A search for relativistic ejecta in a sample of ZTF broad-lined Type Ic supernovae

All photometric observations presented in this work were conducted with the Palomar Schmidt 48-inch (P48) Samuel Oschin telescope as part of the ZTF survey (Bellm et al., 2019; Graham et al., 2019), using the ZTF camera (Dekany et al., 2020). In default observing mode, ZTF uses 30 s exposures, and survey observations are carried out in $r$ and $g$ band, down to a typical limiting magnitude of $\approx 20.5$ mag. P48 light curves were derived using the ZTF pipeline (Masci et al., 2019), and forced photometry (see Yao et al., 2019). Hereafter, all reported magnitudes are in the AB system. The P48 light curves are shown in Figures 1-2. All the light curves presented in this work will be made public on the Weizmann Interactive Supernova Data Repository (WISeREP¹¹1https://www.wiserep.org/).

Preliminary spectral type classifications of several of the SNe in our sample were obtained with the Spectral Energy Distribution Machine (SEDM) mounted on the Palomar 60-inch telescope (P60), and quickly reported to the Transient Name Server (TNS). The SEDM is a very low resolution ($R\sim 100$) integral field unit spectrograph optimized for transient classification with high observing efficiency (Blagorodnova et al., 2018; Rigault et al., 2019).

After initial classification, typically further spectroscopic observations are carried out as part of the ZTF transient follow-up programs to confirm and/or improve classification, and to characterize the time-evolving spectral properties of interesting events. Amongst the series of spectra obtained for each of the SNe presented in this work, we select one good quality photospheric phase spectrum (Figures 3-4; grey) on which we run SNID (Blondin & Tonry, 2007) to obtain the best match to a SN Ic-BL template (black), after clipping the host emission lines and fixing the redshift to that derived either from the SDSS host galaxy or from spectral line fitting (H$\alpha$; see Section 3 for further details). Hence, in addition to the SEDM, in this work we also made use of the following instruments: the Double Spectrograph (DBSP; Oke et al., 1995), a low-to-medium resolution grating instrument for the Palomar 200-inch telescope Cassegrain focus that uses a dichroic to split light into separate red and blue channels observed simultaneously; the Low Resolution Imaging Spectrometer (LRIS; Oke & Gunn, 1982; Oke et al., 1995), a visible-wavelength imaging and spectroscopy instrument operating at the Cassegrain focus of Keck-I; the Alhambra Faint Object Spectrograph and Camera (ALFOSC), a CCD camera and spectrograph installed at the Nordic Optical Telescope (NOT; Djupvik & Andersen, 2010). All spectra presented in this work will be made public on WISeREP.

For 9 of the 16 SNe presented in this work we carried out follow-up observations in the X-rays using the X-Ray Telescope (XRT; Burrows et al., 2005) on the Neil Gehrels Swift Observatory (Gehrels et al., 2004).

We analyzed these observations using the online XRT tools²²2See https://www.swift.ac.uk/user_objects/., as described in Evans et al. (2009). We correct for Galactic absorption, and adopt a power-law spectrum with photon index $\Gamma=2$ for count rate to flux conversion for non-detections, and for detections (two out of ninw events) where the number of photons collected is too small to enable a meaningful spectral fit (one out of two detections). Table 2 presents the results from co-adding all observations of each source.

We observed the fields of the SNe Ic-BL in our sample with the VLA via several of our programs using various array configurations and receiver settings (Table LABEL:tab:data).

The VLA raw data were calibrated in CASA (McMullin et al., 2007) using the automated VLA calibration pipeline. After manual inspection for further flagging, the calibrated data were imaged using the tclean task. Peak fluxes were measured from the cleaned images using imstat and circular regions centered on the optical positions of the SNe, with radius equal to the nominal width (FWHM) of the VLA synthesized beam (see Table LABEL:tab:data). The RMS noise values were estimated with imstat from the residual images. Errors on the measured peak flux densities in Table LABEL:tab:data are calculated adding a 5% error in quadrature to measured RMS values. This accounts for systematic uncertainties on the absolute flux calibration.

