Searching for Stellar and Planetary Emission in Large Field-of-view Radio Sky Surveys

Detection of low-frequency (≤1.4 GHz) radio emission from stellar and planetary systems can lead to new insights into stellar activity, extrasolar space weather, and planetary magnetic fields. In this work, we investigate three large field-of-view surveys at 74 MHz, 150 MHz, and 1.4 GHz, as well as a myriad of multiwavelength ancillary data, to search for radio emission from about 2600 stellar objects, including about 800 exoplanetary systems, 600 nearby low-mass stars, and 1200 young stellar objects located in the Taurus and Upper Scorpius star-forming regions. The selected sample encompasses stellar spectral types from B to L and distances between 5 and 300 pc. We report the redetection of five stars at 1.4 GHz, one of which also shows emission at 150 MHz. Four of these are low- and intermediate-mass young stars, and one is the evolved star α Sco. We also observe radio emission at the position of a young brown dwarf at 1.4 GHz and 150 MHz. However, due to the large astrometric uncertainty of radio observations, a follow-up study at higher angular resolution would be required to confirm whether the observed emission originates from the brown dwarf itself or a background object. Notably, all of the selected radio sources are located in nearby star-forming regions. Furthermore, we use image stacking and statistical methods to derive upper limits on the average quiescent radio luminosity of the families of objects under investigation. These analyses provide observational constraints for large-scale searches for current and ongoing low-frequency radio emissions from stars and planets.


Introduction
Astrophysical stellar and planetary low-frequency (<5 GHz) radio emission is a largely underexplored frontier. Primarily limited by sensitivity, we can nonetheless extrapolate the type of emissions we expect to find while simultaneously pushing the capabilities of current instrumentation. Although the dozens of low-frequency radio detections of stars (Pritchard et al. 2021) and planetary-mass objects (Kao et al. 2018;Zic et al. 2019) are a relatively small number, theoretical models suggest that we are approaching the detection threshold for emission due to plasmas interacting with planetary magnetic fields (Grieβmeier 2017). Starting at stellar-mass scales and working our way down, we can peruse the emission mechanisms and what can be learned from observations along the way.
As evidenced by our Sun, stars are sources of low-frequency radio emissions (Dulk 1985). The most common low-frequency radio emission mechanism on our Sun is thermal free-free, effective at high density and chromospheric temperatures, typically on the order of 10 4 K (White 2002). In fact, practically all of the nonflaring emission at and below 1.4 GHz from the Sun comes from optically thick free-free emission.
Incoherent gyrosynchrotron emission is most prominent at centimeter and millimeter wavelengths, while flares and outbursts producing coherent plasma emission play the dominant role at decimeter and meter wavelengths (Bastian et al. 1998). Specifically, at these longer wavelengths, solar type II and III bursts are often associated with radio emissions, as well as coronal mass ejections (CMEs; Cairns et al. 2003;Reid & Ratcliff 2014;Winter & Ledbetter 2015). Currently, there is only a single possible detection of a CME on another star (Argiroffi et al. 2019), so subsequent monitoring and detection of radio emissions from stars could provide much-needed additional evidence for these episodes of high stellar activity.
It is important to note that our Sun is considered a relatively inactive star (Boro Saikia et al. 2018). In particular, other stellar populations that are far more active include young stars (Lehtinen et al. 2016) and, to a lesser extent, M dwarfs (Stelzer et al. 2013). For these more active stars, flares and CMEs may be more severe and common, which would produce the associated radio emissions at greater amplitudes and more often. The high-energy particles released can incite planetary radio emissions as well. Thus, stars with far more activity, such as young stars, may be attractive targets to probe for stellar low-frequency radio emission.
Additionally, substellar-mass objects, specifically brown dwarfs, have been definitively detected at gigahertz (Hallinan et al. 2007(Hallinan et al. , 2008Kao et al. 2018) and megahertz (Zic et al. 2019;Vedantham et al. 2020b) frequencies. Some of these objects were found to have extraordinarily strong kilogauss magnetic fields and provide interesting constraints on understanding the magnetic fields on extrasolar bodies. These may serve as an intermediate regime between planetary and stellar radio emissions. Interestingly enough, the primary lowfrequency emission mechanism observed on brown dwarfs so far appears to be the same as that expected for planets.
Further down on the mass scale, planets in our solar system have been shown to emit low-frequency radio emissions. Chief among them, Jupiter has been observed to emit very strong low-frequency (peaking at 22.2 MHz) radio emissions, the strongest of which are referred to as Jupiter's decametric radiation (DAM) at amplitudes comparable to those of free-free emission at higher frequencies (i.e., a few gigahertz) from the quiescent Sun (Burke & Franklin 1955). Although there are a variety of mechanisms that create radio emissions within our solar system, all planets with large-scale magnetic fields in our solar system exude some form of low-frequency radio waves analogous to Jupiter's magnetospheric emissions ). In fact, on Earth, low-frequency radio emissions were detected in association with aurorae, being classified as auroral kilometric radiation, and later shown to be generated by the same cyclotron maser instability (CMI) process that is responsible for Jupiter's DAM (Pritchett 1984).
The CMI occurs when input charged particles provide energetic electrons that become trapped and gyrate about magnetic field lines, giving off radio emissions. In our solar system, the electron source is typically the solar wind, though in the case of Jupiter, it is supplemented by free-streaming particles from its moon Io. Under CMI and assuming a dipolar magnetic field, radio waves are emitted with maximum frequency near the local gyrofrequency (Farrel et al. 1999), where B max is the maximum polar component of the planetary magnetic field, which, in the case of Jupiter, is approximately 14 G (Acuna & Ness 1976). An important note is that lowfrequency, ground-based radio observations are limited by the Earth's ionosphere, which becomes opaque at frequencies under about 10 MHz as a result of the ionospheric cutoff frequency (Kivelson & Russell 1995, p. 586). This prevents the ground-based detection of CMI radiation from planets with magnetic fields much weaker than Jupiter's.
An illustration of the spectra from the main processes that generate low-frequency radio emissions in our solar system is presented in Figure 1. Solar processes typically dominate the gigahertz range, while planetary CMI peaks at comparable fluxes at megahertz frequencies.
The radiative flux that CMI produces is determined by the amount of input power, whose principal contributor varies for different physical situations. Empirical observational relationships arise when comparing the input power to the output emission within our own solar system, something referred to as the "radiometric Bode's Law" (Lazio et al. 2004). Using this relationship, modeling work has been conducted investigating the effects of these different input sources of magnetic power (Desch & Kaiser 1984;Grieβmeier et al. 2006), including predictions of the radio flux of previously known exoplanets (Grieβmeier et al. 2007;Grieβmeier 2017). These predictions, however, show that only the most optimistic models would produce emissions in a handful of objects that could be individually detected with current instruments. Nevertheless, observational constraints are needed to probe the validity of these predictions. By pushing the limits of lowfrequency radio detections, emissions associated with CMI may help reveal the nature of exoplanetary magnetic fields, which have implications for planetary characterization (such as atmospheric loss) and habitability.
When discussing planetary habitability, the requirements are often simplified to the general public as being the region around a star where stellar radiative flux is sufficient for maintaining liquid water on the surface of a rocky planet with a basic atmosphere (Kasting et al. 1993;Borucki et al. 2008). However, there are many more considerations when evaluating the viability of a planet sustaining life that are often left out of habitable-zone discussions (Lammer et al. 2009; National Academies of Sciences, Engineering, & Medicine 2018). One such consideration is the role that magnetic fields play in affecting planetary habitability, especially around host stars of increased stellar activity, such as M dwarfs (Lopez-Morales et al. 2011;Vidotto et al. 2013) and young stars (Wolk et al. 2005). Impacts on habitability due to impinging high-energy stellar particles may be exacerbated among more active stars. The presence of a magnetic field around a planet could shield it from high-energy particles and atmospheric stripping and could be an important constraint for habitability (Airapetian et al. 2020).
The observation of low-frequency radio emissions due to CMI would give robust information regarding exoplanetary magnetic fields through Equation (1). This valuable information has implications for habitability, as previously mentioned, but would also help to characterize exoplanets and perhaps reveal details of planet formation (Johansen 2009). This question has led to numerous efforts to investigate the feasibility of observing the emission (Grieβmeier et al. 2007(Grieβmeier et al. , 2011 and to directly search for the emission itself (Bastian et al. 2000;Lazio et al. 2010Lazio et al. , 2016O'Gorman et al. 2018). Detection of low-frequency radio emission with planetary origins would also add a new technique to the roster of exoplanet discovery methods, most notably as a direct detection mechanism. At the moment, we are on the cusp of breaking into this new regime of exoplanetary characterization.
A recent detection of 120-167 MHz radio emission with the Low-Frequency Array (LOFAR) provides a tentative first discovery of planetary-induced CMI (Vedantham et al. 2020b). The high degree of circular polarization, as well as temporal variability, of this source provides evidence that this detection may be caused by planet-star interaction akin to the Jupiter-Io system. LOFAR has also been used to independently discover a cold brown dwarf (Vedantham et al. 2020b), as well as potential exoplanet detections (Turner et al. 2021). Another state-of-the-art facility that is probing the skies at megahertz frequencies is the Murchison Widefield Array (MWA), which has produced the GaLactic and Extragalactic All-sky MWA (GLEAM) catalog of radio sources south of +30°decl. (Wayth et al. 2015).
In recent years, a number of improvements have been made that may increase the likelihood of observing low-frequency radio emission. Software improvements to existing instruments (Lane et al. 2012), as well as innovative hardware for new facilities (Kocz et al. 2015), are pushing the limits of radio astronomy. Additionally, hardware upgrades and new facilities, such as the Owens Valley Long Wavelength Array (OVRO-LWA; Anderson et al. 2019), will soon improve the sensitivity and field of view of radio observations. Lastly, the sample size of potential objects to probe continues to grow. To date, thousands of exoplanets have been discovered and confirmed, many of which are the result of satellite missions like Kepler (Borucki et al. 2010) and the ongoing Transiting Exoplanet Survey Satellite (TESS; Ricker et al. 2015).
Our project searched for low-frequency radio emission from a sample of young stars, exoplanet-hosting systems, and nearby field stars. In Section 2, we describe the data products used for this project, including radio sky surveys and follow-up observations of a subset of candidate detections. In Section 3, we cover the sample populations that were selected for study. Our point-by-point matching results are presented in Section 4, while Section 5 covers the image stacking and statistical analyses that establish upper limits on ensemble emissions. Discussion of our results and our conclusions are covered in Sections 6 and 7, respectively.

Data
This project primarily utilized publicly released data from low-frequency radio sky surveys in order to identify radio sources of interest. Further investigation into these objects of interest made use of the SIMBAD database, the VizieR catalog access tool, and follow-up observations with the Very Large Array (VLA).

