Overall Objectives
Research Program
Application Domains
New Software and Platforms
New Results
Partnerships and Cooperations
XML PDF e-pub
PDF e-Pub

Section: New Results

Development of a novel minimization method

Participants : Clement Beitone, Stephane Redon.

Finding the optimized configuration of a system of particles so that it minimizes the energy of the system is a very common task in the field of particles simulation. More precisely, we are interested in finding the closest atomic structure located at a minima on the Potential Energy Surface (PES) starting from a given initial configuration. Achieving faster but reliable minimizations of such systems help to enhance a wide range of applications in molecular dynamics. To improve the efficiency of the convergence some authors have proposed alternative methods to the steepest descent algorithm; for example, the conjugate gradient technique or the Fast Inertial Relaxation Engine (FIRE).

In this work, we are developing a novel method that helps to increase the efficiency and the reliability of existing optimizers, e.g. FIRE and Interactive Modelling (IM).

We have implemented the modified versions of these algorithms along with others optimization algorithms like L-BFGS and Conjugate Gradient as state updaters in SAMSON. To assess the efficiency of the proposed approaches we have developed an App in SAMSON that allows us to reliably and conveniently probe several criteria during the minimization process (Figure 3).

Figure 3. Comparison of different optimizers with the proposed methods on the fullerene C240. For this experiment the force field used to model the interactions between the atoms is the Brenner potential.