Online material

Appendix A: Observations

Appendix A.1: Photometric observations

We conducted comprehensive photometric observations of the afterglow and host galaxy of GRB 130603B from a range of facilities described below. These data include the original discoveries of the afterglow, track the afterglow behaviour over the first ~ 24 h and place stringent limits at later times. We also include the ultraviolet and optical observation from the UVOT telescope onboard the Swift satellite (de Pasquale & Melandri 2013). Details of each of the observations are given below.

In order to separate the contribution of the host galaxy from the afterglow light, we performed image subtraction as described in the next section, resulting in the photometry provided in Table 1. The analysis performed on the Swift/UVOT data is significantly different from the ground based photometry. Due to this, we display the results in separate tables (Tables A.2 and A.3).

	Fig. A.1 Finding chart showing the location of the afterglow (in blue) and of the stars used for the photometry (in red), as indicated in Table A.1. The field of view is 3.8′ × 2.5′.
Open with DEXTER

Appendix A.1.1: Swift-UVOT observation and analysis

Swift’s Ultraviolet Optical Telescope (UVOT; Roming et al. 2000, 2004, 2005) began settled observations of the burst approximately 62 s after the BAT trigger. A faint source at the location of the afterglow was detected in all 7 UVOT filters.

The photometry was obtained from the image lists using the Swift tool uvotsource. When the source is faint and the count rate low, it is more accurate to use a small source aperture to extract the photometry (Poole et al. 2008; Breeveld et al. 2010). Therefore we used a 3′′ source region to extract the count rate of the source. In order to be consistent with the UVOT calibration, these count rates were then corrected to 5′′ using the curve of growth contained in the calibration files. Background counts were extracted using a circular region of radius 20′′ from a blank area of sky situated near to the source position. The background subtracted count rate was then converted to AB magnitudes using the UVOT photometric zero points (Breeveld et al. 2011). The analysis pipeline used software HEADAS 6.13 and UVOT calibration 20130118. To improve the signal to noise, the count rates in each filter were binned using Δt/t = 0.5. The result of this photometry is given in Table A.2.

To improve the signal to noise of the UVOT data, a single filter light curve was computed from the 7 UVOT filters (see Table A.3). The count rates in each filter were normalised to the U filter and then the resulting light curve was binned using Δt/t = 1.0 (Oates et al. 2009).

Appendix A.1.2: 1.23 m CAHA

The 1.23 m telescope is an f/8 optical system located in the observatory of Calar Alto, Almería, Spain. The observations were carried out in the Johnson V-band and a DLR-MKIII camera. The DLR-MKIII camera is based on a 4k × 4k e2v CCD231-84-NIMO-BI-DD detector. The field of view achieved is 21.5′ × 21.5′. The data were acquired using a 2 × 2 binning read-out, providing a pixel size of 0.63′′ pix^-1.

Appendix A.1.3: NOT

Observations in r-band were obtained with the MOSCA instrument at the Nordic Optical Telescope (NOT) beginning 0.25 days after the burst discovery. Additional observations were obtained on 5 June 2013 and used as a template for subtraction.

Appendix A.1.4: WHT

We observed the location of GRB 130603B with the William Herschel Telescope (WHT) on La Palma on two occasions on 3 June 2013 and 6 June 2013, with the first observations beginning ~0.25 days post burst. At each epoch we obtained observations in g, i and z using the Auxiliary Port Camera (ACAM). The images were debiased and flat-fielded in the standard way within IRAF, with the second epoch subsequently subtracted from the first to provide clean subtractions for photometric observations.

Appendix A.1.5: GTC

Optical imaging was carried out with the Gran Telescopio Canarias (GTC), a 10.4 m telescope located at the observatory of Roque de los Muchachos on La Palma (Canary Islands, Spain), and equipped with the OSIRIS instrument. The observations were obtained with 2 × 2 binning mode, yielding a pixel scale of 0.26′′ pix^-1 and a field of view 7.8′ × 8.5′.