Radio Sky Surveys
We employed three radio sky surveys with which we crossreferenced our samples of interest: the VLA Low-Frequency Sky Survey redux (VLSSr) at 73.8 MHz (Lane et al. 2014), the TIFR GMRT Sky Survey (TGSS) at 150 MHz (Intema et al. 2017), and the NRAO VLA Sky Survey (NVSS) at 1.4 GHz (Condon et al. 1998). Each of these surveys has cleaned final images available as FITS files for public use, along with catalogs of identified radio sources.
Here we will briefly describe each of these surveys, with the critical details summarized in Table 1.

VLSSr
The VLSSr (Lane et al. 2014) observed approximately 75% of the sky above −30°decl. at an averaged frequency of 73.8 MHz with an angular resolution of 75″. Based on an improved data reduction pipeline, the VLSSr offers images with a lower average rms noise of 100 mJy beam −1 compared to the average 130 mJy beam −1 in the earlier VLSS (Cohen et al. 2007). Observations were conducted between 2001 and 2007. The VLSSr identified 92,964 radio sources at a 5σ detection over the local rms noise level. Although many of the detections likely originate from distant radio galaxies and quasars, many others are still unclassified and could be galactic in origin.
At 74 MHz, exoplanetary radio emissions would indicate magnetic field strengths a few times that of Jupiter's, assuming CMI is the emission mechanism. We believe this is well within the realm of possibility for exoplanets. The VLSSr is the lowest-frequency survey employed in our study, so these are the most likely data for which exoplanetary emissions, if present, may be detected. However, as evidenced by the lowfrequency spectrum in our own solar system, emission from stellar outbursts could also occur at this frequency.

TGSS
The TIFR GMRT Sky Survey Alternative Data Release (TGSS ADR, hereafter TGSS; Intema et al. 2017) from the Tata Institute of Fundamental Research (TIFR) and Giant Metrewave Radio Telescope (GMRT) observed 99.5% of the sky above −53°decl. at 150 MHz. It detected approximately 620,000 radio sources at a 7σ significance above a median rms noise of 3.5 mJy beam −1 with an angular resolution of 25″. Observations were conducted between 2010 and 2012. Data products include a searchable repository of observed radio sources, as well as 5°× 5°mosaic images of the sky at 150 MHz.
The CMI emission at 150 MHz would require maximum polar magnetic field strengths approximately an order of magnitude larger than that of Jupiter, thus around 100 G. While this could be feasible for exoplanets with extraordinarily strong magnetic fields, it is likely to be near the higher-frequency end of planetary-based CMI emission. In fact, this is the same frequency as the current lowest-frequency radio detection of a brown dwarf at 144 MHz (Vedantham et al. 2020b). On the other hand, free-free stellar emission, bursts, or even stellarbased CMI (rather than planet-based) may begin to dominate at these frequencies.

NVSS
The NVSS (Condon et al. 1998) observed the entire sky above −40°decl. at 1.4 GHz at an angular resolution of 45″. Over 1.8 million radio sources were detected with an average rms noise of 0.45 mJy beam −1 in Stokes I total intensity. This survey provides a catalog of detected radio sources, as well as 4°× 4°cleaned sky images. The data were taken between 1993 and 1996.
A frequency of 1.4 GHz is likely well outside the frequency range for detecting CMI emission from exoplanets. However, numerous brown dwarfs have been detected at low-gigahertz frequencies (Kao et al. 2018;Zic et al. 2019). Additionally, this is a very active frequency for a variety of stellar emission processes. In our own Sun, thermal bremsstrahlung is the dominant mechanism for quiescent emissions at this frequency (though gyrosynchrotron dominates for flaring events). Therefore, detections of galactic sources with the NVSS will likely probe stellar emissions, rather than planetary.

Ancillary Data
In addition to the main observational surveys presented above, our analysis makes use of several ancillary observations that either cover smaller regions of the sky in the radio, use higher radio-frequency ranges where stellar chromospheric and photospheric emission might be dominant, or are in completely different wavelength regimes that serve to establish the known positions of galactic objects of interest. The supplementary radio data are not used in our primary source matching but are instead mined to elucidate the nature of our low-frequency emission candidates. An overview of each ancillary data product is briefly discussed below.

VLASS
The VLA Sky Survey (VLASS; Lacy et al. 2020) is an ongoing sky survey of the entire sky above −40°decl. at 2-4 GHz. The final survey will have an intended rms noise of 70 μJy beam −1 in Stokes I total intensity. The data are expected to be taken between 2017 and 2024, though quicklook images with an angular resolution of approximately 2″ from the pilot survey and two observing sessions from the first epoch of observation were considered in this project. However, these images are accompanied by a whole host of systematic uncertainties that users must be aware of when utilizing (Lacy et al. 2019). The source position error in these quick-look images is approximately 0 5 at declinations >−20°. Additionally, when imaging objects with known flux density values, peak fluxes are systematically lower by about 15%, total flux densities are lower by about 8%, and there exists a systematic scatter of around 8% for flux densities below 1 Jy. Above 1 Jy, the measured flux densities are reportedly unreliable and should not be used.
Although we did not utilize the VLASS for source matching and localization, its data products were used to illustrate our candidate detections in Figures 4 and 5 due to the survey's higher angular resolution than those of our primary radio sky surveys. We also use flux density values from the VLASS to calculate spectral indices for certain candidate detections, though we emphasize the aforementioned flux uncertainties and incorporate those into our error calculations.
The 2-4 GHz frequency range, like the 1.4 GHz NVSS, will likely probe quiescent stellar free-free or outbursting gyrosynchroton emission if galactic stellar sources are detected.

VLA Follow-up Observations
An early subset of 16 source matches had follow-up observations obtained in the VLA Semester 2019A project 19A-446 (PI: Jason Ling). These objects were selected out of the initial point-by-point matching results (see Section 4.1) as those with the highest proper motion. Given the nearly 20 yr since their NVSS observations, this proposal sought to detect shifts in the position of each object, as well as simply redetecting their radio emission. Most of the 16 observed sources were ruled out as extragalactic sources through the course of this project, either through literature investigation or through the results of these follow-up observations, leaving only three of the observations associated with the final set of candidate galactic detections.
Observations were conducted in the A-array configuration at the 1.4 GHz L band. The total integration time for each source was 2 minutes. The typical FWHM beam size was 2″, and the average rms noise across the images was 0.6 mJy beam −1 , compared to the 25″ resolution and 3.5 mJy beam −1 noise of the NVSS at the same frequency. Data reduction was performed through the VLA's native pipeline, and imaging was performed with the tclean method found in the Common Astronomy Software Applications (CASA) package (McMullin et al. 2007).

Other Source Catalogs
The vast majority of the additional resources referenced below were accessed via VizieR (Ochsenbein et al. 2000).
The Westerbork in the Southern Hemisphere (WISH) survey looked between -26°and -9°decl. at 352 MHz. Its observations were taken between the fall of 1997 and spring of 1998 and identified over 73,000 sources (De Breuck et al. 2002). Each pointing was integrated for 20 s for an average rms noise of approximately 3.2 mJy beam −1 and a characteristic synthesized beam of 40″ × 150″. For the purposes of our project, WISH was used as a supplementary radio survey to help bridge the gap in frequency space between the 150 MHz TGSS and the 1.4 GHz NVSS. Due to its decl. limits, like all other subsequent radio surveys mentioned in this section, the majority of our sample objects are not in the field of view of WISH.
The GLEAM survey covers the entire sky south of 30°decl. at frequencies between 72 and 231 MHz (Wayth et al. 2015). Observations were taken between 2013 August and 2015 July, with higher-frequency observations (250-310 MHz) taken afterward up to 2016 July. The rms noise is 6-10 mJy beam −1 with a typical angular resolution of 100″. It offers an alternate look at many of the same areas of sky at the same frequencies as the VLSSr and TGSS but with better sensitivity than the VLSSr, though worse angular resolution than both.
The Faint Images of the Radio Sky at Twenty cm (FIRST) survey is a VLA program to observe a relatively small area of sky at 1.4 GHz centered around the north Galactic pole (Becker et al. 1994). The rms noise is approximately 0.2 mJy beam −1 with an angular resolution of 5″. It provides a more sensitive look at a smaller area of sky, which allowed us to compare with potential matches through our primary surveys.
The Gould Belt's VLA survey was a large study of hundreds of young stellar objects (YSOs) in numerous nearby starforming regions, including the Taurus-Auriga complex (Dzib et al. 2015). Observations were conducted in two subbands centered on 4.5 and 7.5 GHz in the VLA's B and BnA configurations. The synthesized beam and rms noise for each subband were approximately 1″ × 1″ with 30-40 μJy beam −1 at 4.5 GHz and 0.7″ × 0.6″ with 30 μJy beam −1 at 7.5 GHz.
We also utilized a number of optical and infrared catalogs for the purpose of comparing objects with better localization to the radio emission from the sky surveys. First, the Gaia space satellite's optical positions and proper motions in Data Release 2 have been crucial for establishing the "known" locations of our sample objects (Gaia Collaboration et al. 2018). The APM-North catalog is a digitization using the SERC Automated Plate Measuring (APM) machine of the optical National Geographic Society Palomar Observatory Sky Survey (McMahon et al. 2000;Minkowski & Abell 1963). It covers most of the sky above −33°decl. and was observed between 1949 and 1958. The Spectrograph for INtegral Field Observations in the Near Infrared (SINFONI) was an instrument installed on the VLT that obtained high-resolution spectra with adaptive optics (Eisenhauer et al. 2003;Bonnet et al. 2004). The Panoramic Survey Telescope and Rapid Response System (Pan-STARRS) is a wide-field optical/infrared imaging system located at Haleakala, Hawaii (Chambers et al. 2016 (Mainzer et al. 2011), is yet another source of known positions for our sample objects. The UKIRT Infrared Deep Sky Survey (UKIDSS) uses the United Kingdom Infrared Telescope in Hawaii to conduct a series of five sky surveys, two of which are focused on galactic sources (Lawrence et al. 2007).
A small number of high-energy X-ray or gamma-ray results were likewise consulted for sources located near our investigated radio emissions. Public results from XMM-Newton (Jansen et al. 2001), ROSAT (Truemper 1982), and the Fermi space telescope (Atwood et al. 2009) were utilized.

Samples
In order to test whether any of the radio emission identified by the radio sky surveys discussed above originated from stellar sources, we cross-referenced the position of the emissions to known exoplanetary systems, nearby stars, and objects present in the Taurus and Upper Scorpius star-forming regions. The characteristics of these three different samples are discussed below.

