首页

Euclid's NISP instrument performs slitless spectroscopy between 1206 and 1892 nm, dispersing the light of every source in the field with no entrance slit to isolate individual objects. This makes spectral extraction difficult. The spectra of neighbouring sources overlap, extended objects produce blended and spatially broadened spectra, and the dispersed light of nearby sources contaminates the spectrum of interest. The problem is further complicated by the background properties and the spectral trace, both of which vary with position on the detector. This talk presents preliminary tools that use deep learning to extract one-dimensional spectra from two-dimensional images. We describe the training data and the network architecture, and discuss the validation required before such tools could complement a conventional extraction pipeline.

BIO

W. Gillard is a researcher and lecturer at the Centre de Physique des Particules de Marseille (CPPM; IN2P3/CNRS, Aix-Marseille Université) and is coordinator of the Renoir team (Recherche de l’Énergie NOIRe) at CPPM. He is the NISP Instrument Scientist within the Euclid Consortium and a member of the ESA Euclid Science Advisory Committee. Involved in NISP since the start of its development in 2014, he took responsibility for instrument performance and calibration, from the ground test campaigns through in-flight commissioning and the calibration campaign. His work connects detector- and instrument-level understanding to cosmological inference: from the statistical characterisation of the instrument to the measurement of large-scale structure observables, such as baryon acoustic oscillations and the power spectrum. Earlier in his career, he also worked on indirect dark matter searches with cosmic rays.

Host: Cheng Zhao