A popular model for the formation of the Milky Way bulge, as well as of elliptical galaxies and spiral bulges, is the so-called "monolithic collapse", according to which these structures formed all at once from a single collapsing cloud. This is, however, still largely debated in the literature, also because growing body of evidence suggests that cosmic structures primarily form bottom-up through "hierarchical merging" processes, meaning that larger and larger objects are generated by successive mergers of smaller structures. This scenario is indeed adopted to explain the existence and the observed properties of the massive galaxies (and even, galaxy clusters) populating our Universe. Of course, the same process could also hold for spheroids and, in fact, the "merging picture" proposes that galaxy bulges are generated through the merging of primordial massive clumps of gas and stars, likely formed by the instability and fragmentation of star-forming disks, as those observed at high redshift in the so-called "clumpy galaxies". Indeed this scenario appears to be extremely successful in describing the bulge assembling in external galaxies, but no direct observational evidence was ever obtained for the Galactic bulge, although the high-redshift data confirm that such massive fragments existed at the epoch of the Milky Way formation. This naturally raises questions as the following: Did the Galactic bulge also form in this way? If so, why no fossil relics of that assembling process have ever been found? 

These are the key questions that guided our research already 15 years ago, when we started this project. In fact, although the vast majority of the primordial fragments is expected to dissolve to form the bulge, simulations show that a few of them could survive the total disruption and be still present in the inner regions of the host galaxy, grossly appearing as massive globular clusters. At odds with genuine globular clusters, however, these fossil relics should have been massive enough to retain the iron-enriched ejecta of supernova explosions, and possibly experienced multiple bursts of star formation. As a consequence, they are expected to host multi-iron and multi-age sub-populations.

We followed this route and, indeed, we were very lucky! The first candidate fossil relic in the Galactic bulge was identified by our group back in 2009. This stellar system, named Terzan 5, had all the expected properties and was like no other star cluster known so far: in spite of its globular cluster appearance, it contains both a stellar population remarkably similar the most ancient objects in the Milky Way, and another population of much younger and iron-enriched stars. This discovery initiated a newline of investigation, with the aim to disclose other similar structures that could be still hidden in Galactic bulge.

Indeed we promptly started this hunting, but we had to face the problem that the bulge is the most inaccessible region of our Galaxy, where thick clouds of dust severely dim star light, especially at optical wavelengths (indeed, a huge extinction, reaching up ~10 magnitudes, affects optical observations). Fortunately, however, infrared radiation can travel across these clouds and bring us direct information on the emitting sources.

Thus, in collaboration with a Chilean group mainly working in Concepción and Antofagasta, we started a systematic study of massive star clusters in the Galactic bulge at near- infrared wavelengths (in the J and K filters), by taking advantage of a technical jewel named GeMS (for "Gemini Multi-conjugate adaptive optics System"), in combination with the powerful infrared camera GSAOI (for "Gemini South Adaptive Optics Imager") at the 8-m telescope Gemini South on Cerro Pachón, in Chile. The superb performances of this instrument were exactly what we needed to successfully conduct this unprecedented search. In fact, not only it operates at near-infrared wavelengths (where the dimming effect of star light due to dust is largely reduced), but it also provides very high angular resolution, meaning that individual stars can be distinguished even in the overcrowded central region of the system, as it happens in images taken from the Hubble Space Telescope (HST) in orbit around our planet. The latter is possible thanks to a multi-conjugate adaptive-optics system, an innovative instrument able to efficiently remove the distortions (blurriness) that the Earth's turbulent atmosphere inflicts on astronomical images: to compensate for the degradation effects of the Earth atmosphere, the GeMS system uses three natural guide stars, a constellation of five laser stars and multiple deformable mirrors. By taking advantage of this system we explored the stellar populations of several massive star clusters in the Bulge, searching for Terzan5 twins. Finally in 2015, we obtained superb images of one of the most extincted globular clusters in the Galaxy, one of our top-priority targets: Liller 1. Indeed, the acquired images were of unprecedented sharpness: in the best K-band exposures of Liller 1, stellar images have an angular dimension of only 75 milli-arcseconds, just slightly larger than the theoretical limit (diffraction limit) of the Gemini's 8-meter mirror, meaning that GeMS performed an almost perfect correction of the atmospheric distortions.

This allowed the first detailed inspection of the stellar population in Liller 1. Not surprisingly, we found a very old population (with an age of 12 Gyr), demonstrating that this stellar system formed very early in the history of the Galaxy. But something was strange in that colour-magnitude diagram (CMD): we noted the presence of a blue sequence of stars, mimicking a young population. Just some suspicious clues… nothing more than this, because the system is located deep into the Bulge, where it is really hard to distinguish the stars belonging to Liller 1, from those orbiting in the Galaxy and being just aligned along the same line of sight. So we had no ways to draw any firm conclusion from these observations alone. Anyway, a doubt (together with a hope) naturally raised: could Liller 1 be an object similar to Terzan 5?

