Point Process Algorithm: A New Bayesian Approach for TPF-I Planet Signal Extraction

Abstract

TPF-I capability for planetary signal extraction, including both detection and spectral characterization, can be optimized by taking proper account of instrumental characteristics and astrophysical prior information. We have developed the Point Process Algorithm (PPA), a Bayesian technique for extracting planetary signals using the sine/cosine chopped outputs of a dual nulling interferometer. It is so-called because it represents the system being observed as a set of points in a suitably defined state space, thus providing a natural way of incorporating our prior knowledge of the compact nature of the targets of interest. It can also incorporate the spatial covariance of the exozodi as prior information which could help mitigate against false detections. Data at multiple wavelengths are used simultaneously, taking into account possible spectral variations of the planetary signals. Input parameters include the sigma of measurement noise and the a priori probability of the presence of a planet. The output can be represented as an image of the intensity distribution on the sky, optimized for the detection of point sources. Previous approaches by others to the problem of planet detection for TPF-I have relied on the potentially non-robust identification of peaks in a “dirty” image, usually a correlation map. Tests with synthetic data suggest that the PPA provides greater sensitivity to fainter sources than does the standard approach (correlation map + CLEAN), and will be a useful tool for optimizing the design of TPF-I.