The optical light curve of GRB 221009A: the afterglow and the emerging supernova

Introduction

Long-duration gamma-ray bursts (lGRBs) are typically associated with the signature of a broad-lined type Ic supernova (SN) in their light curves and spectra, as the afterglow fades and an SN rises within 10-20 days. Since the discovery of SN 1998bw/GRB 980425 and SN 2003dh/GRB 030329 [33] (Hjorth et al. 2003) more than 40 probable GRB supernovae have been observed [e.g. 12] (Hjorth & Bloom 2012). However, a supernova signature is not always detected for nearby lGRBs [22, 32], leading to speculation that massive stellar deaths are not the source of all lGRBs (Leśniewska et al. 2022; Rastinejad et al. 2022).

GRB 221009A was first detected and announced by the Neil Gehrels Swift Observatory via the Gamma-ray Coordinates Network (GCN) circulars [23]. A hard X-ray source and an optical counterpart at Swift-UVOT white light magnitude 16.63 were reported, locating it within the Galactic plane at a latitude of $b = 4.322^\circ$. The lGRB nature was confirmed by Kennea et al. (2022), while the Gamma-ray Burst Monitor (GBM), on-board the Fermi spacecraft, reported a detection 1hr before the Swift trigger time (Veres et al. 2022), noting it was the brightest GRB ever detected by Fermi-GBM instrument. Fermi’s Large Area Telescope (LAT) further reported the detection of the GRB and recorded its highest-energy photon at 7.8 GeV [8]. The first ground-based detections of the afterglow in the optical by Perley (2022) and in the radio by Bright et al. [10] were followed by a multitude of GCN circulars reporting measurements across the electromagnetic spectrum. Interest in this extraordinary event further increased when the HLAASO experiment reported the detection of more than 5,000 very high energy (VHE) photons with energies up to 18 TeV (Huang et al. 2022). The Carpet-2 experiment reported a possible 251 TeV photon detection [25] which would be remarkable if proven reliable. Spectra

of the afterglow were obtained by de Ugarte Postigo et al. [21] and Castro-Tirado et al. [15], both of which reported a red continuum with absorption features that correspond to Ca II, Ca I, Na I D at a redshift of $z = 0.151$. A host was later confirmed, and a similar redshift of $z = 0.151$ was determined by Izzo et al. (2022) through the identification of host galaxy absorption and emission lines. At this extragalactic distance, such VHE photons should be absorbed through pair-production when they scatter off the extragalactic background light, raising the possibility of axion-like particle production [14, 6]. The remarkably high fluence, luminosity and detection of VHE photons make GRB 221009A an object of broad interest.

In this Letter, we present the detection of a supernova signature in the fading afterglow of GRB 221009A, the brightest GRB known to date. We present an extensive photometric data set primarily from the Pan-STARRS2 Observatory, supported by multiple other facilities. The search for a supernova signal is complicated due to the high, and uncertain, foreground extinction, the bright afterglow and the uncertainty in host galaxy contribution. Throughout this Letter we assume a GRB detection time of $T_0 = 59861.55347$ (09 October 2022 at 13:16:59.99 UT) from Fermi (Veres et al. 2022), a Hubble Constant of $H_0 = 70\,\mathrm{km\,s^{-1}\,Mpc^{-1}}$ and the redshift of $z = 0.151$ [21, 15] (Izzo et al. 2022). This corresponds to a luminosity distance of $D_L = 718\,\mathrm{Mpc}$ and distance modulus $\mu= 39.28$ assuming a flat Universe with $\Omega_{\mathrm{M}} = 0.3$.

Observational data

The Pan-STARRS (PS) system comprises two 1.8 m telescope units located at the summit of Haleakala on the Hawaiian island of Maui [16]. The Pan-STARRS1 (PS1) telescope is fitted with a 1.4 Gigapixel camera (GPC1) with 0\farcs26 arcsec pixels providing a 3.0 degree diameter focal plane, corresponding to a field-of-view area of 7.06 sq deg. The Pan-STARRS2 (PS2) telescope is fitted with a similar but larger 1.5 Gigapixel camera (GPC2) resulting in a slightly wider field-of-view. Both telescopes are equipped with an SDSS-like filter system, denoted as $grizy_{P1}$. The Pan-STARRS1 Science Consortium (PS1SC) 3$\pi$ Survey produced $grizy_{P1}$ images of the whole sky north of $\delta= -30^{\circ}$ [16]. Multi-epoch observations spanning 2009-2014 have been stacked and a public data release provides access to the images and catalogues [30]. These data provide reference images for immediate sky subtraction, which allows the discovery of transients and accurate photometry with host galaxy removal. Pan-STARRS data from both telescopes are processed in real-time as described in Magnier et al. (2020a), Magnier et al. (2020b) and Waters et al. (2020) at the University of Hawaii and the transient sources are selected, filtered and classified by the Transient Science Server at Queen’s University Belfast (Smith et al. 2020). In normal survey and discovery mode, these feed science programs such as the Young Supernova Experiment (Jones et al. 2021) and the Pan-STARRS Search for kilonovae (McBrien et al. 2021).

