The early radio afterglow of short GRB 230217A

Binary neutron star (BNS) mergers are the key to understanding the behaviour of nuclear matter at extreme densities (Lasky et al., 2014; Ravi & Lasky, 2014; Abbott et al., 2018) and are likely the creation sites of the heaviest elements in the Universe (Pian et al., 2017; Levan et al., 2024). Short-duration Gamma-ray bursts (GRBs) have long been thought to arise from BNS mergers (Lattimer & Schramm, 1976; Eichler et al., 1989; Narayan et al., 1992; Mochkovitch et al., 1993), with the near-simultaneous detection of gravitational wave (GW) event GW170817 and GRB 170817A conclusively establishing this link (Abbott et al., 2017). Short GRBs (SGRBs) are classified based on their prompt gamma-ray emission lasting $<2$ s and being spectrally hard, while the other main class are long GRBs, which last $>2$ s and are spectrally soft (Kouveliotou et al., 1993). This emission arises from highly collimated relativistic jets driven by a black hole or rapidly spinning neutron star from the BNS merger (Rees & Meszaros, 1992; Ackermann et al., 2010; Rezzolla et al., 2011). SGRBs only represent $\sim 10-20$% of the well-localised GRB population discovered by instruments such as the Neil Gehrels Swift Observatory (hereafter Swift; Gehrels et al., 2004) Burst Alert Telescope (BAT; Barthelmy et al., 2005), making them quite rare with $\lesssim 10$ detected per year with Swift.

The jet launched during an SGRB can result in a multi-wavelength afterglow as predicted by the fireball model (Rees & Meszaros, 1992; Kobayashi et al., 1997; Sari & Piran, 1997; Piran, 1999; Kobayashi & Sari, 2000). In this model, the relativistic ejecta interacts with the circumburst medium (CBM), producing a forward shock that accelerates electrons and generates synchrotron emission detectable from radio to X-ray (and sometimes high-energy gamma-ray) wavelengths or energies from days to years post-burst. A reverse shock is also generated, which propagates back into the shocked ejecta and can be detected as a distinct and much faster-evolving synchrotron component. This rapid evolution means that the reverse shock usually fades below detectability at X-ray and optical wavelengths within a few minutes (e.g. the optical reverse shock detection $<90$ s post-burst from the possibly short GRB 180418A; Becerra et al., 2019), making the radio band the most viable option for detecting and tracking its evolution (e.g. van der Horst et al., 2014; Bright J. S. & Rhodes L. et al., 2023).

Radio observations are crucial to broadband modelling, which can map the afterglow flux density to the physical parameters of the explosion, such as its energetics and magnetic field strength as well as the density and profile of the CBM (Granot & Sari, 2002). Currently, only 17 SGRBs have been detected in the radio band (not including GW170817; Berger et al., 2005; Soderberg et al., 2006; Fong et al., 2014, 2015, 2021; Lamb et al., 2019; Troja et al., 2019; Laskar et al., 2022; Anderson et al., 2023a; Rhodes et al., 2023; Chastain et al., 2024; Schroeder et al., 2024a, b), representing $\sim 13\%$ of the full SGRB population with X-ray afterglows (Schroeder et al., 2024b). However, Schroeder et al. (2024b) showed that if only those events detected following the VLA upgrade are considered (the instrument from which most detections are made) the detected fraction is much higher at $\sim 34$%, illustrating the importance of sensitivity for detecting SGRB radio afterglows.

Of this sample of 17 SGRBs, 11 were detected in the radio band within $1$ day post-burst between $4-10$ GHz, and in several cases within a few hours. Interestingly, 8 events had faded below detectability by the time of their second or third observation between 4-9 days post-burst. This indicates that $\sim 50$% of the radio-detected SGRB population may fade within a few days post-burst in the observer frame. Meanwhile, 6 SGRBs have shown radio emission beyond 10 days post-burst, including GRB 211106A and GRB 230205A that were detected up to $\sim 100$ days (Laskar et al., 2022; Schroeder et al., 2024b). Of these, GRB 210726A is unusual as it went undetected until 11.2 days post-burst, at which point a radio flare lasting 50 days was detected with both MeerKAT and VLA, the likely explanation being late-time energy injection or a reverse shock from a shell collision (Schroeder et al., 2024a).