To answer this question we needed some additional and complementary observations, similarly sharp and deep. Surprisingly, we noted that no deep optical images of Liller 1 had been acquired with the HST over its 30-year activity. Incredible! To remedy this lack, we promptly asked to secure HST ultra-deep observations of Liller 1 in the V and I bands, and we got time. It was the first time that the sharp eyes of Hubble observed Liller1 at optical wavelengths!

At this point we had in hand the best observational dataset of Liller 1 obtainable with the current generation of instruments, the best ever collected both in the optical and in the near-infrared bands. However, there were still a lot of problems to deal with! The detailed analysis of the two datasets clearly unveiled a weak point (a weak band, indeed) in both of them: at optical wavelengths, the V-band exposures were much shallower than those in the I-band, while in the near-infrared, the observations were limited by the J-band images. Then the optimal solution could be just one: combining the two deepest exposures, those in the K and the I filters. The analysis had to be completely repeated and the combination of the two sets of images was long and complex. Special care was devoted to guarantee that all the sources originally detected only in the I images were then searched and analysed also in the K exposures, and vice versa, thus maximizing the information provided by the two filters separately. The resulting hybrid (near-IR/optical) CDM was unprecedentedly deep and accurate, providing the ideal tool for the study of the stellar population in Liller 1. However, we still had to deal with differential reddening, which spreads out and distorts the evolutionary sequences in the CMD, making difficult their nterpretation. For this reason we constructed a detailed reddening map in the direction of Liller 1 and we used it to correct the CMD for the effects of differential reddening. Finally, a very well-defined CMD was available, and the suspicious blue sequence of stars (that we dubbed "Blue Plume") was still there, sharper and better outlined than ever. A real population of young stars, or a weird effect of contamination from Galactic field intruders?

The last step: proper motions!

Fortunately, the Gemini and HST observations were acquired with a temporal distance of more than 6 years and with comparable sharpness, thus allowing the accurate determination of stellar proper motions. We thus could clearly distinguish Liller 1 members from field interlopers, obtaining a wonderful cleaning of the CMD from Galactic disk and bulge stars, and leading to the final confirmation: even after proper motion decontamination, the Blue Plume was still present! The wonderful CMD, corrected for differential reddening and decontaminated from field stars, clearly showed the presence of two components, both belonging to Liller1! We also found that the stellar population along the Blue Plume is even more segregated toward the centre of Liller 1 than the old population! No more doubts: the stars along the Blue Plume do belong to Liller1. Hence this system hosts two stellar populations with a huge difference in age: the oldest is 12 Gyr old (it formed at the epoch of the Bulge assembling), while the second could be as young as 1-2 Gyr only!!!

Another non-genuine globular cluster in the Bulge (after Terzan 5) was finally discovered!

The properties of the old stellar populations observed in Liller 1 and Terzan 5 demonstrate that both systems formed very early: 12 Gyr ago, at the epoch of the Galaxy assembly. On the other hand, the properties of the young populations, which appear to be more metal-rich and more centrally segregated than the old populations (in agreement with what is expected in a self-enrichment scenario), confirm that the progenitors of both Liller 1 and Terzan 5 were massive enough to retain supernova ejecta. The available observational evidence thus defines these stellar systems as the likely remnants of primordial massive structures that formed in situ and contributed to generate the Bulge ~12 Gyr ago.

Indeed we discovered a new class of objects, that we named "Bulge Fossil Fragments (BFFs)". These systems (1) are indistinguishable from genuine globular clusters in their appearance, (2) have metallicity and abundance patterns compatible with those observed in the Bulge field stars, (3) host an old stellar population (testifying that they formed at an early epoch of the Galaxy assembling), (4) host a young stellar population, several to many Gyrs younger than the old one (testifying their capacity of triggering multiple events of star formation).

Like archaeologists, who dig through the dust piling up on top of the remains from past civilizations and unearth crucial pieces of the mankind history, we have been gazing through the thick layers of interstellar dust obscuring the Bulge of the Milky Way, unveiling a new class of extraordinary cosmic relics. As for archaeological remains, we now have to study the BFFS in detail because in fragments like these, it is written the story of the formation of the first cosmic structures (as the Galactic bulge) at the time when the Universe was a “baby”, just one billion year old.

More info at the web page of the Cosmic-Lab project: Liller 1 - BFF