The Pan-STARRS afterglow observations of GRB22100A presented here were all taken with PS2. After the discovery of the GRB and the optical afterglow, we triggered PS2 to observe in $r_{\mathrm{PS}}$, $i_{\mathrm{PS}}$, $z_{\mathrm{PS}}$ and $y_{\mathrm{PS}}$, with a nightly sequence between 11th October and 4th December 2022 (Huber et al. 2022), depending on Moon phase and weather. Photometry was carried out on the difference images with the Pan-STARRS1 3$\pi$ sky survey data [16] used as references. The 3$\pi$ data are made of stacks of short exposures and the total exposure times and depths at the position of the afterglow (which we measure at RA = $288.26459^{\circ}$ DEC = $+19.77341^{\circ}$) are listed in Table 1. The photometric zeropoints on the PS2 target images were set with the Pan-STARRS1 3$\pi$ catalogue [30]. We typically used 100-200 sec exposures and stacked images on any one night (from 1 to 12 images, depending on target magnitude and sky brightness).

Filter	Exposure time (sec)	$3\sigma$ depth AB Mag	$3\sigma$ depth Flux ($\mu$Jy)	$A_{\lambda}^{MW}$ mag	$A_{\lambda}^{Tot}$ mag	$\lambda_{\mathrm{obs}}$ (nm)	$\lambda_{\mathrm{rest}}$ (nm)
$r_{\mathrm{PS}}$	636	$>22.24$	$<4.61$	3.497	4.64	617	536
$i_{\mathrm{PS}}$	930-1020	$>22.01$	$<5.70$	2.590	3.44	752	653
$z_{\mathrm{PS}}$	540-570-600	$>21.14$	$<12.71$	2.036	2.70	866	752
$y_{\mathrm{PS}}$	620	$>20.34$	$<26.54$	1.674	2.22	962	836

Two methods were used to measure the flux on the difference images. A difference image was created from each individual exposure, and a point-spread-function (PSF) was forced at the GRB 221009A afterglow position (measured from the early, bright epochs). We statistically combined the measured PSF flux from each difference image through a weighted average, using a small temporal bin size of 0.125 days (3 hours). For the later Pan-STARRS epochs ($t - T_{0} > 34$ days) we increased the bin size to four days in the $zy$-filters to enhance the signal-to-noise.

Alternatively, an image stack on each night was created, and a difference image produced from the stack. Again, a PSF was forced at the GRB afterglow position and flux measurement used. All fluxes and magnitudes quoted here are in microJanskys ($\mu$Jy) and AB mags. The results from image stacking were used instead of the weighted average of fluxes only when the object fell on a masked chip within the camera CCD which prevented the typical pipeline processing of target images described in Chambers et al. [16]. Regardless of the method, the resulting flux measurements were calibrated carefully to the Pan-STARRS1 DR2 3$\pi$ reference catalogue [30] using approximately 1000 field stars visible within the target frames.

While most of the data here are provided by PS2, we gathered other important photometric data with the 0.4-meter SLT R-C Telescope [18] and 1-meter LOT Cassegrain Telescope at the Lulin Observatory, Taiwan; the Dark Energy Camera (DECam) on the 4-meter Telescope at the Cerro Tololo Inter-American Observatory, Chile; the 4.1-meter Southern Astrophysical Research (SOAR) Telescope, Chile; the 1-meter Swope Telescope, Chile; the IO:O on the 2.2-meter Liverpool Telescope (LT), La Palma; and MegaCam on the 3.6-meter Canada-France-Hawaii Telescope, Hawaii.

Eight epochs of DECam observations were conducted between 16th October 2022 (MJD 59868.01) and 31st October 2022 (MJD 59883.011) taking between 2 − 5 × 100 sec exposures in the filters $r$ and $i$. The data were reduced and photometrically calibrated with the photpipe package (Rest et al. 2014) using the images from 16th October 2022 as templates. These were subtracted from all subsequent images and a PSF was forced at the afterglow position on the difference images. Since the template contains transient flux and we were not able to get a final set of templates in which the afterglow had faded, we applied an offset to match the DECam $r$-filter flux measured on MJD 59880.01 to the SOAR epoch on MJD 59880.02, and the DECam $i$-filter flux measured on MJD 59875.01 to the PS2 epoch on MJD 59875.23. These offsets were subsequently applied to all the DECam difference images in the respective filters. Data from Swope were subjected to difference imaging, using the Pan-STARRS1 3$\pi$ references as templates and forced photometry was implemented thereafter.

