A method for extracting maximum resolution power spectra from galaxy surveys

Abstract

The power spectrum estimated from a galaxy redshift survey is the real spectrum convolved with a window function, so when estimating the power on very large scales (for small k), it is important that this window function be as narrow as possible. A method that achieves this is presented. The optimal estimate of P(k) is found to be the Fourier transform of the number density fluctuations (n/nbar - 1) weighted by a function psi_0, and the optimal psi_0 is found to be the ground-state solution of the Schoedinger equation, with the inverse selection function as the potential. This quantum mechanics analogy occurs basically because we want the weight function to be narrow both in Fourier space (to give a narrow window) and in real space (to minimize the variance from shot noise). An optimal method for averaging the estimates at different k-vectors is also presented, generalizing the standard procedure of averaging over shells in k-space. Finally, a discrete version of the method is presented, dividing space into ``fuzzy pixels", which has the advantage of being able to handle redshift distortions in a straightforward way.