Force field parameterization techniques using genetic algorithm evolutionary strategies
Abstract/Contents
- Abstract
- A novel evolutionary strategy in the form of a steady-state genetic algorithm (GA) is presented for the derivation of an accurate parameterization of the shell model potential as applied to a metal oxide perovskite. Using barium titanate, (BaTiO$_3$ or BTO), as a testbed material, the GA technique is implemented by fitting to a density functional theory (DFT) rendered landscape of the four phases of the material, representing a wide range of unit cell configurations. Using the stochastic processes of recombination and mutation which are fundamental to the GA approach, a parameter set of the shell model potential has been identified that not only matches the energetics of the reference database accurately, but is also able to directly capture the implicit elasto-mechanical characterization of the material that is embedded in the DFT analysis. The evolutionary strategy is described in detail, and it is noted that in its current form it is directly applicable to a diverse number of other metal oxides for which similar analysis is entirely compatible. The method is then compared to the conjugate gradient technique whereby it is seen that while the gradient-based method faces significant challenges in identifying globally optimal parameterizations despite repeated initiations at distinct starting points for the search, the GA method is able to continuously explore the rugged parametric solution space to identify those regions which produce the greatest fidelity to the DFT energies. The validation of the technique includes renditions of equilibrium lattice lengths, unit cell volume and bulk modulus for the four phases of BTO using static energetic calculations. In addition, using molecular dynamics (MD) in both the canonical and isobaric-isothermal ensembles, the elastic stiffness constants, coefficient of thermal expansion, bulk modulus, and average electronic polarization for the cubic phase have been derived. These results are in excellent agreement with empirical analysis presented in the literature.
Description
Type of resource | text |
---|---|
Form | electronic; electronic resource; remote |
Extent | 1 online resource. |
Publication date | 2012 |
Issuance | monographic |
Language | English |
Creators/Contributors
Associated with | Solomon, Jose Enrique |
---|---|
Associated with | Stanford University, Department of Mechanical Engineering |
Primary advisor | Darve, Eric |
Thesis advisor | Darve, Eric |
Thesis advisor | Cai, Wei |
Thesis advisor | Srivastava, Deepak |
Advisor | Cai, Wei |
Advisor | Srivastava, Deepak |
Subjects
Genre | Theses |
---|
Bibliographic information
Statement of responsibility | José Enrique Solomon. |
---|---|
Note | Submitted to the Department of Mechanical Engineering. |
Thesis | Thesis (Ph.D.)--Stanford University, 2012. |
Location | electronic resource |
Access conditions
- Copyright
- © 2012 by Jose Enrique Solomon
- License
- This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).
Also listed in
Loading usage metrics...