Share Email Print
cover

Proceedings Paper

Exploiting physical constraints for multi-spectral exo-planet detection
Format Member Price Non-Member Price
PDF $14.40 $18.00

Paper Abstract

We derive a physical model of the on-axis PSF for a high contrast imaging system such as GPI or SPHERE. This model is based on a multi-spectral Taylor series expansion of the diffraction pattern and predicts that the speckles should be a combination of spatial modes with deterministic chromatic magnification and weighting. We propose to remove most of the residuals by fitting this model on a set of images at multiple wavelengths and times. On simulated data, we demonstrate that our approach achieves very good speckle suppression without additional heuristic parameters.

The residual speckles1, 2 set the most serious limitation in the detection of exo-planets in high contrast coronographic images provided by instruments such as SPHERE3 at the VLT, GPI4, 5 at Gemini, or SCExAO6 at Subaru. A number of post-processing methods have been proposed to remove as much as possible of the residual speckles while preserving the signal from the planets. These methods exploit the fact that the speckles and the planetary signal have different temporal and spectral behaviors. Some methods like LOCI7 are based on angular differential imaging8 (ADI), spectral differential imaging9, 10 (SDI), or on a combination of ADI and SDI.11 Instead of working on image differences, we propose to tackle the exo-planet detection as an inverse problem where a model of the residual speckles is fit on the set of multi-spectral images and, possibly, multiple exposures. In order to reduce the number of degrees of freedom, we impose specific constraints on the spatio-spectral distribution of stellar speckles. These constraints are deduced from a multi-spectral Taylor series expansion of the diffraction pattern for an on-axis source which implies that the speckles are a combination of spatial modes with deterministic chromatic magnification and weighting. Using simulated data, the efficiency of speckle removal by fitting the proposed multi-spectral model is compared to the result of using an approximation based on the singular value decomposition of the rescaled images. We show how the difficult problem to fitting a bilinear model on the can be solved in practise. The results are promising for further developments including application to real data and joint planet detection in multi-variate data (multi-spectral and multiple exposures images).

Paper Details

Date Published: 26 July 2016
PDF: 10 pages
Proc. SPIE 9909, Adaptive Optics Systems V, 99091R (26 July 2016); doi: 10.1117/12.2233672
Show Author Affiliations
Éric Thiébaut, Univ. Lyon, Univ. Lyon 1, ENS de Lyon, CNRS, Ctr. de Recherche Astrophysique de Lyon (France)
Nicholas Devaney, National Univ. of Ireland, Galway (Ireland)
Maud Langlois, Univ. Lyon, Univ. Lyon 1, ENS de Lyon, CNRS, Ctr. de Recherche Astrophysique de Lyon (France)
Kenneth Hanley, National Univ. of Ireland, Galway (Ireland)


Published in SPIE Proceedings Vol. 9909:
Adaptive Optics Systems V
Enrico Marchetti; Laird M. Close; Jean-Pierre Véran, Editor(s)

© SPIE. Terms of Use
Back to Top