Electronic Thesis and Dissertation Repository

Thesis Format

Integrated Article

Degree

Doctor of Philosophy

Program

Applied Mathematics

Supervisor

Shantanu Basu

2nd Supervisor

David Jeffrey

Co-Supervisor

Abstract

This thesis aims to study the magnetic and evolutionary properties of stellar objects from the prestellar phase up to and including the late protostellar phase. Many of the properties governing star formation are linked to the core’s physical properties and the magnetic field highly dictates much of the core’s stability.

The thesis begins with the implementation of a fully analytic magnetic field model used to study the magnetic properties governing the prestellar core FeSt 1-457. The model is a direct result of Maxwell’s equations and yields a central-to-surface magnetic field ratio in the equatorial plane in cylindrical coordinates. The model additionally gives estimates to the field direction and the relative magnitude at all points within the core. Through non-linear fitting, the single plane magnetic field is fit to observationally measured near-infrared polarization segments resulting form the integrated properties of the magnetic field and dust grains. The model is back tested by deploying the radiative transfer code POLARIS, used to simulate synthetic polarization maps resulting from the integrated scattering and emission properties of dust grains.

This study is further extended to the protostellar phase in the core’s evolution. In order to accurately model the dense, star forming regions of protostellar cores, non-ideal magnetohydrodynamic (MHD) simulations are used by solving the resistive MHD equations on a nested grid at a series of length scales. The simulation is used to study three instances in time for the core’s evolution represented in terms of the core’s mass. This methodology enables one to resolve the central region on small scales and is more computationally efficient than the adaptive grid method. In doing such, the simulation outputs are served as inputs to the POLARIS code used to simulate synthetic polarization maps at all length scales within the protostellar core allowing for robust, high resolution outputs of the polarization segments within the small scale regions of the core. Additionally, the synthetic polarization maps are used to model the polarization state of the Orion Source I cloud. Through superimposition and an analysis on the mean field direction, the emission or absorption properties of the observational measurements from ALMA can be inferred from both the POLARIS and MHD simulations.

The late protostellar evolution phase involves studying the dynamics of the episodic behavior of mass accretion. By implementing an optimized form of recurrent neural networks utilizing the reservoir computing framework, one can produce robust, model-free predictions to uncover and extract important, underlying temporal features governing mass accretion in the late evolution period. Through carefully pre-processing data, the network is run on a series of hydrodynamic and stellar evolution simulations in an attempt to successfully recover the expected behavior of the simulations whilst introducing an effective method for performing predictive analysis on data that has no governing model.

Summary for Lay Audience

Stars have long been known to be an important factor in the evolution of the universe as a whole, including in its constituent parts like galaxies and planets. In this thesis, the formation of stars is studied in different evolutionary epochs from the prestellar phase up to and including the formation and evolution of protostars. A critical component in star formation is the magnetic field. A large part of this thesis aims to study the magnetic field that governs both prestellar and protostellar cores. Using theoretical simulations in conjunction with observational data, robust models are developed that give insight to some important stellar properties governing the cores. This thesis additionally provides an introduction to forecasting time series using a new form of neural networks.

Creative Commons License

Creative Commons Attribution 4.0 License
This work is licensed under a Creative Commons Attribution 4.0 License.

Available for download on Wednesday, September 15, 2021

Share

COinS