Improving mid-infrared thermal background subtraction with principal component analysis

Context. Ground-based large-aperture telescopes, interferometers, and future extremely large telescopes equipped with adaptive optics (AO) systems provide angular resolution and high-contrast performance superior to space-based telescopes at thermal infrared wavelengths. Their sensitivity, however, is critically limited by the high thermal background inherent to ground-based observations in this wavelength regime. Aims. We aim to improve the subtraction quality of the thermal infrared background from ground-based observations using principal component analysis (PCA). Methods. We used data obtained with the Nulling-Optimized Mid-Infrared Camera on the Large Binocular Telescope Interferometer as a proxy for general high-sensitivity AO-assisted ground-based data. We applied both a classical background subtraction-using the mean of dedicated background observations-and a new background subtraction based on a PCA of the background observations. We compared the performances of these two methods in both high-contrast imaging and aperture photometry. Results. Compared to the classical approach for background subtraction, PCA background subtraction delivers up to two times better contrasts down to the diffraction limit of the LBT's primary aperture (i.e., 350 mas in N-band), that is, in the case of high-contrast imaging. An improvement factor between two and three was obtained over the mean background retrieval within the diffraction limit in the case of aperture photometry. Conclusions. The PCA background subtraction significantly improves the sensitivity of ground-based thermal infrared imaging observations. When apply to LBTI's nulling interferometry data, we expect the method to improve the sensitivity by a similar factor of two to three. This study paves the way to maximizing the potential of future infrared ground-based instruments and facilities, such as the future 30m-class telescopes.