4.1. The light curve
Along with the light curves in Fig. 2, we have also plotted the decay rate of 56Co; 0.98 mag (100d)-1. This is the slope the bolometric late light curve of a radioactively powered supernova would have if it trapped all of its -rays. Most SNe II follow this decay rate closely (Turatto et al. 1990; Patat et al. 1994). It is clear from Fig. 2 that the light curves of SN 1996N decline substantially faster.
SNe Ib/c seem to be able to display a variety of light curves. For example, SN Ib 1984L followed the 56 Co decay rate for at least 500 days (Schlegel & Kirshner 1989), whereas SN 1994I, a Type Ic, faded even faster than SNe Ia, at least up to about 70 days after maximum (Richmond et al. 1996b). A well studied supernova with a fast decay rate is SN 1993J. In Fig. 4 we compare the V and R light curves of SN 1996N with the light curves of SN 1993J, taken from the La Palma archive (cf. Lewis et al. 1994). It is clear from this figure that the light curve slopes of these supernovae are rather similar at late times.
It has been argued (Clocchiatti & Wheeler 1997) that there exists a homogeneous photometric group of SNe with light curve slopes of about 1.9 mag (100d)-1 after 150 days. This group contains SN IIb 1993J as well as SN Ib 1983N and SN Ic 1983V. These SNe have similar late time slopes and peak to tail ratios. The similar photometric behaviour could perhaps indicate similar progenitors, where the early spectra, and thus the classification, is determined by small differences in the thin outer layers of the progenitors (Clocchiatti & Wheeler 1997).
The late time decline rate for SN 1996N is similar, but somewhat faster, than the decline of SN 1993J. A chi-square fit to the data of SN 1993J from the La Palma archive and from Richmond et al. (1996a) gives slopes for the V, R and I light curves of 1.530.03, 1.360.06, and 1.430.12 mag (100 d)-1 respectively, between 170 and 340 days after explosion.
As the filter light curves of SN 1996N all have the same slope, within the errors, a linear scaling to bolometric luminosity seems justified. For the three epochs where we have V, R and I photometry the summed fluxes show a decline rate of 1.750.25 mag (100d)-1. The fast decline of the optical light curves of SN 1996N thus indicates leakage of -rays from the 56 Co decay.
A very simplified model of -ray deposition in an expanding homogeneous sphere predicts a slope of 1.09 (111.3-1 + 2 t- 1) magnitudes per day for late times, when the optical depth 1 (Clocchiatti & Wheeler 1997). Here, 111.3 days is the e-folding time for the radioactive decay of 56 Co . For the epochs of observation for SN 1996N, this gives a slope of 1.9 mag (100d)-1, very similar to the measured value of 1.7 mag (100d)-1. The similarity of the light curve slopes of many supernovae at these late epochs may thus simply reflect the asymptotic behaviour of the -deposition, reached when the optical depth for -rays becomes very low (Clocchiatti & Wheeler 1997).
If the optical depth for -rays is in fact very low, the contribution from positrons must be taken into account. For a simplified model with a central radioactive source (Sollerman et al. 1998) the bolometric luminosity decays as (1 - 0.965 ), where 111.3 days is again the decay time of 56 Co , 96.5 of the energy is mediated by -rays and the rest by positrons. The positrons are assumed to deposit their energy locally while the -rays have an optical depth =(t/)-2, where is the time when =1. In this model, a slope of 1.7 mag (100 days)-1 between 179 and 337 days past explosion can be achieved for =63 - 218 days. In fact, the steepest decline is 1.67 mag (100d)-1 for =121 days, this slope is also shown in Fig. 2. Within this model, it seems that SN 1996N was declining as fast as it could.
As seen from the absolute light curves in Fig. 4, SN 1996N appears to be fainter than SN 1993J. The distance to SN 1993J is well determined from Cepheids as 3.6 Mpc (Freedman et al. 1994) and estimates for the reddening of SN 1993J ranges from =0.08 to 0.4 (Barbon et al. 1995). Here we have adopted =0.19 from Lewis et al. (1994). The distance to NGC 1398 is about 22.0 Mpc for =65 km s-1 Mpc- 1(Kraan-Korteweg 1986), and we have assumed in Fig. 4 that there was no intrinsic extinction for SN 1996N. An error of 0.5 magnitudes due to current uncertainties in , and another 0.25 mag allowing for an uncertainty of two weeks in the date of explosion for SN 1996N, is not enough to resolve the difference shown in Fig. 4.