Exoplanet-hosting Systems
To investigate the possible exoplanetary origins of lowfrequency radio emission, the known population of exoplanets (as of 2019 August 13) was investigated. The list of exoplanets and their characteristics was obtained from exoplanet.eu (Schneider et al. 2011). By constraining the sample to those with distances less than 300 pc away and within the field of view of the radio sky surveys, around 850 confirmed planetary systems were identified. This distance restriction is primarily justified by the 1/d 2 falloff for flux with distance, assuming an isotropic source. However, the chosen distance limit of 300 pc is somewhat arbitrary and was selected due to its effect in creating a comparable sample size to the young and nearby field stars. These planets are primarily randomly distributed across the sky, with a minimum distance from Earth of 1.8 pc. The spectral types of the exoplanets' host stars are shown in the top panel of Figure 2, while the distance distribution of the sample is shown by a cumulative distribution function in the top panel of Figure 3.
Most of the host stars are Sun-like or lower-mass dwarfs. It is important to note that through Zeeman broadening observations, active M dwarfs have been shown to have magnetic field strengths in excess of a few kilogauss, while more Sun-like stars are fractions of a kilogauss (Reiners 2012). The increased magnetic activity on later-type stars likely causes increased flares and other stellar activity, as they are driven by the surface magnetic fields. For CMI considerations, increased magnetic activity from a host star could translate to higher emitted power if it follows the radiometric Bode's Law (Lazio et al. 2004). Thus, M dwarfs could induce CMI signals stronger than those of Sun-like stars on their planetary companions, although perhaps only by a factor of a few, rather than orders of magnitude.
The planets themselves are a diverse population overall but with some notable trends. Like the overall exoplanet population, there are many hot Jupiters on one hand and super-Earths on the other. Should planetary radio emission exist, the radiation environments would likely be very different between various phenomenological regimes. For example, hot Jupiters can potentially be magnetically connected to their host stars and may induce CMI on the star itself, while further-out planets would likely solely depend on charged particles coming from stellar winds to incite planetary-based CMI (Zarka 2007).

Young Stars
Our young-star sample consists of objects from two starforming regions: Taurus and Upper Scorpius. These are the only two regions utilized, since they are the closest wellstudied clusters of young stellar populations within the field of view of all three of the radio sky surveys used for our project. Our interest in pursuing radio emission from young stars is due to the fact that they are much more magnetically active than main-sequence stars (Johns-Krull 2007; Lehtinen et al. 2016) and thus may directly generate more radio emission, as well as induce a greater amount of emission onto potential planetary companions. Young stars may, in fact, be the most likely class of stellar objects for which we may detect low-frequency radio emissions.
With an age of 2-3 Myr and distance between 130 and 160 pc from Earth (Galli et al. 2019), the Taurus star-forming region is one of the youngest and nearest large star-forming regions, making it a prime target for searching for radio emission. Its young stellar population was extensively investigated with the Spitzer Space Telescope and consists of 348 class I, II, and III sources with spectral types varying between B and late M (Luhman et al. 2010; see Figure 2).
At approximately 5-10 Myr in age (Pecaut et al. 2012;Barenfeld et al. 2016), the Upper Scorpius star-forming region is a subgroup of the larger Scorpius-Centaurus association, which, at a distance of about 145 pc (Preibisch & Mamajek 2008, p. 235), is the nearest OB association from the Earth. This sample draws on a survey of the known members of the Upper Scorpius star-forming region, which also includes members from the adjacent and overlapping ρ Oph starforming region (though to simplify, we will henceforth refer to this particular sample as the "Upper Sco YSOs"), and amounts to 863 objects (Luhman & Mamajek 2012). However, due to its relatively more advanced age in the context of star formation and the presence of more massive stars, the Upper Scorpius sample does include a significant number of stars in later evolutionary stages in addition to YSOs. As such, we explicitly identify the stellar evolutionary stage of any candidate detections in our study. Since the region is located at a low decl. relative to observatories in the Northern Hemisphere, only a subset of 117 sample objects were in the field of view of the VLSSr. However, a larger fraction of the total number of identified objects were visible in the TGSS and NVSS (see Table 1). The distribution of the stellar spectral types of the entire sample is shown in Figure 2.

Nearby Stars
Our final sample consists of nearby star systems identified by the Research Consortium on Nearby Stars (RECONS). Specifically, we use the sample identified by the RECONS astrometry program at the Cerro Tololo Inter-American Observatory (Henry et al. 2018). There are 821 entries in this sample, the majority of which are M dwarfs. The sample is reportedly 90% complete inside the area covered by the survey  within 10 pc of Earth. A plot of the stellar spectral types represented by this sample is shown in the bottom panel of Figure 2, while the distance distribution is shown by a cumulative distribution function in the bottom panel of Figure 3.
As previously detailed, our Sun emits low-frequency radio emissions through a wide range of mechanisms. Most nearby stars are also mature, main-sequence stars and thus represent analogs to the stand-alone Sun. This sample best serves as a potential benchmark for radio emission from main-sequence stars without the effect of known exoplanet companions or the highly increased activity associated with young stars.

Point-by-point Matching
The search for radio emission from our sample objects used two distinct approaches. In this section, we discuss results from cross-matching the optical source catalog with the radio sky survey catalogs, while in the next section, we present findings from performing stacking analysis and statistical significance tests on the remaining nondetections.
The main challenge in performing a point-by-point matching was that the location of radio sources is known with a lower accuracy (from a few to tens of arcseconds, depending on the specific survey and intensity of the radio emission) compared to the corresponding optical sources. This could lead to severe confusion between the source of interest and background sources at close angular separation. That consideration, among others, is accounted for in our matching procedure, which we will describe below.
As a first step, we corrected the coordinates of our targets to account for the proper motion that occurred between the objects' position measurements and the survey's observation. Overall, these corrections were very small compared to the angular resolution of the radio observations, even for high proper motion objects.
Second, we measured the position of each radio source near our sample objects by performing a 2D Gaussian fit using CASA and performed an initial search adopting a threshold radius equal to the synthesized beam FWHM of the radio surveys (see Table 1). This step returned 42 potential associations.
As a third step, we refined our search using the formalism detailed in Pineau et al. (2011Pineau et al. ( , 2017, which established the matching probability between sources in two or more catalogs, taking into consideration the respective astrometric uncertainties. In our case, we adopted the two-catalog procedure where we independently compare each radio survey to the catalog comprising Gaia coordinates of our sample objects. Because the latter have uncertainties between 3 and 5 orders of magnitude smaller than our radio coordinates (Gaia Collaboration et al. 2018), in the following analysis, we only consider the astrometric precision of radio observations, which is usually expressed as (Lindegren 1978;Cameron et al. 2009) where S/N is the signal-to-noise ratio of the radio emission. However, this expression is approximate in that it does not account for systematic errors that could dominate the astrometric precision in the limit of high S/N (low δ) and introduce systematic offsets in the source positions.
To investigate these points, in Appendix A, we present a study of the astrometric precision of TGSS and NVSS observations based on the analysis of 150 VLA calibrators. In summary, we find that NVSS and TGSS observations achieve a maximum astrometric precision of about 0 6 and 1 3, respectively. Furthermore, we find no systematic offsets in the positions of radio calibrators. Based on these results, we adopt the following astrometric precision for NVSS and TGSS sources: Pineau et al. (2017), we reevaluated the 42 possible associations found in the previous step using the recommended radius of 3.44δ s , where s is either the NVSS or TGSS. The factor 3.44 corresponds to a 3σ (99.73%) tolerance for completeness in associations between 2D ellipsoids. That is, for every 10,000 real associations between two catalogs, 27 would be missed by using this criterion. Applying this criterion, we are left with nine candidate detections among our sample objects. All of them are detected in NVSS, and two are also detected in TGSS.
The spatial association between optical and radio emission does not guarantee that they arise from the same source. Indeed, there is a nonzero probability of random alignment between our galactic sources and, for example, faraway radio galaxies and quasars. To investigate this point further, we estimate the number of expected chance associations assuming that all of the radio sources in TGSS and NVSS are unrelated to our sample objects and distributed uniformly across the sky. This corresponds to a sum of individual false association probabilities over all of the radio sources with our association criterion (again, assuming the optical position uncertainties are negligible by comparison). A rigorous derivation for this exercise is provided in Pineau et al. (2017), and we convert their Equation (78) into our notation below. Thus, the estimated number of chance associations, N f s , is given by where s i d is the astrometric precision for each source i in each radio survey, k (=3.44) is the association threshold, Ω s is the total field of view of the given survey s, N s is the total number of radio sources detected in that survey (see Table 1), and N o is the number of optical sample sources we are considering. We calculated N f s using the peak flux density and rms noise for all TGSS and NVSS sources. The estimated number of chance associations is 10 for the NVSS and 0.4 for the TGSS. Regarding the fact that these numbers are comparable to the number of candidate detections, we have attempted to determine whether our matches can be disqualified based on being background chance alignments. For each of our nine candidate detections, we consulted available X-ray, optical, infrared, and other radio survey data (see Section 2.2.3) to investigate the presence of known extragalactic sources within the matching radius. 3 One of our candidate detections was found to have an offset between the sample object and the radio emission, while a known quasar was coincident with said radio emission and, for this reason, was rejected as a match. Another two candidate detections were affected by source confusion with objects classified as extragalactic and were also removed from the list of matches, though, for completeness, we list them in Appendix B. Our remaining six candidates do not have any currently identified extragalactic counterpart that would immediately disqualify them from further consideration. Clearly, this is not a sufficient condition to claim that these radio sources are physically associated with our target object, but, for the reasons discussed below, we believe that five of them are redetections of previously known stellar radio sources. The sixth candidate might be associated with a young brown dwarf, but confirmation via higher angular resolution observations is required. Table 2 lists the main properties of the six candidate detections of radio emission from our sample objects. Figure 4 shows three objects for which we have obtained VLA followup observations, which provide a more precise localization of the radio emission (see Section 2.2.2), along with their corresponding quick-look VLASS images and contours from the NVSS. Figure 5 depicts the remaining three objects, along with background VLASS images and NVSS and TGSS (if detected) contours, but for which there are no recent follow-up observations. It is important to note that although we depict the corresponding VLASS quick-look images and use their fluxes to coarsely calculate spectral indices in our subsequent discussions, these data products are not final and may be subject to change (Lacy et al. 2019).