The low number of radio-detected SGRBs may be due to the observed short radio afterglow lifetime as seen from $\sim 50$% of the detected population. This rapid evolution indicates there should be significant variability on minute-to-hour timescales from reverse shock emission or scintillation that is likely missed by the $\sim 1$ h duration snapshot observations often obtained in the radio band (Fong et al., 2015; Schroeder et al., 2024b). To ensure radio detections and comprehensive modelling of the radio afterglow, we require rapid follow-up with observations lasting for several hours post-burst.

To increase the number of radio-detected short GRBs and probe the early-time radio parameters space, we initiated a follow-up program using the Australia Telescope Compact Array (ATCA) in 2018 that utilises the rapid-response triggering system (Anderson et al., 2021). When a short GRB is detected by Swift-BAT, the rapid-response system automatically observes the event as soon as it is above the horizon provided it is at a Declination $<+20$ deg (with an exclusion zone of $-5<$Dec$<+5$ deg necessary for East-West arrays). This ensures an observation occurs $<1$ day post-burst, with integrations up to 12 h. We then continue to observe the GRB three or four more times, which are logarithmically spaced in time, up to $\sim 20$ days post-burst. The first ATCA rapid-response observation was of short GRB 181123B, which began 12.6 h post-burst when the source had risen above the horizon (Anderson et al., 2021). A further four events were triggered as part of this program between 2020 and 2022, three of which were observed $<1$ day post-burst with three additional observations in the following $\sim 2$ weeks, all resulting in non-detections. These observations were combined with quasi-simultaneous observations performed by the South African Square Kilometre Array precursor MeerKAT telescope, which enabled tight constraints to be placed on the gamma-ray efficiency and environmental density of 8 SGRBs (Chastain et al., 2024).

Here we present the first ATCA rapid-response radio detection of an SGRB. ATCA began observing GRB 230217A just 32 minutes post-burst for 7 hours resulting in the earliest radio detection of any GRB to date. Further observations with ATCA and the Karl G. Jansky Very Large Array (VLA) tracked the rapid decline of this radio emission, which faded below detectability between $5-10$ GHz in less than $2$ days post-burst. In the following, we describe our observations and processing in Section 2 and radio analysis in Section 3. The GRB 230217A radio afterglow is then interpreted in the context of the SGRB population with radio detections $<1$ day post-burst in Section 4. Throughout this paper we use the following power law representation for flux density $S\propto t^{\alpha}\nu^{\beta}$ and assume a cosmology of $H_{0}=67.4$ and $\Omega_{m}=0.315$ (Planck Collaboration et al., 2020).

The short GRB 230217A was initially reported by both the Swift-BAT and the Fermi Gamma-ray Space Telescope Gamma-ray Burst Monitor (Meegan et al., 2009) at 21:53:10 UT on 2023 Feb 17 (Moss et al., 2023; Fermi GBM Team, 2023). The Swift X-ray Telescope (XRT; Burrows et al., 2005) identified the X-ray afterglow (Capalbi et al., 2023a, b). Unfortunately, the majority of XRT data was unreliable thus preventing a robust analysis of the X-ray light curve properties¹¹1https://www.swift.ac.uk/xrt_curves/01154967/ (Evans et al., 2007, 2009), which is likely due to a Swift Moon constraint (Moss et al., 2023) and the delay in identifying its X-ray afterglow (Capalbi et al., 2023a, b). The optical afterglow was also very faint (O’Connor & Troja, 2023; Gillanders et al., 2023; Swift/BAT trigger et al., 2023) with only a single detection reported with magnitude $I\sim 24.5$ at 2.5 d (D’Avanzo et al., 2023). The radio afterglow was detected with both ATCA (Anderson et al., 2023a) and the VLA (Schroeder et al., 2024b) $<1$ day post-burst and marginally detected in a single MeerKAT observation at 5 days (Chastain et al., 2024). The data processing and results of the ATCA and VLA observations are described in the following sections.

