Unraveling the origin of the Initial Mass Function

Despite the tremendous progress achieved over the past decade, the study of stellar formation is far from complete. We have not yet measured the minimum mass for star formation, nor the shape of the IMF down to the least massive free-floating planets, or know how universal this shape is. Although clusters are the building blocks of galaxies, little is known about their early dynamical evolution and dispersal into the field. The main culprit for this state of affairs is the high level of contamination and incompleteness in the sub-stellar regime, even for the best photometric and astrometric surveys.
COSMIC-DANCE aims at overcoming these drawbacks and revealing the shape of the IMF with a precision and completeness surpassing current and foreseeable surveys of the next 15 years. Using a novel methodology taking advantage of archival and new deep wide-field ground based observations, we will measure proper motions with an accuracy comparable to Gaia but 5 magnitudes deeper, reaching the planetary mass domain, and, critically, piercing through the dust obscured young clusters inaccessible to Gaia’s optical sensors. These proper motions measurements and associated the multi-wavelength photometry will be fed to innovative hyper-dimensional data mining tools to securely identify cluster members within the millions of sources present in the catalogues, complemented by Gaia at the bright end, to obtain the final census over the entire mass spectrum. These measurements will provide conclusive empirical constraints over a broad parameter space unaccessible to current state-of-the-art surveys on the much debated respective contributions of evolutionary effects (dynamics, feedback and competitive accretion) and initial conditions (core properties) to the shape and bottom of the IMF, the most fundamental and informative product of star formation, with essential bearings on many areas of general astrophysics.

Work Package 1: the entire chain of data processing has been improved. Concerning WP1.1 the public archives queries are now automatized, thanks to the work of the PhD student (N. Miret Roig) with the help of an undergraduate trainee, making it very easy for any team member to access the wealth of data available in the most important astronomical archives. Concerning WP1.2 telescope time requests have been submitted every semester to the best facilities worldwide (ESO, Subaru, La Palma, Calar Alto, NOAO, HST, OHP), and our approval rate has been close to 80%, ensuring that we had plenty of very high quality data to achieve our objectives. The junior team members (PhD and Postdocs) have been trained by the P.I. to prepare and lead the observations at the telescope, and are now experienced astronomers. Concerning WP1.3 we could not hire an engineer to do the update of our image processing software due to the lack of qualified candidates. Fortunately a PhD student co-funded by the university of Bordeaux and the CNES joined the team to develop image processing modules based on artificial intelligence. Combined with the software used by the P.I. it improves significantly the final data quality and so we were able to go on with the data processing. A powerful computing facility has been acquired as planned early in the project to be able to process the large quantity of data in a most efficient way. This facility is absolutely essential and allows our team to achieve our goals in a timely manner. The original hardware design was slightly modified with the addition of GPU cards, which proved absolutely revolutionary and decreased some parts of the calculations by a factor of 50 to 100, but also allowed to use newly developed software based on deep learning methods to perform the data processing.

Work Package 2: The upgrade of the astrometric software included in WP2.1 could not be done, but did not compromise our ability to analyse the data. WP2.3 has led to the design by one of our external collaborator of a novel data processing method to correct the atmospheric turbulence to produce high-spatial resolution images using modern CMOS cameras. WP2.4 has been fully implemented and improved performances (speed) by as much as 30% to 40% in the case of very large dataset. WP2.5 has been implemented as soon as the Gaia DR2 was released, and leads to significant improvements in the quality of the astrometric solution, with residuals dropping by half in the final solution.

Work Package 3: Concerning WP3.1 the first version of a hierarchical model has been fully tested and published in Olivares et al. (2018). The performances were excellent and the results impressive. Concerning WP3.2 the ultimate parallelization has been achieved using GPU cards, with spectacular speed improvements absolutely crucial for the rest of the development. WP3.3 is complete: an external module (called Kalkayotl) has been developed to analyse the spatial distribution. The treatment of binarity in WP3.4 has been completed and is integrated in the hierarchical model. Concerning WP3.5 the treatment of extinction is now fully implemented and functional with already very interesting and exciting results in a number of embedded young star forming regions. Finally, concerning WP3.7 we have been taking advantage of the Gaia DR2 release to compare our results and cross-validate the methods. This allowed us to confirm that our strategy is well defined and performs as expected, and also allowed us to tune and optimize some parameters of the hierarchical model.

Work Package 4: The work planned for WP4 depends directly on the results obtained and delivered by the WP 1 to 3. We have successfully conducted WP4.4 and WP4.3 by gathering spectroscopic data at various world class facilities (OHP, GTC, HST, La Silla) to study the atmosphere of ultracool objects, search for planetary mass companions and measure radial velocities. Part of these have been published and more publication will come in the future as the data are acquired and analysed.

Work Package 5: The team members participated to various outreach events, from seminars for the general public or undergraduate students to podcats, live-streaming, publications in magazines as well as high-school student visits and visit to primary schools.

On the technical side, the use of very modern techniques such as GPU high performance computing and artificial intelligence, puts us at the frontier of software development in astrophysics, and was acknowledged and rewarded by the GPU card manufacturer NVidia, one of the biggest industrial actors of the market.

On the scientific side, the hierarchical models developed for the project have no equivalent so far and outperform all currently published methods in performances and reliability, delivering not only the best possible censuses and mass functions, but also the associated uncertainties derived in the most robust and rigorous way. These measurements will allow us to make meaningful and quantitative comparisons between stellar clusters and associations, but also with the theoretical and numerical predictions. The IMF measurements delivered by COSMIC-DANCE shall remain references for a long time to come. The unique samples of free-floating planets will certainly become targets for many surveys and studies to come, with impact in star formation, planet formation and planet atmosphere studies.

Periodic Reporting for period 4 - COSMIC-DANCE (Unraveling the origin of the Initial Mass Function)

Verwandte Dokumente

Diese Seite teilen

Herunterladen