T Tauri
Of particular note is the clear detection of T Tauri (T Tau), a well-studied triple protostellar system and the eponym of the T Tauri class of variable stars (left panel of Figure 4). It has historically been observed as a variable radio source at 4.84, 8.44, and 14.94 GHz, including circularly polarized emission (Skinner & Brown 1994). Specifically, the emission has been observed to come from the southern component of the T Tau system, T Tau S, which itself is a binary system (Johnston et al. 2004, and references therein). The angular separation between T Tau N and T Tau S is 0 6 (Dyck et al. 1982), while the projected separation of the two components of T Tau S is 13 au (Duchene et al. 2002), meaning we are unable to resolve any of the individual components due to the coarse angular resolution of the radio sky surveys. Additionally, a 149 MHz detection of T Tau was recently made by LOFAR, with a peak flux of 0.96 ± 0.20 mJy beam −1 (Coughlan et al. 2017). The T Tau system is clearly detected (S/N > 10) in the 1.4 GHz NVSS, our follow-up 1.4 GHz VLA observations, and the 2-4 GHz quick-look VLASS image, though it is not detected in the 150 MHz TGSS. The Gaia position of T Tau is 4″ ± 3 9 away from the centroid of the NVSS emission. The centroid of our follow-up VLA observations and VLASS is 0.9 ± 0 2 away from T Tau N and 0.4 ± 0 2 from T Tau Sa/b (Schaefer et al. 2020), indicating that the latter might be responsible for the observed radio emission. By assuming a power-law relationship for the flux (i.e., S = ν α ), the spectral index between 1.4 and 3 GHz is α 1.4−3 GHz = −0.69 ± 0.14. Between 1.4 GHz and 150 MHz, a spectral index lower limit of α 0.15-1.4 GHz −0.62 can be estimated by adopting an upper flux limit at 150 MHz equal to 17.5 mJy, corresponding to 5× the rms noise of the TGSS.

ρ Ophiuchi
The source ρ Ophiuchi (ρ Oph) is a well-studied stellar system consisting of at least five bright stars: ρ Oph A, ρ Oph B, ρ Oph C, and the close separation binary system ρ Oph DE. Stars ρ Oph A and B have a projected separation of 344 au (3 1) and are located approximately 17,000 au from ρ Oph C (136″) and 19,000 au from ρ Oph DE (141″; Cordiner et al. 2013). Thus, A and B are unresolved in the utilized radio sky surveys, but their separation from C and DE is resolved. Recently, 2.1 GHz radio emission has been detected from ρ Oph A, which has been associated with auroral processes , as well as at 1.6 GHz for ρ Oph C ). Our analysis has detected 1.4 GHz radio emission from both ρ Oph A and ρ Oph C (right panel in Figure 4). Both stars have no other identified Simbad objects within the FWHM of each survey's respective synthesized beams (with the exception of the unresolved ρ Oph B near ρ Oph A). There are also no listed VizieR observations closer to the detected radio emission. The centroid of the 1.4 GHz radio emission is nearly coincident with the position of ρ Oph C, separated by just 2″ ± 2 7, while ρ Oph A is also separated by 2″ ± 2 7 from its associated radio emission. The 3 GHz VLASS detects sources at the exact locations of ρ Oph A and C, coincident with a detection made in our VLA follow-up observations. Our follow-up observations have separations and astrometric precisions of 0.88 ± 0 21 for ρ Oph A and 0.87 ± 0 05 for ρ Oph C (these uncertainties are purely flux-based in accordance with Equation (2) and do not account for any systematics). The spectral index between 1.4 and 3 GHz is α ρ Oph C = 0.36 ± 0.081 and α ρ Oph A = −0.29 ± 0.079. The lower limits on the spectral indices between 150 MHz and 1.4 GHz are α ρ Oph A −0.26 and α ρ Oph C −0.29. Together, T Tau, ρ Oph A, and ρ Oph C present benchmark cases of radio emission of YSOs, complete with follow-up observations.

HD 283447
The source 2MASS J04141291+2812124 (HD 283447) is shown in the top left panel of Figure 5. It is a spectroscopic binary system of weak-lined T Tauri stars with a modeled semimajor axis of approximately 0.3 au (Welty 1995). These stars are magnetically active and have been previously detected in the radio at 5 and 8.4 GHz (Feigelson et al. 1994;Torres et al. 2012). In fact, previous radio observations reveal variable linear and circularly polarized emission, along with radio structures around the inner binary stars extending out to ∼0.3 au, resolved through very long baseline interferometry (VLBI; Phillips et al. 1996).
The 1.4 GHz NVSS radio emission is also within the search radius of two other sample objects: 2MASS J04141188 +2811535 and 2MASS J04141358+2812492 (FM Tau). However, at a separation of 3″ ± 3 7, HD 283447 is the closest to the center of the radio emission. There are no other Simbad objects within the search radius of any of the sky surveys. UKIDSS detects a stellar source within 2″ of the NVSS detection. The spectral index between 1.4 and 3 GHz is α 1.4-3 GHz = 0.024 ± 0.112, and the lower limit on the spectral index between 150 MHz and 1.4 GHz is α 150 MHz-1.4 GHz −0.53. Since HD 283447 was previously detected at radio frequencies and fulfills our association threshold, we designate this as a direct redetection of the sample object.

α Sco
The star α Sco (Antares) is a multiple system with no other identified Simbad objects nearby and is shown in the top right panel of Figure 5. The primary component of this binary system is an evolved red supergiant, while the secondary is a main-sequence B-type star. It is a confirmed member of the Upper Scorpius star-forming region, and the evolved nature of one of its members is consistent with a 10 Myr cluster (Ohnaka et al. 2013). It is important to highlight this evolved nature, marking a sharp contrast with the evolutionary stages of the majority of other objects in the YSO sample.
The 1.4 GHz radio emission is located 2 6 ± 3 8 from the optical position of α Sco. There is also a 150 MHz signal, though it is relatively low significance, only achieving an S/N of about 4.1, and offset a bit at a separation of 11 4 ± 6 1 from the optical position of α Sco. The calculated spectral indices are α 1.4-3 GHz = −0.79 ± 0.128 and α 150 MHz-1.4 GHz = −0.44 ± 0.119. A known radio star, α Sco has been studied for its free-free emission and detected at multiple frequencies, such as 1.456, 4.9, and 15 GHz (Hjellming & Newell 1983;Florkowski et al. 1985). It is thus a clear redetection among our candidate detections.  Table 1 for survey quantities and Table 2 for observational quantities. All subpanels are displayed in relative coordinates. Table B1 lists the coordinates of both optical and radio sources.

A Tentative New Detection
The TGSS and NVSS observations reveal the presence of radio emission at the position of 2MASS J16044075-1936525 at 150 MHz and 1.4 GHz, respectively (bottom panel of Figure 5). The separation between the optical position and the centroid of the radio emission recorded at 150 MHz and 1.4 GHz is 4 4 ± 1 7 and 2 7 ± 1 2, respectively. By fitting a 2D Gaussian to TGSS and NVSS emission, we find that the radio emission is unresolved. The calculated spectral indices are α 1.4-3 GHz = −1.09 ± 0.061 and α 150 MHz-1.4 GHz = −0.40 ± 0.038. Additionally, inspection of the VLASS quick-look image at the position of this object shows a radio source with seemingly extended emission. Fitting this emission, we find that the source area is 16% larger than the synthesized beam. However, this might be explained by phase decorrelation, as discussed on the VLASS quick-look image web page. We once again emphasize that these quick-look images are not science-ready and contain clear noise artifacts.
Although this source satisfies our matching criterion, the previously discussed number of chance associations is comparable to the number of positive matches, which means that we are unable to statistically claim that the emission definitively originates from 2MASS J16044075-1936525. Therefore, additional follow-up observations are needed to ascertain the true origin of the radio emission. Nonetheless, fitting the radio source in the TGSS and NVSS reveals that it is pointlike, and we will discuss and analyze the sample object in the same manner as the previous redetections.
The sample object in question, 2MASS J16044075-1936525, is a brown dwarf with a spectral type of M6.5, as derived from low-resolution near-infrared spectroscopy (Martin et al. 2004). At a distance measured by Gaia of 159 ± 4 pc, it is a confirmed member of the Upper Scorpius star-forming region. Comparing stellar properties with evolutionary models from Baraffe et al. (1998), Kraus & Hillenbrand (2007) inferred a mass of 0.066 M e . Spitzer observations did not detect any infrared excess at 24 μm, while UVES/VLT observations measured a  Table 1 for survey quantities and Table 2 for observational quantities. All subpanels are displayed in relative coordinates.
narrow Hα emission line. These observations indicate the lack of both stellar accretion and warm circumstellar material (Scholz et al. 2007). Finally, this object was observed by Pan-STAARS, which measured photometric fluxes consistent with the DENIS survey by Martin et al. (2004).
Interestingly, Itoh et al. (2020) recently classified this radio source as a blazar candidate purely based on the spectral index measured between 150 MHz and 1.4 GHz. However, we suggest that this is a misclassification resulting from not accounting for Gaia and spectroscopic observations in the optical and NIR.

Summary of the Point-by-point Matching
All six candidate detections correspond to objects in the Taurus and Upper Scorpius star-forming region samples. All were detected in the NVSS, resulting in a 1.4 GHz detection rate in star-forming regions of 0.5%. Of the five sample redetections, four are associated with low-and intermediate-mass YSOs, and one is a young but evolved massive star confirmed to be in the star-forming association. The final positive match is at the position of a young brown dwarf candidate, but additional follow-up is needed to confirm the origin of the radio emission.
Two of our six matches, namely, at the position of the brown dwarf 2MASS J16044075-1936525 and the evolved star α Sco, were observed at 150 MHz, corresponding to a detection rate of 0.17%. Among the six detections, the two objects in Taurus, T Tau and HD 283447, are classified as class II protostars. However, none of the Upper Scorpius detection YSOs have definitive classifications through infrared excess measurements provided in the literature (Luhman & Mamajek 2012;Esplin et al. 2018). None of the discovered matches were detected in the VLSSr, establishing an upper limit on the 74 MHz detection rate of <0.24% at a flux density limit of 2.25 mJy beam −1 .
Considering our six detections, we calculate that the YSO sample has a detection rate of 0.5% at 1.4 GHz and 0.17% at 150 MHz (out of over 1200 objects), compared to the upper limits of <0.37% at 1.4 GHz and <0.34% at 150 MHz for the exoplanet sample and <0.19% at 1.4 GHz and <0.15% at 150 MHz for the RECONS nearby stars sample.

Ensemble Analysis of Nondetections
Point-by-point matching reveals radio emission from nearby stars only if the corresponding flux density is above the detection threshold at the time of the observations. This yielded the six matches discussed above. Instead, in this section, we sought to establish observational constraints on the overall behavior of stellar and planetary radio emission through image stacking and statistical significance tests of all of the available data.