Appendix A.1.6: VLT

We observed the field of GRB 130603B with the Very Large Telescope (VLT) utilising both optical (FORS) and infrared (HAWK-I) observations. Our first photometric observations were obtained as part of the spectroscopic acquisition. Each observation was reduced with the appropriate instrument pipeline via esorex, to create dark/bias subtracted, flat-fielded images which were subsequently combined into final images. Photometric calibration in the optical is given relative to SDSS for g-band observations and to our own V-band calibration for the V-band acquisition.

Appendix A.1.7: Gemini

We obtained two epochs of optical (griz) imaging of GRB 130603B with Gemini-N using the GMOS-N instrument over the first ~2 days post burst. An additional epoch was obtained with Gemini-S (using GMOS-S which has an almost identical instrument setup) on 15 June 2013. Image subtraction was performed between the second epoch of Gemini-N and the Gemini-S observations, confirming no transient emission over this timescale. Given this, and the relative ease (and cleanliness) of the subtractions we use the second epoch of Gemini-N observations as a template for subtraction for the first.

Appendix A.1.8: TNG

We imaged the field of the short GRB 130603B with the Italian 3.6 m Telescopio Nazionale Galileo (TNG), located in La Palma, Canary Islands. Optical observations with the r and i SDSS filters were carried out with the DOLoReS camera on 2013 Jun 16 and 24. All nights were clear, with seeing in the range 1.1′′–1.2′′. However, on June 24 the observations were performed under bright Moon conditions.

Image reduction was carried out following the standard procedures: subtraction of an averaged bias frame, division by a normalised flat frame. Astrometry was performed using the USNO B1.0 catalogue³. Aperture photometry was made with the DAOPHOT task for all objects in the field. The photometric calibration was done against the SDSS catalogue. The flux limits on the presence of an optical transient were estimated using small apertures centred on the optical afterglow position. In order to minimise any systematic effect, we performed differential photometry with respect to a selection of local, isolated an not-saturated reference stars visible in the field of view.

Appendix A.2: Spectroscopy

Spectroscopic observations were obtained using different telescopes and instruments as shown in the observing log (Table 2) and detailed in the following paragraphs. The slits of the spectrographs were positioned at different position angles, providing different coverage of the underlying host galaxy. A diagram with the slit position of the most relevant observations is shown in Fig. 2.

Appendix A.2.1: GTC/OSIRIS

We acquired spectroscopic data using the OSIRIS imager and spectrograph at the GTC. The slit was oriented with a position angle (measured from North to East) of 51°. In order to determine the host contribution of the first observation and to be able to isolate the afterglow emission we performed a second observation on day 5, when the afterglow contribution was negligible. The slit was positioned at the same angle and the seeing conditions were similar.

Appendix A.2.2: VLT/X-shooter

We acquired a medium resolution spectrum with the X-shooter spectrograph mounted at the ESO/VLT, covering the range from 3000 to 24 800 Å. Observations began at 00:00:36 on 04 June, 2013 and comprise of 4 nodded exposures, having an individual integration time of 600 s in the UVB, VIS and NIR arms (the mid exposure time is 8.555 hr post burst; Xu et al. 2013). Due to technical problems with the atmospheric dispersion correctors, we set the position angle to the parallactic angle (being 170.1° at that epoch), see Fig. 2, i.e. the X-shooter spectrum does not cover other parts of the host galaxy. The slit widths were 1.0′′, 0.9′′, and 0.9′′ in the UVB, VIS, and NIR, respectively. For this given instrument setup, the resolution is 5100, 8800, and 5300 in the UVB, VIS and NIR, respectively.