The radio afterglow of GRB 230217A was detected up to 1.2 day post-burst with ATCA and VLA in the $5.5-6$ GHz radio band. Having split the ATCA rapid-response observation into three and two images at 5.5 and 9 GHz, respectively, the first flux density measurements has a logarithmic central time of 1 hr post-burst (or 1.3 hr linear central time, see stars plotted in Figure 1), which is the earliest radio detection of any GRB to date.

We first investigate whether the flux density measurements of GRB 230217A could be affected by interstellar scintillation (ISS; Goodman, 1997) as has been seen in the radio afterglows of several long GRBs (Frail et al., 1997, 2000; Waxman et al., 1998; Chandra et al., 2008; van der Horst et al., 2014; Alexander et al., 2019; Rhodes et al., 2022; Anderson et al., 2023b). The all-sky model for refractive interstellar scintillation (RISS19; Hancock et al., 2019) returns a transition frequency of $\nu_{0}=8.9$ GHz, with a scattering screen distance of $D=1.3$ kpc, a scattering measure $SM=9\pm 1\times 10^{-4}$ kpc m^-20/3, and first Frenzel zone size of 2.38 $\mu$as at $\nu_{0}$. This places our $5.5-6$ GHz observations in the strong refractive regime and our 9 GHz observations near the transition frequency. Following the equations in table 1 of Granot & van der Horst (2014), the predicted modulation index is up to $m=0.8$ and $m=1$ at 5.5 and 9 GHz, respectively, for GRB 230217A. The radio afterglow may also experience intrinsic variability due to the passage of characteristic frequencies in the afterglow emitting region(s) (Granot & Sari, 2002).

To test for evidence of short timescale variability within the first 1.2 days post-burst, we performed a $\chi^{2}$ test on the detections of GRB 230217A, returning a probability of $P=0.15$ that the radio afterglow is consistent with a steady source over this time frame (we require $P<0.001$ to claim variability, with $0.01<P<0.001$ being undetermined; Gaensler & Hunstead, 2000; Bell et al., 2015). A lack of detectable variability is not surprising given that splitting the first ATCA observation reduces the sensitivity and necessarily increases the size of the error bars. Similarly, we also perform a $\chi^{2}$ test to search for intra-observational variability within the first VLA observation when split into four $\sim 9$ min images, returning a probability of $P=0.5$ that the flux densities are consistent with a steady source within the 36 minute observation. While scintillation may affect the GRB radio afterglow, it is not significant enough to be distinguishable in our dataset. Similarly, it is also not possible to ascertain if a characteristic frequency is evolving through the observing band within this $\sim 1$ day time frame.

The same $\chi^{2}$ test on the full $5.5-6$ GHz GRB light curve, which includes the detections up to 1.2 days post-burst and the force-fitted measurements for the non-detections at later times returns a probability of $P=0.0016$ of being consistent with a steady source and thus a more likely indication of variability. This is also compared to $P=0.12$ derived from the $5.5-6$ GHz measurements of the check source, which is therefore steady over the 22 days of monitoring.

