Exoplanet spectroscopy – data analysis
The main focus of my research is the characterisation of exoplanetary atmospheres, through multi-wavelength observations of their transits and eclipses. Data detrending is necessary to achieve the required 10^-5–10^-4 precision in spectro-photometric measurements of the transit/eclipse depth. In the past, parametric methods have extensively been used to remove the instrumental systematics, causing many debates regarding the use of different parametric choices to the removal of systematic errors. The main reason is that the correction of the instrumental effects is usually beyond the well known calibration, as the current surveys, except Kepler, were not designed to achieve the 10^-5–10^-4 precision. In addition, stellar activity may be a significant source of astrophysical noise.During my PhD at UCL, I developed some non-parametric data detrending algorithms based on Independent Component Analysis (ICA), i.e. a blind source separation technique. As such, ICA is used to disentangle the different source signals, instrumental and astrophysical, which are mixed in the observations, without using any prior knowledge of the instrumental systematics or the astrophysical signals. Therefore, the pipelines are not specific to one instrument, and can be easily optimised for different instrument designs.

The pixel-ICA algorithm, that I have pioneered, is able to disentangle the different signals in a single photometric observation by using the individual pixel time series as input for the ICA. It has been successfully applied to Spitzer/IRAC datasets, that were previously debated in the literature, to constrain the effect of stellar variability in transit depth over multiple epochs. Based on the results of the IRAC Data Challenge 2015, organised by the Spitzer Science Center, pixel-ICA is one of the best data detrending pipelines among those currently adopted for Spitzer/IRAC datasets, and, in particular, is the most reliable and repeatable.

Recently, I developed and used the stripe-ICA algorithm to detrend spectroscopic datasets taken with HST/WFC3 in the scanning mode technique.


Stellar limb darkening and activity
Stellar limb darkening is the wavelength-dependent decrease in specific intensity from the center to the edge of the sky-projected stellar disk. Due to this effect, the fraction of stellar light blocked by a transiting planet is not exactly equal to the planet-to-star area ratio, but also depends on the location of the occulted area on the stellar disk. The accurate estimation of this effect is necessary to infer the transit depth with the 10^-5–10^-4 precision.

I have experience in using stellar modelling tools to compute ad hoc stellar atmospheric models of the exoplanet host stars and obtain the theoretical limb darkening coefficients integrated over the specific passband observed. In addition, I simulate second-order effects such as stellar spots, oblateness and gravity darkening. These effects are important at the precision level that will be achieved by JWST and other next-generation surveys.

More about empirical limb darkening (will be updated soon).


Machine learning
Starting with a 6-months IPAC Fellowship, I am working on the automated classification of astronomical objects contained in large catalogs. The specific project consists in the development of machine learning algorithms (k-NN, SVM, ANN, SOM, etc.) to select candidate Wolf-Rayet stars with the highest hit rate.