We checked all of our detections (peak flux density above $3\sigma$) for the source morphology (extended versus point-like), and ratio between integrated and peak fluxes using the CASA task imfit. All sources for which these checks provided evidence for extended or marginally resolved emission are marked accordingly in Table LABEL:tab:data. For non detections, upper-limits on the radio flux densities are given at the $3\sigma$ level unless otherwise noted.

Our first ZTF photometry of SN 2018etk (ZTF18abklarx) was obtained on 2018 August 1 (MJD 58331.16) with the P48. This first ZTF detection was in the $r$ band, with a host-subtracted magnitude of $19.21\pm 0.12$ mag (Figure 1), at $\alpha$=15^h17^m02.53^s, $\delta=+03^{\circ}56^{\prime}38\farcs 7$ (J2000). The object was reported to the TNS by ATLAS on 2018 August 8, who discovered it on 2018 August 6 (Tonry et al., 2018a). The last ZTF non-detection prior to ZTF discovery was on 2018 July 16, and the last shallow ATLAS non-detection was on 2018 August 2, at $18.75\,$mag. The transient was classified as a Type Ic SN by Fremling et al. (2018a) based on a spectrum obtained on 2018 August 13 with the SEDM. We re-classify this transient as a SN Type Ic-BL most similar to SN 2006aj based on a P200 DBSP spectrum obtained on 2018 August 21 (see Figure 3). SN 2018etk exploded in a star-forming galaxy with a known redshift of $z=0.044$ derived from SDSS data.

Our first ZTF photometry of SN 2018hom (ZTF18acbwxcc) was obtained on 2018 November 1 (MJD 58423.54) with the P48. This first ZTF detection was in the $r$ band, with a host-subtracted magnitude of $16.60\pm 0.04$ mag (Figure 1), at $\alpha$=22^h59^m22.96^s, $\delta=+08^{\circ}45^{\prime}04\farcs 6$ (J2000). The object was reported to the TNS by ATLAS on 2018 October 26, and discovered by ATLAS on 2018 October 24 at $o\approx 17.3\,$mag (Tonry et al., 2018b). The last ZTF non-detection prior to ZTF discovery was on 2018 October 9 at $g>20.35\,$mag, and the last ATLAS non-detection was on 2018 October 22 at $o>18.25\,$mag. The transient was classified as a SN Type Ic-BL by Fremling et al. (2018b) based on a spectrum obtained on 2018 November 2 with the SEDM. SN 2018etk exploded in a galaxy with unknown redshift. We measure a redshift of $z=0.030$ from star-forming emission lines in a Keck-I LRIS spectrum obtained on 2018 November 30. We plot this spectrum in Figure 3, along with its SNID template match to the Type Ic-BL SN 1997ef. We note that this SN was also reported in the recently released ASAS-SN bright SN catalog (Neumann et al., 2022).

Our first ZTF photometry of SN 2018hxo (ZTF18acaimrb) was obtained on 2018 October 9 (MJD 58400.14) with the P48. This first detection was in the $g$ band, with a host-subtracted magnitude of $18.89\pm 0.09$ mag (Figure 1), at $\alpha$=21^h09^m05.80^s, $\delta=+14^{\circ}32^{\prime}27\farcs 8$ (J2000). The object was first reported to the TNS by ATLAS on 2018 November 6, and first detected by ATLAS on 2018 September 25 at $o=18.36\,$mag (Tonry et al., 2018c). The last ZTF non-detection prior to discovery was on 2018 September 27 at $r>20.12\,$mag, and the last ATLAS non-detection was on 2018 September 24 at $o>18.52\,$mag. The transient was classified as a SN Type Ic-BL by Dahiwale & Fremling (2020a) based on a spectrum obtained on 2018 December 1 with Keck-I LRIS. In Figure 3 we plot this spectrum along with its SNID match to the Type Ic-BL SN 2002ap. SN 2018etk exploded in a galaxy with unknown redshift. We measure a redshift of $z=0.048$ from star-forming emission lines in the Keck spectrum.