This indicates that SN 1996N was underluminous compared to SN 1993J, or that the intrinsic extinction for SN 1996N was as high as =0.80.3. As these supernovae have similar light curve declines, the faintness of SN 1996N could be due to a substantially lower mass of ejected 56 Ni (cf. Sollerman et al. 1998).
4.2. The mass of oxygen
The mass of oxygen in the supernova is potentially interesting as the amount synthesized in supernova models is rather sensitive to the core mass. For example, in a model with a helium core of 3.3 , 0.22 of oxygen is synthesized, while a core of 6 gives 1.5 (Thielemann et al. 1996).
Assuming that the electron density 106, which is quite reasonable for these epochs (Schlegel & Kirshner 1989), one can use the luminosity of the [O i] 6300, 6364 lines to estimate the mass of oxygen (or rather of neutral oxygen), using =102 F([O i]) (Uomoto 1986). Here is the mass of oxygen in solar masses, is the distance to the supernova in Mpc, F is flux in erg s-1 cm-2 and is the temperature in units of 104 K. This estimate assumes these emission lines to be optically thin. We were not able to constrain the density from the [O i] 6300, 6364 lines, although a normal 3:1 ratio seems to give the best deconvolution of the blend. Nevertheless, a lower limit to the oxygen mass can be achieved using this method.
The temperature can be constrained by the [O i] 5577 [O i] 6300, 6364 ratio. Assuming that all the emission seen at 5530 Å in our earliest spectrum is due to [O i] 5577, we find a ratio 0.08. This is rather similar to the values found for SN 1985F and SN 1983N, 0.05 and 0.04 respectively (Gaskell et al. 1986). This constrains the temperature for =108 - 109 cm-3 to be T 5000 - 4400 K, which in turn gives 0.11 - 0.21 , for a distance of 22.0 Mpc. The lower value of the oxygen mass corresponds to the lower density. The O i 7774 line also indicates the presence of ionized oxygen, as this line presumably comes from recombination (Begelman & Sarazin 1986). This would increase the estimated total mass of oxygen.
These estimates are, however, rather sensitive to uncertainties in the distance and reddening. The distance above comes from Kraan-Korteweg (1986), using a value of 65 km s-1 Mpc-1 for . An uncertainty of 15 km s- 1 Mpc-1 in that number transforms to for =108 and for =109.
4.3. The spectra - evidence for hydrogen?
The spectra of SN 1996N at late times are similar to those of other SNe Ib/c in the nebular phase (Filippenko et al. 1990; Filippenko 1997). In particular, they resemble the spectra of SN 1993J (Fig. 5). Note the broad feature at the red side of the [O i] 6300, 6364, which in SN 1993J was attributed to . Here we want to discuss if the same identification can be made for SN 1996N.
SN 1993J showed hydrogen in its early spectra but underwent a spectroscopic metamorphosis to a SN Ib/c in nebular phase, where the rather weak, broad remained the only evidence for the SN II origin. A similar transformation was seen in SN 1987K (Filippenko 1988) and more recently in SN 1996cb (Garnavich 1997), which displayed a spectral evolution similar to that of SN 1993J.
If the emission redward of [O i] 6300, 6364 in SN 1996N is to be interpreted as broad , as it was in SN 1993J, one must thus address the question why no was seen in the early spectrum of SN 1996N. Remembering that the absorption in SN 1993J was very weak, it is tempting to assume that this feature could be totally lost for SNe with even less hydrogen. Perhaps the only early spectrum for SN 1996N was taken at an epoch when the thin hydrogen layer had already recombined.
However, the faint, broad 6600 Å feature seen in the late spectrum of SN 1983N (Gaskell et al. 1986) indicates that this might be a more common scenario. SN 1983N was rather well studied at early times (Harkness et al. 1987), and showed no prominent absorption in its early spectra, hence the Type Ib classification.
Spectral modeling by Wheeler et al. (1994) showed the early spectrum of SN 1983N to be consistent with the presence of small amounts (0.005 M) of hydrogen. For SN 1993J, Swartz et al. (1993) concluded that 0.04 M of hydrogen could reproduce the early spectra. This indicates that only very small amounts of hydrogen could in fact be hidden in the early spectra of SNe Ib/c. However, none of these studies did investigate a broad range of hydrogen abundances or distributions.
