Abstract
Interstellar dust affects many astronomical observations through absorption and reddening, yet this extinction is also a powerful tool for studying interstellar matter in galaxies. Three-dimensional (3D) reconstructions of dust extinction and density in the Milky Way have suffered from artefacts such as the fingers-of-god effect and negative densities, and have been limited by large computational costs. Here, we aim to overcome these issues with a novel algorithm that derives the 3D extinction density of dust in the Milky Way using a latent variable Gaussian process in combination with variational inference. Our model maintains non-negative density and hence monotonically non-decreasing extinction along all lines-of-sight, while performing the inference within a reasonable computational time. Using extinctions for hundreds of thousands of stars computed from optical and near-infrared photometry, together with distances based on Gaia parallaxes, we employ our algorithm to infer the structure of the Orion, Taurus, Perseus, and Cygnus X star-forming regions. A number of features that are superimposed in 2D extinction maps are clearly deblended in 3D dust extinction density maps. For example, we find a large filament on the edge of Orion that may host a number of star clusters. We also identify a coherent structure that may link the Taurus and Perseus regions, and we show that Cygnus X is located at 1300–1500 pc, in line with very-long-baseline interferometry measurements. We compute dust masses of the regions and find these to be slightly higher than previous estimates, likely a consequence of our input data recovering the highest column densities more effectively. By comparing our predicted extinctions to Planck data, we find that known relationships between density and dust processing, where high-extinction lines-of-sight have the most processed grains, hold up in resolved observations when density is included, and that they exist at smaller scales than previously suggested. This can be used to study the changes in size or composition of dust as they are processed in molecular clouds.
Highlights
Interstellar dust affects how we view the Universe by absorbing and scattering starlight, leading to interstellar reddening and extinction
We applied our algorithm to four well-known star-forming regions (SFRs) in the Milky Way: Orion, Cygnus X, Perseus, and Taurus
Our predicted 3D dust densities improve on previous work by recovering features in 3D density, for example filaments in Orion and condensations in Taurus, that are otherwise lost in 2D extinction maps
Summary
Interstellar dust affects how we view the Universe by absorbing and scattering starlight, leading to interstellar reddening and extinction. One of the most comprehensive works was by Schlegel et al (1998), who used IRAS and COBE far-infrared emission to produce an all-sky 2D dust extinction map This was followed by two decades of improvements fuelled by advances in machine. The authors used 2MASS, Gaia, and Pan-STARRS data to carry out important sampling on a parameter grid assuming a GP prior on the logarithm of the dust reddening density These maps have become a go-to resource for estimating interstellar extinction along lines-of-sight as they are accessible. One is the so-called fingers-of-god effect, in which the density distribution is elongated along the line-of-sight due to tangential accuracy being higher than radial accuracy It can be avoided if highaccuracy distances are available and/or if correlations between points in 3D space are incorporated explicitly rather than as individual lines-of-sight coupled together in the plane of the sky. Both packages are built upon PyTorch (Paszke et al 2019), an open-source machine learning framework
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