Our first ZTF photometry of SN 2018jex (ZTF18acpeekw) was obtained on 2018 November 16 (MJD 58438.56) with the P48. This first detection was in the $r$ band, with a host-subtracted magnitude of $20.07\pm 0.29$ mag, at $\alpha$=11^h54^m13.87^s, $\delta=+20^{\circ}44^{\prime}02\farcs 4$ (J2000). The object was reported to the TNS by AMPEL on November 28 (Nordin et al., 2018). The last ZTF last non-detection prior to ZTF discovery was on 2018 November 16 at $r>19.85\,$mag. The transient was classified as a SN Type Ic-BL based on a spectrum obtained on 2018 December 4 with Keck-I LRIS. In Figure 3 we show this spectrum plotted against the SNID template of the Type Ic-BL SN 1997ef. AT2018jex exploded in a galaxy with unknown redshift. We measure a redshift of $z=0.094$ from star-forming emission lines in the Keck spectrum.

We refer the reader to Anand et al. (2022) for details about this SN Ic-BL. Its P48 light curves and the spectrum used for classification are shown in Figures 1 and 3. We note that this SN was also reported in the recently released ASAS-SN bright SN catalog (Neumann et al., 2022).

We refer the reader to Anand et al. (2022) for details about this SN Ic-BL. Its P48 light curves and the spectrum used for classification are shown in Figures 1 and 3.

Our first ZTF photometry of SN 2020jqm (ZTF20aazkjfv) was obtained on 2020 May 11 (MJD 58980.27) with the P48. This first detection was in the $r$ band, with a host-subtracted magnitude of $19.42\pm 0.13$ mag, at $\alpha$=13^h49^m18.57^s, $\delta=-03^{\circ}46^{\prime}10\farcs 4$ (J2000). The object was reported to the TNS by ALeRCE on May 11 (Forster et al., 2020). The last ZTF non-detection prior to ZTF discovery was on 2020 May 08 at $g>17.63\,$mag. The transient was classified as a SN Type Ic-BL based on a spectrum obtained on 2020 May 26 with the SEDM (Dahiwale & Fremling, 2020b). SN 2020jqm exploded in a galaxy with unknown redshift. We measure a redshift of $z=0.037$ from host-galaxy emission lines in a NOT ALFOSC spectrum obtained on 2020 June 6. We plot the ALFOSC spectrum along with its SNID match to the Type Ic-BL SN 1998bw in Figure 3.

We refer the reader to Anand et al. (2022) for details about this SN Ic-BL. Its P48 light curves and the spectrum used for classification are shown in Figures 1 and 3. We note that this SN was also reported in the recently released ASAS-SN bright SN catalog (Neumann et al., 2022).

We refer the reader to Anand et al. (2022) for details about this SN Ic-BL. Its P48 light curves and the spectrum used for classification are shown in Figures 1 and 4.

We refer the reader to Anand et al. (2022) for details about this SN Ic-BL. Its P48 light curves and the spectrum used for classification are shown in Figures 1 and 4.

We refer the reader to Anand et al. (2022) for details about this SN Ic-BL. Its P48 light curves and the spectrum used for classification are shown in Figures 1 and 4.

Our first ZTF photometry of SN 2021aug (ZTF21aafnunh) was obtained on 2021 January 18 (MJD 59232.11) with the P48. This first detection was in the $g$ band, with a host-subtracted magnitude of $18.73\pm 0.08$ mag, at $\alpha$=01^h14^m04.81^s, $\delta=+19^{\circ}25^{\prime}04\farcs 7$ (J2000). The last ZTF non-detection prior to ZTF discovery was on 2021 January 16 at $g>20.12\,$mag. The transient was publicly reported to the TNS by ALeRCE on 2021 January 18 (Munoz-Arancibia et al., 2021a), and classified as a SN Type Ic-BL based on a spectrum obtained on 2021 February 09 with the SEDM (Dahiwale & Fremling, 2021). SN 2020jqm exploded in a galaxy with unknown redshift. We measure a redshift of $z=0.041$ from star-forming emission lines in a P200 DBSP spectrum obtained on 2021 February 08. This spectrum is shown in Figure 4 along with its template match to the Type Ic-BL SN 1997ef.

