Resolved imaging confirms a radiation belt around an ultracool dwarf

Radiation belts are present in all large-scale Solar System planetary magnetospheres: Earth, Jupiter, Saturn, Uranus and Neptune1. These persistent equatorial zones of relativistic particles up to tens of megaelectron volts in energy can extend further than ten times the planet’s radius, emit gradually varying radio emissions2–4 and affect the surface chemistry of close-in moons5. Recent observations demonstrate that very low-mass stars and brown dwarfs, collectively known as ultracool dwarfs, can produce planet-like radio emissions such as periodically bursting aurorae6–8 from large-scale magnetospheric currents9–11. They also exhibit slowly varying quiescent radio emissions7,12,13 hypothesized to trace low-level coronal flaring14,15 despite departing from empirical multiwavelength flare relationships8,15. Here we present high-resolution imaging of the ultracool dwarf LSR J1835 + 3259 at 8.4 GHz, demonstrating that its quiescent radio emission is spatially resolved and traces a double-lobed and axisymmetrical structure that is similar in morphology to the Jovian radiation belts. Up to 18 ultracool dwarf radii separate the two lobes, which are stably present in three observations spanning more than one year. For plasma confined by the magnetic dipole of LSR J1835 + 3259, we estimate 15 MeV electron energies, consistent with Jupiter’s radiation belts4. Our results confirm recent predictions of radiation belts at both ends of the stellar mass sequence8,16–19 and support broader re-examination of rotating magnetic dipoles in producing non-thermal quiescent radio emissions from brown dwarfs7, fully convective M dwarfs20 and massive stars18,21.

