Periodic Reporting for period 1 - CHAMELEON (Virtual Laboratories for Exoplanets and Planet Forming Disks)
Okres sprawozdawczy: 2020-06-01 do 2022-05-31
The overall objectives of the CHAMELEON MC ITN EJD are:
A) Scientific objective: Retrieve and predict the chemical composition of planet-forming disks and exoplanet atmospheres using Virtual Laboratories.
B) Technological objective: Knowledge transfer between planet and disk community by the exchange of state-of-the-art codes. Apply and develop models of different complexity as link between big observational and numerical modelling data. Explore models as Virtual Laboratories for exploring parameter spaces that cannot (yet) be reached by observations nor by (laboratory) experiments.
C) Educational objective: Train complex modelling and big-data interpretation. Use fascination for exoplanets and their birthplaces to promote science in the society, in local and wider communities.
Addressing all three objective enables our common understanding for fundamental processes that lead to the formation of planets that orbit different stars, and for the diversity of exoplanets that have been observed. The involved technology development in form of Virtual Laboratories progresses the necessary in-depth knowledge in chemistry and physics as well as that of numerical approaches. Solving complex coupled systems as well as applying machine learning techniques are part of this process. Our Virtual Laboratories enable the interpretation of high-profile space telescope data like from Hubble, James Webb Space Telescope (JWST), as well as PLATO and Ariel in the future, and hence, enable local science return from internationally funded space missions. These results are the base for science promotion within our local and global communities, and for future technology developments in the spacesector.
We have used machine learning algorithms to apply Bayesian retrieval analysis to observational data from planet-forming disks and exoplanets, respectively, based on our complex theoretical models. A few 100000 disk models were calculated to re-do the analysis of the FP7 DIANA spectral energy distribution (SED) datasets for 30 well-known disks, applying both single zone and two-zone models. The training of neural networks now allows us to generate model SEDs in a couple of milli-seconds, making full Bayesian analysis possible, to determine the uncertainties of the various physical stellar, disk, and dust parameters.
New models for the formation of rocky planets in planet-forming disks and for the influence of pebble accretion on the resulting chemical composition of exoplanet atmospheres were developed.
Bayesian and machine learning retrievals as well as self-consistent forward modelling were applied to provide the observational constraints for the chemical composition of disks and exoplanet. They were applied for interpreting transmission spectra and estimating the chemical and other atmospheric parameters, as well as successful computations of irradiated exoplanetary atmospheres and important steps toward self-consistent treatment of the
atmosphere-cloud-surface interaction and kinetic non-equilibrium of ozone formation and its effect on the atmospheric structure.
– 3D chemical kinetic, dynamic and cloud forming/evolving models for planets,
– 3D chemical and dynamical models for planet-forming disks,
– rigorous treatment of complex physical and chemical systems, and their interplay,
– understanding and exploiting large and combined codes as Virtual Laboratories,
– handling massive synthetic and heterogeneously observed datasets to draw robust conclusions,
– using complex and wide-ranging observational data simultaneously to isolate the behaviour of physical processes and translate that into analytical approaches,
– establish complex models as base tool for data interpretation and strong link to observations,
– use of neural networks/deep learning to train the exoplanet and disk model retrieval algorithms.