We used the three epochs of Hubble Space Telescope (HST) data that is publicly available through the DDT program of Levan et al. (2022). The WFC3 passbands of the F625W, F775W and F098M filters are similar to that of the $r_{\mathrm{PS}}$, $i_{\mathrm{PS}}$ and $y_{\mathrm{PS}}$ filters respectively (see § 4). PSF magnitudes were measured on the HST target images using the DOLPHOT package [24]. Observations conducted by SLT, LOT, SOAR, LT and MegaCam between 10th October and 8th November were not subjected to any form of image subtraction, instead, a PSF was forced onto the target images using python packages: Astropy and Photutils¹ and the resulting flux measurements were calibrated against Pan-STARRS1 3$\pi$ survey field stars.

Since no difference imaging was applied to the SLT, LOT, SOAR or LT one may be concerned by late-time host-galaxy flux contamination. This is particularly concerning when the measured flux of the transient is comparable to, or fainter than, the limits we can put on the host galaxy from the Pan-STARRS1 3$\pi$ data (see Table 1). However the photometry between different instruments (with and without difference imaging) is consistent within the statistical uncertainties. A probable faint host galaxy is visible in the deepest F625W and F775W HST images, approximately $0.005$ offset to the North-East. However, this is too faint to contribute significantly to the $r$ and $i$ photometry and our $y_{\mathrm{PS}}$ data are all image subtracted. Hence we make the assumption that there is no host galaxy flux contributing to the non-differenced images in the filters $r$, $i$ or $z$. This can only be confirmed with deep observations in the next observing season. We list our measurements also in fluxes (microjanskys) within Appendix § 8 so that a future correction can be applied should that be necessary.

Analysis of the x-ray and optical afterglow

The X-ray counterpart to GRB 221009A was observed by the Neil Gehrels Swift Observatory X-ray telescope (XRT) starting 0.9 hours after the Fermi trigger (Veres et al. (2022)), and is still observing at the time of writing. We downloaded the Swift XRT data to date from the Swift Burst Analyser (Evans et al. [26, 28] (2009)). A single decaying power-law can best describe the XRT light curve.

We fit a power-law component to the first 60 days of flux data, use the fluxes and not flux densities so as not to introduce any spectral bias, and found the light curve is described with $f(t) \propto t^{-1.556\pm0.002}$. There is no evidence of any breaks in the light curve which could result from either a spectral break (e.g. the cooling break) passing through the band or a jet break. There is also no evidence of a change in the X-ray photon index (related to the spectral index), indicating no significant spectral evolution occurring over the first 60 days post-burst. The Swift Burst Analyser quotes an X-ray photon index of $1.78\pm0.01$ which corresponds to a spectral index ($S(\nu) \propto\nu^{\alpha}$) of $\alpha=-0.78\pm0.01$. The measured X-ray light curve decay and spectral index indicate the X-ray emission originates from optically thin synchrotron radiation, where the synchrotron cooling break has a frequency that is higher than that of the observing band.

Given that the optical light curve is also showing a decay, and that only optically thin synchrotron radiation produces a decaying light curve within the fireball model, the optical emission should be on the same branch of the afterglow spectrum as the X-ray band and thus the decay rates should be identical ([34] (Sari et al. 1998)). A power-law decay also best describes the measured optical decay. However, the difference between the optical and X-ray decay rates is not large enough to be consistent with the presence of a spectral break between the two bands. The difference in light curve decay rates should be $1/4$ if the cooling break is between the optical and X-ray bands and accompanied by a spectral index difference of $1/2$ ([34]).

In addition to differing light curve decay rates, the optical flux densities do not agree with what is predicted from the X-ray spectrum (Figure 1). This is the result of the line of sight extinction through the Galactic plane being high at $A_V = 4.223$ ((Schlafly & Finkbeiner 2011)). We confirm that the extinction is high by measuring an optical spectral index of ($\alpha\approx-4$) which is notably steeper than the expected and measured X-ray spectral index of the synchrotron emission,

