Abstract:
Equipped with the rich data from recent CMB missions, in particular, Planck, it has become possible to reconstruct the Primordial Power Spectrum in a model-Independent non-parametric manner allowing us to relax the theoretically motivated assumptions on the form of the power spectrum of primordial density fluctuations. This project can broadly be divided into two parts.
In the first part, we demonstrate the efficacy of the Richardson Lucy method in reconstruct-
ing the Primordial power spectrum directly from the Planck 2015 temperature data. As-
suming the best-fit values for the remaining four background parameters viz. baryonic
matter density(Ωbh2), cold dark matter density(Ωch2), optical depth(τ) and acoustic scale
parameter(θ), we are able to recover the form of PPS by applying Improved Richard-
son Lucy Deconvolution method. Our results show features that are absent in the scale-
invariant power-law form and were consistent with results from WMAP data. We analyze
the reliability of these features and the artifacts of the iRLD method by using smoothing
methods. In the latter part of this project, we relax the best-fit values for the background
parameters. We explore the 4-dimensional parameter space using a Monte-Carlo sampler
called CosmoMC while simultaneously doing the iRLD reconstruction at each point. We
obtain new distributions for the background parameters and compared them with the
distributions given by Planck collaboration where the Power law model was assumed for
the PPS. The results of our analysis indicate changes to the best-fit values of some of the
important parameters such as a higher preferred value for the value of Baryon density
and the acoustic scale parameter, while the other two parameters, the cold dark mat-
ter density and optical depth at reionization tend to prefer a value lower than expected.
Our method removes any bias introduced by the assumptions of the model for PPS and
thereby allowing us to take an important step in testing predictions of several cosmolog-
ical models against observations without extraneous theoretical bias.
