Detection of Faint Stars near Gaia Objects

The upcoming ESA space astrometry mission Gaia, planned to be launched in early 2012, has been designed to scan the whole sky (flux limited, corresponding to magnitude 20 in the Gaia passband) during a five-year observational phase. It will measure the astrometric parameters (positions, parallaxes and proper motions) of more than one billion stars, complemented by multi-epoch spectrometric and radial-velocity observations. The astrometric accuracy ranges from 10 μarcsec for bright objects to 200 μarcsec for the faintest. The science goals cover a wide range of problems in modern astronomy, with a focus on Galactic structure, kinematics and evolution, and stellar astrophysics. To achieve the science goals, it is important that possible disturbing effects on the astrometric measurements are well understood. This is necessary both in order to design the instrument so that these effects can be minimized, and to design appropriate algorithms of data analysis for detecting and correcting remaining effects. For example, the measurements of a target star may be disturbed by faint stars, apparently close to the target and therefore not normally detected, because they fall below the flux limit or are hidden behind the main target’s flux. Because of the high accuracy aimed for, even very faint stars could produce significant errors if they remain undetected and therefore not taken into account in the data analysis. In this work I investigate the detectability of faint stars by means of computer simulations of a two-dimensional image reconstruction technique. Using the actual scanning law proposed for the Gaia, one-dimensional (1D) images are first simulated for the whole fiveyear mission and for several random positions on the sky. Then, the two-dimensional (2D) images are produced by stacking the 1D images. The quality of the produced 2D images is very important for the successful detection of very faint stars. Because the pixel geometry of the images is rather peculiar, several well-known stacking algorithms (e.g., shift-and-add, sub-sampling and drizzle methods) are examined. Based on these techniques, an improved stacking algorithm is proposed. The resulting 2D images are strongly affected by artifacts from the original pixel geometry and the non-uniform distribution of scanning directions, and must be corrected by deconvolution. The CLEAN method, which is a well-known deconvolution technique in radio astronomy, presents itself as the best solution for detecting point sources in the simulated Gaia images. Using this iterative method, the positions as well as the G magnitudes of the point-source objects down to a computed threshold level are precisely obtained. Using Monte Carlo simulations, it is investigated how the detection probability depends on the apparent (angular) separation of the two stars. This probability is a function of the signal-to-noise ratio, which depends on the object’s magnitude and the number of scans across the object according to the Gaia scanning law. Minimum and maximum separation limits are therefore computed for different conditions of magnitude difference and number of scans for some random positions on the sky. I also investigate how the detection probability depends on the positional angle of the faint star with respect to the target, and the precision of the resulting position and magnitude of the faint star. Finally, an easy-to-use graphical user interface (GUI) called Gaia FSDtool is described. This is a standalone executable MATLAB code written by the author, which allows the user to choose the input parameters and immediately see the simulated images, stacking and detecting processes, as well as the final results on the detected objects. All of the output of this GUI is computed by the same MATLAB functions that were used for the investigations described in this work