Image Stacking
In the cases where objects emit just below the detection threshold, combining the radio emissions recorded at the positions of all of the sources of interest could result in a positive signal. This combination, called image stacking, was performed by first extracting small, postage stamp-like subimages centered on the objects of interest from the sky survey FITS files. All sample objects that did not yield a match with a detected radio source and did not have a bright radio source within the subimage field of view were considered for this exercise. The subimages were approximately 15′ × 15′ in size and further resized as needed in order to perform the image arithmetic.
For each of our four target samples (exoplanets, Taurus YSOs, Upper Sco YSOs, and nearby stars), images were stacked by calculating the weighted average ) is calculated from the local rms noise σ measured in the vicinity around the object of interest, though excluding the object itself directly. Indeed, while Table 1 lists the average rms noise value for each radio survey, we find that σ can vary substantially across the imaged field due to different observing conditions and deconvolution errors originating from bright sources. The rms noise of the resulting stacked images scales as where N is the number of sources in each survey belonging to each source sample (see last four columns of Table 1). However, the final stacked image of all objects did not yield any indication of underlying emission for any of our samples. For the 74 MHz VLSSr, the rms noise in the stacked images ranged from 4.5 to 17 mJy beam −1 across our various samples, and the center pixel never yielded an S/N higher than 1.4. The 150 MHz TGSS results in stacked rms noise between 0.11 and 0.17 mJy beam −1 and a center pixel S/N of 1 or less. Finally, the 1.4 GHz NVSS results in stacked rms noise between 0.021 and 0.034 mJy beam −1 and center pixel S/Ns of less than 0.1 in the final stacked images. Specific stacked image rms noise values for each permutation of surveys and samples are listed in the third column of Table 3.
A stacking analysis of an earlier and noisier data reduction of the 74 MHz VLSS (Cohen et al. 2007) was performed by Lazio et al. (2010). The reference source sample was comprised of several hundred nearby (<130 pc) F, G, and K main-sequence stars. The stacked images were characterized by a 3σ noise level between 13 and 33 mJy and delivered no detections. We note that there is minimal overlap between Lazio's sample and our exoplanet sample and no overlap with our YSO sample.
Despite our null result, our stacked images allow us to place upper limits on the average radio luminosity of the sample objects. As a first step, we repeat the image stacking after injecting a constant flux density at the position of each sample object. 4 Although this exercise does not correspond to a physical situation (i.e., a constant flux density would imply that all objects emit at the same intensity as seen from Earth, regardless of distance), it allows us to verify that the noise in the stacked image scales as N 1 N s µ even if the rms noise in our subimages is not constant. The fluxes required to obtain a 5σ detection in the stacked image are listed in the fourth column of Table 3.
To test a more physically motivated situation, we repeat the flux injection by assigning the same radio luminosity to all of our target objects and scaling by the object's distance to retrieve the corresponding flux measured on Earth, assuming isotropic emission. This process is repeated by adjusting the input radio luminosity until a 5σ detection in the stacked images is achieved. The results are summarized in the last two columns of Table 3 and include a scaling to the rotationaveraged radio luminosity of Jupiter during active periods at 22 MHz (L J = 6.18 × 10 18 erg s −1 ; ). The required luminosity varies over 4 orders of magnitude For all three surveys, the RECONS sample requires the least amount of injected luminosity before a detection is made and therefore provides the most stringent upper limits. This result is explained by the overall smaller distance of objects in the RECONS sample (see Figure 3).
For completeness, we investigated the effects that the true distance distributions of our samples have on image stacking. For each source group, we binned the sample objects by distance, injected some value of radio luminosity converted to flux, then stacked the binned subimages together until a 5σ detection was made. Comparing across all bins, we analyzed how the sensitivity limit varies with the distance. The results are presented in Figure 6 and show a clear trend that closer distances require far less injected luminosity before achieving a significant detection in the stacked image. In particular, it is noticeable that the luminosity required to achieve a detection in the stacked image of the nearest objects in each group (see leftmost bin in each panel) is lower than the sensitivity achieved by stacking all of the objects together (solid horizontal line). This contrasts with the naive assumption that for a uniform number density, an increasing distance and therefore volume would scale the number of included objects as ∝d 3 , which may outpace the falloff of flux with distance (∝1/d 2 ).
Lastly, rather than assuming that every object has the same radio luminosity, we consider the case in which only a subset of objects emit a flux just below the single-image 5σ detection threshold and investigate how many objects would be required to achieve a 5σ detection in the image obtained by stacking all of the sample objects together. This is essentially an analog of considering that during the observations, only a few objects were in an active state nearly capable of producing observable radio emission, while all of the others were quiescent. By randomly choosing the objects in the active state and repeating the experiment many times, we calculate that the number of active objects needed to produce a stacked signal 5σ varies between 2% and 5% of each sample. The takeaway is that only a small fraction of samples need to emit just below the individual detection limit in order to make a clear impact on the stacking analysis.

Two-sample Nonparametric Statistical Test
Image stacking works best when the distributions being analyzed are Gaussian in nature. However, given the potential for instrumental artifacts in the final cleaned images and the possibility of nonrandom underlying physical processes, we sought out a more generalized approach to derive statistical significances.
In order to independently verify the results of our stacking analysis, we employed the nonparametric two-sample Anderson-Darling (AD) test to derive analogous sensitivity limits for each sample. This test is nonparametric in that it does not assume that there are parameters to be modeled from the distribution in question. Specifically, the AD test was utilized because it compares several populations and is more sensitive to the tails of the distribution compared to other statistical tests, like the commonly used Kolmogorov-Smirnov test (Scholz & Stephens 1987). The AD test thus better accounts for outliers, which in our case may represent sample object fluxes that are systematically higher than the median but fall just short of the detection threshold.
This method analyzes the distribution of independent pixel values for all subimages without any inherent assumptions on the overall shape of those distributions. We compare the center pixel values (where the sample objects are located) to other relative pixel positions to evaluate whether the values appear to be drawn from equivalent or differing distributions. To ensure the comparison of independent elements, each relative pixel position is spaced apart by an FWHM of the synthesized beam of the corresponding sky survey. Satisfying the null hypothesis that all pixels are from equal distributions would suggest that the center pixels, which contain the sample objects, are consistent with background noise. Note. The third column lists the rms noise of the resulting stacked images (excluding the candidate detections). The fourth column presents the necessary artificially injected flux values needed to obtain a significant detection in the resulting stacked image. The last two columns show the average magnitude of radio luminosity each source would need in order to make a significant detection in the resulting stacked image in erg s -1 and scaled to Jupiter's average radio luminosity during active periods.
After taking the median results of the AD test statistics for each survey and sample, we find that there is no statistical evidence that the sample object pixel values are drawn from a different distribution compared to background pixels. However, like with image stacking, we can gauge the sensitivity and establish upper limits on the ensemble emission by artificially injecting flux into the subimages and repeating the statistical tests. In this case, we inject flux until the output significance is equivalent to a 5σ detection. The injection is performed in two schemes, directly paralleling our image stacking procedure: first by injecting a constant flux across all subimages, and second by specifying a given radio luminosity and converting into a corresponding flux received on Earth. The results of these flux injection tests are listed in Table 4 and included in our sensitivity ranges shown in Figure 7.
The upper limits on emission sensitivity derived using AD test comparisons of pixel values (Table 4) can be directly compared to the results from stacking subimages ( Table 3). The AD testderived upper limits are systematically higher, ranging from approximately 10% greater with the NVSS YSO samples to about 100% greater with the VLSSr and TGSS exoplanet samples. Regardless, the AD test results offer independent verification of the image stacking upper limits, with a maximum deviation of only a factor of 2. Because of the close agreement in results, we do not repeat the additional tests detailed in Section 4.2, such as injecting subsamples or the effect of distance binning in deriving upper limits, since those questions have already been investigated with image stacking.
The results of our ensemble analyses of the remaining nondetected objects establish upper limits on the quiescent radio luminosity of each of our samples at each of the survey frequencies. They rule out quiescent radio luminosities 3-7 orders of magnitude greater than Jupiter's average during active periods (Tables 3 and 4). Our distance-binned analysis also shows to what extent the stacking sensitivity is determined by the closest objects ( Figure 6).

Discussion
This work has produced six matches of low-frequency radio emissions from radio sky surveys with galactic stellar sources, all in the vicinity of star-forming regions. We have also provided an in-depth analysis of the upper limits of quiescent emission through image stacking and statistical tests. In this section, we discuss the implications of these results in the context of stellar and planetary emission.

