DL-FIND: an open-source geometry optimizer for atomistic simulations.

Area of Science:

• Computational Chemistry
• Materials Science
• Solid-State Physics

Background:

• Geometry optimization is crucial for computational chemistry, surface science, and solid-state physics, consuming significant CPU time.
• Efficient geometry optimization is essential for simulations using classical force fields and quantum mechanics.

Purpose of the Study:

• To introduce DL-FIND, a versatile geometry optimization software designed for integration into atomistic simulation codes.
• To provide a flexible and efficient tool for various computational tasks including structure optimization and transition state searches.

Main Methods:

• DL-FIND supports optimization in Cartesian, redundant internal, and hybrid-delocalized internal coordinates, as well as general functions.
• It incorporates steepest descent, conjugate gradient, quasi-Newton, L-BFGS, and damped molecular dynamics for minimization.
• Transition-state searches utilize partitioned rational function optimization, a modified dimer method, and nudged elastic band approaches.

Main Results:

• The software is implemented with coordinate transformation independence, allowing diverse algorithmic applications.
• Includes methods for conical intersection optimization (penalty function, gradient projection, Lagrange-Newton).
• Features stochastic search methods like genetic algorithms for global/local minimization, with parallel processing capabilities.

Conclusions:

• DL-FIND offers a comprehensive suite of optimization algorithms for diverse computational needs.
• The open-source nature (GNU LGPL) promotes accessibility and further development in computational science.
• Its flexibility and range of methods make it a valuable tool for researchers in chemistry, physics, and materials science.