Galactic archaeologists find the ancient heart of the Milky Way

Astronomers identify the "poor old heart" of the Milky Way, finding a group of stars left over from the earliest history of our galaxy.
Paul Ratner
An illustraiton of the heart of the galaxy
An illustraiton of the heart of the galaxy


  • Scientists identify the "poor old heart of the Milky Way" – the oldest stars in our galaxy.
  • The identified stars are metal-poor, as exhibited by their metallicity, which points to old age.
  • AI machine learning was instrumental in the discovery.

Taken against the human lifespan, the age of our home galaxy — the Milky Way — is mind-boggling, at about 13.6 billion years. Understanding how the galaxy was created and what was there before has been the task of galactic archaeologists, with a new study coming out that claims to have found “the poor old heart of the Milky Way.” What the scientists from the Max Planck Institute for Astronomy (MPIA) spotted were the oldest stars in our galaxy — its ancient core that remained from the earliest days of the galaxy's formation. 

Galactic archaeologists find the ancient heart of the Milky Way
Heart of the Milky Way galaxy

The team of astronomers, led by Hans-Walter Rix of MPIA, discovered a concentration of around 18,000 stars whose so-called metallicity — the amount of chemical elements heavier than helium contained in its atmosphere — places them as the oldest stars in the Milky Way. To achieve this research feat, the team of scientists used data from ESA’s Gaia satellite and the LAMOST (Large Sky Area Multi-Object Fiber Spectroscopic Telescope) spectral survey and trained an AI neural network to analyze the data. Gaia has been gathering very accurate position and motion information, as well as distances, for over a billion stars since 2014.

AI aids galactic detective work

To find the oldest stars, astronomers utilized machine learning. The neural network was trained on data from the APOGEE Survey (Apache Point Observatory Galactic Evolution Experiment), which includes high-resolution spectral observations of thousands of the Milky Way’s red giant stars. The advantage of starting with this dataset was that the metallicity of stars was already known for half of the data, allowing the scientists to fine-tune the neural network. Once the network could recognize what characteristics to look for, it was let loose on the complete set of Gaia data, producing a huge data set of 2 million stars with accurate metallicities. 

It is within the resulting data set that the researchers found the ancient “heart of the Milky Way” of stars older than 12.5 billion years old. This group of approximately 18,000 centrally-located stars exhibited low metallicity, which correlates to being of older age since being born later in the Universe’s lifespan likely means the stars have more metals. Therefore higher metallicity suggests younger stars, while “metal-poor” stars would be the oldest. 

Further research into the chemical composition of these stars might yield information about which of them belong to progenitor galaxies — those in existence even before the Milky Way. 

Interesting Engineering (IE) reached out to Professor Rix for more details on their work. The following conversation has been lightly edited for clarity and flow.

Interesting Engineering: What is the significance of the galactic archaeology work to retrace the past of the galaxy? What can we find out?

Professor Rix: One of the magical facts about our cosmos is that it appears that the beautiful world of galaxies that we see today arose from truly random fluctuations (white noise) in the mass distribution in the early Universe.

We want to find out how. Specifically: when did gas turn into stars, and when were the orbits of stars arranged into the present disk-like, spiral shape. For that, it would be great to look in the past and see how a galaxy like the Milky Way arose.

We can’t do that directly, but in two indirect ways: we can look very far away, as JWST does now. As light takes time to travel, a look into the distance is a look into the past. Unfortunately, it is a look into the past “somewhere else” (Fortunately, the past somewhere else should statistically be the same as our past). The other approach is to look in our own Milky Way that we see star-by-star.

We can date these stars either directly or by their element composition (which astronomers call “metallicity). That is a measure for how polluted (by earlier generations of stars) the material was from which these stars formed.

This is what we do in the Milky Way. With our “poor old heart” work, we could show that most of the oldest (and hence earliest formed) stars live right at the center of the galaxy. This is in some ways expected, as this is where matter first collapsed under its own gravitational attraction to form stars. But it’s the first time we’ve seen clearly that it is true. And the fact that the first stars still live so centrally concentrated means that after their formation, the Milky Way never got too jostled (by a merger with another galaxy) to shake up the stars in the (original) center to more extended orbits.

IE: How important was machine learning in your study? How does AI aid your galactic detective work?

That was crucial. We used “data-driven modeling” to get the metallicities of millions of stars. How does this work:

The Gaia Mission took low-resolution spectra for 10s of millions of stars. These spectra are hard to model “physically” to get the element composition — the metallicity. But we know the “information should be in the spectra. For a few percent of the stars, we however know the element abundance (metallicity) from much higher resolution spectra. We then train an algorithm saying:

Here are 100.000 examples of Gaia spectra where we know the “answer” (=metallicity) from better information: learn how to extract the metallicity from Gaia spectra. Then we apply this to 10 Bio Gaia spectra. And it works beautifully!

IE: What is next for your research? Can this work lead all the way back to the Big Bang?

Not quite to the Big Bang. But now that we have found the oldest stars at the heart of the Milky Way, we can look (with better spectra) at their element abundances (across dozens of elements in the periodic table). This will teach us what processes forged the first elements heavier than helium, and in particular heavier than Iron. So, we have justified hope to learn about “the origin of the periodic table”.

Read the study “The Poor Old Heart of the Milky Way," in the Astrophysical Journal.


Our Milky Way should host an ancient, metal-poor, and centrally concentrated stellar population, which reflects the star formation and enrichment in the few most massive progenitors that coalesced at high redshift to form the proto-Galaxy. While metal-poor stars are known to reside in the inner few kiloparsecs of our Galaxy, current data do not yet provide a comprehensive picture of such a metal-poor "heart" of the Milky Way. We use information from Gaia Data Release 3, especially the XP spectra, to construct a sample of 2 million bright (GBP < 15.5 mag) giant stars within 30° of the Galactic center (GC) with robust [M/H] estimates, δ[M/H] ≲ 0.1. For ∼1.25 million stars we calculate orbits from Gaia Radial Velocity Spectrometer velocities and astrometry. This sample reveals an extensive, ancient, and metal-poor population that includes ∼18,000 stars with −2.7 < [M/H] < −1.5, representing a stellar mass of ≳5 × 107 M⊙. The spatial distribution of these [M/H] < −1.5 stars has a Gaussian extent of only  around the GC, with most orbits confined to the inner Galaxy. At high orbital eccentricities, there is clear evidence for accreted halo stars in their pericentral orbit phase. Most stars show [α/Fe] enhancement and [Al/Fe]–[Mn/Fe] abundances expected for an origin in the more massive portions of the proto-Galaxy. Stars with [M/H] < −2 show no net rotation, whereas those with [M/H] ∼ −1 are rotation dominated. These central, metal-poor stars most likely predate the oldest disk population (τage ≈ 12.5 Gyr), which implies that they formed at z ≳ 5, forging the proto-Milky Way.

Add Interesting Engineering to your Google News feed.
Add Interesting Engineering to your Google News feed.
message circleSHOW COMMENT (1)chevron
Job Board