Hubble Space Telescope and Ground-based Observations of SN 1993J and SN 1998S: CNO Processing in the Progenitors
ABSTRACT Ground-based and Hubble Space Telescope observations are presented for SN 1993J and SN 1998S. SN 1998S shows strong, relatively narrow circumstellar emission lines of N III-V and C III-IV, as well as broad lines from the ejecta. Both the broad ultraviolet and optical lines in SN 1998S indicate an expansion velocity of ~7000 km s-1. The broad emission components of Lyα and Mg II are strongly asymmetrical after day 72 past the explosion and differ in shape from Hα. Different models based on dust extinction from dust in the ejecta or shock region, in combination with Hα from a circumstellar torus, are discussed. It is concluded, however, that the double-peaked line profiles are more likely to arise as a result of optical depth effects in the narrow, cool, dense shell behind the reverse shock than in a torus-like region. The ultraviolet lines of SN 1993J are broad, with a boxlike shape, coming from the ejecta and a cool, dense shell. The shapes of the lines are well fitted by a shell with inner velocity ~7000 km s-1 and outer velocity ~10,000 km s-1. For both SN 1993J and SN 1998S a strong nitrogen enrichment is found, with N/C ≈ 12.4 in SN 1993J and N/C ≈ 6.0 in SN 1998S. From a compilation of all supernovae with determined CNO ratios, we discuss the implications of these observations for the structure of the progenitors of Type II supernovae.
- SourceAvailable from: articles.adsabs.harvard.edu
Article: Supernovae and Gamma-Ray Bursts[show abstract] [hide abstract]
ABSTRACT: I review the observational status of the Supernova/Gamma-Ray Burst (GRB) connection. Available data suggest that Supernovae (SNe) associated with GRBs form an heterogeneous class of objects including bright and faint hypernovae (Hyp) and perhaps also `standard' Ib/c events. Current estimates of SN and GRB rates and beaming angles yield ratios GRB/SNe-Ibc ~2% and GRB/Hyp ~25%. In the few SN/GRB associations so far discovered the SN and GRB events appear to go off simultaneously, although data do not exclude that the SN may precede the GRB by a few days.Chinese Journal of Astronomy and Astrophysics 01/2007; 6(S1):315. · 0.89 Impact Factor
arXiv:astro-ph/0409439v1 17 Sep 2004
Hubble Space Telescope and Ground-Based Observations of SN
1993J and SN 1998S: CNO Processing in the Progenitors1
Claes Fransson2, Peter M. Challis3, Roger A. Chevalier4, Alexei V. Filippenko5, Robert P.
Kirshner3, Cecilia Kozma2, Douglas C. Leonard6,7, Thomas Matheson3, E. Baron8, Peter
Garnavich9, Saurabh Jha5, Bruno Leibundgut10, Peter Lundqvist2, C. S. J. Pun11, Lifan
Wang12, and J. Craig Wheeler13
Ground-based and Hubble Space Telescope observations are presented for SN
1993J and SN 1998S. SN 1998S shows strong, relatively narrow circumstellar
emission lines of N III-V and C III-IV, as well as broad lines from the ejecta.
Both the broad ultraviolet and optical lines in SN 1998S indicate an expansion
velocity of ∼ 7,000 km s−1. The broad emission components of Lyα and Mg II
are strongly asymmetrical after day 72 past the explosion, and differ in shape
from Hα. Different models based on dust extinction from dust in the ejecta or
1Based in part on observations obtained with the Hubble Space Telescope, which is operated by AURA,
Inc., under NASA contract NAS 5-26555.
2Department of Astronomy,
Stockholm University,AlbaNova, SE–106 91 Stockholm,Sweden;
3Harvard–Smithsonian Center for Astrophysics, 60 Garden St., Cambridge, MA 02138.
4Department of Astronomy, University of Virginia, P.O. Box 3818, Charlottesville, VA 22903.
5Department of Astronomy, University of California, Berkeley, CA 94720–3411.
6Five College Astronomy Department, University of Massachusetts, Amherst, MA 01003-9305.
7Department of Astronomy, 105-24 Caltech, Pasadena, CA 91125.
8Department of Physics and Astronomy, University of Oklahoma, 440 W. Brooks, Norman, OK 73019-
9Department of Physics, University of Notre Dame, 225 Nieuwland Science Hall, Notre Dame, IN 45656.
10European Southern Observatory, Karl-Schwarzschild-Strasse 2, D-85748 Garching, Germany.
11Department of Physics, University of Hong Kong, Pokfulam Road, Hong Kong.
12Institute for Nuclear and Particle Astrophysics, E. O. Lawrence Berkeley National Laboratory, Berkeley,
13Department of Astronomy, University of Texas, Austin, TX 78712.
– 2 –
shock region, in combination with Hα from a circumstellar torus, are discussed.
It is concluded, however, that the double-peaked line profiles are more likely to
arise as a result of optical depth effects in the narrow, cool, dense shell behind the
reverse shock, than in a torus-like region. The ultraviolet lines of SN 1993J are
broad, with a box-like shape, coming from the ejecta and a cool dense shell. The
shapes of the lines are well fitted with a shell with inner velocity ∼ 7,000 km s−1
and outer velocity ∼ 10,000 km s−1. For both SN 1993J and SN 1998S a strong
nitrogen enrichment is found, with N/C ≈ 12.4 in SN 1993J and N/C ≈ 6.0 in
SN 1998S. From a compilation of all supernovae with determined CNO ratios, we
discuss the implications of these observations for the structure of the progenitors
of Type II supernovae.