The constraining $3\sigma$ upper limits from the later observations (see Figure 1) demonstrate the overall radio light curve trend of GRB 230217A is declining. We fit a single power law ($S\propto t^{\alpha}$) to the $5.5-6$ GHz and 9 GHz light curves using Sherpa (Freeman et al., 2001). For both frequencies, we included the detections and force-fitted flux densities (the latter shown as open circles in Figure 1). At both frequencies, we performed two power-law fits to the light curve: one that included a single flux density measurement over the full integration of the first ATCA observation (‘Average’) and the other using the flux densities measured from splitting this first observation into 3 and 2 images at 5.5 and 9 GHz, respectively (‘Split’). The resulting power law index and corresponding reduced $\chi^{2}$ statistic for the two $5.5-6$ GHz fits are listed in Table 2. The fits to the 9 GHz light curves failed as they resulted in an unphysical reduced $\chi^{2}$ statistics near zero. We plot the power-law fit to the $5.5-6$ GHz light curve that includes the average flux density from the first ATCA observation on Figure 1 where $\alpha=-0.3\pm 0.1$, which we favour given its reduced $\chi^{2}\sim 1$ and it is consistent with the $3\sigma$ upper limit from the final VLA observation. Note that the faint GRB afterglow detections meant we could not derive meaningful information on the spectral index between 5.5 and 9 GHz.

Given the limited multi-wavelength data available for GRB 230217A (see Section 2), we restrict our investigation to just the radio afterglow. This includes an interpretation of the emission mechanism and a comparison to other early-time radio detections of SGRBs.

This analysis of GRB 230217A highlights the impact of the ATCA rapid-response program for GRB studies as it probes a previously unexplored radio parameter space of $<0.1$ days and tracks the early afterglow evolution. Using the ATCA rapid-response mode, we observed the short GRB 230217A just 32 minutes post-burst, obtaining the earliest radio detection of a GRB to date. Combined ATCA and VLA observations showed a fading radio light curve consistent with reverse shock emission. This makes GRB 230217A the fifth SGRB with radio detections attributed to a reverse shock at early times ($<1$ day post-burst). Using brightness temperature arguments, we have also placed the tightest constraint on the minimum Lorentz factor in the radio band of $\Gamma_{\rm{min}}>50$, assuming an average radio-detected SGRB redshift of 0.5.

While rapid follow-up with long integrations is key for detecting SGRB radio reverse shocks, we have not tracked their frequency-dependent evolution, which is necessary for probing the thickness of the ejecta material, the density profile of the CBM, and the energetics of the outburst, as has been possible for a small number of long GRBs (e.g. Bright J. S. & Rhodes L. et al., 2023). This ATCA program already demonstrates the advantage of sensitivity and several-hour integrations over the previous GRB rapid-response program (Staley et al., 2013; Anderson et al., 2018) performed with the Arcminute Microkelvin Array Large Array (Zwart et al., 2008), which pioneered this functionality for GRB science. These results encourage installing similar rapid-response programs on other sensitive radio telescopes. For example, a single 36 min VLA observation can reach an RMS of $\sim 5\mu$Jy/beam (see Table 1), which was successfully split into four 9 min images (see Section 3) with an RMS of $\sim 9\mu$Jy/beam (see $3\sigma$ limit drawn as dotted lines in Figure 2, which is expected to be similar for the C(4.0-8.0 GHz) and X(8.0-12.0 GHz) bands in the B array configuration³³3https://obs.vla.nrao.edu/ect/). Square Kilometre Array Mid is expected to reach an RMS of $\sim 2\mu$Jy/beam and $\sim 6\mu$Jy/beam at the position of GRB 230217A in a 9 min and 1 min observation, respectively, assuming an elevation of 45 deg and a Briggs weighting scheme with Robust set to 0 in both the 4.6-8.5 GHz and 8.3-15.4 GHz bands (see $3\sigma$ limits drawn as dashed and dot-dashed lines in Figure 2).⁴⁴4https://sensitivity-calculator.skao.int/mid A rapid-response trigger within 12 hr post-burst that lasts for $\gtrsim 3$ hrs with either instrument would allow us to monitor the rapidly varying early-time radio emission on minute timescales with competitive sensitivity for deep and high cadence tracking of short and long GRB reverse shocks.