A needle in a haystack: Towards the best strategy to find the first stars

Data: 
08/05/2013 - 14:00 - 15:00
Local: 
Auditório 1 do IAG

A needle in a haystack: Towards the best strategy to find the first stars

Rafael S. de Souza (KASI, Korea)
 
During the talk, I will give a brief review about the possibility to observe the first stars by looking their final fate as gamma-ray bursts and  supernova (de Souza, R. et al. 2011, 2012). In special, I will discuss one of my recent projects explained bellow. 
 
With the next generation of optical and near-infrared (NIR) surveys, the possibility to observe primordial stars becomes real. A fraction of these first   stars, with masses between ~ 140 and 260 solar mass, are expected to die as pair-instability supernovae (PISN). Using  the state of art of cosmological simulations, we  infer the predicted rate of PISN as a function of redshift. From   radiation hydrodynamical simulations, we are capable to model the spectral energy distribution of PISN in realistic circumstellar environments with Lyman absorption by the neutral intergalactic medium.  We then constructed a synthetic survey  of supernovae  light curves as they should be observed by James Webb Space Telescope (JWST), including all important characteristic of the experiment, as exposure time, filter transitivity, dust map of Milky Way and so on.  We are capable to provide a sample of  FITS file that can be treated as the same way as  a real data. Future supernova surveys are expected to observe much more supernovae than can be confirmed spectroscopically. Given the rare nature of these primordial events, it's imperative to rely in some good photometric predictor, in order to find suitable candidates for posterior spectroscopically confirmation. Thus,  using our synthetic sample, we perform a comprehensive study of the best strategy capable to find these objects,  estimate the redshift and classify  them photometrically. To do so, we are using a combination of the  best learning methods available, in special kernel principal components analysis (KPCA),   gaussian process and genetic algorithms. Our  methodology based in KPCA combined with k-nearest neighborhood already proved to have a higher accuracy only relying in photometric information in the problem of Type Ia identification (Ishida, E.   and de Souza, R, MNRAS 2013). We  are developing the most realistic and complex framework, from far, to study the best way to find the first supernovae. Being one of the most promising approaches to drive the future searches of these objects.