Radiation belts are present in all large-scale Solar System planetary magnetospheres: Earth, Jupiter, Saturn, Uranus, and Neptune [1].These persistent equatorial zones of trapped high energy particles up to tens of MeV [1] can produce bright radio emission [2][3][4] and impact the surface chemistry of close-in moons [5].Recent observations confirm planet-like radio emission such as aurorae from large-scale magnetospheric current systems [6][7][8] on very low mass stars and brown dwarfs [9][10][11].These objects, collectively known as ultracool dwarfs, also exhibit quiescent radio emission hypothesized to trace stellar coronal flare activity [12] or extrasolar radiation belt analogs [11,[13][14][15][16]].Here we present high resolution imaging of the ultracool dwarf LSR J1835+3259 demonstrating that this radio emission is spatially resolved and traces a long-lived, double-lobed, and axisymmetric structure similar in morphology to the Jovian radiation belts.Up to 18 ultracool dwarf radii separate the two lobes.This structure is stably present in three observations spanning >1 year.We infer a belt-like distribution of plasma confined by the magnetic dipole of LSR J1835+3259, and we estimate ∼15 MeV electron energies that are consistent with those measured in the Jovian radiation belts [4].Though more precise constraints require higher frequency observations, a unified picture where radio emissions in ultracool dwarfs manifest from planet-like magnetospheric phenomena has emerged.
Historically, stars with strong magnetic heating in their atmospheres have informed interpretations of ultracool dwarf quiescent radio emissions [12].In particular, a tight correlation between the non-thermal quiescent radio and thermal X-ray luminosities from magnetically active stars [17] also holds for solar and stellar flare emission [18], suggesting that flare-accelerated electrons produce quasi-steady quiescent radio emission from active stars.However, ultracool dwarf quiescent radio luminosities do not follow empirical flare relationships [11,12].Instead, a mechanism entirely different from low-level flaring may drive the quiescent emission.
As an alternative, Jovian radio emissions have inspired recent studies [11,[14][15][16].The three main sources of Jovian radio emission are thermal cloud emission from the photosphere [19,20], aurorae manifested as circularly polarized and rotationally periodic bursts powered by the electron cyclotron maser instability [21,22], and more gradually varying quiescent synchrotron emission from high-energy electrons populating radiation belts extending up to 13 Jupiter radii (R J ) in the Jovian dipole field [3].
A search for companions around an auroral ultracool dwarf marginally resolved 8.5 GHz emission [27], which we propose hints at a possible radiation belt like Jupiter's.However, at 10.6 pc distant, current instrumentation cannot conclusively resolve the spatial distribution of its emitting electrons.
LSR J1835+3259 straddles the substellar boundary.Its M8.5 spectral type [29] corresponds to a 2316 ± 51 K effective temperature [30].Modeling finds that for an age ≥500 million years (Myr), it will have a mass of 77.28 ± 10.34 Jupiter masses (M J ) near the hydrogen burning limit that differentiates between low-mass stars and massive brown dwarfs [30].These correspond to a radius R UCD = 1.07 ± 0.05 Jupiter radii (R J ) that, together with its 2.845 ± 0.003 hr optical rotation period and projected surface velocity v sin i = 50 ± 5 km s −1 [31], imply that LSR J1835+3259 is edge-on relative to our line of sight with a rotation axis inclined at an angle i ≈ 90 • .
Using the High Sensitivity Array (HSA) of 39 radio dishes spanning the USA to Germany, we searched for and imaged extended quiescent radio emission at 8.4 GHz from LSR J1835+3259 from a large-scale magnetospheric plasma structure as evidence of an extrasolar analog of Jovian radiation belts.Our observing campaign consisted of three 5-hour epochs from 2019 to 2020 (Table 2).
We find that quiescent radio emissions from LSR J1835+3259 persist throughout each epoch and exhibit a double-lobed morphology that is stable for more than one year (Figure 1).Up to 18.47 ± 2.20 R UCD separate its radio lobes, which have no detectable circular polarization (Tables 3 and 2).These data constitute the first resolved radio imaging of an ultracool dwarf magnetosphere.
Auroral bursts appear centrally located between the two lobes in Epoch 2, which has the highest-quality data.Figure 2 shows an aurora from Epoch 2 separately imaged and then overlaid on the quiescent emission contours from that same epoch.Epoch 1 cannot detect the faint extended emission observed in later epochs due to missing antennas, so aurorae appear coincident with the right quiescent radio lobe (Figure 1).We simulate an observation of the Epoch 2 image using the Epoch 1 antenna configuration and find that the lobe separation in Epoch 1 is consistent this simulated Epoch 2 observation.The relative locations between aurorae and quiescent radio lobes are also consistent.In Epoch 3, aurorae are too faint to confidently locate.
8.4 GHz aurorae originate in 3 kiloGauss magnetic fields near the surface of LSR J1835+3259 [33].From Epoch 2 (Figure 2), we infer that lobe centroids sit at 6−9 R UCD from the ultracool dwarf (with 7-10% uncertainties; Table 3), while their outer extents reach at least 16 − 18 R UCD .The structure may μJy/beam ) contours of quiescent 8.4 GHz emission from LSR J1835+3259 in Epoch 2 with its right circularly polarized aurora overlaid in grey scale.The synthesized beam (black ellipse) sets the resolution element for the aurora and is determined by the array configuration.Figure 1 shows synthesized beams for quiescent emission.The aurora appears centrally located with respect to the double-lobed morphology of the quiescent emission.Coordinates are for midnight in International Atomic Time.
be even larger; individual epochs may not be sensitive to fainter and more extended emission, as is the case for Epoch 1.
At these large extents, dipole magnetic fields decaying with radius as B ∝ r −3 will dominate higher-order magnetic fields (B ∝ r −4 or more rapid decay) of similar surface field strengths inferred for LSR J1835+3259 with multiwavelength spectra [36,38].Indeed, the persistent double-lobed and axisymmetric morphology observed is consistent with a stable dipole magnetic field, and theoretical treatments assuming such can explain radio aurorae observed from LSR J1835+3259 and other ultracool dwarfs [6][7][8].
Our observations present compelling evidence for the first known analog of Jovian radiation belts outside of our Solar System, consisting of a longlived population of relativistic electrons confined in a global magnetic dipole field [39].To explore implications of the lobe separation for electron energies, we consider a ≥3 kiloGauss surface dipole field.At the lobe centroids, the field strength and corresponding non-relativistic electron cyclotron frequency ν c = eB/2πm e c would be 2 Gauss and 6 MHz.