Unfortunately, it is not trivial to estimate if such a small amount of hydrogen is able to produce the observed emission at late times. A simple-minded way is to assume that all the emission comes from case B recombination with T=10 000 K. Then
is the mass of ionized hydrogen in solar masses, is the filling factor, V is the maximum velocity of the hydrogen shell in km s-1, t is time since explosion measured in days, f is the flux of in erg s-1 cm-2, d is the distance to the supernova in Mpc and ( is the ratio of protons to electrons. We estimated the flux in the line to be 810-15 erg s-1 cm-2 in our earliest spectrum, by measuring the red unblended part of the line and assuming that it is symmetric. The line extends to 10 000 km s-1, assuming it is blueshifted by the same amount as [Ca ii] 7291, 7324. These numbers give, for an epoch of 179 days, a required mass of ionized hydrogen of 0.5 ()0.5 . If the hydrogen is really uniformly distributed this estimate is quite high, and seem difficult to reconcile with the lack of hydrogen in the early spectrum. The hydrogen is, however, likely to be distributed in a narrow shell, and the distribution is probably very clumpy. If the hydrogen would be distributed in a shell between 8000 - 10 000 km s-1, with a filling factor =0.01, we would instead get =0.02 . Such a low mass of hydrogen is perhaps not in conflict with early time spectra.
If the discussed feature is indeed , it must somehow be excited at these late epochs. An obvious suggestion is ionization by the X-rays from the shock between the ejecta and the CSM. After all, SNe Ib/c and transition objects like SN 1993J are believed to be core-collapse supernovae with progenitors which lost much of their envelopes before the explosion. Evidence for circumstellar interaction comes from radio observations; SN 1993J, SN 1996cb as well as SNe Ib 1983N and 1984L were, just as SN 1996N, detected at radio wavelengths (Weiler et al. 1998). Furthermore, the broad line in SN 1993J was powered by circumstellar interaction after 250 days (Houck & Fransson 1996; Patat et al. 1995).
However, for a constant mass loss rate, CSM density profile and ejecta density profile, the X-ray luminosity from a radiative reverse shock, and thus the emission, is expected to stay rather constant with time, as was seen after 300 days in SN 1993J. In SN 1996N this is not observed.
An alternative suggestion for the excitation of in the late time spectra of SN 1996N is line blending with [O i] 6364. This was in fact shown to be the most important mechanism for populating n=3 in SN 1993J at this epoch (Houck & Fransson 1996).
If the discussed emission line is not , we must postulate the existence of a broad blend of emission lines at 6600 Å. This was suggested by Patat et al. (1995) for SN 1993J, were this blend possibly contributed 30 of the emission at the position of . A broad blend, attributed to Fe ii, is seen in early time spectra of SN Ia. This feature is indeed sometimes incorrectly identified as (Filippenko 1997). We would then be left with the annoying fact that very similar spectral features can be due to very different physics. We must also try to understand why the 6600 Å bump is so prominent in SN 1996N, compared to other SNe Ib/c.
The ongoing discussion on the nature of SNe Ib/c, and especially of their progenitors, has been focused on the presence or absence of hydrogen (and helium) in their early spectra, see the review by Filippenko (1997). Based on the resemblance of the spectra of SN 1996N and SN 1993J, we can speculate whether small amounts of hydrogen might go unnoticed in early spectra of SN Ib/c. Later on, when the SN continuum has faded, this hydrogen shell might be reionized by CSM interaction, or excited via line blending, thus revealing its existence. Late time spectra might then be the best way to resolve this issue.
4.4. The blueshifts of the emission lines
As indicated in Table 5, the emission lines of SN 1996N appear to be blueshifted with respect to the parent galaxy, the velocities inferred are 1000 km s-1. This is, again, similar to the case of SN 1993J, were the blueshifts of the oxygen lines attracted attention by several authors. Many different models were put forward to explain the blueshifts in SN 1993J: Wang & Hu (1994) proposed that in a clumpy distribution, only the approaching clumps would be seen due to their proximity to the photosphere. Similarly, Filippenko et al. (1994) proposed that the blueshifts were due to optically thick ejecta, were we view only the approaching side. A different explanation was suggested by Spyromilio (1994), who interpreted the lineshifts as indications for large scale asymmetries in the distribution of the ejecta. This scenario was also suggested by Lewis et al. (1994). Finally, Houck & Fransson (1996) proposed that the lineshifts were simply an effect of line blending.