Radio Emission from Young Stars
The higher detection rate of YSOs may be physically motivated by the fact that pre-main-sequence stars exhibit high levels of magnetic activity resulting from enhanced rotation and deep convective layers (Feigelson 2010), and accretion from the disk may also instigate flare-like events Figure 6. The 5σ sensitivity level (scaled to Jupiter's peak low-frequency radio luminosity) as a function of distance-binned subsets of sample objects. The downwardfacing triangles are the required injected flux to achieve a 5σ detection in the stacked image for each distance bin. The width of each bin is denoted by the horizontal error bars and is 15 pc for the exoplanet sample and 12.5 pc for the RECONS sample. The sensitivity limits from all nonstacked single images in each bin are shown by the lighter-colored vertical ranges, which overlap with the stacked results in bins with only a single element. Lastly, the sensitivity limit obtained by stacking all objects in that particular sample is given by the horizontal black line (from Table 3). Results are shown for 74 MHz VLSSr and 150 MHz TGSS exoplanet and RECONS samples. (Tofflemire et al. 2017). Studies of pre-main-sequence and zero-age main-sequence stars have revealed that the stellar magnetic activity, as probed by the ratio between X-ray and bolometric luminosity, increases during the first few million years as circumstellar disks dissipate and decays between the ages of about 10 Myr and 1 Gyr (Gudel et al. 1997;Flaccomio et al. 2003). While X-ray observations are the typical tracer for stellar activity, there are limitations, such as the saturation for fast stellar rotators (Pizzolato et al. 2003;Wright et al. 2018). Meanwhile, radio observations of different emission processes may reveal additional information and further our understanding of stellar activity outside of our solar system. For our Sun, an increase in magnetic activity leads to an increased occurrence and intensity of stellar flares and associated energetic particle events such as CMEs (Aarnio et al. 2011). However, this trend may not necessarily hold when considering the behavior of CMEs outside of the solar system. To date, there has only been a single tentative detection of an exosolar CME (Argiroffi et al. 2019), so radio observations of stars with increased magnetic activity, like YSOs, may prove to be crucial in making some first strides toward understanding stellar activity as a whole (Vourlidas et al. 2020). The CMEs are accompanied by distinct radio signatures, which are caused by shocked plasma moving through the corona (type II radio burst; Figure 7. Flux values plotted against emission frequency for predicted and observed radio emission. Blue symbols are predicted exoplanet radio emission for a rotation-dependent, magnetic energy model at the 0.1 mJy significance (Grieβmeier 2017; see main text). The light blue shaded region represents the ionospheric cutoff. The black triangles denote the 5σ single-image detection threshold for each of the megahertz-range radio surveys. Upper limits on the average emission derived from our stacking analyses are represented by the colored rectangles (VLSSr: purple; TGSS: green); see Tables 3 and 4 for exact values. Lastly, our candidate detections at 150 MHz are represented by red stars. Note. The third column presents the necessary artificially injected flux values needed to obtain the equivalent of a 5σ statistical variation among the subimages. The last two columns show the magnitude of radio luminosity each source would need in order to have a 5σ statistical variation in erg s -1 and scaled to Jupiter's average radio luminosity during active periods.
Claβen & Aurass 2002) and accelerated particles streaming along open magnetic field lines (type III radio burst; Reid & Ratcliff 2014). Although these bursts occur at lower frequencies in the solar system (100 MHz) and require time series data to identify them based on dynamic spectra, the increased magnetic activity in young stars provides an intriguing avenue to investigate and definitively classify exosolar CMEs based on radio emissions. Another mechanism capable of generating radio emissions is the reconnection of magnetic field lines in the stellar corona. Radio emissions at hundreds of megahertz have been associated with such reconnection events, which are produced by large flares and accompanied by a host of X-ray activity (Manoharan et al. 1996). A potential caveat is that young stars with stronger magnetic fields may actually correspond to those with lower observed activity, as stronger magnetic fields may suppress the bending and breaking that cause reconnection (Yang & Johns-Krull 2011).
A possible radio emission mechanism exclusive to YSOs arises from accretion shocks in protostellar disks. Models of stellar formation predict a shock front as the primordial gas cloud collapses into a protostellar disk, which reradiates emission at long wavelengths, including radio (Winkler & Newman 1980;Cassen & Moosman 1981). Additionally, as circumstellar disks dissipate and planets are directly exposed to the high-energy stellar radiation, star-planet interaction generating strong auroral emission at low radio frequencies is possible. Planets with close-in orbits and/or around stars with larger magnetic fields may be entrained within the Alfvén surface of their host star, which would generate CMI emission on the star itself, with the star-planet essentially acting as a Jupiter-Io analog (Kavanagh et al. 2021).
More broadly speaking, due to their enhanced activity, premain-sequence stars are capable of producing impressive flares with energies that can exceed 10 36 erg, equivalent to more than 10 4 × the strongest solar flares (e.g., the Carrington event; Jackman et al. 2018;Carvalho et al. 2021). Such violent space weather phenomena and the accompanying production of X-rays are expected to heat the circumstellar material to thousands of degrees, resulting in a variety of transient events including enhanced disk photoevaporation (Gorti & Hollenbach 2004), enhanced formation of molecular ion and radical chemical reactions (Semenov et al. 2004), increased ionization-induced turbulence leading to faster inward planet migration (Nelson & Papaloizou 2004), and flash melting of dust grains possibly producing the chondrules and calciumaluminum inclusions found in meteorites (Shu et al. 2001). All of these processes not only affect the agglomeration of planets in the terrestrial region between 0.5 and 5 au but also control the chemical compositions of these planets and their atmospheres. For all of these reasons, determining the physical origins of radio emissions from pre-main-sequence stars could be key to further understanding the environment and energy budget involved in the formation and early evolution of planets.
The first step in determining the physical mechanism responsible for the observed emission is to analyze its temporal and spectral behavior. Through our (single-epoch) VLA follow-up observations of ρ Oph A, ρ Oph C, and T Tau, we find flux density values at 1.4 GHz comparable to well within an order of magnitude of what was detected in the NVSS (see Table 2). Although this represents only two temporal data points, it does suggest that the observed emission might be quiescent rather than transient bursts. The calculated spectral indices for ρ Oph A, ρ Oph C, and T Tau between 1.4 and 3 GHz are moderately valued at −0.29, 0.36, and −0.69, respectively. These values are either within or near the thermal range of −0.1 to 2 (Condon & Ransom 2016). The detection of HD 283447 may also fall within this regime, though we do not yet have additional follow-up observations.
We can compare the brightness temperature of these radio emissions to what would be needed to explain these emissions as free-free. At these low radio frequencies, the Rayleigh-Jeans approximation can be used to calculate the brightness temperature, where c is the speed of light, k is the Boltzmann constant, Ω is the solid angle of the beam, and S is the flux density. Averaged over the entire beam of their respective surveys, the brightness temperatures of our detections range between 5-20 K at 1.4 GHz and 5500-15,000 K at 150 MHz. We can then relate these average brightness temperatures to a simplified free-free temperature by factoring in the ratio of the solid angles of the emission region and the beam. That is, with the solid angle of the emission, σ, being defined as where r is the emitting radius of the emission, and d is the distance to the system. We can estimate the required free-free emitting radius by imposing a characteristic physical temperature of 10 4 K, though in reality, the spectrum of radio freefree emission depends on many factors, such as electron density and mass-loss rate (Panagia & Felli 1975). The results of this simplified free-free model suggest radii of emission between 45-100 au at 1.4 GHz and 1000-1500 au at 150 MHz. These large emitting radii, along with the fact that the flux densities at 150 MHz are higher than those at 1.4 GHz, suggest that thermal free-free is likely not the dominating mechanism of these emissions.
A more comprehensive spectral analysis at additional frequencies is certainly needed, as this would provide a significant indication for what type of physical process is generating this emission. For example, CMI has a distinct spectral behavior, as the emission falls off sharply above the local cyclotron frequency. A logical next step would be to search for circular polarization from our discovered sources, which is a characteristic of various plasma emission processes such as CMI. Furthermore, the temporal monitoring of the observed low-frequency radio emission would also provide key information about the nature of the emission.
In regard to our reported detections possibly being chance associations, we present two lines of reasoning to argue that our results should not be dismissed. First, many of our detections were already known radio emitters at similar, if not the same, frequencies: T Tau, ρ Oph A, HD 283447, and α Sco. These are clearly redetections of stellar objects. Second, we can begin to account for false positives in various ways. The literature review of our candidate detections narrowed our matches from nine to six, since three were found to have nearby extragalactic sources more likely to be responsible for the emission. Additionally, during the course of this review, we discovered another three cases where alternate stellar objects not originally present in our samples did fulfill the match criterion with radio emission (detailed in Appendix C). These three alternate matches arose from reviewing an early pool of 42 potential associations and suggest that there may be many more qualifying matches that may or may not be real if one uses more comprehensive stellar samples.

Radio Emission from Brown Dwarfs
Although additional follow-up is needed to confirm the young brown dwarf 2MASS J16044075-1936525 as the origin of radio emission, it brings up an important discussion of this subclass of radio emitters. Observations of brown dwarfs at low radio frequencies provide important constraints on their magnetic field properties and convective dynamo generation mechanisms (Kao 2017). Furthermore, because of their similarities with the most massive planets, the study of the magnetic properties of brown dwarfs might shed light on the generation of planetary magnetic fields.
To date, roughly 10 brown dwarfs have been detected at or below 1.4 GHz, namely, at 144 (Vedantham et al. 2020b) and 610-1300 (Zic et al. 2019) MHz, so our discovery may add another important entry to this small but ever-growing cohort of objects. Detections at higher frequencies, up to about 18 GHz, are more plentiful (Hallinan et al. 2015;Kao et al. 2018), totaling roughly 30 radio detections of ultracold brown dwarfs (Lynch et al. 2016;Metodieva et al. 2017;Zic et al. 2019).
Object 2MASS J16044075-1936525 has a spectral type of M6.5 D, which fits among the observed set of M-, L-, and T-type brown dwarfs that have been observed in the radio. However, it is a confirmed member of the Upper Scorpius starforming region and has an estimated age between 5 and 10 Myr, which would make it the youngest brown dwarf detected at radio frequencies.
Additionally, while all of the other brown dwarfs were detected within a single band, the emission at the location of 2MASS J16044075-1936525 was detected at three widely separated frequencies, namely, 150 MHz, 1.4 GHz, and 3 GHz. This allowed us to measure spectral indexes of −0.4 and −1.1 between 150 MHz and 1.4 GHz and between 1.4 and 3 GHz, respectively. Taken at face value, these results might suggest a steepening of the radio emission between 1.4 and 3 GHz reminiscent of the sharp cutoff that characterizes CMI emission at a frequency near the electron cyclotron fundamental frequency (Section 1). Assuming that this is the case, our measurement would suggest a magnetic field intensity between about 0.5 and 1 kG, which is consistent with typical values of a few kilogauss for cool dwarfs (Reiners & Basri 2006;Kao et al. 2018). However, it is important to note that these data are not contemporaneous, so this type of rough estimate may not apply if these sources have a large degree of time variability in their emission fluxes.
Another difference from previous low-frequency radio detections of brown dwarfs would be the large intrinsic luminosity of the emission at the location of 2MASS J16044075-1936525. For example, Vedantham et al. (2020b) detected the cold T6.5 dwarf BDR J1750+3809 (d = 65 pc) at 144 MHz with a radio spectral luminosity of 5 × 10 15 erg s −1 Hz −1 , corresponding to a brightness temperature T b ∼ 2.8 × 10 14 K · x −2 , where x is the emitter radius in units of the brown dwarf characteristic radius of 7 × 10 9 cm. This discovery was notable for having a 100× higher radio spectral luminosity and brightness temperature compared to similar objects at comparable frequencies. Applying a similar analysis to our match, we measure a radio spectral luminosity at 150 MHz of ∼1.3 × 10 18 erg s −1 Hz −1 and a brightness temperature T b ≈ 6.5 × 10 16 K · x −2 . Adopting a radius of 0.5 R e (∼3.5 × 10 10 cm) for a 5 Myr 0.07 M e object (Baraffe et al. 2015), we calculate a brightness temperature T b ∼ 2.6 × 10 15 K. In conclusion, the 150 MHz luminosity and brightness temperature of the emission at the location of 2MASS J16044075-1936525 are about 260× and 10× higher than those of BDR J1750+3809, respectively. Such a large difference in intrinsic luminosity may be difficult to explain if the emission observed from both objects is solely due to a sub-Alfvénic planetary companion analogous to the Jupiter-Io system. For a companion with the same Mach number as Io and the same geometric configuration, the Poynting flux scales as R 2 vB 2 , where R is the radius of the companion, v is the differential velocity between the brown dwarf's magnetic field and the companion's orbit, and B is the magnetic field strength of the brown dwarf at the position of the companion (Zarka 2007;Vedantham et al. 2020b). Both old and young brown dwarfs have rotation periods of a few hours (Scholz et al. 2015;Kao et al. 2018), as well as both cold L types and active M types having magnetic field strengths of a few kilogauss (Berdyugina et al. 2017;Kao et al. 2018). Thus, to recover such a large factor of difference in emission would require other arguments.
Instead, a potentially more viable explanation for such a large difference in intrinsic luminosity among brown dwarf radio emitters would be differences in the plasmas of the brown dwarfs. Although there is limited evidence that their activity significantly varies by age, M-type brown dwarfs have indeed been found to be more active, as traced by Hα emission, than later-typed L dwarfs (Schmidt et al. 2007). Plasma processes that can take place within brown dwarf atmospheres likely contribute to observed radio emission; one of these may be Alfvén ionization, as described by Stark et al. (2013). In such cases, the strength of the seed mechanism is determined by local plasma properties, including the plasma density and temperature. Thus, a more active M-type brown dwarf may be able to sustain a larger degree of radio emission through its more energized local plasmas than a comparatively less active T dwarf. However, it is difficult to quantify such a difference without more information about the specific local plasma conditions, so a definitive conclusion remains elusive.
In contrast to other nearby brown dwarfs studied in the radio (see, e.g., the source sample in Kao et al. 2018), we do not know if the radio emission of our match is polarized and variable in time. Furthermore, little is known about the properties of our sample brown dwarf itself, like its rotation period. Therefore, the nature of the radio emission observed at the location of this object and, consequently, our estimate of its magnetic field are highly uncertain.

