2 Ground based observations

2.1 Late-time optical photometry

The photometry discussed here is based on the late observations presented in Sollerman et al. (2000; hereafter S00). S00 obtained data at eight epochs between 32 and 503 days past explosion, using the ESO 3.6 m telescope on La Silla and the Very Large Telescope (VLT) on Paranal. Here we assume the date of explosion to be 980425.9, the moment of detection of GRB 980425 (Soffitta et al. 1998).

To the S00 data set we have added the late time observations presented in Patat et al. (2001; hereafter P01). These data were primarily obtained using the ESO 3.6 m telescope and are added in particular to include the B-band photometry. We have also added previously unpublished photometry obtained at the VLT. The details of the ground-based optical observations are shown in Table 1.

 

 

Table 1: Log of ground based observations.

Date	MJDa	Epochb	Telescopec	Bands	Source
13 Sep. 98	1069	140	3.6 m	$V\!RI$	S00
26 Nov. 98	1143	214	3.6 m	$V\!RI$	S00
18 Mar. 99	1255	326	3.6 m	$V\!RI$	S00
18 Mar. 99	1255	326	0.9 m	B	P01
08 Apr. 99	1276	347	3.6 m	$BV\!R$	P01
12 Apr. 99	1280	351	3.6 m	I	P01
17 Apr. 99	1285	356	VLT	$V\!RI$	S00
17 Apr. 99	1285	356	VLT	B	This work
21 May 99	1319	390	3.6 m	$BV\!RI$	P01
13 June 99	1342	413	VLT	$V\!RI$	S00
17 June 99	1346	417	3.6 m	$BV\!RI$	P01
09 Aug. 99	1399	470	3.6 m	$V\!RI$	S00
11 Sep. 99	1432	503	VLT	$V\!RI$	S00
11 Sep. 99	1432	503	VLT	B	This work
16 Oct. 99	1467	538	VLT	$BV\!RI$	This work
01 Apr. 00	1635	706	VLT	$BV\!RI$	This work
Oct./Nov. 00	$\sim$1836	$\sim$907	VLT	$BV\!RI$	This work

^a MJD = Julian Date-240000.5.
^b Epoch in days past 980425.9.
^c VLT = UT1 + FORS1, 3.6 m = ESO-3.6 m + EFOSC2, 0.9 m = ESO-Dutch-0.92 m.

As discussed in S00 and P01, the photometry of SN 1998bw is fairly complicated. This is because the supernova is superposed on an H II region, and sits in a rather complex region of the host galaxy. This was strikingly revealed by the HST/STIS image published by Fynbo et al. (2000), where several objects can be seen within $1\arcsec$ of the position of the supernova. This is here shown in detail in Fig. 1.

Figure 1: HST/STIS CLEAR images of SN 1998bw and its environment. The images were obtained in 2000 June (upper left), 2000 November (upper right), and 2001 August (lower left). The field of view is merely $1\farcs5 \times 1\farcs5$. Lower right shows a subtraction between the first and the final frame. The supernova is clearly identified. The apparent differences in resolution are due to differences in the dithering procedure, as explained in the text.

This means that even sophisticated point-spread function (PSF) fitting photometry may not be adequate to accurately measure the SN brightness at late phases. S00 used DAOPHOT to estimate the errors due to background subtraction, while P01 used the SNOOPY software. Although these methods were consistent, the very complex background of SN 1998bw as observed with HST called for a revision of the light curve. We decided to try another method to establish the late time light curve of this important supernova.

We therefore obtained deep B, V, R, and I imaging of the region containing the supernova using FORS1 on the VLT. These observations were obtained in service mode under excellent atmospheric conditions in early October and in early November 2000, some 900 days after the supernova explosion. In each filter band 6-9 frames were coadded to achieve a deep combined image. The B and V frames have a total exposure time of 90 min each, while the R and Iframes have 120 min. All combined frames have good image quality, with a FWHM less than one arcsecond.

In these images, no sign of the supernova could be detected, i.e., the background on the position of the supernova appeared smooth. These frames were therefore used as template frames to subtract the background from all the previous images (see e.g., Filippenko et al. 1986 for an early account of this method). The detailed procedure used is outlined in Schmidt et al. (1998), and we have used the same software as used by the High-z Supernova Search Team for galaxy subtraction, magnitude measurements and error estimates.

After the usual bias subtraction and flat fielding, all images were aligned with the template image in the same filter band. The image with the best seeing was then convolved to match the seeing of the other image. Thereafter the image was scaled to the same intensity as the template, and the template was subtracted from the image in a region around the supernova. The program DOPHOT (Schechter et al. 1993) was used to measure the magnitudes of the supernova and of the local standard stars in the subtracted image.

A PSF was constructed from stars in the field using DAOPHOT (Stetson 1987). It was used to add ten artificial stars with the same flux as the supernova to user-specified locations on the original frame. The magnitudes of these stars were then retrieved using the same technique as for the supernova. This method was used to assess the accuracy of our measurements, and the standard deviations in the differences in the retrieved magnitudes of the artificial stars were adopted as the errors on the supernova magnitudes in Table 2.

 

 

Table 2: Supernova magnitudes.

