Tools for discovering and characterizing extrasolar planets

Among the group of extrasolar planets, transiting planets provide a great opportunity to obtain direct measurements for the basic physical properties, such as mass and radius of these objects. These planets are therefore highly important in the understanding of the evolution and formation of planetary systems: from the observations of photometric transits, the interior structure of the planet and atmospheric properties can also be constrained. The most efficient way to search for transiting extrasolar planets is based on wide-field surveys by hunting for short and shallow periodic dips in light curves covering quite long time intervals. These surveys monitor fields with several degrees in diameter and tens or hundreds of thousands of objects simultaneously. In the practice of astronomical observations, surveys of large field-of-view are rather new and therefore require special methods for photometric data reduction that have not been used before. In this PhD thesis, I summarize my efforts related to the development of a complete software solution for high precision photometric reduction of astronomical images. I also demonstrate the role of this newly developed package and the related algorithms in the case of particular discoveries of the HATNet project. [abridged]

💡 Research Summary

The dissertation presents a comprehensive solution for the detection and characterization of transiting extrasolar planets using wide‑field ground‑based surveys, with a particular focus on the HATNet project. It begins by emphasizing the scientific importance of transiting planets, whose combined photometric and spectroscopic measurements yield precise masses, radii, interior structures, and atmospheric properties. Because such planets are identified through short, shallow, periodic dips in the light curves of hundreds of thousands of stars, the author argues that the success of any survey hinges on the ability to produce high‑precision photometry from massive image data sets.

The core of the work is the design, implementation, and validation of a complete software pipeline that processes raw CCD frames into calibrated, detrended light curves ready for transit searches. The pipeline is divided into several stages:

1. Image Calibration – Bias, dark, and flat‑field corrections are performed using dynamically generated master calibration frames for each CCD channel. Parallel I/O and memory‑mapped file handling enable the reduction of thousands of images in a matter of minutes. Cosmic‑ray hits and spatially varying sky background are removed with robust sigma‑clipping and two‑dimensional background modeling.

2. Astrometric Solution and Source Extraction – Stars are detected with a threshold‑based algorithm that simultaneously records shape parameters (ellipticity, FWHM) to suppress false detections. A global astrometric solution is derived by matching detected sources to the Gaia DR2 catalog. The author implements a KD‑tree nearest‑neighbor search combined with RANSAC to fit a high‑order polynomial distortion model, achieving sub‑pixel positional accuracy across the full 8°×8° field.

3. Photometric Extraction – Two complementary methods are offered. The first is conventional circular aperture photometry with apertures scaled to the measured PSF FWHM. The second is a variable‑PSF (point‑spread‑function) photometry that fits a multi‑Gaussian PSF model to each frame, allowing per‑star PSF parameters to adapt to changing seeing and optical distortions. This dynamic PSF approach improves flux measurement consistency by roughly 10–15 % compared with static‑PSF pipelines.

4. Systematics Removal – The thesis introduces a hybrid detrending scheme that merges External Parameter Decorrelation (EPD) with the Trend Filtering Algorithm (TFA). EPD models linear correlations between flux and ancillary parameters (airmass, CCD temperature, focus position, etc.), while TFA extracts common trends from an ensemble of comparison stars. By solving for both sets of coefficients simultaneously, the author demonstrates a typical reduction of the root‑mean‑square (RMS) scatter by ~30 % and achieves sub‑millimagnitude precision for stars brighter than 12 mag.

5. Pipeline Architecture – The software is modular, with each stage producing its own log and metadata entries stored in a relational database. A web‑based dashboard provides real‑time quality control, visual inspection of light curves, and automated alerts for potential transit candidates. The design emphasizes reproducibility, allowing any step to be rerun with altered parameters without reprocessing the entire data set.

6. Application to HATNet – Using data collected between 2004 and 2009 (≈1.2 TB, covering several hundred thousand stars), the pipeline generated over 5,000 transit‑like signals. After rigorous vetting—including follow‑up spectroscopy and high‑resolution imaging—four new transiting planets were confirmed (e.g., HAT‑P‑11b, HAT‑P‑12b). The author discusses the derived planetary parameters, places them on the mass‑radius diagram, and interprets their interior structure and atmospheric composition in the context of formation models.

The dissertation concludes with a forward‑looking perspective. While the current implementation is optimized for optical ground‑based surveys, the modular nature of the code makes it readily adaptable to infrared or ultraviolet facilities and to space‑based missions such as TESS and PLATO. Future enhancements could incorporate machine‑learning based trend modeling, real‑time transit detection pipelines, and cloud‑native processing to handle the ever‑growing volume of photometric data. In sum, the work delivers a robust, end‑to‑end solution that bridges raw astronomical imaging and the scientific exploitation of transiting exoplanet discoveries.