Our first ZTF photometry of SN 2021epp (ZTF21aaocrlm) was obtained on 2021 March 5 (MJD 59278.19) with the P48. This first ZTF detection was in the $r$ band, with a host-subtracted magnitude of $19.61\pm 0.15$ mag (Figure 1), at $\alpha$=08^h10^m55.27^s, $\delta=-06^{\circ}02^{\prime}49\farcs 3$ (J2000). The transient was publicly reported to the TNS by ALeRCE on 2021 March 5 (Munoz-Arancibia et al., 2021b), and classified as a SN Type Ic-BL based on a spectrum obtained on 2021 March 13 by ePESSTO+ with the ESO Faint Object Spectrograph and Camera (Kankare et al., 2021). The last ZTF non-detection prior to discovery was on 2021 March 2 at $r>19.72\,$mag. In Figure 4 we show the classification spectrum plotted against the SNID template of the Type Ic-BL SN 2002ap. SN 2021epp exploded in a galaxy with known redshift of $z=0.038$.

Our first ZTF photometry of SN 2021htb (ZTF21aardvol) was obtained on 2021 March 31 (MJD 59304.164) with the P48. This first ZTF detection was in the $r$ band, with a host-subtracted magnitude of $20.13\pm 0.21$ mag (Figure 1), at $\alpha$=07^h45^m31.19^s, $\delta=46^{\circ}40^{\prime}01\farcs 4$ (J2000). The transient was publicly reported to the TNS by SGLF on 2021 April 2 (Poidevin et al., 2021). The last ZTF non-detection prior to ZTF discovery was on 2021 March 2, at $r>19.88\,$mag. In Figure 4 we show a P200 DBSP spectrum taken on 2021 April 09 plotted against the SNID template of the Type Ic-BL SN 2002ap. SN  2021htb exploded in a SDSS galaxy with redshift $z=0.035$.

Our first ZTF photometry of SN 2021hyz (ZTF21aartgiv) was obtained on 2021 April 03 (MJD 59307.155) with the P48. This first ZTF detection was in the $g$ band, with a host-subtracted magnitude of $20.29\pm 0.17$ mag (Figure 1), at $\alpha$=09^h27^m36.51^s, $\delta=04^{\circ}27^{\prime}11\farcs$ (J2000). The transient was publicly reported to the TNS by ALeRCE on 2021 April 3 (Forster et al., 2021). The last ZTF non-detection prior to ZTF discovery was on 2021 April 1, at $g>19.15\,$mag. In Figure 4 we show a P60 SEDM spectrum taken on 2021 April 30 plotted against the SNID template of the Type Ic-BL SN 1997ef. SN 2021hyz exploded in a galaxy with redshift $z=0.046$.

We refer the reader to Anand et al. (2022) for details about this SN Ic-BL. Its P48 light curves and the spectrum used for classification are shown in Figures 1 and 4.

We confirm the SN Type Ic-BL classification of each object in our sample by measuring the photospheric velocities (${\rm v_{ph}}$). SNe Ic-BL are characterized by high expansion velocities evident in the broadness of their spectral lines. A good proxy for the photospheric velocity is that derived from the maximum absorption position of the Fe II ($\lambda 5169$; e.g., Modjaz et al., 2016). We caution, however, that estimating this velocity is not easy given the strong line blending. We first pre-processed one high-quality spectrum per object using the IDL routine WOMBAT, then smoothed the spectrum using the IDL routine SNspecFFTsmooth (Liu et al., 2016), and finally ran SESNSpectraLib (Liu et al., 2016; Modjaz et al., 2016) to obtain final velocity estimates.

In Figure 5 we show a comparison of the photospheric velocities estimated for the SNe in our sample with those derived from spectroscopic modeling for a number of SNe Ib/c. The velocities measured for our sample are compatible, within the measurement errors, with what was observed for the PTF/iPTF samples. Measured values for the photospheric velocities with the corresponding rest-frame phase in days since maximum $r$-band light of the spectra that were used to measure them are also reported in Table LABEL:tab:opt_data.

We note that the spectra used to estimate the photospheric velocities of SN 2019xcc, SN 2020lao, and SN 2020jqm are different from those used for the classification of those events as SNe Ic-BL (see Section 3 and Figure 3). This is because for spectral classification we prefer later-time but higher-resolution spectra, while for velocity measurements we prefer earlier-time spectra even if taken with the lower-resolution SEDM.

