A Search for Transiting Exoplanets

APT The Automated Patrol Telescope at Siding Spring Observatory, NSW, Australia.

Over 270 planets are now known to be orbiting stars other than the Sun. Most of these planets were detected using the radial velocity method, which involves measuring (using spectroscopy) the "wobble" of the star due to the planet's gravitational pull. An alternative used by an increasing number of planet searches is the transit method (see figure below). This involves measuring the small decrease in apparent brightness (of order 1%) of the star, as the planet eclipses (or transits) it. This method is thus only sensitive to planets with edge-on orbits. However, if such a planet is detected, its size, mass and orbital characteristics can be determined (with follow-up spectroscopy), constraining models of its structure and formation.

High-precision spectroscopic observations of a host star during a known transit may reveal something about the chemical composition of the planet's atmosphere. The first such observation was recently made by Charbonneau et al. (2001). Under highly favourable circumstances, it may even be feasible to search for the presence of atmospheric oxygen (Webb & Wormleaton, 2001). For the transit method to be effective, two main difficulties need to be overcome. Firstly, the probability that a given star hosts a planet in an edge-on orbit and the transit occurs while the star is being observed is low. Thus, a reasonable detection rate can only be achieved by continuous photometric monitoring of a large sample (tens of thousands) of stars. A wide-field automated telescope is ideal for this task. The second challenge is the high level of photometric precision required. Changes in brightness on timescales of a few hours need to be measured to better than 1% precision. This is difficult to achieve with CCD photometry on ground-based telescopes (the only feasible method for most transit searches). Above the fundamental limits set by Poisson statistics (since measurement involves counting photons in the image), systematic errors become important.

APT photometry Lightcurve of transit of the planet of HD209458, observed on 26 Jul2000 at the 0.9-m Sierra Nevada telescope (IAA, Granada) by Deeg & Garrido, taken from www.iac.es/galeria/hdeeg/.

Photometric Precision - The RMS magnitude variation versus approximate V magnitude over a night of observations (150sec exposures) using the raster-scan technique. The red dashed line shows the Poisson limit due to star and sky flux. Stars brighter than V~9 are saturated.)

The Automated Patrol Telescope

The Automated Patrol Telescope (APT) is a 0.5m telescope owned and operated by the University of New South Wales, and located at Siding Spring Observatory, Australia. The previous CCD camera images a 2x3 degree field with 9.4 arcsecond pixels. The telescope is entirely computer-controlled, with the possibility of remote or fully automated observation. A new camera is currently being installed on the APT.

An important drawback of using a wide-field telescope is that the images tend to be undersampled (as is the case with the APT). Because the sensitivity of a CCD is not uniform over the surface of a single pixel, the total brightness measured for a star is dependent on where the star falls relative to pixel boundaries. This becomes significant in undersampled images, where the light from a star is spread over only a few pixels. For the APT, this effect proved to be the limiting factor, causing photometric errors of several percent. We have developed a new observing technique specifically aimed at minimising the effects of intra-pixel variations. We systematically move the telescope during intergration, in a raster-type scan covering 1x1 or 2x2 pixels. Although this does broaden the PSF slightly, it effectively eliminates the intra-pixel variation problem. The resulting PSF is fairly flat topped with a very rapid falloff. We therefore developed an optimised aperture photometry package to process the resulting images, with co-located apertures for each object positioned to better than 1/100th of a pixel repeatability in each frame. Using this system we are obtaining differential photometric precision of ~2 mmag down to V = 11 in 150 sec exposures. The lightcurves are further processed to find and remove periodic signals from variable stars and searched for transit signals using a variant of the Gregory-Loredo algorithm optimised for transit detection.


We are seeking PhD students to work with our team. We are installing a new CCD camera and need people to work on this exciting project! If you are interested, contact Prof John Webb or Prof Michael Ashley in the People section.