Link to the University of Pittsburgh Homepage
Link to the University Library System Homepage Link to the Contact Us Form

Quantum Monte Carlo and Molecular Fragmentation Methods for the Treatment of Electron Correlation in Molecules and Solids

Dumi, Amanda (2023) Quantum Monte Carlo and Molecular Fragmentation Methods for the Treatment of Electron Correlation in Molecules and Solids. Doctoral Dissertation, University of Pittsburgh. (Unpublished)

Updated Version

Download (22MB) | Preview


The quantum mechanical treatment of molecular systems in computational chemistry offers insight into the nature of chemical bonding, reaction mechanisms, and many experimental observables. An improved treatment of the molecular electronic structure is tied to an increase of computational resources.
This work explores two approaches that attempt to describe the electronic structure with two families of approximations which reduce the computational cost without significantly sacrificing accuracy.

One approximation explored is the application of Quantum Monte Carlo methods, which stochastically solve the Schrödinger equation.
Here, the Diffusion Monte Carlo formulation is used to provide insight into chemical systems that are not well described by the commonly utilized \acrlong{dft}.
Two systems are explored: a hydrogen atom chemisorbed to the surface of graphene and a model non~valence correlation-bound anion.
The Diffusion Monte Carlo approach is systematically improvable for most approximations, except for the fixed node error which can often be addressed through a careful choice of trial wave functions.
Trial wave functions composed of multi-Slater determinants as generated by a selected configuration interaction procedure are investigated as they produce a compact determinant expansion through selecting the most important determinants for a specific system in an iterative fashion.

The second approach is the development of a fragment selection scheme through unsupervised machine learning approaches.
Fragmentation approaches are motivated by the short range nature of correlation effects.
The full system is approximated by subsystems which are each treated at a certain level of theory and an estimate of the interactions between them.
For this approach to return valuable results, the chemical domains need to capture the most important physics of the desired problem.
In this work an unsupervised machine learning based method is developed which will allow the systematic identification of important chemical domains with minimal quantum mechanical data which can improve transferability and automation of fragmentation approaches


Social Networking:
Share |


Item Type: University of Pittsburgh ETD
Status: Unpublished
CreatorsEmailPitt UsernameORCID
Dumi, Amandaaed63@pitt.eduaed630000000226938537
ETD Committee:
TitleMemberEmail AddressPitt UsernameORCID
Committee ChairJordan,
Committee MemberHutchison, Geoffreygeoffh@pitt.edu0000-0002-1757-1980
Committee MemberLaaser, Jenniferj.laaser@pitt.edu0000-0002-0551-9659
Committee MemberLambrecht,
Date: 10 May 2023
Date Type: Publication
Defense Date: 8 December 2022
Approval Date: 10 May 2023
Submission Date: 12 January 2023
Access Restriction: No restriction; Release the ETD for access worldwide immediately.
Number of Pages: 195
Institution: University of Pittsburgh
Schools and Programs: Dietrich School of Arts and Sciences > Computational Modeling and Simulation
Degree: PhD - Doctor of Philosophy
Thesis Type: Doctoral Dissertation
Refereed: Yes
Uncontrolled Keywords: Quantum Monte Carlo, electronic structure, molecular fragmentation, graphene, non-valence correlation-bound
Date Deposited: 10 May 2023 17:13
Last Modified: 22 Feb 2024 18:07


Monthly Views for the past 3 years

Plum Analytics

Actions (login required)

View Item View Item