Radio Emission from Planets
The detection of radio emission from exoplanets is a highly sought-after goal, as planetary radio emission would reveal a wealth of information currently unavailable through other means. In particular, low-frequency radio emissions would probe planetary magnetic field characteristics, with further implications for atmospheric loss and habitability. Though none of the radio emission reported in this work can be unquestionably associated with planets, both the level of radio flux detected from young stars and the upper limits derived from the statistical analysis of nondetections inform us about planetary radio emission.
We begin our discussion of exoplanetary radio emission by analyzing our results in the context of other searches. Sirothia et al. (2014) performed a similar search for exoplanetary radio emission using the NVSS and an early data release of the TGSS and proposed a number of objects as potential matches. A couple of these, 61 Vir and 1RXS 1609, were also found in our initial matching efforts but have since been eliminated from consideration due to classification as extragalactic in the literature (Murphy et al. 2015) or, in the case of 1RXS 1609, thanks to better astrometric precision from our follow-up VLA observations (see Appendix B for more details). All but one of their other proposed matches do not meet our 5σ threshold to be considered a radio detection. Their final proposed match, HD 43197, was originally not included in our exoplanet sample due to its low decl. not being observed in the VLSSr field of view; however, we have not found any evidence to rule out this match and can even report that the final TGSS products reveal a 45 ± 2.9 mJy beam −1 source located about 22″ away in addition to the original NVSS match claimed by Sirothia et al. (2014). Meanwhile, recent low-frequency exoplanet detections made by LOFAR (Vedantham et al. 2020a;Turner et al. 2021) are too low in both frequency and sensitivity for our project to make any meaningful contributions.
There exist efforts to predict exoplanetary radio emission based on estimating planetary magnetic moments and stellar winds and assuming different physical mechanisms for radio emission (Grieβmeier et al. 2007;Grieβmeier 2017). The different models consist of the kinetic energy model, where the input radio power is dominated by the kinetic energy flux of impinging stellar winds; the magnetic energy model, in which the magnetic energy flux of the interplanetary magnetic field determines the output power; the CME model, similar to the kinetic energy model but where CMEs are the main driver of emission; and the unipolar interaction model, in which an unmagnetized planet provides energetic particles to produce emission closer to the stellar surface, analogous to a scaled-up version of the Jupiter-Io system (Zarka 2007). Out of these possibilities, the magnetic energy model appears to provide the most favorable predictions for radio emissions, with fluxes up to a few janskys and frequencies up to hundreds of megahertz.
Predictions for radio fluxes from the known population of exoplanets as of 2017 using the magnetic energy model are plotted in Figure 8 (adapted from Grieβmeier 2017). Although the emission of each object would produce continuous emission over multiple frequencies, the values plotted are the radio fluxes at the peak emission frequency, which also corresponds to the cutoff frequency for CMI emission. The main characteristic of this model is that it predicts that no planets would emit at frequencies above about 400 MHz. About 77.5% of the planets would have peak emission frequencies below 70 MHz, and 23% would peak at frequencies below the ionospheric cutoff at 10 MHz. Detection of the latter would therefore require observations from space. Regardless of peak emission frequency, only 34% of exoplanets are predicted to produce a flux density of 0.1 mJy or more. Based on this model, only a minor fraction of the exoplanet sample taken into consideration would emit in the frequency range covered by our observations. Objects whose predicted emission peaks above 70 MHz with fluxes above 0.1 mJy are labeled in the figure. For comparison, we also plot the detected fluxes at 150 MHz of the objects listed in Table 2 (red stars), as well as the 5σ single-image sensitivity of the VLSSr and TGSS at 74 and 150 MHz, respectively (black triangles), and the upper limits for 5σ detection derived from our image stacking and statistical distribution analyses (colored rectangles; see Sections 5.1 and 5.2).
At 150 MHz, the radio fluxes measured from young stars in Upper Scorpius and Taurus are bracketed by the predicted emission of a few massive hot Jupiters and low-mass brown dwarfs. For example, HD 41004 B b has a mass of 18.4 ± 0.2 M J and an orbital period of 1.3 days, and it orbits a 0.4 M e star with an estimated age of 1.6 Gyr at a distance of 41.6 pc (Zucker et al. 2004). Scaling its predicted flux of about 1 Jy to the distance of Upper Scorpius (145 pc), HD 41004 B b could have been clearly detected by the TGSS with a flux of about ∼70 mJy, i.e., comparable to that measured toward 2MASS J16044075-1936525. Interestingly, the HD 41004 system was recently targeted by dedicated GMRT observations aimed at detecting the exoplanet emission (Narang et al. 2021). No emission was detected down to a 3σ limit of 1.8 mJy at 150 MHz and 0.12 mJy at 400 MHz. The 150 MHz detection limit of these observations is about 500 times lower than the flux predicted by Grieβmeier (2017). Narang et al. (2021) mentioned a number of reasons why no emission was detected from the HD 41004 system. These include the key physical process driving the emission being different from the predicted magnetic energy model, the emission peaking at a different frequency due to a higher or lower planetary magnetic field strength, or unfavorable viewing geometry due to the radiation being beamed and/or time variable. Indeed, CMI radiation is highly anisotropic, and its beaming pattern strongly influences its visibility (Louis et al. 2019, and references therein). It is also associated with auroral emissions, which have highly time-variable amplitudes primarily driven by transient outbursts. It is therefore possible that HD 41004 B b was observed in a period of low activity. In summary, the nondetection of HD 41004 B b provides few constraints on planetary radio emission outside of the observationally derived upper limits on emission.
In this context, the approach of searching for planetary radio emission from hundreds of planetary systems via large field-ofview surveys should be more robust against time variability and radiation beaming, even though these surveys might not be able to achieve a comparable sensitivity of targeted observations of single systems. Regarding our detections at 150 MHz, there still remains the possibility that they may be due (in part) to CMI from undiscovered planetary-mass objects. However, additional evidence is needed before asserting this claim, likely in the form of circularly polarized dynamical spectra similar to those obtained for the Jupiter-Io system (Marques et al. 2017).
The 5σ sensitivity limit corresponding to image stacking and statistical analysis of 150 MHz observations (indicated in Figure 7 with a green rectangle) is substantially lower than the flux of five planets predicted to emit at this frequency. On one hand, this sensitivity limit sets constraints on the average quiescent radio luminosity of the targeted exoplanet systems. As shown in the bottom left panel of Figure 6, the strongest constraints are achieved for the nearest planetary systems and correspond to a radio luminosity of 4 × 10 3 L J . On the other hand, our sensitivity limit could, in principle, constrain the fraction of planets that might have been in outburst phase at the time of the observations. To this end, one of the statistical tests discussed in Section 5.1 involved injecting artificial emission with intensity equal to the 5σ single-image sensitivity limit (i.e., about 17.5 mJy at 150 MHz) into a subset of images. We found that if 2%-5% of our sample objects were to emit at this level, the image stacking would have returned a detection. Unfortunately, this fraction is larger than the fraction of planets expected to emit at 150 MHz, and, as a result, our analysis does not provide any useful constraints on the number of planets in outburst phase. However, more sensitive future surveys of a larger sample of exoplanets might return valuable upper limits, even in the case of nondetections.
As a final note, even if more planets are expected to be observable at 74 MHz, the low sensitivity of VLSSr observations does not allow us to place any stringent constraints on the planetary emission at this frequency. The brightest planets, the hot Jupiters Tau Boo b and WASP-18 b, have predicted fluxes of only about twice the rms noise of the VLSSr, and only three planets have a predicted flux higher than or comparable to even that of the 5σ sensitivity resulting from image stacking.