$\alpha\approx-0.8$. We correct the optical data for the Galactic extinction across the specific rizyPS-filters as reported in the NASA Extragalactic Database (NASA/IPAC Extragalactic Database (NED) 2019), which uses the maps of Schlafly & Finkbeiner (2011) and the Fitzpatrick [29] reddening law (see Table 1). Figure 1 shows the X-ray to optical SED. Once corrected for Galactic extinction (only), the optical flux densities are still too faint to agree with the extrapolation of the X-ray data using a spectral index of $S(\nu) \propto\nu^{-0.78\pm0.01}$. Additional extinction is required and the amount required to reconcile the optical and X-ray fluxes varies across the three epochs shown in Figure 1. The optical spectral index (Figure 1) is also lower than the expected $\alpha\simeq-0.8$, further supporting a requirement for a source of extra extinction. The earliest epoch (1-2 days post-burst) requires 1.5 magnitudes of extra extinction, reducing to 1.0 (at 2-3 days) and 0.8 magnitudes (4-5 days). It’s not clear why the implied extra extinction would vary across the three epochs but time-variable extinction has been proposed before (e.g. GRB 190114C; Campana et al. [11]; Melandri et al. 2022). We use the value of 0.8 magnitudes (which is the average value across the $rizy_{\rm PS}$ bands) as this is the closest value in time to the possible emergent SN signal. There are three possibilities for this additional line of sight extinction. The Galactic dust structure may not be accurately captured by the low resolution Milky Way extinction maps of Schlafly & Finkbeiner (2011) or it is possible that $R_V > 3.1$, which may be plausible in this high density region. The third

possibility is additional absorption in the host. This may not be surprising given four of the five GRBs with very high energy detections ($E_{\gamma} > 100$ GeV) also have evidence of strong dust contamination as mentioned in Rhodes et al. (2022). The VHE GRBs appear to have an increased likelihood of dust absorption compared to the rest of the long GRB population where only about 25% of events have significant optical extinction (Chandra & Frail [17]).

We plot our optical data in AB magnitudes in Figure 2, including all points with $\geq2$-sigma significance. We first measured the optical decay rate in all filters independently and across all epochs ($T_0+1 < t \leq T_0+56$). The decay rate in the $ri$-filters follow $f(t) \propto t^{-1.43\pm0.02}$, the $z$-filter follows $f(t) \propto t^{-1.54\pm0.02}$ and the $y$-filter follows $f(t) \propto t^{-1.21\pm0.01}$. There isn’t a single decaying term that can describe the fade in all filters, and, with the exception of the $z$-filter, the decay terms derived are substantially shallower than that of the X-ray slope. Using epochs from ($1 \leq t-T_0 \leq4.7$ d) in the $izy_{\mathrm{PS}}$ filters, we also fit the X-ray derived slope of $f(t) \propto t^{-1.556\pm0.002}$ to these points. In the $r_{\mathrm{PS}}$-filter, we lack data at 2-3 days so we extend the temporal baseline to include the 6.7 d PS2 point. We find the early data are well explained by the model, as evidenced by the measured normalized chi-squared value of $\chi^2_{\mathrm{n}}/dof = 0.4/6.0$.

The dissimilar decaying rates between the optical filters, and the deviation of the optical light curve from the X-ray light curve starting at $t-T_0 > 7$ days are suggestive of another component appearing in the data. The light curve deviation is most pronounced in the $y_{\mathrm{PS}}$ filter at $t-T_0 = 21$ days, and we discuss this in the next section. Figure 3 illustrates the data quality and detections around the peak of the flux excess in the PS2 images, for which we have deep reference images from the 3$\pi$ survey (Chambers et al. [16]).

Interpretation of the excess flux as a supernova signature

To determine if the excess flux observed starting from $\sim7-8$ days is consistent with a Type Ic Broad-Lined supernova (Ic-BL) contribution, we compare the excess to the $riz$ light curve of SN 2016jca from Cano et al. [13] and the $griz$ light curve of SN 2017iuk from Izzo et al. (2019). SN 2016jca was a type Ic-BL that emerged in GRB 161219B (Cano et al. [13]), at a redshift of $z = 0.1475$. This redshift is almost identical to GRB 221009A, and as such, the $riz$ light curve can be compared directly. SN 2017iuk was another type Ic-BL that emerged in GRB 171205A. It is the only type Ic-BL with published rest frame $z$-band filter coverage, to compare to the observer frame $y_{\mathrm{PS}}$ data of GRB 221009A. At this redshift, the GROND $griz$ SDSS filters (Fukugita et al. [31]) correspond to the observer frame of the Pan-STARRS system $rizy_{\mathrm{PS}}$-filters at $z = 0.151$ to a good approximation (see Table 1). In particular the restframe $z$-band transforms to the observed $y_{\mathrm{PS}}$ filter and since the $y_{\mathrm{PS}}$ data apparently show the strongest excess, the $z$-band restframe data are essential.

