105. JACOBIAN-ENHANCED NUDGED ELASTIC BAND SOLVER FOR MICROMAGNETICS

Department: Electrical & Computer Engineering
Research Institute Affiliation: Center for Magnetic Recording Research (CMRR)
Faculty Advisor(s): Vitaliy Lomakin

Primary Student
Name: Marco Antonio Escobar Acevedo
Email: maescoba@ucsd.edu
Phone: 858-534-4479
Grad Year: 2013

Abstract
An important new component of the presented micromagnetic nudged elastic band (NEB) solver is the use of partial or full Jacobian for accelerating the stepping algorithm. In this work we present for first time an analytical expression for the Jacobian times an arbitrary vector and we perform a numerical study of its efficiency in terms of computation time and iterative algorithm steps. The NEB method relies on the ability to compute the system energy and its gradients. The energy is computed using the effective field evaluation methods similar to those used in the high-performance FastMag simulator. The magnetostatic field is evaluated using the non-uniform grid interpolation method (NGIM) at O(N) operations or non-uniform Fourier transform method at O(NlogN) operations. These methods are implemented on massively parallel Graphics Processing Unit (GPU) computing architectures as well as on multi-CPU systems, which helps to speed up the solver. The calculation of the thermal stability requires the estimation of transition rates, such calculations require the knowledge of the most probable path between an initial and a final state also known as minimum energy path (MEP). Understanding thermal stability and reversal is of key importance for new technologies such as BPM, HAMR and MRAM. We recently analyzed the reversal of complex BPM systems with this new implementation of the NEB.

Related Links:

  1. http://cem.ucsd.edu/
  2. http://cem.ucsd.edu/index_files/fastmag.html

« Back to Posters or Search Results