QUASI-QUIESCENT RADIO EMISSION FROM THE FIRST RADIO-EMITTING T DWARF

Although there is now abundant evidence that very low mass stars and brown dwarfs (collectively, ultracool dwarfs: UCDs) can host significant magnetic activity, the origins and detailed nature of this activity are not well understood. While there is consensus that the αΩ dynamo that powers the solar magnetic field cannot operate in these fully convective objects, modeling of turbulent dynamos yields a wide range of predictions for the strength, morphology, and time dependence of UCD fields (e.g., Durney et al. 1993; Chabrier & Küker 2006; Dobler et al. 2006; Browning 2008). These poorly constrained yet basic physical properties affect the internal structure, atmospheric conditions, and dynamics of low-mass stars, brown dwarfs, and giant planets.

The evidence for UCD magnetic activity comes from a variety of tracers. Standard indicators such as X-ray and Hα emission (e.g., Rutledge et al. 2000; West et al. 2004) have been augmented by Zeeman-Doppler imaging (Semel 1989; Donati et al. 2008; Morin et al. 2008) and FeH spectroscopy (Reiners & Basri 2006, 2007). A particularly noteworthy discovery was the detection of UCD radio emission (Berger et al. 2001), which is much stronger than would be expected from stellar scaling relations (Güdel & Benz 1993; Benz & Güdel 1994). Radio emission is an increasingly important observable because other tracers become much more difficult to exploit for the very latest-type dwarfs, in which rapid rotation obliterates Zeeman signals, the overall optical luminosity decreases greatly, and X-ray and Hα emission fall off precipitously (Stelzer et al. 2006; Berger et al. 2010; Gizis et al. 2000; West et al. 2004). This last effect is possibly due to increasingly neutral atmospheres (Mohanty et al. 2002), enhanced trapping of energetic electrons, centrifugal stripping of the corona, or other effects decreasing the efficiency with which the coronal plasma is heated (Berger et al. 2010).

A deeper understanding of the convective dynamo will likely be built upon a clearer picture of the magnetic phenomenology of UCDs across the widest possible range of temperatures and Rossby numbers. Measurements of activity in the latest dwarfs, which are extreme in both these parameters, are thus vital. Until recently, the latest-type UCD to be detected in the radio was the L3.5 object 2MASS J00361617+1821104 (Berger 2002; Berger et al. 2005; Hallinan et al. 2008). However, in the past year, two later-type sources have been detected. The L5e+T7 binary 2MASS J13153094−2649513AB (2M 1315−26 hereafter), known to have strong long-lived Hα emission (Hall 2002; Burgasser et al. 2011), was shown by Burgasser et al. (2013) to have quiescent radio emission. Even more surprisingly, the T dwarf barrier was broken by Route & Wolszczan (2012, hereafter RW12), who detected circularly polarized radio bursts from the T6.5 dwarf 2MASS J10475385+2124234 (2M 1047+21 hereafter) with the Arecibo observatory. At a distance of 10.6 pc and a temperature of a mere ∼900 K (Vrba et al. 2004), 2M 1047+21 represents a new benchmark in the study of cool, magnetically active objects.

In this Letter, we report the detection of 2M 1047+21 with the Karl G. Jansky Very Large Array (VLA) with properties significantly different from those of the flares detected by RW12 (Section 2). We discuss its implications as well as some potential paths forward (Section 3).

2M 1047+21 was spectroscopically identified as a T dwarf by Burgasser et al. (1999). It was later assigned spectral types of T7 (optical; Burgasser et al. 2003a) and T6.5 (near-IR; Burgasser et al. 2006). Burgasser et al. (2003a) also made a marginal (2.2σ) detection of Hα emission, measuring a flux of f_Hα = (5.9  ±  2.7) × 10⁻¹⁸ erg cm⁻² s⁻¹. Astrometric monitoring revealed a distance of 10.6 ± 0.4 pc and a proper motion of 1728.4 ± 7.7 mas yr⁻¹ (Vrba et al. 2004). Vrba et al. (2004) also computed a bolometric luminosity of ℓ_bol = −5.36 for 2M 1047+21, where ℓ = log₁₀(L/L_☉). Combining these results, ℓ_Hα − ℓ_bol = −5.3  ±  0.2, well above the trend for typical UCDs (Berger et al. 2010). There is no evidence for 2M 1047+21 being a binary system, with Burgasser et al. (2003b) ruling out companions at separations ≳4 AU with mass ratios ≳0.4 and Carson et al. (2011) finding weaker constraints. Before the radio detection by RW12, Berger (2006) reported a flux density upper limit of 45 μJy at 8.46 GHz (ℓ_r − ℓ_bol < −5.52, where L_r = νL_ν).¹

Using the astrometric parameters of Vrba et al. (2004), the predicted location of 2M 1047+21 at the time of our VLA observations is R.A. = 10:47:52.03, decl. = +21:24:16.7 (FK5 J2000). We use a simple Monte Carlo approach to estimate the 1σ confidence ellipse in this position to have major and minor axes of 019 and 009 at a position angle of 77°. This is much smaller than the ∼1'' astrometric uncertainties of our observation in the VLA C configuration at ∼6 GHz and much larger than the difference between the FK5 and ICRS frames.