Subject headings: stars: circumstellar matter — stars: mass loss — stars: evo-
lution — nucleosynthesis, abundances — supernovae: individual: SN 1993J, SN
The nature of the progenitors of the different types of supernovae (SNe) is still debated.
One of the main unknowns is the amount of mass loss suffered by the progenitor before the
explosion. The extent of the mass loss, driven either by stellar winds in single stars or by
binary evolution, causes widely different progenitor structures. Because the supernova light
curve and spectrum to a large extent are functions of the mass of the hydrogen and helium
envelope, the mass-loss history is a key ingredient in understanding the various observational
signatures, manifested in the many separate classes of core-collapse SNe. Mass loss affects
the chemical composition of the ejecta; hence, detailed spectral studies of the different types
of SNe and their environments can shed some light on this issue.
A large number of SNe have now shown various types and degrees of circumstellar inter-
action; see Filippenko (1997) for a recent review. Both of the “Type II-linear supernovae”
(SNe II-L) 1979C and 1980K showed evidence for circumstellar interaction in their radio
emission (Weiler et al. 1986), as well as in the ultraviolet (UV) emission in the case of SN
1979C (Panagia et al. 1980; Fesen et al. 1999). The circumstellar medium most likely origi-
nates from the dense, slow superwind of the red supergiant progenitor. Type Ib and Ic SNe
are believed to have lost most of their hydrogen envelope prior to exploding. They in general
show both radio emission and X-ray emission caused by circumstellar interaction. Because
of the high wind speed (? 1,000 km s−1) of the Wolf-Rayet (WR) progenitor, the wind in
this case is considerably less dense compared to that of SNe II-L and II-P.
– 3 –
SN 1993J provided a link between SNe II and SNe Ib in that it underwent a transition
from a Type II spectrum at early epochs to a Type Ib-like spectrum at ∼ 300 days (Filip-
penko, Matheson, & Ho 1993; Filippenko, Matheson, & Barth 1994; Finn et al. 1995; Barbon
et al. 1995); see Filippenko & Matheson (2004) for a recent review. At early epochs it already
showed the typical signatures of circumstellar interaction at radio (Van Dyk et al. 1994),
UV (Fransson, & Sonneborn 1994), and X-ray (Zimmermann et al. 1994) wavelengths. After
about one year, the optical spectrum became dominated by emission lines excited by the
circumstellar interaction (Filippenko, Matheson, & Barth 1994; Patat, Chugai, & Mazzali
1995; Matheson et al. 2000a,b).
Type IIn supernovae (e.g., Filippenko 1997) have optical spectral signatures of a circum-
stellar medium (CSM) present already at early phases. Although they show considerable
variation from object to object, the common characteristics are strong, relatively narrow
emission lines after a few weeks, or perhaps earlier; these lines are thought to be produced
by the interaction between the supernova ejecta and the CSM. In extreme cases (e.g., SN
1988Z; Turatto et al. 1993; Chugai & Danziger 1994) the CSM is clearly very dense and
extensive, and the interaction lasts for many years. Occasionally (e.g., SN 1994W; Chugai
et al. 2004) there is evidence that much of the CSM was produced just a short time before
the explosion. Also, the light curves of SNe IIn have a large variation in terms of the rela-
tive importance of the radioactive heating and energy input from circumstellar interaction.
However, the nature and relation of the SNe IIn to other SN types are not clear (for a dis-
cussion, see Nomoto, Iwamoto, & Suzuki 1995). SN 1998S represents one of the best-studied
examples of this type of supernova.
In contrast to SN 1993J, the optical spectrum of SN 1998S was dominated by circum-
stellar interaction even at early epochs. Bowen et al. (2000) found that many low-ionization
UV lines had a narrow P-Cygni component, which they interpret as a wind with velocity
∼ 50 km s−1, as well as a more highly ionized component with velocity ∼ 300 km s−1.
This was later confirmed from ground-based observations by Fassia et al. (2001). Chugai
(2001) argued that the narrow core and smooth high-velocity wings of the Hα line at early
epochs could be understood as a result of electron scattering in a dense circumstellar shell
extending from the supernova. While the observations during the first months showed fairly
symmetrical line profiles, the lines changed character after ∼ 100 days: Hα, as well as He I
λ10830, displayed a highly asymmetrical triple-peaked structure (Gerardy et al. 2000). In-
frared observations showed evidence for CO formation at ∼ 95 days, and dust formation
at ∼ 225 days (Fassia et al. 2000; Gerardy et al. 2002). The supernova also showed strong
X-ray emission at late epochs (Pooley et al. 2002). No radio emission was seen at early
epochs. However, by 310 days emission was detected at 8.46 GHz, and subsequently also at
1.47 GHz and 4.89 GHz (Pooley et al. 2002). The late turn-on and its frequency dependence
– 4 –
may be understood as a result of strong free-free absorption by the circumstellar gas.
