The stars, like dust, encircle me In living mists of light; And all of space I seem to see In one vast burst of sight
Isaac Asimov - 1951: The Stars, like dust
Stars are the building blocks of all structures in our Universe. They are responsible for producing the more complex chemical elements, spreading them through space, igniting the formation of planets and ultimately, forming the necessary environment for the development of life. Understanding the earlier stages of stellar evolution is the first step towards a better comprehension of how our own Sun was formed and may provide important clues on which regions of our Milky Way have the potential to host planetary systems similar to our own.
Throughout history, astronomers have devoted a lot of effort to identify and carefully study regions of star formation which can help solve parts of this puzzle. However, since stars are mainly formed in very dense gas clouds, these stellar nurseries pose a very complicated observational problem and demand a time consuming analysis to be correctly cataloged. Consequently, most of the studies on young stellar populations concentrate in small regions of the Milky Way. For a long time, such small area searches were our only possibility to confront theories of stellar evolution against real observations.
Nevertheless, in order to have a general view of how the different types of stars are formed, taking into account their diverse masses, sizes, temperatures and environment, we need larger surveys.
The Spitzer Space Telescope devoted significant time to scanning large areas of our Galaxy in a hunt for young stellar objects (YSOs). Our Galaxy is shaped like a disk, with both our Sun and star-forming regions located inside the disk, meaning that most star-forming regions can be found in a thin strip that circles the sky. During an observing campaign named GLIMPSE, Spitzer took high resolution images of this strip revealing tens of millions of stars. However, this posed another very difficult question: how to find young stars among the tens of millions of objects present in such a large data set?
Telescopes like Spitzer were built by astronomers to solve important astronomical issues, but as a consequence, generated other big data problems which astronomers cannot solve alone. In order to identify the YSOs in the ocean of data, it is necessary to combine astronomical domain knowledge with modern techniques developed in the realm of statistics and computer science. In order words, interdisciplinary collaboration between scientists with different views and complementary expertise was mandatory.
In August 2019 a group of scientists gathered in the French alps, and under the shadow of the Mont Blanc, decided to approach this problem. During the COIN Residence Program #6, they got together to develop a strategy in three steps: 1) use standard methods of analysis to classify a small number of YSOs (build a training sample); 2) use this sample to train automatic machine learning algorithms and obtain automatic classification for the remaining larger part of the data and 3) carefully analyse the resulting YSOs, the groups they belong to and the environments in which they live.
As a result, they were able to construct a detailed census of stellar nurseries throughout the inner regions of the Milky Way. With the help of another space telescope, Gaia, that measures distances to objects, this census has been turned into a 3D map. The final catalog, named SPICY (Spitzer/IRAC Candidate YSO Catalog), contains more than 110,000 YSO candidates (90,000 of which had never been identified before) and is publicly available to anyone who wishes to study the first stages of stellar development.
Text by the CRP6 team
Reference: Kuhn et al., 2020 - submitted to ApJS