First, the spatial extent of the quiescent radio emission is consistent with synchrotron emission from relativistic electrons.An electron gyrating about a magnetic field emits at multiple harmonics s of its relativistic cyclotron frequency, ν c /γ: where γ is the Lorentz factor of the electron described by its speed, and ν is the observed frequency [40].Since γ > 1, 8.5 GHz emission corresponds to s ≥ 1500 in the lobe centroids.Higher order magnetic fields fall off more rapidly in strength than dipole fields, resulting in higher harmonics.These high harmonics rule out gyrosynchrotron emission from mildly relativistic electrons, which typically emit at harmonics s ≈ 10 − 100 [40].Instead, such high harmonics indicate synchrotron emission from very relativistic electrons, which cannot produce strong circular polarization.This is consistent with stringent ≤8 ± 2% constraints on the integrated 8.44 GHz circular polarization from a previous 11-hr observation of LSR J1835+3259 [33].Indeed, we do not detect circular polarization in its resolved radio lobes in any epoch (Table 3).For the less resolved and brighter quiescent emission in Epoch 1, our noise floor gives a 95% confidence upper limit of ≤8.8% and ≤15.5% circular polarization in the left and right lobes, respectively.Synchrotron emission instead produces linear polarization [40] that has been observed at the 20% level for Jupiter [3].Similar measurements would confirm a synchrotron interpretation for LSR J1835+3259, and higher frequency observations can further constrain electron energies [40].
For synchrotron emission, we can estimate electron energies because a single electron emits at a narrow range of frequencies.Its power spectrum peaks at the critical frequency where α is the pitch angle [40].For ν crit ≈ 8.5 GHz, electrons at the centroids of our target's resolved radio lobes will have γ ≈ 30.These high Lorentz factors correspond to 15 MeV and are comparable to Jovian radiation belt electron energies up to tens of MeV [3,4]. 2 Quiescent emission: left lobe, right lobe 3 No circularly polarized emission was detected.We give 95% confidence upper limits for the absolute value of percent circular polarization in the peak emission calculated with σrms from the Stokes V image.
Relativistic electrons lose energy as they emit synchrotron radiation, giving a cooling time in seconds [40] τ ≈ 6.7 For LSR J1835+3259, we estimate τ ≈ 65 days, yet the double-lobed structure that we observe persists for over a year.For the left lobe, its integrated flux varies from 189 ± 34 to 342 ± 25 µJy between all epochs, and the right lobe varies from 164 ± 40 to 225 ± 35 µJy (Table 3).Although unresolved, this level of quiescent emission at the same observing frequencies was also present over a decade ago [32][33][34].
Flares can impulsively accelerate electrons [40], and ultracool dwarfs similar in spectral type to LSR J1835+3259 such as Trappist-1 can flare at optical wavelengths as frequently as once per ∼day for the lowest-energy flares or once per ∼month for higher energy flares [41].However, Trappist-1 lacks detectable radio emission [42], as do most ultracool dwarfs of similar spectral type to LSR J1835+3259 [25].Curiously, the flare star UV Ceti (M5.5 spectral type) temporarily displayed a double-lobed structure at 8.4 GHz during a radio flare, but that structure disappeared within hours [43].Double-lobed flares are unlikely to explain persistent quiescent emission from that star or LSR J1835+3259.
Instead, radiation belts have been observed for decades to maintain MeV electrons in equatorial magnetosphere regions of Solar System planets.They offer an analogy for interpreting LSR J1835+3259's double-lobed quiescent emission.In contrast to impulsive acceleration from flares, radiation belt electrons undergo sustained adiabatic heating as they encounter stronger magnetic fields during inward radial diffusion in Jupiter's magnetosphere [4].Intriguingly, recent radiation belt modeling for massive stars can also reproduce 8.4 GHz quiescent radio luminosities from LSR J1835+3259 [15].The doublelobed and axisymmetric geometry observed from its quiescent radio emission is similar to the radio morphology of Jupiter's radiation belts [2] and consistent with a belt-like structure about the magnetic equator for an edge-on system like LSR J1835+3259 (Figures 1 and 2).Jupiter's GHz radiation belts trace its highest energy electrons and are more compact than its MHz radiation belts [2,3,44].97.5 GHz quiescent emission from LSR J1835+3259 raises the possibility that its 8.5 GHz emission may similarly trace less energetic electrons at more extended distances in its magnetosphere, calling for comparisons to resolved imaging at higher frequencies.
We conclude that quiescent radio emission around LSR J1835+3259 exhibits properties consistent with an extrasolar analog to the Jovian radiation belts: its long-lived double-lobed structure (1) is morphologically similar to the Jovian radiation belts, (2) with an ≈18 R UCD lobe separation implying ∼MeV electrons confined in a magnetic dipole field and (3) suggesting acceleration mechanisms, distinct from flare activity, that rely on rapidly rotating magnetic dipole fields [39] to produce a belt-like structure of relativistic electrons in its magnetospheric equatorial regions.Our results support broader re-examination of the role that rotating magnetic dipoles may play for the non-thermal quiescent radio emission from massive stars [15] and fully convective M dwarfs [17], for which growing evidence points to the prevalence of large-scale magnetospheres.
Many open questions remain, including: what is the source of ultracool dwarf radiation belt plasma?Ongoing searches for their predicted planets and moons [45,46] may help demonstrate that volcanism from such companions seed ultracool dwarf magnetospheres in a manner similar to Io in Jupiter's magnetosphere [47].Additionally, unlike Jupiter, flares on ultracool dwarfs [41] may provide a seed population of electrons that are later accelerated to the high energies that we infer.Variability on days-long timescales such as what we observe for LSR J1835+3259 are also observed from radiation belts around Jupiter and Saturn.They are attributed to changes in radial diffusion tied to solar weather [48,49].For isolated ultracool dwarfs such as LSR J1835+3259, we postulate that their flaring activity may similarly perturb radial diffusion of radiation belt electrons while contributing to their population.
Beginning with the discovery of ultracool dwarf radio emission first announced in this Journal [50] and the later confirmation of aurorae occurring on ultracool dwarfs [9], our result completes a paradigm in which planetarytype radio emissions emerge at the bottom of the stellar sequence as stellar-like flaring activity subsides.