One of the most important indicators of the extent of mass loss is the relative abundances
of the CNO elements. Depending on the mass lost and the degree of mixing, CNO burning
products may be seen in either the circumstellar medium or the outer parts of the progenitor,
and therefore in the supernova (e.g., Meynet & Maeder 2000; Heger & Langer 2000; Wellstein
& Langer 1999). Evidence for such CNO processing has earlier been seen in SN 1979C
(Fransson et al. 1984), SN 1987A (Fransson et al. 1989), and SN 1995N (Fransson et al.
2002). In this paper we add two more cases.
SN 1993J and the Type IIn SN 1998S are among the best-observed circumstellar inter-
actors to date. In this paper we report on UV observations of these two SNe, and on their
implications for the abundances of the CNO elements. Early-epoch Hubble Space Telescope
(HST) observations of both SNe have previously been discussed, but mainly in the context
of the outer regions of the supernova ejecta (Jeffery et al. 1994; Houck & Fransson 1996;
Lentz et al. 2001). In this paper we concentrate on the late-epoch observations, where most
of the indicators of circumstellar interaction are present. In addition to the HST observa-
tions, we also add new ground-based observations. Based on the complete evolution of the
line profiles, we discuss implications of the shapes of these for the physics of the interaction
region. The analysis in this phase is simplified by the basically nebular conditions of the
SN 1993J was discovered on 1993 March 28.9 (Ripero et al. 1993), while SN 1998S
was discovered on 1998 March 2.68 UT (Li et al. 1998; Qui et al. 1998). Here we adopt
explosion dates of 1993 March 27.5 for SN 1993J (Lewin et al. 1994), and 1998 March 2 for
SN 1998S. The true explosion date of SN 1998S was probably somewhat earlier, of course,
but by a negligible amount given the late-time phases studied here. Richmond et al. (1994)
discuss the reddening of SN 1993J and argue for a most likely value of EB−V = 0.2 mag,
which we adopt. For SN 1998S Leonard et al. (2000) find a reddening of EB−V = 0.23 mag.
The distance to SN 1993J is 3.63 Mpc (Freedman et al. 1994) and the recession velocity
is −135 km s−1(Vladilo et al. 1994). The recession velocity of SN 1998S is 846.9 km s−1
(Fassia et al. 2001), and we adopt a distance of 17 Mpc (Tully 1988).
The HST observations in this paper were obtained with the Faint Object Spectrograph
(FOS) and the Space Telescope Imaging Spectrograph (STIS). Tables 1 and 2 give the journal
of observations of SN 1993J and SN 1998S (respectively), including exposure times, gratings
– 5 –
used, their dispersion, and the wavelength ranges covered. The FOS spectra were obtained
with the G160L, G270H, and G400H gratings, while the STIS spectra were obtained with
gratings G140L, G230L, and G430L. All spectra were calibrated by the FOS and STIS
pipelines at the Space Telescope Science Institute. The Lyα region in the FOS spectra of
SN 1993J on days 1063 and 1399 is dominated by the geocoronal Lyα, so it is not used
in this paper. In the STIS spectra on days 1792–2585 this component could be subtracted
To complement the HST observations, optical low-dispersion spectra of SN 1998S for
days 29, 75, 257–258, 431, and 440 were obtained with the FAST spectrograph (Fabricant et
al. 1998) on the 1.5-m Tillinghast telescope at the Fred L. Whipple Observatory (FLWO).
The FAST spectrograph employs a 2688 × 512 pixel Loral CCD with a spatial scale of
1.′′1 pixel−1in the binning mode used for these observations. Details of the exposures are
given in Table 3. The data were reduced in the standard manner with IRAF1and our own
routines. Wavelength calibration was accomplished with He-Ne-Ar lamps taken immediately
after each supernova exposure. Small-scale wavelength adjustments derived from night-sky
lines in the supernova frames were also applied. Spectrophotometric standards used in the
reductions are listed in Table 3. We attempted to remove telluric lines using the well-exposed
continua of the spectrophotometric standards (Matheson et al. 2000a). The FLWO spectra
were observed, in general, within ∼ 10◦of the parallactic angle to minimize losses from
atmospheric dispersion (Filippenko 1982).
To complete the temporal coverage of the optical spectra of SN 1998S we have also
included spectra up to day 494 from Lick Observatory, published by Leonard et al. (2000);
see Table 3. In addition to these, two spectra on days 643 and 653 were also obtained at the
Keck II 10-m telescope with LRIS (Oke et al. 1995), and one very late-time spectrum (day
2148) was obtained with the MMT 6.5-m telescope using the Blue Channel spectrograph
(Schmidt, Weymann, & Foltz 1989). All data were reduced in a manner similar to that
described above for the FLWO spectra. For the spectra on days 643 and 2148 the conditions
were such that a proper background subtraction was difficult. These spectra may therefore
be contaminated by the galaxy background and the fluxes may be correspondingly uncertain.
Note that our spectrum from day 653 is also shown in the Hα compilation in Pozzo et al.
Details of the Keck observations of SN 1993J on days 670, 976, 1766, and 2454, included
1IRAF is distributed by the National Optical Astronomy Observatories, which are operated by the As-
sociation of Universities for Research in Astronomy, Inc., under cooperative agreement with the National
– 6 –
in this paper, are given by Matheson et al. (2000a). In addition, we have included one
spectrum of SN 1993J taken with MMT on 1997 April 8 (day 1473).
