 |
Center for Nonlinear Studies |
Order of the talks will vary
Mohira Rassel
T-CNLS/CCS-2
Title:Kilonova Emission – Particle-In-Cell Simulations of Mildly Relativistic Outflows
Abstract: One of the unsolved problems in high energy astrophysics is the relationship between amplification of magnetic fields and the process of particle acceleration. The aim of this project is to implement VPIC(vector particle-in-cell) kinetic simulations for collisionless plasma for mildly-relativistic shock simulations over range of parameters including kilonova outflows. Effects of various plasma conditions were observed in magnetic fields and particle energies. Additionally, validity of using different and often unrealistic proton to electron mass ratios in VPIC simulations was tested.
Coleman Kendrick
New Mexico Tech
Title: Simulating Climate on Slowly Rotating Exoplanets
Abstract: A pseudo 3D exoplanet climate model is presented in which the ocean and atmosphere are simulated via stacked 2D meshes with a one dimensional vertical coupling. Both the ocean and atmosphere are simulated using a hydrodynamic solution in spherical polar coordinates, vertically coupled to climate models for water evaporation, condensation, transport, solar heating, and ice melting/formation. Simulation results will be presented which show how climate evolves on an Earth-like tidally locked planet. These results investigate how the climate changes and ice distribution forms on a tidally locked planet with varying parameters such as planet size, orbital distance and period, rotational period, and ice/water content. In the future, this model can be applied to various exoplanets to help determine how atmospheric composition affects climate and habitability.
Phuong Chau
Smith College
Title: Machine Learning in Molecular Dynamics Simulation
Abstract: Molecular Dynamics (MD) simulation is useful for understanding the molecular structure and dynamics of complex system. However, parameterization for MD is still a burden because it is computationally expensive, not very transferable, and only applicable for non-reactive system.To overcome these problems with parameterization, we are creating an innovative method that is similarly accurate and much faster than quantum mechanic approaches. Specifically, parameters for point-charge all-atom MD are generated by a machine learning network that is trained on quantum mechanical calculations. We validate the method by comparing the experimental chemical thermodynamic properties to those properties calculated by standard MD and our machine learned parameters.This research has the potential to revolutionize the field of MD simulation by eliminating computational expense and improving accuracy in parameterization. Our framework for MD parameterization will accelerate the research efficiency in various fields including but not limit to drug design, materials development, and industrial catalysis.
Host: David Métivier