In our analysis we correct all ZTF photometry for Galactic extinction, using the Milky Way (MW) color excess $E(B-V)_{\mathrm{MW}}$ toward the position of the SNe. These are all obtained from Schlafly & Finkbeiner (2011). All reddening corrections are applied using the Cardelli et al. (1989) extinction law with $R_{V}=3.1$. After correcting for Milky Way extinction, we interpolate our P48 forced-photometry light curves using a Gaussian Process via the GEORGE³³3https://george.readthedocs.io package with a stationary Matern32 kernel and the analytic functions of Bazin et al. (2009) as mean for the flux form. As shown in Figure 1, the colour evolution of the SNe in our sample are not too dissimilar with one another, which implies that the amount of additional host extinction is small. Hence, we set the host extinction to zero. Next, we derive bolometric light curves calculating bolometric corrections from the $g$- and $r$-band data following the empirical relations by Lyman et al. (2014, 2016). For SN 2018hxo, since there is only one $g$-band detection, we assume a constant bolometric correction to estimate its bolometric light curve. These bolometric light curves are shown in the bottom panel of Figure 2.

We estimate the explosion time ${\rm T}_{\rm exp}$ of the SNe in our sample as follows. For SN 2021aug, we fit the early ZTF light curve data following the method presented in Miller et al. (2020), where we fix the power-law index of the rising early-time temporal evolution to $\alpha=2$, and derive an estimate of the explosion time from the fit. For most of the other SNe in our sample, the ZTF $r$- and $g$-band light curves lack enough early-time data to determine an estimate of the explosion time following the formulation of Miller et al. (2020). For all these SNe we instead set the explosion time to the mid-point between the last non-detection prior to discovery, and the first detection. Results on ${\rm T}_{\rm exp}$ are reported in Table LABEL:tab:opt_data.

We fit the bolometric light curves around peak ($-20$ to 60 rest-frame days relative to peak) to a model using the Arnett formalism (Arnett, 1982), with the nickel mass ($M_{\rm Ni}$) and characteristic time scale $\tau_{m}$ as free parameters (see e.g. Equation A1 in Valenti et al., 2008). The derived values of $M_{\rm Ni}$ (Table LABEL:tab:opt_data) have a median of $\approx 0.22$ M_⊙, compatible with the median value found for SNe Ic-BL in the PTF sample by Taddia et al. (2019), somewhat lower than for SN 1998bw for which the estimated nickel mass values are in the range $(0.4-0.7)$ M_⊙, but comparable to the $M_{\rm Ni}\approx 0.19-0.25$ M_⊙ estimated for SN 2009bb (see e.g., Lyman et al., 2016; Afsariardchi et al., 2021). We note that events such as SN 2019xcc and SN 2021htb have relatively low values of $M_{\rm Ni}$, which are however compatible with the range of $0.02-0.05$ M_⊙ expected for the nickel mass of magnetar-powered SNe Ic-BL (Nishimura et al., 2015; Chen et al., 2017; Suwa & Tominaga, 2015).

Next, from the measured characteristic timescale $\tau_{m}$ of the bolometric light curve, and the photospheric velocities estimated via spectral fitting (see previous Section) we derive the ejecta mass ($M_{\rm ej}$) and the kinetic energy ($E_{\rm k}$) via the following relations (see e.g. Equations 1 and 2 in Lyman et al., 2016):

where we assume a constant opacity of $\kappa=0.07$ g cm^-2.

We note that to derive $M_{\rm ej}$ and $E_{\rm k}$ as described above we assume the photospheric velocity evolution is negligible within 15 days relative to peak epoch, and use the spectral velocities measured within this time frame to estimate ${\rm v}_{\rm ph,max}$ in Equation 1. However, there are four objects in our sample (SN 2018hom, SN 2018hxo, SN 2020tkx, and SN 2021hyz) for which the spectroscopic analysis constrained the photospheric velocity only after day 15 relative to peak epoch. For these events, we only provide lower limits on the ejecta mass and kinetic energy (see Table LABEL:tab:opt_data).