3.1. SN 1993J
HST observations of SN 1993J at early phases were discussed in Jeffery et al. (1994) and
Baron, Hauschildt, & Branch (1994). In this paper we limit ourselves to the nebular phase,
when the spectrum is dominated by the circumstellar interaction, and the line profiles are
easier to identify and analyze.
Figure 1 shows the HST spectra of SN 1993J from January 1995 to April 2000 (days
670 to 2585), where we have noted the positions of some of the strongest spectral features.
As will be discussed below, however, some of these are likely to be blends of several lines.
Note that the overlap region of the short-wavelength and long-wavelength UV detectors in
the range 1600–1700˚ A is affected by an increased noise level.
During this whole time in the nebular phase, the spectrum is remarkably constant. The
far-UV region is dominated by the strong Lyα emission line, probably blended with N V
λλ1238.8,1242.8. The region shortward of ∼ 2000˚ A forms a pseudo-continuum of blended
lines of highly ionized nitrogen, carbon, and oxygen, which we discuss in more detail below.
The well-defined line at ∼ 2140˚ A is most likely due to N II] λλ2139.7,2143.5. The presence
of this prominent line indicates a strong nitrogen enrichment, as we will confirm later. The
plateau longward of ∼ 2280˚ A is probably formed mainly of Fe II resonance lines, although
a blend of C II] 2323.5–2328.1, O III] 2320.9–2331.4, and Si II] λ2334.6 is likely to explain
the feature at ∼ 2332˚ A. Finally, the near-UV region is dominated by the prominent Mg II
The most interesting temporal change is in the shape of the Mg II line. This evolution
is similar to that seen in Hα (Matheson et al. 2000a). In Figure 2 we compare the Lyα, Hα,
and Mg II line profiles. The Hα profiles are taken from Matheson et al. (2000a). From the
figure we see that the shapes of both the Hα and Mg II lines change in concordance with
each other. On day 670 both lines are clearly asymmetric, with the blue side of Mg II a
factor of ∼ 2.2 stronger than the red. The asymmetry of Hα is considerably smaller, with
a blue-to-red intensity ratio of ∼ 1.1. By day 1063 this asymmetry has decreased for both
lines, and for the later epochs the two sides have nearly equal strength. The Lyα line is
strongly affected by geocoronal emission, which makes a comparison with the other lines
– 7 –
As has been previously noted (e.g., Filippenko, Matheson, & Barth 1994; Finn et al.
1995; Patat, Chugai, & Mazzali 1995; Matheson et al. 2000b), the line profiles after ∼ 1 year,
in particular that of Hα, can be best described as box-like. This strongly indicates that the
emission comes from a relatively thin shell in the expanding gas. Matheson et al. (2000b)
found during this period a maximum velocity of ∼ 9,000–10,600 km s−1for the blue edge
of the Hα line, and a slightly lower velocity for the red edge.
Of the UV lines, only the Mg II and [N II] λλ2139.7,2143.5 lines are unblended and
have sufficiently high signal-to-noise ratio (S/N) to warrant a detailed line fit. As is seen
from Figure 1, the UV spectrum does not change appreciably from January 1996 to February
1997. To increase the S/N, we therefore average the day 1063 and day 1399 spectra. We then
fit these line profiles, as well as the Hα profile from the Keck spectrum on day 976, with a
shell having inner velocity Vinand outer velocity Vout; see Fransson (1984a) for details. The
variation of the emissivity within the shell is of minor importance; as long as the shell is
reasonably thin, it only affects the wings of the lines (for V > Vin).
In Figure 3 we show a fit for the three lines with Vin = 7,000 km s−1and Vout =
10,000 km s−1. Given the simplicity of the model, the fits are surprisingly good, especially
that of Hα. There is a minor asymmetry in the Hα line, where the red side would be better
fit with Vin= 6,000 km s−1and Vout= 10,000 km s−1. There is some indication that the
[N II] line has a somewhat smaller width than Hα, with Vin≈ 6,000 km s−1. Because of the
lower S/N compared to Hα, however, this is hardly significant.
Part of deficit on the red side of Mg II is caused by the interstellar Mg I λ2852 absorption
line from the host galaxy of SN 1993J (M81) and the Milky Way (see de Boer et al. 1993).
This is, however, unlikely to explain the entire asymmetry. Additional interstellar Fe II
absorption lines may possibly contribute, but these should not be stronger than the Mg I
absorption, and therefore they only marginally affect the line. It is also difficult to understand
the red deficit as a result of an intrinsic asymmetry of the Mg II-emitting region, since this
should coincide with the Hα line, which does not show such a pronounced effect. In Section
4.2 we discuss a different explanation based on optical-depth effects in the emitting region.
Of the other UV lines, only N IV] λ1486 is isolated enough for blending not to be
important. The noise, however, makes the profile of this line uncertain. With this caveat,
the line can be fit with the same line profile as for Hα.
For the region below 2000˚ A, which is of most interest for the CNO abundance analysis,
the blending of the lines requires a detailed spectral synthesis to derive accurate line fluxes.