VLT/X-shooter data were reduced with the X-shooter pipeline v2.0 (Goldoni et al. 2006)⁴. The wavelength binning was chosen to be 0.2 Å  pix^-1 in the UVB and VIS, and a 0.5 Å pix^-1 binning in the NIR. All spectra were flux calibrated with the spectrophotometric standard star LTT3218 and scaled with our photometric data to correct for slit losses. Furthermore, data were corrected for Galactic extinction (E(B − V) = 0.02 mag). We transformed the wavelength solution to vacuum. No attempt was made to correct for telluric absorption lines. This has no implications on our analysis.

Appendix A.2.3: VLT/FORS2

We obtained further spectroscopy with the Focal Reducer and Spectrograph (FORS2) on the VLT. Three spectroscopic exposures of each 600 seconds exposure time were obtained, starting at 23:57:43 UT on 3 June 2013 (i.e. midtime of the observation was 8.40 hr after burst), at average airmass of 1.4. We used a 1′′ slit width and the 300V grism without order separation filter (to extend the wavelength range, at the expense of some order overlap). The CCDs were binned 2 × 2, resulting in a spatial scale along the slit of 0.25′′ pix^-1. Data were reduced using IRAF routines, using calibration data (flatfields and arc lamp exposures) taken the same night, the three exposures were combined before source extraction. We measure a FWHM spectral resolution R = 590 at 7000 Å. The slit position angle (58°) was chosen such that the slit covers both the afterglow and the host nucleus, giving the largest possible spatial separation of the afterglow spectral trace from that of the brightest part of the host galaxy. We are therefore able to extract the afterglow spectrum and host spectrum with minimised contamination from one to the other, while using both traces to obtain an accurate trace position fit. We used an observation of the spectrophotometric standard star BD 25 4655 to correct for instrument response, and photometric observations were used to bring the data to an absolute flux scale.

Appendix A.2.4: WHT/ACAM

We obtained spectroscopy using ACAM on the WHT immediately following the identification of the afterglow in the imaging data. A single exposure with 900 s of integration was obtained, centred on the afterglow, at an average airmass of 1.9, starting at 23:54:36 UT on 3 June 2013 (i.e. the mid time of the observation was 8.22 hr after the burst). The slit position angle was set to parallactic. Data were reduced using IRAF routines, using calibration data (flatfields and arc lamp exposures) taken the same night. ACAM uses a 400 lines/mm volume phase holographic grating, and we used a 1′′ slit width; we measure a FWHM spectral resolution R = 530 at 7000 Å. The trace shows a strong continuum. Several emission lines from the host are also detected, somewhat offset from the afterglow trace photocenter and at a redshift consistent with the data from VLT and GTC; absorption lines are only marginally significant.

Appendix B: The afterglow

Appendix B.1: Photometry of the afterglow

To obtain photometric measurements of the afterglow in each band we performed, when possible, image subtraction with the public ISIS code (Alard 2000). For clean subtractions we selected a later time (afterglow free) image from each telescope as a template and subtract this from the earlier data. Photometric calibration of these subtracted images was obtained by creating artificial stars of known magnitude in the first image, with the errors estimated from the scatter in a large number of apertures (of radius approximately equal to the seeing) placed within the subtracted image. The placement of artificial stars close to the limiting magnitude within the image confirms that these can be recovered, and so the given limiting magnitudes are appropriate. However, we do note that the limiting magnitudes are based on the scatter in photometric apertures placed on the sky, not on the relatively bright regions of the host directly underlying the GRB. The log of optical/NIR photometry of the afterglow is given in Table 1.

We used the stacked UVOT data (A.3), where the last epoch was subtracted from the rest, assuming that this last one had only emission from the host galaxy. These data are important, as they cover the early phase of the optical light curve and allow us to identify a flattening in the optical light curve which significantly differs from the X-ray emission.

The X-Ray Telescope (XRT) onboard Swift began observing the GRB field on 2013 Jun 03 at 15:49:13.945 UT. We used photon counting (PC) mode data for the X-ray spectral analysis. X-ray data were obtained with Swift/XRT using the reductions provided by the Swift Burst Analyser (Evans et al. 2010) and transformed to 2 keV. The light curves obtained with all these observations are plotted together in Fig. B.1.