Epoch	B	$B_{\rm err}$	V	$V_{\rm err}$	R	$R_{\rm err}$	I	$I_{\rm err}$
140	-	-	17.38	0.027	17.06	0.019	16.77	0.007
214	-	-	18.76	0.025	18.14	0.032	17.95	0.051
326	21.21	0.05	20.80	0.011	19.87	0.007	19.72	0.020
347	21.56	0.011	21.15	0.025	20.21	0.018	-	-
351	-	-	-	-	-	-	20.10	0.024
356	21.66	0.069	21.30	0.037	20.27	0.017	20.03	0.025
390	22.19	0.010	21.78	0.029	20.98	0.030	20.95	0.068
413	-	-	22.12	0.029	21.18	0.016	21.03	0.029
417	22.57	0.05	22.16	0.025	21.32	0.030	21.13	0.036
470	-	-	22.86	0.032	22.21	0.052	22.08	0.099
503	23.53	0.043	23.07	0.067	22.59	0.041	22.50	0.103
538	-	-	23.43	0.33	23.05	0.082	23.25	0.465
706	>25.1	-	>25.1	-	>25.1	-	>24.7	-

In all cases, the success of the subtraction has been manually checked. Indeed, very clean results were achieved, where only the supernova remains on the subtracted frame. This is due to the fact that many of these observations have been obtained with the same telescope/instrument/CCD setup. Similar seeing conditions are also important, and in particular a high quality template frame is crucial.

In Fig. 2 we show an example of the region of the supernova in our template frame, and in our frame from day 503. This was the last epoch used in S00 and P01. Also shown is the subtraction, which clearly shows the supernova.

Figure 2: Illustration of the template subtraction method. Upper left shows the supernova and its galaxy as observed with the VLT in 1999 September, 503 days past explosion. Upper right is the template image. North is up and East to the left; the field of view is about $52\arcsec\times52\arcsec$. The subtraction is shown in the lower left. Only the supernova, as well as some residuals from stars that are saturated in the template frame, are seen. Lower right shows a blowup of the supernova region, before and after subtraction. Note how the galaxy is completely subtracted.

The instrumental magnitudes were finally converted using local standards measured by Galama et al. (1998). This worked well for the B-band, where four standard stars showed a standard deviation of $\le$0.03 mag. Most of this scatter is in fact due to the neglect of color terms, which, however, do not affect the supernova as we chose a primary standard with similar color as the SN itself. For the other bands, all Galama standards were saturated in the long exposures required at these late phases. We then established secondary, faint local standards. The magnitudes of four of these local standards are presented in Table 3, together with their angular offsets from the supernova. The standard deviations for these stars measured over all the epochs were never higher than 0.03 mag. Adding this to the uncertainty reported by Galama et al. for their standards, we consider the accuracy of the secondary standards to be about 0.06 mag.

2.2 Late upper limits

At our last epoch of VLT observations prior to the template images, i.e., at 706 days past explosion, the supernova is not detected in the subtracted images. We have used these difference images to estimate upper limits on the supernova magnitude in the following way. We measured the fluxes in a number of apertures located next to the supernova. From these measurements we estimated the standard deviations in a 5 pixel radius aperture, corresponding to the image FWHM. This estimate was then used to acquire $3\sigma$upper limits within the aperture. These limits are 25.1, 25.1, 25.1,

 

 

Table 3: Magnitudes of faint local standards.

Offsets from SN	V	R	I
$46\arcsec$ N, $17\arcsec$ W	19.77	19.03	18.45
$50\arcsec$ N, $27\arcsec$ W	21.13	19.95	18.69
$17\arcsec$ N, $26\arcsec$ W	21.64	20.44	19.12
$30\arcsec$ S, $17\arcsec$ E	21.24	20.37	19.63

and 24.7 in B, V, R, and I, respectively. This procedure is more conservative than a simple measurement of the statistics in the region of the supernova. We also verified that a R=25.1 supernova would indeed have been detected by adding such a star to the original frame, and recovering the magnitude in the subtracted frame. This would thus provide conservative upper limits on the flux of the supernova at this epoch. The main assumption is that the supernova was indeed absent in the template frame. This is further elaborated below.

2.3 The near-infrared

Here, we present the last infrared (IR) observations of SN 1998bw of which we are aware.

SN 1998bw was observed using ISAAC on the VLT on 1999 May 1, 370 days after the explosion. The observations were obtained in Autojittermode, offsetting the telescope within a $60\arcsec$ wide box. Five frames were obtained in the J-band (1.11-1.39 $\mu$m), each one composed of a stack of five 30 s exposures. The total exposure time in the J-band was thus 750 s.

In the H-band (1.50-1.80 $\mu$m) five frames composed of five 20 s exposures yielded a total exposure time of 500 s. In addition, an infrared spectrum was obtained at this epoch, but the signal-to-noise ratio of this spectrum is low, and we will not discuss it further. Instead, we will use the two acquisition images to determine the IR contribution at these late phases.

The reductions were done using IRAF. From each image we subtracted a mean sky, constructed from combining all the other frames with suitable clipping to remove objects. Flat field corrections were applied using dome flat-fields. The images were then stacked and combined. The FWHM in the combined frames were $0\farcs4$ and $0\farcs55$ for J and H, respectively.

For the calibration of the photometry, we used the zero points determined for this specific night at the VLT, from observations of photometric standards. To check the stability of the zero point, we also re-measured three local standard stars from the three earlier epochs obtained with NTT/SOFI (P01). The standard deviation in the standard stars on the three SOFI nights were 0.04 and 0.02 mag in J and H, respectively. The magnitudes agreed with the ISAAC measurements to better than 0.05 mag, with a scatter of 0.02 mag.

In the combined frames, the SN magnitude were measured with PSF photometry using DAOPHOT within IRAF. The measured magnitudes are J=20.16 and H=19.54. The aperture correction was obtained from the three local standards, with a negligible scatter. The DAOPHOT measurement error was 0.03 (J) and 0.06 (H) magnitudes. The total error budget for the photometry is therefore certainly better than 0.1 mag in both bands. These results should be treated with some caution, however, as we know from the optical data that PSF fitting may not be the best technique for this supernova.