Home
[Research]
Publications
Software
Graphics
Talks
Links
Contact
My research focuses on the spectra of supernovae in general, and on the study of evolutionary effects in thermonuclear supernovae ("Type Ia", or SN Ia) in particular. I have extensively studied these objects as part of the ESSENCE collaboration.
My research interests include modelling supernova spectra, as well as developing software tools for spectral extraction and cross-correlation.
Please follow the links below for more details on recent and ongoing work:
A tool to determine the type, redshift, and age of SN
spectra
Line profiles in nearby and distant SN Ia
2D Richardson-Lucy restoration of SN spectra
Together with John Tonry I developed an algorithm to identify the type of a supernova spectrum, and determine its redshift and age. This algorithm, based on the correlation techniques of Tonry & Davis, is implemented in the SuperNova IDentification code (SNID). It is used by members of ongoing high-redshift supernova searches to distinguish between type-Ia and type-Ib/c SNe, and to identify "peculiar" SNe Ia. We show that accurate redshifts (with a typical error ≤ 0.01) can be determined for SNe Ia for which a spectrum of the host galaxy is unavailable. Last, the age of an input spectrum is determined with a typical accuracy ≤ 3 days. The SNID code, which is available to the community, can also be used for comparative studies of supernova spectra, as well as comparisons between data and associated models.
The input spectrum is cross-correlated with a library of template supernova spectra of all types (CfA Supernova Archive; SUSPECT database), covering a broad range of ages. For each cross-correlation, we compute a correlation parameter, rlap = r x lap, where r is defined in Figure 1 and lap is the overlap in wavelength between the input spectrum and the template spectrum at the correlation redshift. The higher the rlap, the better the correlation.
| Figure 1: The correlation r-value is the ratio of the height of the correlation peak to the RMS of the antisymmetric component of the correlation function. (from Blondin et al. 2007) |
In Figure 2 we compare the redshift determined using SNID with the spectroscopic redshift of the host galaxy for a sample of 57 high-redshift SN Ia spectra from the ESSENCE survey (Matheson et al. 2005; Foley et al. in prep.; Blondin et al. in prep.). The agreement between the supernova (SNID) redshift and that of the host galaxy is excellent out to z = 0.8, with a dispersion of only ~0.005.
| Figure 2: Comparison of supernova and galaxy redshifts. (from Blondin et al. 2007) |
In Figure 3 we compare the (rest frame) spectral age determined using SNID and the (observer frame) age determined from the light curve, for a sample of 54 high-redshift SN Ia spectra from the ESSENCE survey (Miknaitis et al. 2007) in the redshift range 0.164 ≤ z ≤ 0.587. The light-curve age has been corrected for the (1+z) time-dilation factor expected in an expanding universe.
| Figure 3: Comparison of supernova and light-curve ages. (from Blondin et al. 2007) |
In Figure 4 we illustrate the use of SNID in distinguishing a 91T-like SN Ia around maximum light (upper panel) from other SNe Ia at z=0.5. We show the fraction of templates in the SNID database that correlate with the input spectrum, as a function of the rlap parameter: 91T-like SNe Ia (solid blue line); other SNe Ia (dashed green line); supernovae of other types ( dotted red line). From left to right, we show the effect of having no prior on either age or redshift; a flat +/- 0.01 prior on redshift; a flat +/- 3 day prior on the age; a combined prior on both redshift and age.
At these ages, the spectroscopic differences between 91T-like and other SNe Ia are most apparent, and the distinction is possible for good correlations. At later ages (lower panel), the differences between the different Ia subtypes are less obvious (see the rightmost panel), and the impact on the ability for SNID to distinguish between the different Ia subtypes is severe without a prior on redshift or age.
| Figure 4: Identification of a 91T-like SN Ia at z = 0.5 around maximum light (upper panel) and ~1-2 weeks past maximum light (lower panel). (from Blondin et al. 2007) |
Figure 5 is similar to Figure 4, except now the input spectrum is a SN Ib/c around maximum light. At z = 0.3 (upper panel), the fraction of SN Ib/c templates that correlate with the input spectrum exceeds that of SNe Ia (and other SN types) for rlap ≥ 5. At z = 0.5, the result is essentially unchanged.
| Figure 5: Identification of a SN Ib/c around maximum light at z = 0.3 (upper panel) and z = 0.5 (lower panel). (from Blondin et al. 2007) |
Despite their lower intrinsic luminosity, SNe Ib/c remain potential contaminants in high-redshift SN Ia surveys. Identifying them as such is necessary to prevent a bias in cosmological parameters derived from calibrated SN Ia magnitudes.