We observed 2M 1047+21 in a 2 hr session starting on UT 2012 April 17.02 (project 12A-451). The total correlated bandwidth is 2048 MHz, with two basebands centered at 5000 and 6750 MHz and divided into 512 channels each. Standard calibration observations were obtained, using 3C 286 as the flux density standard and bandpass calibrator and the quasar J1051+2119 (09 from 2M 1047+21) serving as the complex antenna gain calibrator. The total integration time on 2M 1047+21 is 80 minutes. The data were calibrated using standard procedures in the CASA software system (McMullin et al. 2007). Radio frequency interference was flagged manually.

We created a deep Stokes I image of 2048 × 2048 pixels, each 1'' × 1'', using multi-frequency synthesis (Sault & Wieringa 1994) and CASA's multi-frequency CLEAN algorithm with two spectral Taylor series terms for each CLEAN component; this approach models both the flux and spectral index of each source. The synthesized beam is 42 × 36 and the reference frequency is 5.84 GHz, with data taken at frequencies between 4.51 and 7.24 GHz contributing to the imaging process. In the resulting image of the initial Taylor term (i.e., brightness at the reference frequency), we detect a 4.6σ (peak-to-rms) unresolved source at R.A. = 10:47:52.10, decl. = +21:24:16.2 (ICRS J2000) with 1σ positional uncertainties of 09 and 06 in R.A. and declination, respectively. The separation from the IR astrometric prediction is 10 at a position angle of 117°. Taking into account only the uncertainties on the VLA source position, and not the uncertainties in the IR position or the astrometric calibration of the VLA image, the separation is 1.3σ. The relevant portion of the image is shown in Figure 1. Because we search for a source at only one a priori image position, the stringent detection criteria that have been advocated for blind radio source searches (e.g., Frail et al. 2012) do not apply.

Zoom In Zoom Out Reset image size

Fitting the image region near the source with a Gaussian shaped like the synthesized beam, we find a flux density of 16.5 ± 5.1 μJy while the rms residual of the deconvolution process in the neighborhood is 3.6 μJy bm⁻¹, well above the ∼0.2 μJy bm⁻¹ confusion limit for this observing setup (Condon 2002). The uncertainty in the flux density is that reported by the fitting process, which is not generically equal to the background rms due to the two positional parameters. The overall significance of the peak, however, is set by the rms. We identify this source with 2M 1047+21. The likelihood of a spurious association is small: Fomalont et al. (1991) found the density of radio sources brighter than 16 μJy at 5 GHz to be ∼2 × 10⁻⁴ arcsec⁻², making the probability of background source contamination ∼10⁻³ (assigning our search area to be the synthesized beam size). We counted peaks >16 μJy in the central region of our image and found a density of ∼2.4 × 10⁻⁴ arcsec⁻², agreeing well with the Fomalont et al. (1991) value. Our measurement does not account for primary beam attenuation, but the correction for that effect is a factor of a few, not large enough to change the conclusion of a low probability of chance coincidence.

The deconvolution process also determines a nominal spectral index of α = 0.9 ± 1.0 (S_ν∝ν^α). We also imaged the two VLA basebands separately with only one spectral Taylor term, i.e., assuming spectrally flat emission, and detected emission consistent with the position of 2M 1047+21 in both images. The best-fit flux density of the source in the lower-frequency image is 18.0  ±  9.2 μJy, larger than that in the higher-frequency image, 16.7 ± 9.7 μJy; in both cases, the best-fit position is ∼0.5 of a beam away from the best-fit position in the combined image. These fluxes tentatively suggest α < 0, but the uncertainties are large. Combining our results with the upper limit of Berger (2006) at 8.46 GHz implies α ≲ 2.7, if variability is not a factor. Without calibrating the instrumental polarization leakages, we imaged the Stokes V parameter and found no emission at the location of 2M 1047+21. We estimate |V|/I ≲ 80% (3σ).