Target parameters
In addition to the parameter modeling discussed in the Main Text, absorption line modeling for LSR J1835+3259 gives a higher 2800 ± 30 K effective temperature corresponding to a lower mass, young age, and inflated radius (22 ± 4 Myr, 55 ± 4 M J , 2.1 ± 0.1 R J [36]).However, this higher temperature is inconsistent with its spectral type and may be subject to systematic effects in the model atmospheric spectra.Indeed, the young inferred age does not exceed typical M dwarf disk dissipation timescales [51] and no infrared excess indicates the presence of a disk [37].Furthermore, LSR J1835+3259 does not have detectable lithium absorption in its atmosphere [52], indicating that its mass is likely higher than the ≈65 M J mass threshold for lithium depletion to occur and that its age is older than the depletion timescale [53,54].Instead, the properties we adopt (≥500 Myr, 77.28 ± 10.34 M J , 1.07 ± 0.05 R J [30]) are consistent with these multiwavelength observations of LSR J1835+3259.We summarize our adopted target properties in Table 1.

Observations
The HSA combines the Very Long Baseline Array (VLBA, ten 25-m dishes), the Karl G. Jansky Very Large Array (VLA, twenty-seven 25-m dishes) as a phased array, Greenbank Telescope (GBT, single 100-m dish), and Effelsberg Telescope (EB, single 100-m dish).Not all epochs successfully included all telescopes because of weather, equipment failures, and site closings.Additionally, LSR J1835+3259 was visible on the longest baseline from the VLBA dish at Mauna Kea, Hawaii (MK) to EB in Bad Münstereifel, Germany (10328 km) for no more than one hour per observation.We prioritized time on EB to increase observational sensitivity and long baselines to the telescopes on the continental USA.Table 2 summarizes presented HSA observations.
To incorporate the VLA in a very long baseline interferometry (VLBI) observation, all the antennas in the array must be phase corrected and summed (i.e.phased).After this is done, the VLA can be treated as a single element in the VLBI array with a primary beam equal to the synthesized beam of the VLA.Along with the phased-VLA as a single telescope that is correlated with the other HSA telescopes, the phased-VLA data can be used as a regular VLA observation.
We observed in A and B configurations for the VLA, giving phased-VLA primary beams of ∼0.2 and 0.6 (half power beam width), respectively, at our X-band (8.5 GHz) observing frequency.To obtain the full sensitivity of the phased-VLA observations, the position of LSR J1835+3259 needed to be within 0.1 of the center of the field when observing in A configuration and 0.3 of the center of the field when observing in B configuration.
This can pose a challenge for an object with as high proper motion and parallax as LSR J1835+3259 (Table 1).Gaia Data Release 2 measured proper and parallax motions exceeding 2 mas day −1 and 1 mas day −1 , respectively [55].To ensure target capture, a week to a month prior to one or more HSA observations, we obtained a 60 min observation using only the VLA in array mode at X band to image and locate the position of the target.
We also observed the VLBA standard phase calibrator J1835+3241 to calibrate the phase errors from atmospheric fluctuations.This phase calibrator was within 0.33 • of our target.Our phase calibration cycle periods were 4 min and 8 min for HSA and VLA-only observing blocks, respectively.During HSA observing, we also phased the VLA every 10 minutes with this same phase calibrator to maintain coherence across the VLA and observed J1848+3219 every ∼2 hours for fringe-finding.Finally, we observed 3C286 as a flux calibrator for the VLA in each epoch to allow for independent analyses of the phased-VLA data.

