Dissertation Defense of CMSE Ashe Hicks
Department of Computational Mathematics, Science & Engineering
Michigan State University
Dissertation Defense Notice
August 16, 2024 at 9:00 AM
Room: BPS 1300
Zoom: https://msu.zoom.us/j/98756729855
Meeting ID: 987 5672 9855
Passcode: 772577
Algorithms for the Nuclear Many-Body Problem and Beyond
By: Ashe Hicks
Abstract:
The nuclear many body problem allows us to take our fundamental understanding of the
most basic building blocks of the universe and from them build an understanding of
larger and more complicated systems. It is the essential problem of how individual
particles form atoms and larger structures. Its applications are varied, and many
tools have been developed to address this problem. Despite the breakneck pace of computational
development, the nuclear many-body problem still stretches our computational and numerical
methods to and beyond their breaking points. In this work, we introduce two algorithms
which can help in solving the nuclear many-body problem. First, we introduce trimmed
sampling. This is an algorithm which can be used to treat noisy data obtained from
highly sensitive calculations, particularly the generalized eigenvalue problem which
emerges from a number of techniques. We solve a number of example models for which
small errors such as rounding error or statistical noise are sufficient to entirely
destroy any usable results, but see that trimmed sampling is able to recover good
results from these methods. It does so using Bayesian inference, by applying physics-informed
criteria and statistical sampling methods we are able to eliminate any solutions which
are non-physical, leaving a more accurate, physically meaningful result. We show ways
that this algorithm can be further expanded and enhanced, improving sampling statistics,
convergence rate, and accuracy, before demonstrating its performance on the Lipkin
model. In the next section, we describe the Projected Cooling algorithm. This is a
method whereby we use an analogue of evaporative cooling to calculate the ground state
of a system. We show results of projected cooling for several models. Together, this
work provides a description of useful algorithms which can be applied to the nuclear
many-body problem.
Committee Members:
Dean Lee (chair)
Scott Bogner
Huey-Wen Lin
Alexei Bazavov
Andreas von Manteuffel