Gaia low-resolution spectra of nearly 1 million giant stars within 4 kpc (~13,000 lyr) of the Sun (position indicated by ⊙) reveal the global distributions of different metals which encode the formation history of the Milky Way. The colour coding shows the stars’ iron content ([Fe/H], left) and the alpha-to-iron ratio ([α/Fe], right), a compositional property of stars that allows astronomers to tell apart different stellar populations. The Galactic disk is mostly metal-rich and α-poor, while the regions above and below the Galactic midplane become progressively more metal-poor and α-rich. Image created by Alvin Gavel, Andreas Korn, Rene Andrae and Morgan Fouesneau.

The Milky Way evolves both dynamically and chemically, fusing nuclei in thousands of generations of stars. Decades of star counts and spectroscopic follow-up have established the existence of a few major Galactic building blocks called stellar populations: the thin and thick disk, the halo and the central bulge. Gaia’s spectrophotometry records how much starlight we receive from the blue to the red for most stars observed by Gaia.

The analysis presented here shows that astronomers can use Gaia’s low-resolution spectrophotometry to assign stars to stellar populations across most of the Galaxy! Unravelling the assembly history of the Milky Way requires not only knowing the positions and motions of its stars, but also their intrinsic properties, such as their surface temperatures, luminosities, chemical compositions and ages.

Determining these parameters is the task of DPAC Coordination Unit 8 (CU8) responsible for the "Astrophysical Parameters". CU8 uses all of the data available from the Gaia space telescope – astrometry, optical spectrophotometry, near-infrared spectroscopy – to characterize the observed stars (and extragalactic objects). The DPAC Coordination Unit 5 (CU5) deals with the processing of the spectrophotometry – the low-resolution (R~80) optical spectra of stars that encode key stellar parameters like surface temperature and chemical composition. CU5 processed and delivered all the spectra on which this work is based.

In particular, CU5 recently provided spectra for a set of 200,000 stars used to train a stellar-parameter and chemical-composition model based on machine-learning algorithms. This training set was compiled from the Australian-led GALAH (GALactic Archeology with HERMES) survey providing us with known stellar parameters and compositions from their high-resolution (R~30,000) ground-based spectroscopy.

The above figures are Galactic maps showing the application of such a model to 945,384 giant stars within 4 kpc (~13,000 lyr) from the Sun, a distance corresponding to half the distance to the Galactic center. The colour coding represents the stellar iron content (left) and one additional key stellar observable, the alpha-to-iron ratio, labelled [α/Fe] (right). The latter quantity (indirectly) tells astronomers about the time scale on which a given stellar population was formed. [α/Fe] varies across the different Galactic populations. A stellar population that came into existence over a short (t < 1 Gyr) timescale will end up having high values of [α/Fe]. Conversely, a population of stars that formed over several Gyr will have lower [α/Fe] values. The Sun, the product of some 4 Gyr of thin-disk evolution, is a typical low-[α/Fe] star.

Clearly, the vertical regions in which stars of the Galactic thin disk dominate (z < 1 kpc) are α-normal (in reference to our Sun) and metal rich while the regions above and below the disk are markedly α-rich and metal poor. This is what astronomers expect from the dominance of α-rich stellar populations contributing to the Galactic thick disk and to the halo at high Galactic latitudes (z > 1 kpc) which formed early and rapidly in the history of the Milky Way while the metal rich thin disk formed over a long time period. There are other global features that this work qualitatively reproduces, e.g. the fact that the Galactic disk flares at large Galactocentric distances.

In Gaia Data Release 3 CU5 will publish millions of low-resolution spectra of stars all across the sky and CU8 will extract from them as well as from the integrated photometry and the parallaxes various intrinsic properties such as effective temperature, [Fe/H] and surface gravity. This work is a demonstration that the quality of Gaia’s low-resolution spectra allows the community to extract other quantities from these data that are not provided as part of the release, for example the [α/Fe] abundance ratios presented here.

Technical details

The results presented above have been derived from an ExtraTrees model trained on ~200,000 stars taken from the GALAH (GALactic Archeology with HERMES) survey catalog. The model was finally applied to a sample of 3,380,241 stars (6 < G magnitude < 15) observed with Gaia within 4 kpc of the Sun.

From these, 945,384 giant stars were selected via a surface-gravity criterion (logarithmic surface gravity below 3.3). The precision of the surface-gravity inference is high enough to secure a homogeneous sample of luminous stars out to fairly large distances.

α is a placeholder for a group of elements that is created by adding α particles (helium nuclei) to carbon. Prominent members of this group of elements, mostly fused in core-collapse (type II) supernovae of massive stars, are oxygen, magnesium, silicon and calcium.

Iron, the other element discussed above, is instead mostly fused in type Ia supernovae, that is, white dwarfs accreting matter from a companion thereby surpassing a stability mass limit.