There were difficulties in all of these models for SN 1993J. The emission lines did not show the same amount of blueshift. For example, in the spectrum from September 20, 1993, shown in Fig. 5, we measured the blueshift of [O i] 5577 to be 1500 km s-1 whereas [Ca ii] 7308 only shows a shift of 300 km s-1. It is difficult to understand how optical depth effects can affect these lines so differently. Moreover, such scenarios would predict the lineshifts to decrease with time, as the optical depth decreases. On the contrary, the shifts seemed unchanging in nature (Lewis et al. 1994; Spyromilio 1994).
Similarly, the case for large scale asymmetries in the ejecta mass have a problem in explaining the much smaller blueshift of O i 7774, as pointed out by Houck & Fransson (1996). This line seems to be blueshifted by only 500 km s-1, significantly less than the [O i] 5577 line. This led Spyromilio (1994) to suggest that also the distribution of 56 Co was asymmetric, explaining the spatial differences in excitation conditions.
Houck & Fransson (1996) used a detailed spectral modeling code to conclude that no large scale asymmetries were required to reproduce the spectra of SN 1993J. Instead, they argued that the apparent blueshifts could be explained as effects of line blending. For example, the blueshift of the [O i] 6300, 6364 line could be due to blending with fast-moving hydrogen. They further proposed that the shifts of Mg i] 4571 and [O i] 5577 could be explained along the same line. However, the O i line at 7774 Å seems to be rather unaffected by line blending. The blueshift of this line in SN 1993J might thus indicate real large scale asymmetries.
For SN 1996N, blueshifts of the order of 1000 km s-1 were observed for [O i] 6300, 6364, [Ca ii] 7291, 7324 and Mg i] 4571. The actual numbers given in Table 5 are rather uncertain, they merely reflect Gaussian fits to these non-Gaussian lines. These fits do, however, indicate systematic blueshifts of the emission lines, and show that these stay rather constant with time. That the lines are indeed shifted can be seen from Fig. 6, where the above mentioned lines, as well as O i 7774, are plotted in velocity space. Furthermore, as shown in Fig. 7, no evolution in the widths or positions of the emission lines is seen between 179 and 337 days past discovery. As the density is expected to decrease by a factor of 7 between our first and last spectrum, an evolution of the line centers towards the rest positions would have been expected, if the shifts were due to optical depth effects. The scenario of Houck & Fransson may provide some clues to the observed blueshifts in SN 1996N. In particular, if we identify the emission redward of [O i] 6364 as , as suggested in the previous section, the same scattering effect may be at work also in SN 1996N. We do, however, see a significant blueshift also in the [Ca ii] 7291, 7324 lines, which was not noticed in SN 1993J. Furthermore, the O i 7774 line is positioned at 7749 Å in our earliest spectrum. The line is rather weak and broad but is clearly shifted to the blue by 800 - 1000 km s-1 (Fig. 6). Again, this line is supposed to be relatively unblended, and the blueshift we observe might therefore correspond to real asymmetries in the distribution of the ejecta mass.
Another possible explanation for the blueshifts of the emission lines is dust. In SN 1987A, the emission lines slowly departed to the blue after 500 days, an effect attributed to dust formation (Lucy et al. 1989). The shifts in SN 1987A were 600 km s-1. To achieve velocity shifts of km s-1, optical depths for the dust in excess of =0.5 are needed for an ejecta velocity of 5000 km s-1. We do not observe any dependence of the line shifts on wavelengths, as seen for the dust extinction in SN 1987A. In particular, the Na i 5893 line seems to be less blueshifted than all other lines. Also, our first observation is at 179 days, much earlier than the epoch of dust formation in SN 1987A. The temperature may well be too high for dust formation at this early stage. Similarly, in SN 1993J the shifts were observed already 50 days past maximum. We therefore consider this scenario less likely.
Finally we would like to mention the fact that many supernova remnants show emission lines with large (500 km s-1) velocity shifts. Also the high space velocities of pulsars are well known. These phenomena are often suggested to be due to asymmetric supernova explosions. Perhaps this is what we see here, for the supernova itself.
© European Southern Observatory (ESO) 1998
Online publication: August 6, 1998