Because of the higher S/N of the N III]—C III] region compared to the N IV]—C IV region,
we concentrate on the former. Based on line identifications in UV spectra of other supernovae
– 8 –
and other photoionization-dominated objects, we have added lines from all ions in the interval
1600–2000˚ A at the appropriate wavelengths and adjusted the relative fluxes to provide a best
fit. The lines included are He II λ1640.4, O III] λλ1660.8,1666.2, N III] λλ1746.8–1754.0,
Ne III] λ1814.6, Si III] λλ1882.7,1892.0, and C III] λλ1908.7. The relative intensities of the
individual multiplet components are calculated for Te= 15,000 K and ne= 108cm−3, values
typical of the conditions in the outer ejecta. From models of the ejecta one also expects Al III
to have a substantial fractional abundance. Because Al III λλ1854.7,1862.8 are resonance
lines, they in general have high optical depths, causing a scattering of lines blueward of Al III
(especially the N III and Ne III] lines) to longer wavelengths. We calculated this scattering
by a Monte Carlo procedure, taking the doublet nature of the lines into account. For all
lines we assume a boxy line profile determined from N II] λλ2139.68,2143.5 (see above). The
Al III scattering is assumed to take place in the same region.
To increase the S/N in the abundance analysis we again use the averaged day 1063 and
day 1399 spectrum. The result of the best-fit model is shown in Figure 4, and the individual
line fluxes relative to the total flux of the N III] multiplet are given in Table 4. Because of
blending and resonance scattering, the resulting spectrum has a complicated form. We note
that most of the feature at 1620–1680˚ A is due to He II, rather than to O III]. The latter
flux is uncertain, and we can only give a conservative upper limit of ∼ 1.0 times the N III]
flux. The best fit in Figure 4 has an O III] flux only ∼ 0.1 times the N III] flux. The peak
at ∼ 1780˚ A is caused by resonance scattering of Ne III] λ1814.6 by Al III λλ1854.7,1862.8.
The scattered Ne III] flux gives rise to some of the emission at ∼ 1880˚ A. The uncertainties
in the fluxes of the individual lines are difficult to quantify, but we estimate the N III] and
the C III] fluxes to be accurate to ±30%.
Finally, we comment on the N IV] λ1486 and C IV λλ1548.9,1550.8 lines (see Fig. 1).
While the N IV] line is clearly seen, the C IV doublet is unfortunately swamped by the
continuum, as well as affected by the interstellar absorption. We can therefore only give
a lower limit to the (N IV] λ1486)/(C IV λλ1548.9,1550.8) ratio of ∼ 1.5. This limit is
conservative, because we have assumed a lower continuum for C IV than for N IV]; it is
likely that the true ratio is considerably higher.
In Figure 5 we show the full spectral evolution of SN 1998S for the epochs where HST
observations exist, while in Figure 6 the important far-UV region is shown in more detail.
Finally, we show in Figure 7 the spectral evolution in the optical range for the nebular phase.
This complements the observations in Leonard et al. (2000) by adding the very late spectra
– 9 –
at 643 and 2148 days.
The spectrum during the first 1–2 months is mainly a continuum with superimposed
P-Cygni lines, as discussed by Lentz et al. (2001). By day 72 several strong emission lines
have emerged. The lines can be divided into narrow lines with full-width at half-maximum
(FWHM) ? 300 km s−1, and broad lines extending up to ∼ 10,000 km s−1(half-width near
zero intensity; HWZI); they are likely to originate in the circumstellar medium and in ejecta
The strongest of the broad lines are Lyα, Hα, and Mg II λλ2795.5,2802.7. In Figure 8 we
compare their profiles throughout the course of the HST observations. The flux scale is linear
and varies for the different lines and dates to facilitate a comparison. During the first two
epochs only the central absorptions of Lyα and Mg II are visible. These features are shown
in detail in high-resolution spectra published by Bowen et al. (2000), where both a Galactic
and a host-galaxy component are seen, in addition to a ∼ 350 km s−1component from the
circumstellar medium of SN 1998S. Lyα is even more affected by interstellar absorptions.
Because of the large H I column density, these lines show strong damping wings, extending
∼ 10˚ A from the line center.
On day 72 and later, both Lyα and Mg II have developed a strong blue emission com-
ponent. The shapes of the Lyα and Mg II lines are similar within the uncertainties, with
the blue wing at least a factor of 10 stronger than the red. The lines extend on day 72 to
∼ 8,000 km s−1(HWZI). In the later spectra a marginal decrease in this velocity can be
We immediately note the difference between the Hα line profile on the one hand and
Lyα and Mg II on the other hand. While the former line up to the 238–258 day spectrum
shows a fairly strong red component, this is missing for the latter. Further, in the day 72–75
spectrum the Hα line is peaked at ∼ −1,000 km s−1, and extends to zero velocity. The flux
of both Lyα and Mg II, however, drops close to zero already at ∼ 2,000 km s−1. As is seen
from the earlier spectra, one reason for this difference is likely to be interstellar Lyα and
Mg II absorption from our Galaxy and from the host galaxy. Even considering this, the low
fluxes in the red wings of Lyα and Mg II make them distinctly different from Hα.