Fig. B.1 Light curves of GRB 130603B, indicated detections with dots and upper limits (3σ) with arrows. V-band photometry has been scaled and plotted together with the g-band. The vertical lines indicate the times when spectra were obtained. Dotted lines indicate the light curve fits to a power law temporal decay from 0.3 to 3 days after the burst. We include data from the literature (Cucchiara et al. 2013a; Tanvir et al. 2013). The dashed blue line is the expected r-band light curve of a supernova like SN 1998bw, the most common template for long GRBs after including an extinction of AV = 0.9 magnitude. The most constraining limits indicate that any supernova contribution would be at least 100 times dimmer than SN 1998bw in the r-band, once corrected of extinction (blue dashed-dotted line).

Open with DEXTER

The X-ray light curve is characterised by a slow decay during the first 0.1 days, followed by a gradual steepening. The late optical data match reasonably well the steep late decay of the X-ray light curve, which makes the afterglow undetectable after one day even for large telescopes. There is a clear break in the optical light curve at ~ 0.25 days before which the evolution strongly differs from the X-ray one, with the optical being much flatter than the X-rays or even consistent with a brightening until 0.2 days. A direct comparison of optical and X-ray light curves is shown in Fig. B.2.

Fig. B.2 Light curve of GRB 130603B, where all the bands have been scaled to the r-band 0.3 days after the burst. The evolution of all the bands is consistent after ~0.25 days, but before that the X-rays are much stronger than the optical which seem to reach a maximum at around 0.2 days. Overplotted are the spectral slopes of the fits to some of specific segments of the light curve, where Fν ∝ tα.

Open with DEXTER

Appendix B.2: Spectral energy distribution of the afterglow and extinction

In this section we aim to fit the X-ray to optical/NIR SED using the method followed in Zafar et al. (2011, 2012) to derive the extinction in the line of sight of the GRB and determine some spectral parameters. The procedure is briefly explained below.

The flux calibrated spectrum has been analysed after removing wavelength intervals affected by telluric lines and strong absorption lines. We then rebinned the spectrum in bins of approximately 8 Å by a sigma-clipping algorithm. To check the flux calibration of the X-shooter spectrum, we compare the continuum with the flux densities obtained from the extrapolation of the photometry at the time of the spectrum (mid time around 8.56 h).

We include the X-ray spectrum from XRT on board Swift. We used XSELECT (v2.4) to extract spectral files from the event data in the 0.3–10 keV energy band. The X-ray spectrum was extracted in the time interval 10 000 to 52 000 s, resulting in 9.7 ks effective exposure time. The X-ray spectral file was grouped to 15 counts per energy channel. The spectra were fitted within XSPEC (v12.8.0; Arnaud 1996) with a Galactic-absorbed power-law model, with absorption from the Galactic neutral hydrogen column density (fixed to 1.93 × 10²⁰ cm^-2; Kalberla et al. 2005), and absorption from the GRB host galaxy. The best fit parameters are the photon index and the intrinsic column density cm^-2. The resulting fit is good with for 11 degrees of freedom.

We try to fit the data using different extinction curves, namely SMC, LMC, and MW, together with single and broken power-law models. The best fit to the X-ray to optical/NIR SED results in an extinction of A_V = 0.86 ± 0.15 mag and an SMC extinction law. The SED is well fitted with a broken power-law with the optical slope β_O = −0.65 ± 0.09, X-ray slope β_X = −1.15 ± 0.11, and a spectral break at log ν_break/Hz = 15.98 ± 0.76. The resulting fit is good with for 2615 degrees of freedom. We also fit the data with a single power-law model resulting in a poor fit with best fit parameters of A_V = 0.65 ± 0.13 mag and β = −0.89 ± 0.13 and a for 2617 degrees of freedom.

