Correcting for background changes in CoRoT exoplanet data

Abstract

Context. The CoRoT satellite is a highly accurate photometer with 2 channels respectively optimised for asteroseismology and exoplanet finding. The design includes an effective straylight rejection system, however residual straylight reaches the detectors. Aims. We test four different background models in order to apply the best possible background correction to the 12 000 stars observed over the 2 exoplanet CCDs. We identify the best correction method for two types of data reduction – the validation and production data. We also describe a bright pixel correction method and compare background correction quality before and after this correction. Methods. We used jackknifing – a particular example of bootstrapping – to increase our statistical analysis of a small dataset, comparing the background model with a background datapoint. This enabled us to quantify the background correction quality, which would be impossible when applying the correction to star data. Results. Our examination of the in-orbit data from two CoRoT runs shows that they give very different results. The commissioning run had far from optimal background window placement. Both runs demonstrate that the closest window correction is very sensitive to the bright pixel problem. Using three windows increases the chance of including a bright pixel impacted window and does not increase performance. Both median and polynomial fit methods give a good correction in most cases, but the median is overall most efficient. Conclusions. We have shown that a median of all available background windows is the correction method most resistant to bright pixels. It also gives a good background correction for data post-bright-pixel correction. This method has been implemented in the CoRoT pipeline for both validation and production data.