The evolution of the shape of the Hα line has been discussed in several papers (Leonard
et al. 2000; Gerardy et al. 2000; Fassia et al. 2001; Pozzo et al. 2004). Our observations,
however, cover a longer time interval. In Figure 9 we show the temporal evolution of the
– 10 –
Hα line from day 75 to day 2148. From being symmetric during the first months, as seen
in the day 75 spectrum, the line developed a clear asymmetry by day 100, with the blue
wing considerably fainter than the red. By day 200 the line again changed character, now
with a considerably fainter red wing compared to the blue, as seen in the day 258 spectrum
(see also Fig. 4 in Gerardy et al. 2000). This evolution continued, and by day 653 the red
wing was almost gone. In the very late spectrum on day 2148 the blue peak had almost
disappeared, and instead the central peak was strong. The red part of the line is still much
weaker than the blue. The absolute level, however, is uncertain due to the difficulty of an
accurate background subtraction.
A central peak is seen in the spectra at several epochs. It is, however, not clear that
this peak is of supernova origin. The flux of this peak relative to the rest of the line is
sensitive to the subtraction of the background emission from the galaxy, and therefore to the
atmospheric seeing. In particular, the reality at the last epochs is questionable. In these the
background [N II] λλ6548.0,6583.5 lines at −686 km s−1and 960 km s−1, respectively, can
clearly be seen, which shows that there is contamination from the background. The fact that
a peak is not seen in the Mg II line, observed with HST, further strengthens this conclusion.
From Figure 9 we note the clear decrease in the velocity of the blue peak from ∼4,300–
4,700 km s−1on days 108–258 to ∼ 3,900 km s−1on days 494 and 653. The maximum
blue velocity decreases marginally from ∼ 7,300 km s−1, to ∼ 6,800 km s−1. In addition to
the dramatic change in the line profile, the large Balmer decrement is of special interest, as
noted earlier by Leonard et al. (2000), who argue that this strongly implies an origin in gas
of very high density.
By day 72 the Mg II line already showed a strong red-blue asymmetry, which persisted
throughout the duration of the remaining observations, but there is no indication of a red
peak in this line. The Lyα line, however, shows an indication of a red peak in the spectra
after day 72. Gerardy et al. (2000) propose that the fading of the red peak relative to the
blue can be explained as a result of dust formation. The fact that the red wings of Mg II
and Lyα are suppressed earlier than Hα is consistent with this interpretation, as is the dust
signature seen in the infrared (Gerardy et al. 2002). For a standard extinction curve, the
optical depths at Mg II and Lyα are related to that at Hα by τ(Mg II) = 2.4τ(Hα) and
τ(Lyα) = 4.3τ(Hα).
Other broad lines at ∼ 1288, 1344, 1380, 1790, and 1880˚ A are less obvious to identify.
We note, however, the similar profiles of these lines to that of the Mg II line. Based on the
velocity shift of the blue peak, we find that a consistent set of line profiles can be obtained if
we identify these lines as O I λλ1302.2–1306.0, O I λλ1355.6–1358.5, Si IV λλ1393.8,1402.8,
[Ne III] λ1814.6, and C III] λλ1906.7–1908.7, respectively. The presence of broad O I lines
– 11 –
in the UV is not surprising in view of the broad O I λλ7771.9–7775.4 features identified by
Fassia et al. (2001); both sets of lines probably arise as a result of recombination. We return
to the implications of the presence of these lines in §4.1.
In Figure 10 we show the day 72–75 optical and UV line profiles on a common velocity
scale. There is some indication for a broad component of C IV λλ1548.9,1550.8, but this
is marginal. For N V λλ1238.8,1242.8, N IV] λ1486, and N III] λλ1746.8–1754.0 there
is no evidence for any broad components above the noise. From the figure we note that
there is marginal evidence for a somewhat smaller extension of the blue wing of Hα (to
∼ 7,000 km s−1) compared to Mg II. The Lyα, C III], Si IV, and O I lines have velocities
consistent with the Mg II line. The [O I] λλ6300.3,6363.8 lines seem to be considerably
narrower, but their shape is affected by neighboring lines, as well as by the large separation
of the doublet components.
The [O III] λλ4958.9,5006.8 region is similarly complicated by the doublet nature, as
well as by the neighboring Hβ line. In the spectra from 75–137 days, there is a clear feature
at approximately the position of the [O III] lines (Figure 7). The peak is ∼ 2000 km s−1
redward of the rest wavelength, but this is also approximately the case for Hα. To test this
more quantitatively we have used the Hα profile to create a synthetic spectrum, including
Hβ and [O III] λλ4958.9,5006.8, adjusting the line strengths so as to give a best fit to the
4700–5200˚ A region. The result of this, however, gives a peak that is too blue, at ∼ 5040˚ A.
In addition, the [O III] λλ4958.9 component gives a too shallow red wing of the Hβ line. We
therefore conclude that the feature at ∼ 5040˚ A in the early spectra is unlikely to be due to
[O III], but is probably an Fe II line with λ ≈ 5015˚ A. In the spectra later than ∼ 300 days
there is a faint line at the [O III] wavelength, best seen in the last spectrum at 2148 days,
which is most likely from [O III]. The line profile of the [O III] line is, within the errors,
similar to that of Hα, with a heavily absorbed red component.