One of the key diagnostics of where GRBs explode is the absorption in their afterglow spectra, with LGRBs showing very high soft X-ray absorbing column densities (N_HX) on average, associated with the star-forming regions in which they explode (e.g. Galama & Wijers 2001). Oddly, analyses of the small sample of SGRBs obtained over the past decade with Swift have shown that SGRBs can have high X-ray absorbing column densities too (Kopač et al. 2012), and a distribution in this absorption comparable to long GRBs. This result is surprising given the very different expected environments and points to a high density environment for some SGRBs. However, these results are based on small number statistics, are biased towards GRBs closer to their host galaxies (to have firmer associations), and are contaminated with short GRBs with extended emission (50% of the mentioned sample), so the issue is not settled.

Appendix C: GRB 130603B in the context of short GRBs

We select all Swift GRBs with T₉₀ ≤ 2 s, using the standard classification of short GRBs (Kouveliotou et al. 1993), and a firm host galaxy association with a measured redshift. These selection criteria exclude short GRBs with extended emission, which may share a common progenitor to short GRBs but this has not been proven to date. The sample is given in Table C.1. GRB 130603B is an unambiguously short GRB with one of the shortest T₉₀ durations in the sample of short GRBs (with well constrained host galaxy redshifts), spectrally harder than the average short GRB in this sample and with slightly higher than average fluence. To compare with this sample we construct pseudo-bolometric light curves using the gamma-ray and X-ray data from the Swift satellite as explained below. The unabsorbed observed 0.3–10 keV light curves were downloaded from the Swift Burst Analyser using 5σ binning (Evans et al. 2010). When 5σ binning resulted in a non-detection in the BAT observations, we manually bin the BAT data using 3σ significance bins and extrapolate each bin to 0.3–10 keV assuming a simple power-law spectral model. These data points were combined with the XRT light curves obtained from the Swift Burst Analyser. Each light curve was converted into rest frame 1–10 000 keV luminosity light curves using a k-correction (Bloom et al. 2001) giving an approximation to a bolometric light curve. This procedure is also repeated for all long GRBs with photometric and spectroscopic redshifts.

Fig. C.1 Pseudo-bolometric light curve evolution of Swift GRBs from the prompt to the afterglow phase. The density plot was built from over 280 long GRBs with redshift information. Its median light curve is shown in black. Overplotted are in red the 13 short GRBs from Table C.1 and in dark blue GRB 130603B. X-ray data sets naturally have orbit gaps. To correct for this, we estimate the luminosity of long GRBs at missing grid points by interpolating between adjacent data points.

Open with DEXTER

In Fig. C.1 we plot the rest frame light curve of GRB 130603B in comparison to other short GRBs and the distribution of long GRBs. The luminosities of long and short GRBs during the prompt emission should be compared with great caution. We assumed that the spectrum of all bursts are adequately described with a power law from 0.3 to 10 000 keV, however most GRBs have a peak energy of a few hundred keV (Nava et al. 2011). Depending on the intrinsic spectrum, this can result in an uncertainty in the bolometric luminosity of about three orders of magnitude. Despite these uncertainties, it appears that short GRBs can be as luminous as the most luminous long GRBs during the prompt emission. The comparison of the X-ray afterglow is not affected by this limitation, because by the time the X-ray observations begin, the peak of the intrinsic spectrum is not at γ-ray energies anymore but in the mm/sub-mm range. The short GRB afterglows are offset from the long GRBs, as they are fainter and typically fade more rapidly but otherwise show similar behaviour (consistent with the findings of e.g. Nysewander et al. 2009). GRB 130603B is consistent with the rest of the short GRB sample, as it has an average luminosity for a short GRB but is about 1.5 dex less luminous than an average long GRB and only consistent with less than 1–2% of the long GRB sample.