We used the observed data of GRB 171205A and SN 2017iuk listed in Izzo et al. (2019) which is already corrected for host galaxy contribution, and we further corrected for dust extinction to the SN. We do not subtract off any X-ray-derived power-law as the afterglow contribution represented by which is considerably less than the SN contribution as early as three days after explosion. As the light curve of 2017iuk is only $\sim27$ days long, we generated a continuous model light curve fit up to 100 days after explosion using an MCMC Bazin fit (Bazin et al. [7]) for each of the $griz$ data.

To obtain the light curve of SN 2016jca we used the observed data listed in (Cano et al. [13]) and subtracted off an X-ray-derived power-law of $f(t) \propto t^{-0.79}$ fitted to the early-time data ($t-T_0 < 1$ day) to remove the GRB 161219B afterglow contribution. We also corrected for the host galaxy contribution (using the same host flux determined by Cano et al. [13]) and dust extinction to the SN. The light curve of SN 2016jca is well sampled up to 70 days after explosion, and so we did not require a Bazin fit here.

These light curves were then adjusted accordingly for extinction along the line-of-sight to GRB 221009A and time dilation to the GRB redshift of $z = 0.151$. We used the total estimated line of sight extinction implied from the X-ray to optical power-law slope as derived in § 3 (which is higher than that from Milky Way only), explicitly $A_r = 4.64$, $A_i = 3.44$, $A_z = 2.70$, $A_y = 2.22$. These light curve fluxes of SN 2016jca and SN 2017iuk were subsequently added to the X-ray power-law slope derived in Section 3 and overlaid onto the GRB afterglow in Figure 4.

The light curve fluxes of SN 2016jca required moderate (arbitrary) scaling to produce a power-law plus SN component that quite satisfactorily matches the observed data of GRB 221009A. Scaling factors of 2.0 and 1.6 were used for the $r$-filter and $i$-filter respectively. We computed Bayes Factors to compare the continuation of the imposed X-ray power-law to the power-law plus an SN component. We found factors of 3.4 and 7.5 for the $r$-filter and $i$-filter respectively. This favours a model with a power-law plus SN rather than the imposed X-ray power-law only. The light curve shape and peak magnitudes are similar (within the errors) to that of SN 2016jca (Cano et al. [13]; Ashall et al. 2019). We estimate rest-frame, absolute magnitudes of $M_g = -19.7 \pm0.6$ and $M_r = -19.6 \pm0.3$ for a Ic-BL SN component of similar nature to SN 2016jca.

The SN 2017iuk light curve model fluxes in the griz-filters required considerable scaling (arbitrarily) to match the corresponding observed fluxes of GRB 221009A in the rizy$_{\mathrm{PS}}$ bands. We require scaling factors in the restframe grz-filters of 4.8, 1.9, and 4.6, respectively, to match the excess flux observed in the GRB data. This would produce peak magnitudes of $M_g = -19.9 \pm0.6$ and $M_r = -19.3 \pm0.3$ and $M_z = -20.1 \pm0.3$ for the SN component inside GRB 221009A. No consistent single scaling factor can produce the colors and peak magnitudes in all filters. However, the variation in color of type Ic supernovae and the uncertain extinction toward GRB 221009A (both Milky Way and the additional required extinction) may explain why differences are found in each filter. We note that a difference in scaling factors was considered in the Cano et al. [13] interpretation of SN 2016jca.

We computed Bayes Factors (in the same way as for the SN 2016jca comparison) and found 3.4, 7.4 and 3.9 for the grz-filters respectively. We emphasise that this is comparing the imposed x-ray power-law (no SN) to power-law plus SN component after 7 days. Again, this method favours a Ic-BL component, but only with these assumptions.

We do not measure a peak absolute magnitude in the observer frame z-filter for either SN comparison, as the Bayes Factors from such comparisons strongly favoured a “no SN” solution regardless of the scaling factors used. We also considered a scenario without the extra dust extinction suggested by the X-ray analysis. In this, we measure peak absolute magnitudes of $M_g = -18.7 \pm0.6$, $M_r = -18.6 \pm0.3$ and $M_z = -19.6 \pm0.3$ from the SN 2017iuk comparison and $M_g = -18.6 \pm0.6$, $M_r = -18.7 \pm0.3$ from the SN 2016jca comparison.

We subtracted the afterglow (X-ray power-law) from the observed data and measured the peak of the flux excess at two phases: peak brightness ($t - T_0 \sim21$ days) and at late times ($t-T_0 \sim34$ days), to check for a color change which may be indicative of an SN component. Using only measurements made through difference imaging to mitigate host contribution, and discounting dust extinction, we measure observed $i-y$ colors ($r-z$ in the rest frame) of $i-y = 2.77 \pm0.56$ and $i-y = 4.24 \pm1.95$ at peak brightness and late times, respectively. The relatively large errors imply no meaningful color evolution can be interpreted.