The radio emission detected by RW12 comprised flares with flux densities up to ∼2 mJy on timescales of minutes. Thanks to the excellent sensitivity of the VLA, less than a second of integration time would be needed for a 5σ detection of these flares in our data, so superb time resolution of any flares should be possible. We searched for flares and other variability using a visibility-domain analysis. We subtracted the emission of all other detectable sources from the u − v data and summed the residual visibilities after rephasing to the position of 2M 1047+21. Ignoring noise and calibration errors, the real part of the summed rephased visibility is the source flux, while the imaginary part is zero. Given that noise and calibration errors are always present, the variation of the imaginary part provides an assessment of the uncertainty in the real part. The mean real part is 16.5 μJy and the variance associated with the imaginary parts $\sigma _{\rm im}/\sqrt{N_{\rm im}} = 4.6$ μJy, agreeing with the image-domain analysis. Our calibrated data have a 5 s cadence (reduced from the 1 s cadence of the raw data), so any Arecibo-like flares should be apparent as several consecutive ∼12σ bins. We find no variability of the kind observed by RW12, with only two bins of >3σ significance, which is consistent with the expectation from Gaussian noise of 2.6 such events.

We Hanning-smoothed the data over a variety of timescales to search for phenomena of longer duration but lower significance. Figure 2 shows the data smoothed on 30 s and 6.5 minute timescales. There is a rise-and-fall pattern in the 6.5 minute cadence data suggestive of a ∼60 μJy enhancement over a 20 minute timescale. The data are, however, fully consistent with a constant flux density, matching this model with a reduced χ² of 1.05. We imaged a 36 minute span around the event (56, 034.041 ⩽ MJD[TT] ⩽ 56, 034.066) and found a source of flux density 37.2  ±  6.8 μJy with a background rms of 4.8 μJy, yielding a 7.8σ detection. The significance of the difference between this value and the mean flux density is lessened, however, because we searched a range of start times and durations precisely to maximize this difference. The possible flare or enhancement seen in the binned data is a factor of ∼30 times less luminous and evolves on a timescale ∼10 times longer than the RW12 flares. The total energy of the potential event cannot be compared to that of the RW12 flares without knowledge of its spectrum; if it is a broader-band phenomenon (see below), the two event classes could have approximately equal energies.

Zoom In Zoom Out Reset image size

Assuming that the emission from 2M 1047+21 is indeed steady and using the distance reported by Vrba et al. (2004), the radio spectral luminosity is L_ν = (2.2 ± 0.7) × 10¹² erg s⁻¹ Hz⁻¹ or ℓ_rad − ℓ_bol = −6.11 ± 0.14. We compare this value to previously detected UCDs in Figure 3. Not only is 2M 1047+21 by far the coolest UCD to be detected in the radio, but it also has flares that are two orders of magnitude brighter than its quiescent emission, second only to the M8 dwarf DENIS J1048.0−3956 (Burgasser & Putman 2005) in the observed dynamic range of its luminosity. If the emission that we detect is not truly quiescent, the dynamic range must be even larger.

Zoom In Zoom Out Reset image size

If the radio emission of 2M 1047+21 is indeed quiescent and unpolarized, it is suggestive of a gyrosynchrotron process with a brightness temperature:

$$\begin{eqnarray} T_b \! &=& \! \frac{c^2}{2 k_B \nu ^2} \frac{S_\nu }{\Omega } \,{\approx}\, (10^{6.5}\,{\rm K}) \left(\frac{\nu }{{\rm GHz}}\right)^{-2} \left(\frac{S_\nu }{\mu {\rm Jy}}\right)^{} \left(\frac{d}{{\rm pc}}\right)^{2} \left(\frac{r}{R_J}\right)^{-2} \nonumber\\ & \approx & (10^8\,{\rm K}) \left(\frac{r}{R_J}\right)^{-2}, \end{eqnarray} \tag{ 1 }$$

where r is the effective radius of the emitting region and the final line plugs in the relevant values from our observation. Length scales of r ≲ 0.01 R_J would be problematic for this interpretation due to the brightness temperature constraints of gyrosynchrotron emission (Dulk 1985). Gyrosynchrotron emission is broad band with a spectral peak; the location of which depends on local physical parameters. Further characterization of the radio spectrum, ideally paired with measurements in other bands, could yield conditional estimates of the magnetic field strength and electron number density (e.g., Berger 2006).