Other lines in the optical spectrum in Figure 7 are discussed by Gerardy et al. (2000),
Leonard et al. (2000), Fassia et al. (2001), and Pozzo et al. (2004). We only remark that in
our last two spectra on days 643–653 and 2148 we see the same lines as in the earler spectra,
in particular Hβ, He I 5876, [O I] λλ6300,6364, and [O III] 4959,5007.
3.2.2. Narrow Lines
Because of the relative lack of blending, the identification and flux measurements of the
narrow circumstellar lines are considerably simplified compared to the broad lines. This is
particularly true when we compare with the flux measurements of the far-UV lines of SN
– 12 –
1993J, where we had to use a complicated deblending procedure.
In Table 5 we give the fluxes of the narrow lines below 2300˚ A. Because of the strong
continuum and the influence of P-Cygni absorptions of several lines, the day 28 fluxes are
uncertain. The noise level of the days 238 and 485 spectra are such that the fluxes are also
very uncertain for many of the lines. In the subsequent abundance analysis we therefore
restrict ourselves to the day 72 spectrum, which has a suitable combination of high S/N and
well-defined emission lines (see Fig. 6).
In Figure 11 we show the day 72 spectrum shortward of 2000˚ A in greater detail, together
with line identifications. Based on their observed rest wavelengths in the day 72 spectrum,
1394.0˚ A and 1403.1˚ A, we identify these lines as coming from Si IV rather than O IV.
The FWHM of N IV] λ1486 is consistent with being unresolved, noting that the resolution
is ∼ 250 km s−1at this wavelength. High-resolution optical and UV observations during
the first month showed two low-velocity components (Fassia et al. 2001; Bowen et al. 2000).
While [O III] λ5006.8, for example, had a FWHM of 50–60 km s−1, the intermediate-velocity
Hα absorption component extended to ∼ 350 km s−1. Unfortunately, the limited resolution
of our spectra prevents us from relating the narrow emission lines to one or the other of these
components. The C IV λλ1548.9,1550.8 doublet is clearly resolved in the day 72 spectrum.
Also, the Si IV λλ1393.8,1402.8 and the N V λλ1238.8,1242.8 doublets are resolved in the
same spectrum. The intensity ratio (N V λ1239)/(N V λ1243) is ∼ 1.5. In the region
longward of 2000˚ A, narrow lines of [N II] λλ2139.7,2143.5 are clearly present, as well as
C II] λλ2322.7–2328.8.
For the following discussion we note that the uncertainties in the fluxes in the day 72
spectrum are 7–9% for all lines, except for the N III] lines (∼ 20%). For the other dates
the uncertainties are considerably larger. These values only include the noise, estimated in
the neighborhood of each line. In addition, there may be systematic errors originating from
the assumed level of the continuum, as well as from line blending; in contrast to SN 1993J,
however, these errors are likely to be small because of the well-defined line profiles.
As for SN 1993J, the high noise level in the 1600–1700˚ A region prevents us from giving
a reliable O III] λ1664 flux. As an upper limit to the flux we find ∼ 1.5×10−15erg s−1cm−2.
– 13 –
4.1.Origin of the Emission Lines
As has been discussed elsewhere (e.g., Chevalier & Fransson 1994, 2003), the origin of
the broad lines seen in SN 1993J and SN 1998S can be explained as a result of photoionization
by X-rays and UV emission from the radiative reverse shock propagating into the supernova
ejecta. The reprocessed radiation from the low-density, unshocked ejecta emerges mainly as
UV lines of highly ionized species like C III-IV, N III-V, and O III-VI. Because of cooling,
a dense shell is formed behind the reverse shock — that is, between the reverse shock and
the contact discontinuity. The high density in this region causes the reprocessed radiation
to emerge mostly as emission from neutral or singly ionized ions, such as H I lines, Mg II,
and Fe II.
While the box-like line profiles in SN 1993J (Fig. 3) are consistent with those coming
from a narrow, spherical shell, the geometry of the line-emitting region in SN 1998S is less
clear. Gerardy et al. (2000) interpret the double-peaked Hα line profile as a result of the
interaction of the ejecta with a circumstellar ring (see also Chugai & Danziger 1994). The
centrally peaked profile is explained by a separate component — a spherical distribution of
shocked circumstellar clouds. Similar ideas are proposed by Leonard et al. (2000), who relate
this picture to a binary scenario, reminiscent of that discussed for SN 1987A.
Although there is no quantitative discussion of this hypothesis, some features in this
scenario may be understood from the general picture in Chevalier & Fransson (1994). The
broad wings would then represent the emission from the ejecta, while the narrow component
is from the shock propagating into the ring, and possibly also the reverse shock, in case
this is radiative. Most of the UV and optical line emission is then produced as a result of
photoionization of the ejecta, as well as the ring and the cool postshock gas. The ingoing
fraction will mainly be absorbed by ejecta gas close to the reverse shock, explaining the
ring-like emission also from the ejecta component. The density distribution of the ejecta is
likely to be more uniform with respect to polar angle, possibly resembling the non-spherical
models of Blondin, Lundqvist, & Chevalier (1996). We return to this scenario in the next
In SN 1995N (Fransson et al. 2002), there was a clear distinction in the line widths
of the hydrogen and Mg II lines compared with those of the high-ionization UV metal
lines. While the former had smoothly declining line profiles with velocities (HWZI) up
to ∼ 10,000 km s−1, the latter were box-shaped with HWZI ≈ 5,000 km s−1. No such
distinction is obvious either for SN 1993J or SN 1998S.
