BACK to the
The Typing Machine - A brief
Current supernova (SN) classification methods rely almost solely on the analysis of SN spectra
to determine their type. However, spectroscopy may not be possible or practical when SNe are faint, very
numerous, or discovered in archival studies. We present a classification method for SNe based on the
comparison of their observed colors with synthetic ones, calculated from a large database of multi-epoch
optical spectra of nearby events.
Classification of SNe using broad-band colors is a complex problem, and a
full solution (i.e, deriving SN
type, redshift, and age from a few broad-band colors) is probably impossible. The main concept of our
approach to the SN classification problem is that it may be simplified. Type, redshift, and age need not all be
determined simultaneously. Most SNe are associated with host galaxies that are readily detectable. One can
therefore infer the redshift of the SN from that of the host, derived either from spectroscopy (that may be
obtained long after the SN has faded) or using a photometric redshift. For the majority of SNe, the redshift can
be treated as a known parameter. Since for most practical uses, the SN
age is immaterial, we concentrate our
efforts on using multicolor broad-band photometry to find the SN types.
In order to use multicolor broad-band photometry to constrain the type of a SN with an arbitrary
(but known) redshift, we have compiled a large spectral database of nearby,
well-observed SNe, most of which were obtained at the Lick Observatory by Alexei V.
Filippenko and his collaborators.
These spectra are used to calculate synthetic
broad-band colors for the various SN types
through a given filter set at a given redshift. The temporal coverage of our compilation allows us to draw paths
in color space which show the time evolution of each SN type. Inspecting the resulting diagrams, one can then
look for regions which are either populated by a single type of SN, or that are avoided by various SN types. The
observed photometric colors of a candidate SN with a known redshift can then be plotted on the relevant
diagrams, and constraints on its type and age may be drawn. The type of a SN can sometimes be uniquely
determined, but even when this is not the case, the type may still be deduced by supplementing the color
information with other available data on the SN, such as limits on its brightness, and information (even if very
limited) on its variability (e.g., whether its flux is rising or declining).
Our SN classification method is based on colors, determined by the spectral energy distribution of each event.
By definition, such a method can differentiate only between SN subtypes with unique spectral characteristics.
SNe are also a veritable zoo, with many peculiar and even unique events, from luminous hypernovae to faint SN
1987A-like events and SN 'impostors'. Currently, we limit our discussion to SN types that are not extremely rare.
Assuming that the SNe reported in the IAU Circulars are representative of the SN population that is discovered
by current programs, we estimate that collectively, the well-defined subtypes we consider (Ia, Ib, Ic, II-P, IIn,
and IIb) constitute at least 85%, and probably more, of the entire population. SNe II-L have not been well
characterized spectroscopically, as the best spectroscopically studied events (SNe 1979C and 1980K) are
considered to be photometrically peculiar ("overluminous"; Miller & Branch 1990) relative to other SNe II-L. We
therefore consider only SNe II-P, IIn, and IIb as subtypes of the SN II population.
All spectra were deredshifted according to the host-galaxy redshift taken from the NED database, and telluric
lines were removed. The effects of atmospheric absorption (including telluric features) are
accounted for after the spectra are shifted to a chosen redshift.
According to the Galactic coordinates of the observed SN, the spectra were
corrected for Galactic reddening, as given by Schlegel et al. (1998), using the extinction curve of Cardelli,
Clayton, & Mathis (1989). All spectra were either obtained at the parallactic angle, or were calibrated by
simultaneous photometry, to ensure that spectral shape distortions due to wavelength dependent atmospheric
refraction will not be significant.
Using our spectral database, we calculate the synthetic colors of SNe at a given redshifts through the
Johnson-Cousins UBVRI(JHK) and SDSS ugriz filter systems. If a filter's bandpass is not fully covered by a SN
spectrum, the flux in the missing spectral region is extrapolated linearly using the median value of the
spectrum. Each calculated magnitude which includes such an extrapolation is assigned an error that is equal to
the amount of flux in the extrapolated part of the spectrum. The resulting synthetic photometry is then
displayed on color-color diagrams. Note that the choice of colors that gives the best SN type differentiation,
apart from the obvious dependence on the filter system used, also depends on redshift. For optimal results,
for each SN (having a known redshift, and some observed colors) one needs to search for color-color
diagrams that give the maximum information content.
Broad-band photometry through standard Johnson-Cousins UBVRI filters can be useful to classify SNe up to
z ~ 0.6. At higher redshifts, the V band samples the restframe UV, not covered by our spectral database. Thus,
only one color (R-I) remains in the Johnson-Cousins system. The use of Sloan Digital Sky Survey (SDSS) ugriz
filters allows extending our classification method to higher redshifts (z = 0.75), and the use of infrared bands,
to z = 2.5.
For a more complete description see Poznanski,
Gal-Yam et al.
Comments & questions will be most appreciated.
BACK to the