Performances of Wang-Landau algorithms for continuous systems.

Area of Science:

• Computational physics and chemistry
• Statistical mechanics
• Materials science

Background:

• The Wang-Landau (WL) method is a powerful simulation technique for calculating density of states.
• Assessing WL method performance is crucial for its reliable application in complex systems.
• Continuous degrees of freedom present unique challenges for simulation methods.

Purpose of the Study:

• To evaluate the performance of different Wang-Landau method implementations.
• To compare WL method accuracy against parallel tempering Monte Carlo simulations.
• To identify and address limitations of the WL method for specific systems.

Main Methods:

• Simulations of two polypeptides and two Lennard-Jones (LJ) atomic clusters.
• Variations in the multiplicative factor 'f' during WL simulations were analyzed.
• Parallel tempering Monte Carlo simulations were used as a reference.
• Complementary order parameters and 2D joint density of states were calculated.

Main Results:

• WL method achieved comparable accuracy to parallel tempering for the studied systems.
• Significant difficulties were observed in reproducing low-temperature transitions for LJ clusters.
• The LJ(38) system, known for its complexity, showed notable improvement with enhanced WL methods.
• Comparable accuracy to parallel tempering was not achieved for the LJ(31) system using WL multicanonical sampling.

Conclusions:

• The Wang-Landau method is a viable alternative to parallel tempering for certain systems.
• Low-temperature transitions in atomic clusters remain a challenge for standard WL implementations.
• Enhancements like complementary order parameters and 2D density of states significantly improve WL performance.
• Further methodological development is needed for challenging systems like LJ(31).