Considering only the SNe in our sample for which we are able to measure the photospheric velocity within 15 d since peak epoch, we derive median values for the ejecta masses and kinetic energies of $1.7$ M_⊙ and $2.2\times 10^{51}$ erg, respectively. These are both a factor of $\approx 2$ smaller than the median values derived for the PTF/iPTF sample of SNe Ic-BL (Taddia et al., 2019). This could be due to either an intrinsic effect, or to uncertainties on the measured photospheric velocities. In fact, we note that the photospehric velocity is expected to decrease very quickly after maximum light (see e.g. Figure 5). Since the photospheric velocity in Equation (1) of the Arnett formulation is the one at peak, our estimates of ${\rm v}_{\rm ph,max}$ could easily underestimate that velocity by a factor of $\approx 2$ for many of the SNe in our sample. This would in turn yield an underestimate of $M_{\rm ej}$ by a factor of $\approx 2$ (though the kinetic energy would be reduced by a larger factor). A more in-depth analysis of these trends and uncertainties will be presented in Srinivasaragavan et al. (2022).

Based on the explosion dates derived for each object in Section 4.2 (Table LABEL:tab:opt_data), we searched for potential GRB coincidences in several online archives. No potential counterparts were identified in both spatial and temporal coincidence with either the Burst Alert Telescope (BAT; Barthelmy et al. 2005) on the Neil Gehrels Swift Observatory⁴⁴4See https://swift.gsfc.nasa.gov/results/batgrbcat. or the Gamma-ray Burst Monitor (GBM; Meegan et al., 2009) on Fermi⁵⁵5See https://heasarc.gsfc.nasa.gov/W3Browse/fermi/fermigbrst.html..

Several candidate counterparts were found with temporal coincidence in the online catalog from the KONUS instrument on the Wind satellite (SN 2018etk, SN 2018hom, SN 2019xcc, SN 2020jqm, SN 2020lao, SN 2020tkx, SN 2021aug). However, given the relatively imprecise explosion date constraints for several of the events in our sample (see Table LABEL:tab:opt_data), and the coarse localization information from the KONUS instrument, we cannot firmly associate any of these GRBs with the SNe Ic-BL. In fact, given the rate of GRB detections by KONUS ($\sim 0.42$ d^-1) and the time window over which we searched for counterparts ($30$ d in total; derived from the explosion date constraints), the observed number of coincidences (13) is consistent with random fluctuations. Finally, none of the possible coincidences were identified in events with explosion date constraints more precise than 1 d.

Seven of the 9 SNe Ic-BL observed with Swift-XRT did not result in a significant detection. In Table 2 we report the derived 90% confidence flux upper limits in the 0.3–10 keV band after correcting for Galactic absorption (Willingale et al., 2013).

While observations of SN 2021ywf resulted in a $\approx 3.2\sigma$ detection significance (Gaussian equivalent), the limited number of photons (8) precluded a meaningful spectral fit. Thus, a $\Gamma=2$ power-law spectrum was adopted for the flux conversion for this source as well. We note that because of the relatively poor spatial resolution of the Swift-XRT (estimated positional uncertainty of 11.7″ radius at 90% confidence), we cannot entirely rule out unrelated emission from the host galaxy of SN 2021ywf (e.g., AGN, X-ray binaries, diffuse host emission; see Figure 8 for the host galaxy).

For SN 2019hsx we detected enough photons to perform a spectral fit for count rate to flux conversion. The spectrum is found to be relatively soft, with a best-fit power-law index of $\Gamma=3.9^{+3.0}_{-2.1}$. Our Swift observations of SN 2019hsx do not show significant evidence for variability of the source X-ray flux over the timescales of our follow up. While the lack of temporal variability is not particularly constraining given the low signal-to-noise ratio in individual epochs, we caution that also in this case the relatively poor spatial resolution of the Swift-XRT (7.4″ radius position uncertainty at 90% confidence) implies that unrelated emission from the host galaxy cannot be excluded.