To investigate why no apparent excess flux is visible in the observer frame $z$-filter (rest frame $i$-filter), we took the spectrum of SN1998bw and SN 2017iuk at peak and reddened both with the total extinction that we estimated in § 2 ($A_V = 5.3$) and redshifted it to $z = 0.151$. These are plotted in Figure 5 along with the convolution of the spectra with the effective transmission curves of the filters used². At $z = 0.151$, the strong Ca II triplet has the deep P-Cygni absorption precisely at the position of the $z_{\rm PS}$ filter. The $i_{\rm PS}$ filter and the $y_{\rm PS}$ filter cover the emission peaks of the Ca II triplet and the 7500 Å blend. This may explain why there is little-to-no sign of a Ic-BL supernova signature in the $z$-filter but is plausibly detected in the other filters.

We conclude that the excess flux above the extrapolation of the afterglow power-law of $f(t) \propto t^{-1.556\pm0.002}$ can be explained by the emergence of a supernova similar in luminosity and duration to the observed type Ic-BL SNe SN 2016jca and SN 2017iuk. This supernova associated with GRB 221009A would be labelled SN 2022xiw as reported on the IAU Transient Name server [21]. At the later epochs, our PS2 images have a depth and sensitivity close to those of the reference images we use for image subtraction. We demonstrate the reality of the detections compared with the PS1 $3\pi$ stacks at several important epochs in Figure 3.

We compare our $r$ and $i$-filter photometry with original data (not GCN values) in Laskar et al. (2023), Shrestha et al. (2023) and Levan et al. (2023), three papers which suggest there is no strong evidence for a supernova component in the data. Their data are in excellent agreement with ours, even though no image subtraction was undertaken. The $z$-filter data of Laskar et al. (2023) and Shrestha et al. (2023) are slightly brighter than ours after 8-10 days, which may suggest some host galaxy contribution affects their measurements. The converted $y$-filter from the HST data of Levan et al. (2023) are fainter than ours by $\sim0.3$ magnitudes at the respective epochs, which we attribute to the combined galaxy plus PSF fits of Levan et al. (2023). Neither group have extensive $y$-filter or photometric NIR data at $> 10$ days. While Kann et al. (2023) also claim no evidence for (or against) a supernova component, their measurements include only 2 epochs of their own data beyond 10 days and no meaningful comparison is possible with our data set.

It appears that an SN component is only detectable with extensive and accurate photometric coverage beyond 10 days and assuming the X-ray power-law is applied to the optical data (which may be disputed e.g. Laskar et al. (2023)). If we were to assume there is no SN component, and that there is a break in the afterglow SED between the X-ray and optical decline rates, then, by applying similar extinction corrections, we find almost identical agreement with Laskar et al. (2023), Shrestha et al. (2023) and Levan et al. (2023) in the power-law decline rates for the $riz$-filters across all epochs (see top panel of Figure 2), even though our data is better sampled and extends further in time. However, it is not possible to fit our $y$-filter with the same single decaying power-law derived from the $riz$-filters, which would imply the achromatic behaviour in the afterglow. Although, we acknowledge that if we were to replace our HST $y$-filter measurements with that reported in Levan et al. (2023), then we find it possible to describe the $y$-filter decay with the same power-law derived from the $riz$-filters. We measure this single decaying power-law as $f(t) \propto t^{-1.46\pm0.05}$.

Discussion and Conclusions

Given that the observational data of GRB 221009A and the emergence of a supernova (SN 2022xiw) could be explained by the addition of scaled observed fluxes of two type Ic-BL SNe, we can determine the physical parameters of SN 2022xiw. We model the afterglow subtracted SN signal to constrain the ejecta mass ($M_{\rm ej}$), nickel mass ($M_{\rm Ni}$) and ejecta expansion velocity ($v_{\rm ej}$) using MOSFiT, an open-source, one-zone, semi-analytical fitting tool for broadband SN light curves (Guillochon et al. 2018). MOSFiT employs dynesty, a nested sampling technique (Speagle 2020), to estimate posteriors of the fitted model. We use the default Arnett model within MOSFiT, which assumes that the SN is entirely powered by the radioactive decay of ${}^{56}$Ni and ${}^{56}$Co and assumes that the spectral energy distribution can be described by a black body [1] (Villar et al. 2017). Assuming a constant grey opacity $\kappa= 0.07\,\mathrm{cm}^2\mathrm{g}^{-1}$ [19, 13] and flat priors on all fitted parameters, we find the posterior distributions returned are $M_{\rm ej} = 5.6^{+1.0}_{-0.9}M_\odot$, $M_{\mathrm{Ni}} = 0.5^{+0.20}_{-0.1}\,M_{\odot}$ and $v_{\mathrm{ej}} = 35{,}500^{+4{,}300}_{-4{,}600}\ \mathrm{km\,s^{-1}}$ to model the $riz$-band data of SN 2016jca (see Appendix § 9 Figure 6). Errors quoted represent a $1\sigma$ width of the posterior distributions. The inferred explosion energy is then $E_{\mathrm{kin}} \simeq2.7$–$6.3 \times10^{52}$ ergs. These are broadly similar to those derived in the previous analysis of Cano et al. [13] and Ashall et al. [2]. Our $M_{\mathrm{Ni}}$ is higher, which we attribute to the different methods of fitting a bolometric light curve rather than the black body fitting of the selected filters.