– 14 –
The strong O I recombination lines present in SN 1998S resemble those seen in SN
1995N. In Fransson et al. (2002), it was argued that in order to explain their strengths
relative to the hydrogen lines, they had to originate in processed gas with a high abundance
of oxygen. We believe the same is true for SN 1998S as well, which indicates that some of
the gas giving rise to the broad metal lines comes from processed gas. SN 1993J also has
strong metal lines in the UV, as well as in the optical, which grow stronger with time relative
to Hα (Matheson et al. 2000b). Hence, there cannot be a large mass of hydrogen present
between the reverse shock and the oxygen-rich core. It is likely that SN 1993J, SN 1995N,
and SN 1998S had lost most of their hydrogen-rich gas, and the helium and oxygen core is
therefore efficiently illuminated by the X-rays from the reverse shock.
The origin of the relatively narrow lines seen in SN 1998S, and in other SNe IIn, is less
clear. Given their small velocity width, they must arise in circumstellar gas. This can either
be unshocked gas in front of the shock, or dense regions shocked by the blast wave.
4.2.Asymmetric Hα Emission and Dust Formation in SN 1998S
As mentioned in §1, there are several indicators of dust present in SN 1998S, both from
the line profiles and from the strong infrared excess (Gerardy et al. 2002; Pozzo et al. 2004).
In addition, the double-peaked line profiles provide a very interesting clue to the nature of
the line-emitting region. Gerardy et al. (2000), in particular, attribute this to emission by a
torus-like region. In this section we discuss the dust and the geometry of the line-emitting
The location and nature of the dust is not clear. An obvious candidate is dust formed
in the heavy-element enriched ejecta. If most of the line emission comes from a shell close
to the reverse shock, the dust in the core will primarily absorb the high-velocity emission
from the receding ejecta, in agreement with the observations. This is the scenario favored by
Gerardy et al. (2000) for the asymmetric line profiles. Gerardy et al. (2002), however, find
that the same dust cannot explain the excess emission seen in the infrared. Instead, they
attribute this to an echo from pre-existing dust.
Another interesting possibility is that the dust absorption, as well as the emission, takes
place in the cool dense shell (hereafter CDS) in the shock region (Deneault, Clayton, &
Heger 2003). This is a favorable place for dust formation because of its high density and low
temperature (Fransson 1984b; Chevalier & Fransson 1994). Whether it is cool enough, given
the X-ray heating from both shocks, remains to be demonstrated quantitatively. The fact
that the gas swept up by the shocks may be enriched by heavy elements may be important
– 15 –
in this context. Indications for metal enrichment in SN 1998S are also presented from X-ray
observations by Pooley et al. (2002). The temperature of the emission should be close to the
condensation temperature, which agrees with the Gerardy et al. (2002) temperature estimate
of the dust emission, 700–1400 K. Recently, Pozzo et al. (2004) find that the luminosity and
temperature of the IR emission gives a blackbody radius corresponding to a velocity of
∼ 4,000 km s−1at 330–400 days. This is comparable to the velocity of the blue and red
peaks of the Hα line at this time. Later, the blackbody velocity of the dust receeds to
∼ 1,000 km s−1, indicating a thinning of the dust.
A problem with this idea is that the dust is likely to condense behind the zone relative
to the reverse shock where Hα, Mg II, and other low-ionization lines are formed. If the
CDS is formed behind the reverse shock (in the Lagrangian sense), this means that both the
front and back side line emission will be absorbed by the dust, contrary to the observations.
A solution to this may come from the fact that hydrodynamical mixing may give rise to a
clumpy CDS (Chevalier & Blondin 1995). This could lead to a covering factor of the dust
less than unity, and allow some of the Hα to penetrate.
An alternative, related possibility, discussed for SN 1993J by Fransson, Lundqvist, &
Chevalier (1996), is that the forward shock is radiative. In this case the line emission from
the receeding part of the shell (red wing) will be absorbed by the dust shell, while the line
emission from the front (blue wing) can emerge freely. Although dismissed for SN 1993J, a
high circumstellar density, as may be the case for SN 1998S, may make this a real possibility.
Alternatively, there may be a combination of ejecta dust absorption and emission from dust
created in the CDS. The latter then has to be either clumpy or optically thin, in order to
not block the line emission from the shell.
Finally, we note that emission from perhaps 2–4 M⊙of cold dust from Cas A has recently
been claimed by Dunne et al. (2003). This estimate of the amount of dust, however, has been
questioned by Dwek (2004), who finds that a more reasonable mass of only 10−4−10−3M⊙
is needed, if the dust is in the form of conducting, metallic needles. This illustrates the
sensitivity of the mass estimate to the nature of the dust. But the most interesting result of
Dunne et al. (2003) is that the dust emission in Cas A is concentrated in the region between
the reverse and forward shocks. This is consistent with the scenario above where the dust is
formed in the CDS during the first years after the explosion.
To study the different scenarios for the dust absorption above in more detail, as well
as the effects of an asymmetric Hα emission, we have calculated several models for the line
profiles. In particular, we have considered the different possibilities discussed above for the
location of the dust-absorbing region.