Actions for Systematic Coarse-Graining of Interfaces

Email RIS file
Bookmark
Report an Issue

Systematic Coarse-Graining of Interfaces

Author
De Lyser, Michael
Published
[University Park, Pennsylvania] : Pennsylvania State University, 2020.
Physical Description
1 electronic document
Additional Creators
Noid, William George
Access Online

Availability

I Want It
I Want It

Finding items...

Graduate Program
Restrictions on Access
Open Access.
Summary
All-atom (AA) molecular dynamics (MD) simulations provide explicit structural details at a resolution inaccessible to experimental methods. Unfortunately, AA MD simulations are inherently limited in their length- and time-scales due to their computational requirements. Although newer and faster computers are constructed every year, there will always be relevant systems to study beyond the capabilities of AA MD. To circumvent these limitations, researchers frequently parameterize coarse-grained (CG) models which represent the relevant system in less detail. CG models are less computationally demanding than their AA counterparts because CG models represent groups of atoms as individual CG sites. In order to be of practical use, CG sites must interact in such a way as to reproduce the behavior of interest from the original AA system. Generally, there are two classes of methods used to paramterize CG models' interactions. Top-down models parameterize simple functional forms such that the CG model reproduces a given thermodynamic property of interest, such as the compressibility or the surface tension. Alternatively, bottom-up models start with an explicit AA simulation and attempt to integrate out high frequency degrees of freedom. This document will focus on this second type of CG model. Bottom-up methods all seek to optimally approximate the many-body potential of mean force (PMF) in order to reproduce specific structural properties of interest. Most bottom-up models suffer from two problems. Bottom-up models are typically only accurate at the state point at which they are parameterized. If you change the temperature, density, or system composition of your system, the CG model will no longer be accurate. This is the transferability problem. Furthermore, although the structural properties are often reproduced accurately, bottom-up models are usually unable to reproduce the proper thermodynamics. This is the representability problem. Both of these problems arise from the fact that the many-body PMF is formally a state point dependent free energy. However, researchers usually only approximate the configuration dependence of the PMF and neglect this state point dependence. In an attempt to treat the volume dependence of the PMF, Das and Andersen, and subsequently Dunn and Noid, included a term in their PMF approximation that was explicitly a function of volume. The resulting CG models reproduced the same structural features as other CG models, but also reproduced the AA equation of state. This volume potential effectively introduces corrections to the average pressure and the compressibility of the CG model, which are then applied directly to the entire system. Accordingly, this method can only be applied to homogeneous systems because, in general, each phase of iia multi-phase CG system requires its own pressure and compressibility correction. This document details my work employing local density (LD) dependent potentials in an attempt to capture the density dependence of the many-body PMF. LD potentials are many-body in nature, but the resulting forces have a convenient pairwise decomposition and therefore computationally scale the same as standard non-bonded pair interactions. First, I demonstrate a straightforward way to turn the volume potential into a LD potential and assess the accuracy of the resulting CG models. Second, I present an analysis of the use of non-bonded pair and LD interactions simultaneously and try to demonstrate the effects on the resulting CG models induced by the choice of the indicator function that defines the local density. Third, I incorporate external potentials into CG models, analyze the effects the different interactions have on each other, and demonstrate how CG models stabilize interfaces. Fourth, I present some results from expanding the LD potential with a gradient term in analogy to classical density functional theory. Finally, I conclude this work and discuss future directions for research.
Other Subject(s)
Genre(s)
Dissertation Note
Ph.D. Pennsylvania State University 2020.
Technical Details
The full text of the dissertation is available as an Adobe Acrobat .pdf file ; Adobe Acrobat Reader required to view the file.
View MARC record | catkey: 31803855