The constraints derived from the Swift-XRT observations can be compared with the X-ray light curves of low-luminosity GRBs, or models of GRB afterglows observed slightly off-axis. For the latter, we use the numerical model by van Eerten & MacFadyen (2011); van Eerten et al. (2012). We assume equal energies in the electrons and magnetic fields ($\epsilon_{B}=\epsilon_{e}=0.1$), and an interstellar medium (ISM) of density $n=1-10$ cm^-3. We note that a constant density ISM (rather than a wind profile) has been shown to fit the majority of GRB afterglow light curves, implying that most GRB progenitors might have relatively small wind termination-shock radii (Schulze et al., 2011). We generate the model light curves for a nominal redshift of $z=0.05$ and then convert the predicted flux densities into X-ray luminosities by integrating over the 0.3-10 keV energy range and neglecting the small redshift corrections. We plot the model light curves in Figure 6, for various energies, different power-law indices $p$ of the electron energy distribution, and various off-axis angles (relative to a jet opening angle, set to $\theta_{j}=0.2$). In the same Figure we also plot the X-ray light curves of low-luminosity GRBs for comparison (neglecting redshift corrections). As evident from this Figure, our Swift/XRT upper limits (downward-pointing triangles) exclude X-ray afterglows associated with higher-energy GRBs observed slightly off-axis. However, X-ray emission as faint as the afterglow of the low-luminosity GRB 980425 cannot be excluded. As we discuss in the next Section, radio data collected with the VLA enable us to exclude GRB 980425/SN 1998bw-like emission for most of the SNe in our sample.

We note that our two X-ray detections of SN 2019hsx and SN 2021ywf are consistent with several GRB off-axis light curve models and, in the case of SN 2021ywf, also with GRB 980425-like emission within the large errors. However, for this interpretation of their X-ray emission to be compatible with our radio observations (see Table LABEL:tab:data), one needs to invoke a flattening of the radio-to-X-ray spectrum, similar to what has been invoked for other stripped-envelope SNe in the context of cosmic-ray dominated shocks (Ellison et al., 2000; Chevalier & Fransson, 2006).

As evident from Table LABEL:tab:data, we have obtained at least one radio detection for 11 of the 16 SNe in our sample. None of these 11 radio sources were found to be coincident with known radio sources in the VLA FIRST (Becker et al., 1995) catalog (using a search radius of 30″ around the optical SN positions). This is not surprising since the FIRST survey had a typical RMS sensitivity of $\approx 0.15$ mJy at 1.4 GHz, much shallower than the deep VLA follow-up observations carried out within this follow-up program. We also checked the quick look images from the VLA Sky Survey (VLASS) which reach a typical RMS sensitivity of $\approx 0.12$ mJy at 3 GHz (Villarreal Hernández & Andernach, 2018; Law et al., 2018). We could find images for all but one (SN 2021epp) of the fields containing the 16 SNe BL-Ic in our sample. The VLASS images did not provide any radio detection at the locations of the SNe in our sample.

Five of the 11 SNe Ic-BL with radio detections are associated with extended or marginally resolved radio emission. Two other radio-detected events (SN 2020rph and SN 2021hyz) appear point-like in our images, but show no evidence for significant variability of the detected radio flux densities over the timescales of our observations. Thus, for a total of 7 out of 11 SNe Ic-BL with radio detections, we consider the measured flux densities as upper-limits corresponding to the brightness of their host galaxies, similarly to what was done in e.g. Soderberg et al. (2006a) and Corsi et al. (2016). The remaining 4 SNe Ic-BL with radio detections are compatible with point sources (SN 2018hom, SN 2020jqm, SN 2020tkx, and SN 2021ywf), and all but one (SN 2018hom) had more than one observation in the radio via which we were able to establish the presence of substantial variability of the radio flux density. Hereafter we consider these 4 detections as genuine radio SN counterparts, though we stress that with only one observation of SN 2018hom we cannot rule out a contribution from host galaxy emission, especially given that the radio follow up of this event was carried out with the VLA in its most compact (D) configuration with poorer angular resolution.

In summary, our radio follow-up campaign of 16 SNe Ic-BL resulted in 4 radio counterpart detections, and 12 deep upper-limits on associated radio counterparts.