Calibrations
For HSA observations, we applied standard phase reference VLBI data reduction methods [56] using the Astronomical Image Processing System AIPS package by the National Radio Astronomy Observatory [57].The target is too faint to self-calibrate, so it was phased referenced to the nearby calibrator J1835+3241.No further calibration to account for atmospheric differences between the target and the calibrator were required due to the excellent proximity of our phase calibrator (0.33 • separation).
The narrow 256 MHz bandwidth of the HSA observations and 8.4 GHz center frequency largely avoids radio frequency interference as a source of noise.Nevertheless, we carefully examine the data from each epoch to identify and remove bad data.No data from MK was collected for Epoch 1, and calibrating the MK-EB baseline in Epochs 2 and 3 proved very difficult.This was because of the short amount of time both telescopes overlapped and the fact that they created a single very long baseline that was not similar to any other baseline in the observations.For this reason, in Epoch 2 we did not include the MK-EB baseline and in Epoch 3 we did not include MK at all in final imaging.
For each epoch, we calculated the combined apparent motion of our target from both proper motion and parallax.For the latter, we assumed a geocentric observer on an orbit with negligible eccentricity and take the location of the Sun as the Solar System barycenter.These assumptions give parallax offsets that are within 2% of the true parallax amplitude [58] and well within our resolving power (Table 2) while greatly simplifying our calculations.LSR J1835+3259 moves 0.4 − 0.6 mas in right ascension and 0.3 − 0.9 mas in declination per 5-hour observation.Our observations can resolve this apparent motion, so we compute an "effective proper motion" by assuming our target takes an approximately linear path on the sky over each observation.We then correct for apparent motion smearing using the AIPS task clcor.
Circular polarization, which is the difference in the right and left circularly polarized data, can distinguish between and characterize electron cyclotron maser, gyrosynchrotron, and synchrotron emissions.Since all telescopes in the HSA use circularly polarized feeds, it is easy for even slightly incorrect amplitude calibration to produce spurious instrumental circular polarization.To ensure that any circular polarization detected was from LSR J1835+3259 rather than errors in the calibration, we checked the circular polarizations of our calibrators.These showed instrumental contamination resulting in ∼7 − 10% spurious circular polarization in the HSA observations.We separately inspect the VLBA-only and VLA-only data on the phase calibrator and find that these data contain 0.1% circular polarization from instrumental contamination and/or circular polarization intrinsic to the calibrator.To correct the amplitude calibration on the HSA, in each epoch we self-calibrated [59] the phase calibrator to reduce instrumental polarization to 1%.We then transferred the resulting amplitude calibrations to our target to reduce any instrumental circular polarization to the ∼1% level.

Timeseries
Radio aurora on LSR J1835+3259 manifest as bright, periodic and strongly circularly polarized bursts every 2.84 ± 0.01 hr [e.g.33] that are clearly evident even in the stand-alone phased-VLA data.LSR J1835+3259 is unresolved in phased-VLA data, so we used this data to produce timeseries (Figure 3) using the AIPS task dftpl, which is specifically designed for unresolved objects.We find that two auroral bursts were partially or fully detected in each epoch for both the right and left circularly polarized data.

