You are here

Home

Geophysics

Physics of the Earth and Planetary Interiors (Gabriele Morra)


In the last decade, Computational Physics has evolved with the arrival of modern Massively Parallel Supercomputers, Artificial Intelligence (CNN, Autoencoders, GANs), Cloud Computing and the new computer languages that allow to connect all of them (e.g. Python and Julia). These new tools allow modeling the formation and evolution of planets at unprecedent resolution and allow adding of astro- and geo-chemistry and Ultra High Pressure (UHP) mineral physics to planetary evolutionary models. In the Geo-and Planetary Physics group at UL, we use supercomputing to combine numerical models with machine learning to understand how the Earth and the other terrestrial (solid) planets form and evolve in our and other solar systems.


Current projects:

  • Study of the initial stages of Terrestrial Planetary Formation, with focus on the modeling the dynamic interaction between the liquid metal (core) and liquid silicates (mantle) during planetary accretion. Physical considerations show that an impactor metallic core (radius~1000 km) can emulsify into extremely small drops down to ~cm size. We develop numerical models to track metal-silicate fluid dynamics using our novel implementation of the Lattice Boltzmann Model, which uses a new formulation for modeling surface tension at extreme conditions. This is then combined with chemical and mineral physics data from collaborating groups at Tulane University and LSU to identify fine-scale metal-silicate scenarios. This is part of a larger project involving Tulane University, Louisiana State University, University of Louisiana at Lafayette and NASA Johnson Space Center. The project is supported by NASA.

Bibliography:

P. Mora, G. Morra, 2021, D.A. Yuen, Optimized surface tension isotropy in the Rothman-Keller colour gradient Lattice Boltzmann Method for multi-phase flow, Physical Review E.

P. Mora, G. Morra, D.A. Yuen, R. Juanes, 2021, Optimal wetting angles in Lattice Boltzmann simulations of viscous fingering, Transport in Porous Media, https://doi.org/10.1007/s11242-020-01541-7

 

  • Inferring the causes of the eruptions in Strombolian Volcanoes, which are the most active volcanoes in the world and provide insights into mechanisms occurring deep below the surface. In this research, we introduce Machine Learning as a method to detect eruptions in millions of infrared images of the magma lake on top of a volcano. All the numerical techniques are developed in-house by the students who work on the project. This allows them to build a unique set of skills that will help them in their future career. For this project, dedicated Convolutional Neural Networks have been designed to obtain compact and fast solutions, as well as developed an approach borrowed by solar physics, the combination of Zernike moments and Support Vector Machine.

Bibliography:

B. Dye and G. Morra, 2020, Machine learning as a detection method of Strombolian eruptions in infrared images from Mount Erebus, Antarctica,  Big Data in Geosciences: From Earthquake Swarms to Consequences of Slab Dynamics, Physics of the Earth and Planetary Interiors, 106508

  • Mud volcanoes on Mars. Presence of water under Mars' surface has been hypothesized, but direct evidence is not available. Based on the observation of tens of thousands of mounds over Mars' surface, it has been proposed that wet granular flow (mud) has emerged to the surface after cataclysmic events such an asteroid impact. Mud volcanoes on Earth exist and are associated with instability in wet granular material that flows towards the surface. We apply Machine Learning tools to spaceborne Mars data aimed at creating an integrated dataset of to reduce the uncertainty associated with their identification by remote sensing. Identifying the location of water on Mars crucial to plan the upcoming Martian explorations.

 

  • Scaling in Thermal Convection. The interior of stars and planets dynamics is mainly driven by thermal convection, called Rayleigh-Benard regime. To better understand the regimes through which habitable planets go through during their evolution after formation, we introduced a 3D parallel implementation of the Lattice Boltzmann Method. The code is entirely written in Python, a clean programming paradigm whose scientific use is rapidly growing and scales linearly on thousands of cores on standard Beowulf clusters. The software builds on Message Passage Interface (MPI) and vectorized operations. Results are organized in the phase space described by macroscopic parameters such as the Nusselt number (Nu), the Reynolds number (Re), the Rayleigh number (Ra) and the Prandtl number (Pr). For very large Ra we explore non-normal-nonlinear transition to turbulence and infer the consequences for Earth and Super Earth's.

Bibliography:

        G. Morra, D.A. Yuen, H.R. Tufo, M.G. Knepley, 2020, Fresh Outlook in Numerical Methods for Geodynamics. Part 1: Introduction and Modeling, Encyclopedia of Geology, 2e. Pages 826-840, ISBN 9780081029091, Also published in Reference Module in Earth Systems and Environmental Sciences. https://doi.org/10.1016/B978-0-08-102908-4.00110-7
        G. Morra, D.A. Yuen, H.R. Tufo, M.G. Knepley, 2020, Fresh Outlook in Numerical Methods for Geodynamics. Part 2: Big Data, HPC, Education, Encyclopedia of Geology, 2e. Pages 841-855, ISBN 9780081029091. Also published in Reference Module in Earth Systems and Environmental Sciences. https://doi.org/10.1016/B978-0-08-102908-4.00111-9
        P. Mora, G. Morra, D. A. Yuen, 2019, A concise Python implementation of the Lattice Boltzmann Method on HPC for geo-fluid flow, Geophysical Journal International, Volume 220, Issue 1, Pages 682–702, https://doi.org/10.1093/gji/ggz423