The Two Wells System: A Benchmark for Monte Carlo Methods in Statistical Mechanics

Undergraduate Thesis Or Project

Public Deposited

Citeable URL: https://ir.library.oregonstate.edu/concern/undergraduate_thesis_or_projects/z603r5709

Descriptions

Attribute Name	Values
Creator	Sprueill, Henry
Abstract	Computational statistical mechanics leans heavily on the class of Markov Chain Monte Carlo Algorithms. Many such algorithms exist, and simulations can run for several weeks, making the selection of algorithms a difficult, but critical task. In this work, we introduce a benchmark system of two quadratic wells with an analytical solution, continuous state space, and phase transition. The system features an adjustable energy barrier: a common inhibitor to Markov Chain Monte Carlo algorithms. We test 4 different algorithms: the replica exchange method–Parallel Tempering, flat histogram methods–1/t Wang and Landau sampling, Stochastic Approximation Monte Carlo with a Dynamic Update Factor, and a novel method–Zeno’s Monte Carlo. Our tests reveal the adverse effect of the energy barrier on these algorithms’ performance and shed light on their sensitivity to energy barriers. The approach can be employed for testing new algorithms before wide-range deployment.
Resource Type	Undergraduate Thesis
Date Issued	2022-05-27
Degree Level	Bachelor's
Degree Name	Bachelor of Science (B.S.)
Degree Field	Physics
Degree Grantor	Oregon State University
Commencement Year	2022
Advisor	Roundy, David J.
Academic Affiliation	Physics
Rights Statement	In Copyright
Publisher	Oregon State University
Peer Reviewed	No
Language	English [eng]

Thumbnail	Title	Date Uploaded	Visibility	Actions
	The-Two-Wells-System.pdf	2022-06-10	Public	Download