Context. Mergers are an important aspect of galaxy formation and evolution. With large upcoming surveys, such as Euclid and LSST, accurate techniques that are fast and efficient are needed to identify galaxy mergers for further study. Aims. We aim to test whether deep learning techniques can be used to reproduce visual classification of observations, physical classification of simulations and highlight any differences between these two classifications. With one of the main difficulties of merger studies being the lack of a truth sample, we can use our method to test biases in visually identified merger catalogues. Methods. A convolutional neural network architecture was developed and trained in two ways: one with observations from SDSS and one with simulated galaxies from EAGLE, processed to mimic the SDSS observations. The SDSS images were also classified by the simulation trained network and the EAGLE images classified by the observation trained network.Results. The observationally trained network achieves an accuracy of 91.5% while the simulation trained network achieves 65.2% on the visually classified SDSS and physically classified EAGLE images respectively. Classifying the SDSS images with the simulation trained network was less successful, only achieving an accuracy of 64.6%, while classifying the EAGLE images with the observation network was very poor, achieving an accuracy of only 53.0% with preferential assignment to the non-merger classification. This suggests that most of the simulated mergers do not have conspicuous merger features and visually identified merger catalogues from observations are incomplete and biased towards certain merger types. Conclusions. The networks trained and tested with the same data perform the best, with observations performing better than simulations, a result of the observational sample being biased towards conspicuous mergers. Classifying SDSS observations with the simulation trained network has proven to work, providing tantalizing prospects for using simulation trained networks for galaxy identification in large surveys.

Aims. We aim to study the statistical properties of dusty star-forming galaxies across cosmic time, such as their number counts, luminosity functions (LF) and dust-obscured star-formation rate density (SFRD). Methods. We use state-of-the-art de-blended Herschel catalogue in the COSMOS field to measure the number counts and LFs at far-infrared (FIR) and sub-millimetre (sub-mm) wavelengths. The de-blended catalogue is generated by combining the probabilistic Bayesian source extraction tool XID+ and informative prior on the spectral energy distributions derived from the associated deep multi-wavelength photometric data. We compare our results with previous measurements and predictions from a range of empirical models and physically-motivated simulations.Results. Thanks to our de-confusion technique and the wealth of deep multi-wavelength photometric information in COSMOS, we are able to achieve more accurate measurements of the number counts and LFs while at the same time probing roughly ten times below the Herschel confusion limit. Our number counts at 250 µm agree well with previous Herschel studies. However, our counts at 350 and 500 µm are considerably below previous Herschel results. This is due to previous Herschel studies suffering from source confusion and blending issues which is progressively worse towards longer wavelength. Our number counts at 450 and 870 µm show excellent agreement with previous determinations derived from single dish observations with SCUBA-2 on the JCMT and interferometric observations with ALMA and SMA. Our measurements of both the monochromatic LF at 250 µm and the total IR LF agree well with previous results in the overlapping redshift and luminosity range. The increased dynamic range of our measurements allows us to better measure the faint-end slope of the LF and measure the dust-obscured SFRD out to z ∼ 6. We find that the fraction of dust obscured star-formation activity is at its highest (> 80%) around z ∼ 1 which then decreases towards both low and high redshift.We do not find a shift of balance between z ∼ 3 and z ∼ 4 in the cosmic star-formation history from being dominated by unobscured star formation at higher redshift to obscured star formation at lower redshift. However, we do find the redshift range 3 < z < 4 to be an interesting transition period as the fraction of the total SFRD that is obscured by dust is significantly lower at higher redshifts.Article number, page 1 of 21 Article number, page 3 of 21

IR spectroscopy in the range 12-230 µm with the SPace IR telescope for Cosmology and Astrophysics (SPICA) will reveal the physical processes that govern the formation and evolution of galaxies and black holes through cosmic time, bridging the gap between the James Webb Space Telescope (JWST) and the new generation of Extremely Large Telescopes (ELTs) at shorter wavelengths and the Atacama Large Millimeter Array (ALMA) at longer wavelengths. SPICA, with its 2.5-m telescope actively-cooled to below 8 K, will obtain the first spectroscopic determination, in the mid-IR rest-frame, of both the star-formation rate and black hole accretion rate histories of galaxies, reaching lookback times of 12 Gyr, for large statistically significant samples. Densities, temperatures, radiation fields and gas-phase metallicities will be measured in dust-obscured galaxies and active galactic nuclei (AGN), sampling a large range in mass and luminosity, from faint local dwarf galaxies to luminous quasars in the distant Universe. AGN and starburst feedback and feeding mechanisms in distant galaxies will be uncovered through detailed measurements of molecular and atomic line profiles. SPICA's large-area deep spectrophotometric surveys will provide mid-IR spectra and continuum fluxes for unbiased samples of tens of thousands of galaxies, out to redshifts of z∼6. Furthermore, SPICA spectroscopy will uncover the most luminous galaxies in the first few hundred million years of the Universe, through their characteristic dust and molecular hydrogen features.