Another candidate emission mechanism is the electron cyclotron maser instability (ECMI; Wu & Lee 1979; Treumann 2006). Stereotypical ECMI emission is highly polarized, narrow band, and time variable, making it a natural explanation for the bursts observed by RW12. Its applicability to the emission we detect is less clear. The emission that we detect shows no evidence of strong polarization, is broad band, and may be variable but clearly does not show the strong spikes seen in 2M 1047+21 by RW12 or 2MASS J07464256+2000321 by Berger et al. (2009). Hallinan et al. (2006, 2008) have argued that continuous particle acceleration in a region of varying magnetic field strength can generate ECMI emission that is broad in frequency and time, and that propagation effects can depolarize it, potentially yielding results that are indistinguishable from gyrosynchrotron. Their arguments are partially motivated by the detection of rapid variability in the unpolarized emission of 2MASS J00361617+1821104 that would violate gyrosynchrotron brightness temperature constraints given the implied size of the emission region. If the emission we detect is really concentrated in one or possibly two weak flares of 10–20 minute duration, then the same brightness-temperature argument might apply to 2M 1047+21. We note that Hallinan et al. (2008) also argue that cooler objects should tend toward more polarized emission, as propagation effects are less significant in their increasingly neutral atmospheres; if the emission we detect is strongly depolarized, it would challenge the ECMI interpretation. On the other hand, the marginal Hα emission (Burgasser et al. 2003a) may suggest that the atmosphere of 2M 1047+21 may not be fully neutral.

RW12 detect three radio flares in ∼26 hr of observing for a rate of ∼0.1 hr⁻¹. If the flares occur at random times, this is consistent with the lack of events in our 2 hr observation. It is also possible that the flares occur periodically, as observed in a significant fraction of radioactive UCDs (e.g., Berger et al. 2005, 2009; Hallinan et al. 2007, 2008), but with only three detections this hypothesis cannot be tested meaningfully. The best candidate periodicity for the three RW12 events is nominally 2.08156  ±  0.00003 hr (M. Route 2012, private communication), slightly longer than the length of time for which 2M 1047+21 is observable with Arecibo. The time baseline between the Arecibo observations and ours is sufficiently long that, even taking this periodicity and its uncertainty at face value, the phasing of our observations relative to the Arecibo flares is virtually unconstrained. We thus cannot shed any light on the regularity of the flares.

Additional observations of this unique object are clearly needed. The most pressing question to resolve is whether the quasi-quiescent emission we detect is truly steady or whether it consists of very small flares as suggested by the binned data in Figure 2. Fortunately, the sensitivity of the fully upgraded VLA makes this a relatively straightforward proposition to test. Longer integrations allow for a more rigorous characterization of the variability of this faint source, constraining both the quasi-quiescent component as well as the duty cycle of the bright flares detected by RW12. A detection of periodicity in either of these could probe the rotation rate of 2M 1047+21, which is currently unknown. If the quasi-quiescent emission is genuinely steady, it would imply that the magnetic field is likewise long-lived and potentially stable on a timescale of ∼years. If the quasi-quiescent emission is variable, the relationship between its modulation and the temporal behavior of the RW12 flares could shed light on the magnetic field topology of 2M 1047+21.

Further observations to pin down the radio spectrum and polarization of 2M 1047+21 are also desirable. A measurement of the magnetic field strength associated with the quasi-quiescent emission could be compared to the 1.7 kG estimate from the flares of RW12 to provide more clues as to the configuration of the radio-emitting regions. Strict limits on the polarization fraction would require a more careful consideration of the level of ionization in the atmosphere of 2M 1047+21 and its relation to the ECMI propagation effects proposed by Hallinan et al. (2008). Simultaneous multiwavelength observations (radio, Hα, optical broad band, X-rays, e.g., Berger et al. 2008, 2010) could probe the relationship between the photosphere, chromosphere, and corona, through monitoring of either a flare event or modulation of the quasi-quiescent emission.

Both 2M 1047+21 and 2M 1315−26 (Burgasser et al. 2013) are exciting objects that may yield a wealth of new insights into the mechanisms of radio emission from UCDs. Both are also prime examples of the discoveries made possible by start-of-the-art, wide-bandwidth radio telescopes. The VLA in particular promises to be a key tool for the radio detection of UCDs well below the bottom of the main sequence. The tenfold increase of its sensitivity over the previous capabilities enabled the unprecedentedly low-luminosity detection presented in this Letter and will make possible the first solid constraints on the UCD radio luminosity function, successful surveys of targeted subsets of the UCD population, and detailed studies of exotic dwarfs that are rare and hence distant.

We thank Matthew Route and Aleksander Wolszczan for insight into the Arecibo results and useful discussion on the manuscript, and the anonymous referee for comments that improved the Letter. We thank the NRAO for granting the Director's Discretionary Time used to make the observations presented in this Letter. We acknowledge support for this work from the National Science Foundation through Grant AST-1008361. The VLA is operated by the National Radio Astronomy Observatory. The National Radio Astronomy Observatory is a facility of the National Science Foundation operated under cooperative agreement by Associated Universities, Inc. This research has made use of NASA's Astrophysics Data System and the SIMBAD database, operated at CDS, Strasbourg, France.

Facilities: VLA - Very Large Array