Appendix D: Comparison of the Fireball model to the Magnetar model

In the fireball model of GRBs, the late time X-ray and optical emission originates from a forward shock interacting with the surrounding medium with a reverse shock propagating back through the jet and, with well sampled optical and X-ray data, it is possible to derive the electron energy distribution and structure of the surrounding medium (Rees & Meszaros 1992; Meszaros & Rees 1997; Wijers et al. 1997). At later times, the light curve is predicted to have an achromatic jet break which can be used to derive values such as the jet opening angle (e.g. Kumar & Panaitescu 2000). This model can be fit to GRB 130603B, assuming that the optical data are consistent with a fireball peaking at ~0.15 days with a jet break at ~0.25 days and a photon index of p = 2.3. However, the X-ray emission would only be consistent with this if there is an additional component prior to ~0.1 days. For the observations to be consistent with the fireball model, the peak of the emission should be reached at unusually late timescales which implies that the initial Lorentz factor was very low. Additionally, the pre-jet break optical and X-ray temporal slopes are inconsistent with the fireball model. Therefore, GRB 130603B has evidence of early-time energy injection in the X-rays.

	Fig. D.1 Rest frame 1–10 000 keV light curve for GRB 130603B. The red data are Swift observations, with the first 2 data points being from the BAT observations and the remainder are XRT observations. The blue data are the XMM observations obtained by Fong et al. (2014). The Swift XRT data are fitted using the magnetar model, as described in Rowlinson et al. (2013), shown by the black line.
Open with DEXTER

An alternative interpretation can be obtained by studying the origin of the early-time energy injection in more detail. In the standard progenitor theory for short GRBs, the compact binary merger of two neutron stars or a neutron star and a black hole will result in a black hole. The majority of accretion will occur within the first seconds (e.g. Rezzolla et al. 2011), which means that there is no large reservoir of material in an accretion disk typically used to explain the long-term energy injection. However, the merger of two neutron stars may instead form a magnetar with millisecond spin periods (Dai & Lu 1998) which can power a plateau phase as it spins down (Zhang & Mészáros 2001). This model has been fit to a large sample of short GRBs with evidence of energy injection and many were found to be consistent with the magnetar central engine model (Rowlinson et al. 2013). By fitting this model to GRB 130603B, we find that the X-ray light curve can be fitted with a newly-formed stable magnetar (with a magnetic field of G and a spin period of ms, where ϵ is the efficiency that the rotational energy is converted to X-rays and θ is the beaming angle of the magnetar emission, using the method described in Rowlinson et al. 2013). The magnetar model is shown in Fig. D.1 and assumes that the GRB is a “naked” GRB with no standard afterglow component. We also plot the two late time XMM observations obtained by Fong et al. (2014), these data are consistent with the magnetar fit obtained using the XRT data. As this model describes the entire light curve, the fireball model is no longer relevant and the possible jet break mentioned previously was a consequence of approximating a smooth curve with a broken power law model.

The early time optical data (before the 0.25 day break) are significantly fainter than expected from the X-ray observations and this discrepancy cannot be explained using absorption, as the absorption is well fitted using late time observations, or using the standard afterglow spectrum. The excess X-ray emission is consistent with that observed in many of the short GRBs previously fitted with the magnetar model, in which the discrepancy is explained as resulting from additional energy injection in the X-ray light curve (Rowlinson et al. 2013). Within the context of the magnetar model the late-time optical emission could be explained as reprocessing of the X-ray emission. For instance, short GRBs are expected to eject a small amount of material which produces a kilonova (e.g. Li & Paczyński 1998; Metzger et al. 2010), this kilonova component would be boosted by additional heating of the ejecta by the magnetar (Kulkarni 2005). Also the newly formed magnetar is expected to produce powerful winds, in some cases reaching relativistic velocities, which interact with the surrounding medium producing synchrotron emission peaking on the deceleration timescale of the magnetar (approximately the plateau duration, Gao et al. 2013).

© ESO, 2014