Characterizing the cold, optically-thick ISM with convolutional neural networks
Abstract
A precise observational accounting of multiphase gas and dust in the Galactic interstellar medium (ISM) is vital both for correcting observations of extragalactic light and for resolving the bottleneck between mass reservoirs and star formation in galaxies. To differentiate between distinct thermal phases of neutral hydrogen (HI), we train a convolutional neural network using synthetic spectra from numerical simulations to predict the true HI column density and the fraction of cold, optically thick HI along the line of sight from 21cm brightness temperature spectra. We validate the model using 100s of HI absorption observations from the literature, finding excellent accuracy. With this model, we construct the highest-resolution all-sky map of optically thick HI in the local ISM using large-area 21cm emission data from the GALFA-HI and the HI4Pi surveys. This map will characterize the structure of neutral gas envelopes to molecular clouds, and to improve dusty Galactic foreground estimation for extragalactic surveys. Via comparison with tracers of dust reddening (E(B-V)), we demonstrate that E(B-V)/N(HI) increases with increasing cold gas fraction, which will be leveraged to produce the the highest-resolution, highest-fidelity E(B-V) map at high latitudes to date.
- Publication:
-
American Astronomical Society Meeting Abstracts #233
- Pub Date:
- January 2019
- Bibcode:
- 2019AAS...23325209M