Search for: All records

High-redshift dusty star-forming galaxies with very high star formation rates (500−3000 M ⊙ yr −1 ) are key to understanding the formation of the most extreme galaxies in the early Universe. Characterising the gas reservoir of these systems can reveal the driving factor behind the high star formation. Using molecular gas tracers such as, high- J CO lines, neutral carbon lines, and the dust continuum, we can estimate the gas density and radiation field intensity in their interstellar media. In this paper, we present high resolution (∼0.4″) observations of CO(7−6), [CI](2−1), and dust continuum of three lensed galaxies from the South pole telescope – sub-millimetre galaxies (SPT-SMG) sample at z  ∼ 3 with the Atacama Large Millimetre/submillimetre Array. Our sources have high intrinsic star formation rates (> 850 M ⊙ yr −1 ) and rather short depletion timescales (< 100 Myr). Based on the L [CI](2−1) / L CO(7 − 6) and L [CI](2−1) / L IR ratios, our galaxy sample has similar radiation field intensities and gas densities compared to other submillimetre galaxies. We performed visibility-based lens modelling on these objects to reconstruct the kinematics in the source plane. We find that the cold gas masses of the sources are compatible with simple dynamical mass estimates using ULIRG-like values of the CO-H 2 conversion factor α CO , but not Milky Way-like values. We find diverse source kinematics in our sample: SPT0103−45 and SPT2147−50 are likely rotating disks, while SPT2357−51 is possibly a major merger. The analysis presented in the paper could be extended to a larger sample to determine better statistics of morphologies and interstellar medium properties of high- z dusty star-forming galaxies. 

ABSTRACT Separating galactic foreground emission from maps of the cosmic microwave background (CMB) and quantifying the uncertainty in the CMB maps due to errors in foreground separation are important for avoiding biases in scientific conclusions. Our ability to quantify such uncertainty is limited by our lack of a model for the statistical distribution of the foreground emission. Here, we use a deep convolutional generative adversarial network (DCGAN) to create an effective non-Gaussian statistical model for intensity of emission by interstellar dust. For training data we use a set of dust maps inferred from observations by the Planck satellite. A DCGAN is uniquely suited for such unsupervised learning tasks as it can learn to model a complex non-Gaussian distribution directly from examples. We then use these simulations to train a second neural network to estimate the underlying CMB signal from dust-contaminated maps. We discuss other potential uses for the trained DCGAN, and the generalization to polarized emission from both dust and synchrotron.