Open Access. Powered by Scholars. Published by Universities.®

Physical Sciences and Mathematics Commons

Open Access. Powered by Scholars. Published by Universities.®

Articles 1 - 2 of 2

Full-Text Articles in Physical Sciences and Mathematics

A Non-Linear Eigensolver-Based Alternative To Traditional Self-Consistent Electronic Structure Calculation Methods, Brendan E. Gavin Jan 2013

A Non-Linear Eigensolver-Based Alternative To Traditional Self-Consistent Electronic Structure Calculation Methods, Brendan E. Gavin

Masters Theses 1911 - February 2014

This thesis presents a means of enhancing the iterative calculation techniques used in electronic structure calculations, particularly Kohn-Sham DFT. Based on the subspace iteration method of the FEAST eigenvalue solving algorithm, this nonlinear FEAST algorithm (NLFEAST) improves the convergence rate of traditional iterative methods and dramatically improves their robustness. A description of the algorithm is given, along with the results of numerical experiments that demonstrate its effectiveness and offer insight into the factors that determine how well it performs.


Reformulation Of The Muffin-Tin Problem In Electronic Structure Calculations Within The Feast Framework, Alan R. Levin Jan 2012

Reformulation Of The Muffin-Tin Problem In Electronic Structure Calculations Within The Feast Framework, Alan R. Levin

Masters Theses 1911 - February 2014

This thesis describes an accurate and scalable computational method designed to perform nanoelectronic structure calculations. Built around the FEAST framework, this method directly addresses the nonlinear eigenvalue problem. The new approach allows us to bypass traditional approximation techniques typically used for first-principle calculations. As a result, this method is able to take advantage of standard muffin-tin type domain decomposition techniques without being hindered by their perceived limitations. In addition to increased accuracy, this method also has the potential to take advantage of parallel processing for increased scalability.

The Introduction presents the motivation behind the proposed method and gives an overview …