This semi-quantitative study uses chactacteristic velocities associated with line profiles in SN Ia optical spectra (P-Cygni shape) to evaluate the degree to which high-z SN Ia are similar/different to their local counterparts. We find no differences in the measurements we present here, namely the velocity at maximum absorption (vabs; see Figure 1 below) and at peak emission (vpeak) for various spectral lines characteristic of SN Ia optical spectra near maximum light.
| Figure 1: Upper panel: Absorption velocities for Ca II λ3945 in local SN Ia, for three different dm15 ranges. If a double-absorption is present, only the redder component is plotted. Lower panel: Same for the high-redshift data. (from Blondin et al. 2006) |
We also find that the two S II λλ5454,5640 features could be used to discriminate between fast- and slow-decliners past maximum light (see Figure 2 below). Both the post-maximum vabs values and the vabs gradient with the supernova phase are clearly different for SN Ia with dm15 > 1.7.
| Figure 2: Upper panel: Absorption velocities for S II λ5454 (left) and S II λ5640 (right) in local SN Ia, for three different dm15 ranges. Lower panel: Same for the high-redshift data. (from Blondin et al. 2006) |
We use the CMFGEN radiative transfer code of Hillier & Miller (1998) to illustrate line formation mechanisms in an expanding medium with steep density gradients, as is the case in SN Ia ejecta (see Figure 3 below). The code has been adapted to core-collapse (Type II) supernova conditions by Dessart & Hillier (2005).
| Figure 3: P-Cygni profiles of Ca II λ3945, S II λ5432, and Si II λ6347 in a CMFGEN model of a low-luminosity SN Ia near maximum brightness, with density exponent n = 7. Lower panels: Grayscale image of the quantity p I(p) as a function of p and classical Doppler velocity v, where p is the impact parameter and I(p) the emergent specific intensity along p (at v). The overplotted thick black curve gives the line-of-sight velocity location where the integrated continuum optical depth equals 2/3. The dotted white lines are for p = [1,1.8]. Upper panels: (solid curve) Line profile flux obtained by summing p I(p) over the range of p; (broken curves) velocity profile of p I(p) for two p-rays, at p = [1,1.8]. The vertical dotted line corresponds to the (continuum) photospheric velocity. (from Blondin et al. 2006) |
In Blondin et al. (2005) we use a new technique to extract the spectrum of a supernova from that of the contaminating background of its host galaxy, and apply it to the specific case of high-redshift Type Ia supernova (SN Ia) spectroscopy. The algorithm is based on a two-channel iterative technique employing the Richardson-Lucy restoration method and is implemented in the IRAF code specinholucy. We run the code both on simulated (SN Ia at z = 0.5 embedded in a bright host galaxy) and observed (SNe Ia at various phases up to z = 0.236) data taken with VLT+FORS1 and show the advantages of using such a deconvolution technique in comparison with less elaborate methods. This paper is motivated by the need for optimal supernova spectroscopic data reduction in order to make meaningful comparisons between the low and high-redshift SN Ia samples.
The restoration proceeds in two channels: one for the point source (i.e. the supernova), the other for the (extended) background source (i.e. the galaxy). Figure 1 we illustrate how well the restoration separates the SN trace from the underlying galaxy. δFtot refers to the mean total flux residuals. δFtot > 1 indicates the restoration has failed, while δFtot ≤ 1 shows the restoration is successful. For the case of SN 2002bo a secondary point source (indicated by an arrow) in its vicinity is successfully restored in the point source channel.
| Figure 1: Upper panel: Normalised average spatial profiles of the input (dashed line) and restored (solid line) 2D background spectra. Two runs of specinholucy were executed, one including the point source in the restored 2D background spectrum and the other excluding it, so as to appreciate how well the background underneath the SN was fit. Lower panel: Wavelength-averaged spatial residuals in units of the statistical noise of the input 2D spectrum. For all cases we have δFtot ≤ 1, meaning the combined SN+background spatial profile is restored to the statistical noise limit over the whole spatial range. (from Blondin et al. 2005) |
Figure 2 on the other hand shows the flux residuals in the dispersion direction, and one can immediately appreciate the successful allocation of (extended) atmospheric absorption features and sky emission lines to the background channel. Moreover one can immediately pick out the spectral regions affected by systematic errors in the restoration (δFtot > 1) and only use those where δFtot ≤ 1 for analysis.
| Figure 2: Upper panel: Normalised restored point source spectra (solid line) and underlying background, both including (dashed line) and excluding (dotted line) the point source. The spectra have been normalised to the integral of the underlying background flux. The circled + symbol denotes the atmospheric A-band (~7600-7630 Angstroms), whilst arrows indicate strong sky emission lines. Lower panel: Spectral residuals in units of the statistical noise of the input 2D spectrum. (from Blondin et al. 2005) |
We choose to compare our method with alternative techniques, namely a standard extraction using IRAF, an iterative Gaussian extractor (gaussextract) and a statistical decomposition of a 1D flux-calibrated spectrum into SN and galaxy contributions (SN-fit, Sainton 2004). The results are summarised in Figure 3 below.
| Figure 3: Upper panel: Restored supernova spectra using specinholucy (solid line) and other methods. The spectra are normalised to the maximum flux value of the specinholucy output. Due to the specificities of SN-fit - namely the restriction of the local SN template spectra to epochs -20 d < t < +20 d, we restrict its application to SNe 2002gr and 2002go (we plot the solution corresponding to the smallest chi2, for which we have not plotted the gaussextract output for the sake of clarity. The circled + symbol and arrows have the same meaning as in Figure 2. Lower panel: Ratios of the various spectra to the specinholucy output. We have deliberately restricted the y-axis range to [0,2] to be able to visualise small ~20% differences in the SN flux. (from Blondin et al. 2005) |
Last updated on: 27 September 2007