Using the same set of priors, we model the $riy$-filter light curve of SN 2022xiw. The posterior distributions return the values $M_{\mathrm{ej}} = 7.1^{+2.4}_{-1.7}\,M_{\odot}$, $M_{\mathrm{Ni}} = 1.0^{+0.6}_{-0.4}\,M_{\odot}$, and $v_{\mathrm{ej}} = 33{,}900^{+5{,}900}_{-5{,}700}\ \mathrm{km\,s^{-1}}$, (see Figure Appendix B). inferring explosion energy of $E_{\mathrm{kin}} \simeq2.6$–$9.0 \times10^{52}$ ergs. These are comparable to the quantities derived for SN 2016jca above but suggest a more energetic explosion and a larger mass of $^{56}\mathrm{Ni}$ is required. This is reflected in the fact that the light curve data of SN 2016jca requires modest scaling to reproduce the flux of SN 2022xiw.

We test the robustness of our modelling by excluding the $y_{\mathrm{PS}}$-band photometry (which may significantly deviate from our black body assumptions) and by excluding late-time observations (which may deviate from our assumption that the SN is in the photospheric phase). We find no statistically significant differences in derived properties from these tests.

The parameters of SN 2022xiw are similar to the more energetic Ic-BL SNe associated with lGRBs, termed hypernovae ((Iwamoto et al. 1998; Mazzali et al. 2003)). The sample of GRB-SNe has been reviewed and summarised by Hjorth & Bloom (2012) and more recently Wheeler et al. (2015) and Cano et al. [12]. The latter studies suggest that GRB-SNe are characterised by the following average values: kinetic energies of $E_{\mathrm{K}} = 2.5 \times10^{52}$ erg ($\sigma_{E_{\mathrm{K}}} = 1.8 \times10^{52}$), ejecta masses of $M_{\mathrm{ej}} = 6M_{\odot}$ ($\sigma_{M_{\mathrm{ej}}} = 4\,M_{\odot}$). and peak photospheric velocities of $v_{\mathrm{ph}} = 20{,}000M_{\odot}$ ($\sigma_{v_{\mathrm{ph}}} = 8000\,M_{\odot}$).

The most luminous GRB-SNe require $^{56}\mathrm{Ni}$ masses of $0.5 \lesssim M_{\mathrm{Ni}} \lesssim0.9$, if the luminosity is powered by radioactivity.

The $^{56}\mathrm{Ni}$ mass we derive ($M_{\mathrm{Ni}} = 1.0^{+0.6}_{-0.4}\,M_{\odot}$), is on the high side of the known distribution of GRB-SNe. Similar to the SN 2016jca case above, we attribute this partly to fitting a computed bolometric light curve, which is made worse by the bright afterglow of GRB 221009A and the high, uncertain extinction creating considerable noise in the SN extracted flux.

Nevertheless, the large uncertainty comfortably brackets GRB-SNe parameters previously derived. Despite these uncertainties, the excess flux above the extrapolated afterglow leads us to conclude that there may be a supernova signature in our $\sim60$ day, well-sampled light curve data of GRB 221009A (denoted SN 2022xiw) which is comparable in both luminosity and ejecta properties to other bright type Ic-BL supernovae events. To confirm, this will require a reanalysis of all multi-wavelength data, to model the expected optical afterglow behaviour.

