STAGE Pixel photometric redshifts for Euclid

In the last decade, precise determination of galaxy distances and physical properties from wide-field broad-band photometry -- photometric redshifts -- have revolutionised observational cosmology. These techniques are at the heart of every major forthcoming cosmology experiment. However, the limitations of current methods to translate imaging data to physical parameters (distances, absolute luminosities and stellar masses) is becoming increasingly apparent. In particular, in today’s era of moderately deep wide-area surveys spanning optical to infrared bands making precise photometric measurements is becoming increasingly challenging. The aim of this stage is to investigate the inference of distance and, thus, physical parameters directly from pixel data using deep inference techniques in the well-explored COSMOS field.
Objectives and expected results :
The SDSS is a large shallow survey of the local Universe. COSMOS, on the other hand, densely samples Redshift space and represents a unique database of the distant Universe. Applying inference techniques to COSMOS with its unique validation data set of spectroscopic redshifts is therefore the necessary next step. In particular, we want to investigate how well physical parameters such as stellar mass can be inferred from pixels.
The Euclid satellite will revolutionize our knowledge of the distribution of dark and luminous matter by making a survey of tens of thousands of square degrees with the largest camera ever to be launched into space. However, astronomical parameter inference will require precise knowledge of the distance to each galaxy in the main survey and for this photometric redshifts are essential. A key element of this are the Euclid deep fields , a smaller survey but one which will still cover twenty times the area of COSMOS. It is important to demonstrate the superiority of pixel-based photometric redshifts over conventional techniques by making key science measurements, such as the measurement of the high-redshift stellar mass function. Demonstrating the effectiveness of this technique in COSMOS would pave the way for future applications Euclid, which could be the subject of a future thesis.

The stage will be co-supervised by H. J. McCracken (director) and Tom Charnock. H. J. McCracken has a long experience in photometric catalogues & photometric redshifts and Tom Charnock is recognized experts on the application of advanced deep learning and statistical techniques in astronomy.