1. Cape Town Reloaded 2012
Searching for Supernovae in
SDSS Galaxy Spectra
Rahman Amanullah
Roger Deane
Ariel Goobar
Michelle Knights
Aleksander Kurek
Bob Nichol
Hadi Rahmani
2. Why Search for Supernovae in SDSS?
To do cosmology we
need light curves not
spectra, so why bother
looking for supernovae
in the SDSS database?
Perlmutter et al. (1998)
3. Why Search for Supernovae in SDSS?
To do cosmology we
need light curves not
spectra, so why bother
looking for supernovae
in the SDSS database?
Type Ia supernova rates
help constrain the time
delay between
progenitor formation
and explosion. This
improves cosmological
constraints.
Perlmutter et al. (1998)
4. Supernova Rates
Ia Supernova rate as a
function of redshift.
Lines show models for
different delay times of
SNe progenitors.
Dahlen et al. (2004)
5. Supernova Rates
A photometrically
selected sample could
yield different SN rates
to a spectroscopic one.
There could be other
surprises once broken up
as a function of host
type, inclination etc.
Dahlen et al. (2004)
6. Dependence on Host Parameters
Rates are seen to depend
on star formation rate
and stellar mass. We
will look for
relationships between
galaxy properties and
SN rates.
Sullivan et al. (2006)
7. Searching for Supernovae
Broad
features
Example galaxy spectrum. Example galaxy spectrum
with supernova.
8. FFT Method to Reduce the Number of Candidates
Periodogram of spectrum
FF
T
10. FFT Method to Reduce the Number of Candidates
Epoch
FFT of SNIa Templates from Hsiao et al. (2007)
11. FFT Method to Reduce the Number of Candidates
These areas have
increased power,
relative to the rest
of the
Epoch periodogram.
FFT of SNIa Templates from Hsiao et al. (2007)
13. Template Fitting
We fit a polynomial
to the residuals of the
spectrum minus the
template to correct
for wavelength
dependent effects.
14. Template Fitting
1) Smooth the spectrum to remove galaxy emission lines (Gaussian filter).
2) Step through all epochs, fitting the template to the spectrum.
3) For each epoch:
* Scale the template appropriately.
* Subtract the template from the spectrum.
* Fit a second order polynomial to the residuals, to remove wavelength-
dependent effects.
* Calculate the χ2 using the template + polynomial as the model.
4) The minimum χ2 indicates the best fit epoch.
5) All spectra with epoch >-20 (at least some light comes from a
supernova) are candidates.
15. Mock Catalogue
To test the efficiency of our methods, we use mock catalogues. These are
generated using randomly chosen galaxy spectra from the SDSS dataset
and inserting some SNIa templates into some of them.
Example spectrum from the
mock catalogue with best fit
template.
16. Summary
Finding supernovae in the SDSS spectral database can constrain
supernova rates and give information about SN progenitors.
With a dataset of nearly one million objects, efficient techniques
must be developed to perform this search in a computationally
feasible way.
An FFT based method has been developed to cut down the
number of candidates. Other methods, such as using
supernovae identifier codes, are also being investigated.
As it is essential to know how efficient a method is before
applying it to the SDSS data, a mock catalogue has been
created.