acknowledgments

Pan-STARRS is a project of the Institute for Astronomy of the University of Hawaii, and is supported by the NASA SSO Near Earth Observation Program under grants 80NSSC18K0971, NNX14AM74G, NNX12AR65G, NNX13AQ47G, NNX08AR22G, 80NSSC21K1572 and by the State of Hawaii. The Pan-STARRS1 Sky Survey data were facilitated by the University of Hawaii, the Pan-STARRS Project Office, the Max Planck Society (MPIA, MPE), Johns Hopkins University, Durham University, University of Edinburgh, Queen’s University Belfast, the Harvard-Smithsonian CfA the Las Cumbres Observatory Global Telescope Network Incorporated, the National Central University of Taiwan, the Space Telescope Science Institute, NASA Grant No. NNX08AR22G, NSF Grant No. AST-1238877, the University of Maryland, Eotvos Lorand University, and the Los Alamos National Laboratory. The Young Supernova Experiment (YSE) and its research infrastructure is supported by the European Research Council under the European Union’s Horizon 2020 research and innovation program (ERC Grant Agreement 101002652, PI K. Mandel), the Heising-Simons Foundation (2018-0913, PI R. Foley; 2018-0911, PI R. Margutti), NASA (NNG17PX03C, PI R. Foley), NSF (AST-1720756, AST-1815935, PI R. Foley; AST-1909796, AST-1944985, PI R. Margutti), the David & Lucille Packard Foundation (PI R. Foley), VILLUM FONDEN (project 16599, PI J. Hjorth), and the Center for AstroPhysical Surveys (CAPS) at the National Center for Supercomputing Applications (NCSA) and the University of Illinois Urbana-Champaign. S. J. Smartt, K. W. Smith and D. Young acknowledge STFC grants ST/P000312/1. L. Izzo and J. Hjorth were supported by a VILLUM FONDEN Investigator grant awarded to J. Hjorth (project number 16599). C. D. Kilpatrick was supported in part by a CIERA Postdoctoral Fellowship. M. Nicholl is supported by the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (grant agreement No. 948381) and by a Fellowship from the Alan Turing Institute. S. Yang has been supported by the research project grant ”Understanding the Dynamic Universe” funded by the Knut and Alice Wallenberg Foundation under Dnr KAW 2018.0067, and the G.R.E.A.T research environment, funded by Vetenskapsrådet, the Swedish Research Council, project number 2016-06012. D. A. Coulter acknowledges support from the National Science Foundation Graduate Research Fellowship under Grant DGE1339067. The UCSC team is supported in part by NASA grant 80NSBC20K0953, NSF grant AST–1815935, the Gordon & Betty Moore Foundation, the Heising-Simons Foundation, and by a fellowship from the David and Lucile Packard Foundation to R. J. Foley. This work includes data obtained with the Swope Telescope at Las Campanas Observatory, Chile, as part of the Swope Time Domain Key Project (PI: Piro, Co-Is: Drout, Phillips, Holoien, French, Cowperthwaite, Burns, Madore, Foley, Kilpatrick, Rojas-Bravo, Dimitriadis, Hsiao). We thank Abdo Campillay and Yilin Kong Riveros for performing the Swope observations. Data from NOIR DECam supported by proposals 2021A-0275 and 2021B-0325. We thank Lulin staff H.-Y. Hsiao, C.-S. Lin, W.-J. Hou, H.-C. Lin and J.-K. Guo for observations and data management. T.-W. Chen thanks D. B. Malesani for photometric measurement on the SLT GCN circular and Y.-C. Cheng for LOT time. We acknowledge use of NASA/IPAC Extragalactic Database (NED), operated by JPL, Caltech under contract with NASA.

Facilities: Pan-STARRS, Swift(XRT), CTIO:4m,DECam, CFHT(MEGACAM), SOAR, HST(WFC3), Swope-1m, Liverpool Telescope (IO:O), Lulin:0.4m-SLT, 1m-LOT.

Software: Astropy ([3, 4, 5]), DOLPHOT ([24]), MOSFiT (Guillochon et al. 2018), Photpipe (Rest et al. 2005; Miknaitis et al. 2007; Rest et al. 2014), Photutils ([9]), Scipy (Virtanen et al. 2020), YSE-PZ ([20])

Appendix

Photometry data table

A CSV file is provided, with all the photometry data for GRB 221009A behind Figure 2 and Figure 4. Measurements are in AB magnitudes and microJanskys. All photometry is of the AG+SN and are uncorrected for galactic and host galaxy dust extinction. Magnitude limits are quoted to 2-sigma. The total exposure time for each night is provided which is typically the sum of sub-exposures combined. The final column notes the method used to measure the fluxes for that particular epoch i.e. stacked fluxes from individual difference images (w-avr-flx), stacked fluxes from individual difference images binned across multiple nights (bw-avr-flx), fluxes from a single stacked difference image (d-stack-flx) or fluxes from a single stacked target image without differencing (target-flx).

Posterior distributions for the physical parameters of SN 2016jca and SN 2022xiw