Artificial Intelligence Assisted Dark Universe Science

In June 2022 the European Space Agency Euclid mission will launch. The objectives of the Euclid mission are to make a high-resolution visible wavelength map of the sky to image over three billion galaxies, and to measure near-infrared spectra for several tens of millions of galaxies. By using this data one can create 3D maps of dark matter and determine the properties of dark energy - these dark components account for 95% of the mass-energy content of the Universe and yet their nature is unknown.

One of the primary ways to determine dark energy properties from Euclid is to measure the shapes of galaxies in order to use a statistical method known as weak lensing. However, the ability to measure galaxy shapes is hampered by the quality of images from the Euclid CCDs. In particular in space CCDs are subject to an effect known as Charge Transfer Inefficiency (CTI) caused by impact of cosmic rays on the Silicon of the detectors. Accounting for and correcting the effect of CTI is required for high quality dark Universe science.

This project will use Machine Learning, deep neural network, methods to correct images for the effect of CTI. It will involve working closely with the Euclid engineering teams to understand the images and the impact of cosmic rays. It will also involve creating a machine learning model that can correct the images whilst maintaining the integrity of the science that can be inferred from the images. This project will then input any findings into an end-to-end pipeline to assess the impact on the inference of cosmological parameters on the ability to correct for CTI.