Imaging: Quiescent Emission
To create images of the quiescent emission, we excised auroral bursts identified in the timeseries (Figure 3) and imaged the remaining data in each epoch (Figure 1).All images presented in this Letter were imaged using the AIPS task imagr with a 0.1 mas pixel size to give 4 − 6 pixels across the narrowest part of the synthesized beam.We used a Briggs robust weighting [60] of 0.0, which balances between uniform and natural weighting to allow both high resolution and sensitivity to non-point sources.
We observe a double-lobed morphology in each epoch (Figure 1).Detailed modeling of the quiescent emission morphology to distinguish between different morphology types is outside of the scope of this Letter.Instead, we measure quiescent emission source sizes and flux densities using the AIPS task jmfit.In each epoch image, we fit two elliptical Gaussians with freely floating centers, sizes, peaks, and integrated flux densities (Tables 2, 3, 4).The left lobe and right lobe are resolved along approximately the east-west axis in the later two epochs.Measured integrated flux densities are consistent with those reported in the literature [32,33,50].
Finally, in order to help distinguish between synchrotron and gyrosynchrotron emission, we imaged the right minus the left circular polarization (total circular polarization; Stokes V) for the quiescent emission from LSR J1835+3259 in each epoch (Table 3).There was no detectable circular polarization above the 12 − 13 µJy/beam noise floors of the Stokes V images.As a further check, we also image our target using data from the VLBAonly and stand-alone phased-VLA.We find no convincing circular polarization emission to an rms noise floor of 37 and 30 µJy/beam respectively.These nondetections are consistent with low integrated circular polarization (∼8 ± 2%) measured in a previous 11 hr VLA observation at 8.44 GHz that also averages over circularly polarized but periodically bursting aurorae [33].For our brightest quiescent lobe, ∼8% circular polarization would be a 2σ source in the circular polarization images made from the HSA data.

Imaging: Auroral Bursts
We imaged auroral bursts in the same way as described in Section 1.5.However, first we removed the quasi-steady quiescent emission.We used the AIPS task uvsub to subtract a model of the quiescent emission obtained from the images for each epoch shown in Figure 1.As a check, we also subtracted this model from our quiescent-only datasets and re-imaged the data to ensure that no flux remained.
We then imaged only time ranges containing right circularly polarized auroral bursts, which are brighter in our data (Figure 3).To obtain a high signal-to-noise, we imaged the brightest 15-20 minutes noted in Figure 3.These shorter time ranges also avoided averaging longer periods of data with very different flux densities, which can cause artifacts in interferometric imaging.
We imaged right circularly polarized auroral bursts from Epochs 1 and 2. In Epoch 3, aurorae were too faint to confidently image, which was further exacerbated by a set of ∼10 min extended calibration scans coinciding with one of the auroral bursts.We also attempted to image left circularly polarized auroral bursts, but these were too faint to be imaged for all epochs.Figure 2 shows the right circularly polarized auroral burst from Epoch 2 and shows that it is morphologically distinct from the quiescent radio lobes.
Finally we fit an elliptical Gaussian with the AIPS task jmfits to measure the spatial extent and location of the right circularly polarized auroral bursts in Epochs 1 and 2 (Tables 4 and 5).In Epoch 1, the auroral burst appears unresolved and associated with the inner part of the right lobe.In Epoch 2, the burst is consistent with both being unresolved or being marginally resolved along approximately the east-west axis with a minor axis of ∼ 0.4 mas (Figure 2).

Epoch 1 :Fig. 3
Fig.3Timeseries for LSR J1835+3259 using only phased-VLA data from each epoch binned into 5 min intervals showing the right circularly polarized emission (magenta), left circularly polarized emission (blue) and excised aurorae (grey).Calibrator and autophasing scans that extended for ∼10 minutes and occurred ∼2 hours apart are evident in the data and partially coincided with an aurora in Epoch 3.

Table 1
Properties of LSR J1835+3259 1Stokes I quiescent radio flux density reported in the literature.When available, listed observing frequencies are the center of the reported observing band.

Table 4
Imaging: Lobe & Aurorae Fits See Tables3 and 5for peak and integrated flux densities.Uncertainties in the least significant digit are given in parentheses for coordinates, which are for midnight in International Atomic Time on the epoch date.Minor axes in Epoch 1 are unresolved.