix the electron index at the value used by Rahaman et al. [22], , which gives a high-frequency slope of . The only free parameters are then the break frequency and the flux density at that break. The model spectrum is normalised so that is the flux of the two asymptotic branches at ; the smooth join peaks slightly below .
We fit the radio spectrum with this broken power-law model using Bayesian statistics. In particular, we use a dynamic nested sampling algorithm through the dynesty python package ([10, 23, 12]). The two parameters have log-uniform priors: in 1–300 GHz and peak flux between 1 and 500 Jy. We find
See Figure 2. The spectral peak luminosity corresponding to is .

Figure 2. Left panel: radio-to-millimetre spectrum of the persistent source associated with FRB 20201124A. Red points are VLA measurements at 6, 15 and 22 GHz from Bruni et al. [3]; the yellow star is the ALMA Band 3 detection at 100 GHz presented in this work. Upper limits from NOEMA (Bruni et al. [3]) are shown as downward arrows. The solid line is the median of the 10 000 best-likelihood spectra, and the shaded blue areas within the dashed lines show the 1, 2 and 3 credible regions. Right panel: corner plot of the two-parameter fit shown in the left panel, made using corner (Foreman-Mackey [8]).
This model fits the data well, but slightly over-predicts the flux at 6 GHz. For completeness we also performed a three-parameter fit that includes the self-absorption frequency as a free parameter. However, this more complex model is not statistically preferred. The Akaike Information Criterion favours the simpler two-parameter model, and the Bayes factor likewise supports the baseline model. We therefore adopt the two-parameter solution as our primary result. This is also consistent with the theoretical expectations of Rahaman et al. [22], who predict that the self-absorption frequency of a magnetar wind nebula should lie below GHz. Such a low value does not affect the detectability of the FRB bursts, which have been observed down to a few hundred MHz.
Conclusions
The centimetre-to-millimetre spectrum of the persistent source associated with FRB 20201124A shows a clear turnover. Bayesian fits with the Granot–Sari broken power law of Rahaman et al. [22] (slow-cooling regime, ) yield a peak frequency GHz and a peak flux Jy. The model is consistent with the theoretical expectations of Rahaman et al. [22], who predict a self-absorption frequency below GHz for a young magnetar wind nebula. At such frequencies the nebula remains optically thin to the FRB emission, in agreement with the detection of bursts down to a few hundred MHz.
In the young-magnetar scenario of Rahaman et al. [22], a rotation-powered magnetar wind nebula can sustain the observed radio luminosity only while the system is still young ( yr). Beyond this age the spin-down luminosity drops too rapidly, and only internal magnetic-field decay can power the brighter PRSs associated with other repeating FRBs. The relatively low luminosity of the PRS of FRB 20201124A is therefore naturally explained if the nebula is rotation-dominated and only a few tens of years old. The measured peak frequency (–60 GHz) and the absence of a detectable self-absorption break in the observed band match the expectations of Rahaman et al. [22] for a compact, still-young nebula whose expansion has been slowed by a relatively massive, low-energy ejecta shell.
With its inverted spectrum extending to tens of gigahertz and the highest peak frequency among known PRSs, the persistent source associated with FRB 20201124A is the most extreme in the current sample in terms of spectral properties. This makes it a strong candidate for the youngest magnetar associated with a repeating FRB observed to date. Concluding, an inverted spectrum extending to tens of gigahertz and a self-absorption break well below the peak are the characteristic signatures of a young, rotation-powered magnetar wind nebula, precisely the regime favoured by our data.
Acknowledgements. The research leading to these results has received funding from the European Union’s Horizon 2020 programme under the AHEAD2020 project (grant agreement no. 871158). Y.P.Y is supported by the National Natural Science Foundation of China (No. 12473047), the National Key Research and Development Program of China (No. 2024YFA1611603) and the Yunnan Key Laboratory of Survey Science (No. 202449CE340002). This paper makes use of the following ALMA data: ADS/JAO.ALMA#2023.1.01090.S. ALMA is a partnership of ESO (representing its member states), NSF (USA) and NINS (Japan), together with NRC (Canada), NSTC and ASIAA (Taiwan), and KASI (Republic of Korea), in cooperation with the Republic of Chile. The Joint ALMA Observatory is operated by ESO, AUI/NRAO and NAOJ. The National Radio Astronomy Observatory and Green Bank Observatory are facilities of the U.S. National Science Foundation operated under cooperative agreement by Associated Universities, Inc. Partly based on observations with the 100-m telescope of the MPIfR (Max-Planck-Institut für Radioastronomie) at Effelsberg.
References
- [1]Bruni, G., Piro, L., Yang, Y.-P., et al. 2026, A&A, 710, L31
- [2]Bruni, G., Piro, L., Yang, Y.-P., et al. 2025, A&A, 695, L12
- [3]Bruni, G., Piro, L., Yang, Y.-P., et al. 2024, Nature, 632, 1014
- [4]Bruno, L., Pelliciari, D., Bernardi, G., et al. 2026, A&A, 709, L1
- [5]CHIME/FRB Collaboration. 2021, ApJS, 257, 59
- [6]Cordes, J. M. & Chatterjee, S. 2019, ARA&A, 57, 417
- [7]Fong, W., Dong, Y., Leja, J., et al. 2021, ApJ, 919, L23
- [8]Foreman-Mackey, D. 2016, The Journal of Open Source Software, 1, 24
- [9]Granot, J. & Sari, R. 2002, ApJ, 568, 820
- [10]Higson, E., Handley, W., Hobson, M., & Lasenby, A. 2019, Statistics and Computing, 29, 891
- [11]Janssen, M., Goddi, C., van Bemmel, I. M., et al. 2019, A&A, 626, A75
- [12]Koposov, S., Speagle, J., Barbary, K., et al. 2024, josspeagle/dynesty: v2.1.4
- [13]Marcote, B., Paragi, Z., Hessels, J. W. T., et al. 2017, ApJ, 834, L8
- [14]Margalit, B. & Metzger, B. D. 2018, ApJ, 868, L4
- [15]McMullin, J. P., Waters, B., Schiebel, D., Young, W., & Golap, K. 2007, in Astronomical Society of the Pacific Conference Series, Vol. 376, Astronomical Data Analysis Software and Systems XVI, ed. R. A. Shaw, F. Hill, & D. J. Bell, 127
- [17]Nimmo, K., Hewitt, D. M., Hessels, J. W. T., et al. 2022, The Astrophysical Journal Letters, 927, L3
- [18]Pelliciari, D., Bernardi, G., Margalit, B., et al. 2026, arXiv e-prints, arXiv:2609.35730arxiv.org/abs/2609.35730
- [19]Petroff, E., Hessels, J. W. T., & Lorimer, D. R. 2019, A&A Rev., 27, 4
- [20]Piro, L., Bruni, G., Troja, E., et al. 2021, A&A, 656, L15
- [21]Planck Collaboration, Aghanim, N., Akrami, Y., et al. 2020, A&A, 641, A6
- [22]Rahaman, S. M., Acharya, S. K., Beniamini, P., & Granot, J. 2025, ApJ, 988, 276
- [23]Speagle, J. S. 2020, MNRAS, 493, 3132
- [24]Spitler, L. G., Scholz, P., Hessels, J. W. T., et al. 2016, Nature, 531, 202
- [25]Yang, Y.-P. 2026, MNRAS, 550, 13325
- [26]Yang, Y.-P., Li, Q.-C., & Zhang, B. 2020, ApJ, 895, 7
- [27]Yang, Y.-P., Lu, W., Feng, Y., Zhang, B., & Li, D. 2022, ApJ, 928, L16