Conclusions
This project focused on searching for radio emission at frequencies between 74 MHz and 1.4 GHz from a sample of about 2600 stellar objects grouped in known exoplanetary systems (∼800 objects), nearby low-mass stars (∼600 objects), and YSOs in the Taurus, Upper Scorpius, and ρ Oph starforming regions (∼1200 objects). The selected sample encompassed stellar spectral types from B to late M and L ( Figure 2) and distances between 5 and 300 pc (Figure 3). The search for radio emission was performed by comparing the position of the sample objects to radio observations from three large field-of-view radio surveys: the VLSSr at 74 MHz (VLA), the TGSS at 150 MHz (GMRT), and the NVSS at 1.4 GHz (VLA). Furthermore, many ancillary observations at wavelengths from the radio, including the still-ongoing VLASS at 2-4 GHz (VLA) and follow-up VLA observations, to the optical were used to assess the nature of the detected emission and, when possible, discriminate between galactic and extragalactic origins (Section 2).
Our analysis reveals six radio source matches at 1.4 GHz within a criterion of 3.44 times the astrometric precision of the radio emissions. Of these six matches, two were also associated with radio emission at 150 MHz ( Table 2). None of our target stellar objects were detected at 74 MHz.
All six radio sources are located in the vicinity of starforming regions. Two of them are located in Taurus, and four are in Upper Scorpius/ρ Oph. All radio sources were detected at 1.4 GHz, resulting in a YSO sample detection rate at this frequency of 0.5% with a flux density limit of 2.25 mJy beam −1 . Two were also detected at 150 MHz, corresponding to a detection rate of about 0.17% at a flux density limit of 17.5 mJy beam −1 . Five of these matches are redetections of previously known radio emitting stars; four of them are lowand intermediate-mass young stars, and one is an evolved star. Our final reported match corresponds to the location of a young brown dwarf, which, if confirmed, would have a luminosity about 260 times higher than the current brightest brown dwarf observed at similar frequencies (Vedantham et al. 2020b). We find no radio emission from exoplanet-hosting systems and nearby stars, resulting in detection rates lower than 0.2%-0.4% between 150 MHz and 1.4 GHz. Thus, we demonstrate that YSOs are far more likely to produce observable radio emission than other populations, such as field stars or mature exoplanethosting systems.
Our results find that the probability of detecting lowfrequency radio emission from our samples of YSOs is at least twice as high as that of exoplanet-hosting systems and nearby stars, despite the greater average distance of star-forming regions. We argue that this might be explained by the stronger magnetic activity that characterizes pre-main-sequence stars (see, e.g., Preibisch & Feigelson 2005). The observed radio emission could arise from stellar bursts and flares, like those observed on the Sun, and/or might be regulated by the accretion of circumstellar material in objects that still retain a primordial disk. Additionally, in Section 6.3, we speculated that the observed emission at 150 MHz could also potentially arise from the magnetic interaction between the central star and possible planetary companions. However, future observations aimed at characterizing the polarization and spectral shape of the radio emission, as well as its temporal variability, are required to constrain the nature of the observed radio emission.
The statistical analysis of the nondetection of radio emissions was used to constrain the average quiescent radio luminosity of the three samples of objects under consideration (Section 5). For the YSOs in Taurus and Upper Scorpius, characterized by a common distance of about 150 pc, we found upper limits of about 10 7 , 10 5 , and 10 6 L J at 74 MHz, 150 MHz, and 1.4 GHz, respectively, where L J = 6.18 × 10 18 erg s −1 . For the exoplanet and RECONS samples, we derived upper limits for the radio luminosity as a function of the distance from Earth ( Figure 6). We found that the most constraining limits are obtained at 150 MHz for the objects closest to Earth and correspond to about 4 × 10 3 and 1 × 10 3 L J for the exoplanets and RECONS sample, respectively.
We conclude by discussing some of the pros and cons of using large field-of-view observations for searching for radio emission from stars and planets, as opposed to targeted observations. One of the main benefits of using the surveys is the fact that they can deliver information about a large number of objects distributed across most of the sky. For example, nearly 85% of the confirmed exoplanet population has a decl. above the southern cutoff of −25°for VLA observations. As already discussed in the previous sections, observing a large sample is important to maximize the probability of detecting strongly collimated emission. The large frequency range of existing radio surveys is another benefit, since they probe different regimes of radio emission. The VLSSr at 74 MHz is only a factor of a few above Jupiter's peak frequency of radio emission (22 MHz), which makes it a good candidate for detecting exoplanetary radio emission, alongside possible low-frequency stellar processes such as plasma emission from bursts. On the other hand, the NVSS at 1.4 GHz is much more likely to detect exclusively stellar radio emission. Lastly, the public availability of uniformly calibrated images is another convenience of existing radio surveys, as it simplifies the statistical analysis.
That being said, existing radio surveys have their share of shortcomings as well. First, their sensitivity is rather mediocre compared to that of dedicated observations. We mentioned the targeted 150 MHz GMRT observation of the exoplanet system HD 41004 B (Section 6.3), which achieved an rms noise about six times lower than that of the TGSS at the same frequency with the same instrument. In particular, the lack of a sensitive survey at frequencies below 100 MHz poses a major limit to the study of planetary radio emission. Second, the snapshot nature of existing radio surveys does not match well with the intrinsic time-variable nature of radio emission from mechanisms such as CMI. In fact, CMI from Jupiter varies on timescales from hours when the primary driver is Io to days when driven by enhanced solar activity (Panchenko et al. 2013).
It is clear that more work needs to be done in the area of detecting low-frequency radio emission from stars and planets. Administration. The National Radio Astronomy Observatory is a facility of the National Science Foundation operated under cooperative agreement by Associated Universities, Inc. We thank the staff of the GMRT that made these observations possible. The GMRT is run by the National Centre for Radio Astrophysics of the Tata Institute of Fundamental Research. This work has made use of data from the European Space Agency (ESA) mission Gaia (https://www.cosmos.esa.int/ gaia), processed by the Gaia Data Processing and Analysis Consortium (DPAC; https://www.cosmos.esa.int/web/gaia/ dpac/consortium). Funding for the DPAC has been provided by national institutions, in particular the institutions participating in the Gaia Multilateral Agreement. This publication makes use of data products from the Wide-field Infrared Survey Explorer, which is a joint project of the University of California, Los Angeles, and the Jet Propulsion Laboratory/ California Institute of Technology, and NEOWISE, which is a Possible Additional Matches In this Appendix, we discuss two additional sources that satisfied our matching criteria but whose association with radio emission is uncertain due to the limited astrometric precision of radio observations. The observational characteristics of the relevant radio and optical observations of all candidate detections (including those presented in Section 4 of the main text) are listed in Table B1.
Object 1RXS J160929.1−210524 was reported as a possible radio source by Sirothia et al. (2014) based on a previous analysis of the then-ongoing TGSS and NVSS data. This object is a pre-main-sequence star at a distance of 140 pc with a 14 M J companion orbiting at a semimajor axis of 330 au (Pecaut et al. 2012). We do not find any radio emission at 150 MHz (TGSS) above our 5σ detection threshold, which is higher than that adopted by Sirothia et al. (2014), but we recover an NVSS radio source at a distance of 7″ ± 5″ toward the southwest from the Gaia position. Furthermore, we note a radio source at 352 MHz (WISH) at 12″ ± 10″ from the optical source. More recent observations show radio emission at 1.4 GHz (our VLA follow-up observations) and 2-4 GHz (VLASS) about 13″ ± 0 2 southwest of the optical source. These new data better constrain the position of the radio emission (at least at 1.4 GHz) and disfavor an association with the 1RXS J1609 system. However, the association between the optical position and the radio emission observed by WISH might be worth follow-up observations at MHz frequencies.
Object 2MASS J15535586-2358410 (HD 142184) is a Be star in Upper Scorpius with no other identified Simbad objects within the search radius. We recover an NVSS radio source at 0 8 ± 2 7 from HD 142184 but also note a quasar identified by the Large Quasar Astrometric Catalog (Gattano et al. 2018) about 0 3 from the optical star. Additional high angular resolution VLA observations are required to investigate the nature of this source. Figure A1. Relative offset between catalog coordinates of 150 VLA calibrators and those measured in NVSS (left) and TGSS (right). The red plus signs and ellipses denote the mean and standard deviation of the offsets, respectively. Alternate Stellar Matches Here we discuss additional cases where alternate stellar objects not originally present in our samples fulfilled the match criterion with radio emission. After obtaining an initial pool of 42 possible matches and before refining our search as discussed in Section 4, we performed a literature search for any other objects, observations, or designations that may rule out the matches as false positives. In three cases, we found archival optical and infrared objects, classified as stellar, that satisfy our matching criterion of 3.44 times the astrometric precision of the radio emission while the original sample objects within the initial search radius did not. Because the nature of these sources is uncertain, we did not include them among the objects listed in Table 2 of the main text. The properties of these three sources are listed in Table C1.
A notable aspect of these serendipitous matches is that a relatively sizable number of them arose from taking a closer look at a small pool of initial objects. That is, whether or not these matches are examples of galactic stellar low-frequency radio emission, it shows that using more complete stellar samples may return many more potential associations with these radio surveys. Thus, the estimated number of chance associations may actually be dwarfed when using more comprehensive input stellar samples.
Object UGCS J043233.36+225730.7 is represented by a cyan dot in the top left panel of Figure C1. The classification of this source is somewhat unclear; the sixth data release of UKIDSS classifies it as a galaxy, while all subsequent releases list it as a star. The source is not detected in other optical/IR surveys, particularly Gaia. This object is separated from the radio emission by 8″ ± 3 2 at 1.4 GHz and 5″ ± 1″ at 150 MHz. The calculated spectral indices are α 1.4−3 GHz = −2.8 ± 0.298 and α 150 MHz−1.4 GHz = −1.07 ± 0.067. This match was discovered when taking a closer look at the objects around the Upper Scorpius YSO sample object 2MASS J04323205+2257266, which is well outside of our astrometric precision tolerance. There are no other identified Simbad objects within the FWHM of each survey's synthesized beams around the original sample object.
Pan-STARRS object ID 72642378644129733 is shown by the cyan dot in the top right panel of Figure C1. This object was detected in Gaia, although no parallax information is available, and it was also included in the TESS Input Catalog as a stellar source. The center of the 1.4 GHz radio emission is 3 5 ± 2 9 away from the sample object. The spectral index between 1.4 and 3 GHz is α 1.4−3 GHz = −0.91 ± 0.131. This match was discovered  Figure C1. Composite images of the three alternate match detections. Background: 3 GHz VLASS quick-look images. Red: 1.4 GHz NVSS; blue: 150 MHz TGSS. The cyan dot shows the alternate stellar object, the green star shows the original sample object, the plus signs show the radio emission centroid of the respective survey, and the black triangle (if present) shows a background galaxy. Contour levels are 3σ, 5σ, 7σ, and 10σ, where σ is the local rms noise. The lower left corner of each subpanel shows the FWHM of each survey's synthesized beam. See Table 1 of the main text for survey quantities and Table C1 for observational quantities. All subpanels are displayed in relative coordinates. Note. These are cases where the radio emission better corresponds in position to an alternate stellar source found through literature searches than to the original sample object. Radio frequency, flux density, and rms noise come from the NVSS and VLASS (and TGSS, when detected).
corresponding to a distance of 73 pc (Gaia Collaboration et al. 2018); and suggest a spectral type of M7 (Reyle 2018). Within Simbad, there is a nearby galaxy (6dFGS gJ155129.4-292812) 33″ from the original sample object, but it is 52″ from our alternate match object. Object UGCS J160133.80-244329.7 is represented by the cyan dot in the bottom panel of Figure C1. This object was observed in numerous surveys, such as UKIDSS and VISTA, and has been classified as "likely stellar." It has a Gaia DR2 parallax of 0.7682 mas, corresponding to a distance of 1.3 kpc, which would place it well outside the Upper Scorpius star-forming region. The center of the radio emission at 150 MHz and 1.4 GHz is 6″ ± 3″ and 13″ ± 2 9 away from the sample object, respectively. The calculated spectral indices are α 150 MHz-1.4 GHz = −0.80 ± 0.065 and α 1.4-3 GHz = −2.65 ± 0.196. This match was discovered when taking a closer look at the objects around the Upper Scorpius YSO sample object 2MASS J16013529-2443365, which, upon closer inspection, fell outside of our astrometric precision tolerance. There are no other identified Simbad objects within the FWHM of each survey's synthesized beams around the original sample object.