Physics & Astronomy ETDs

Publication Date

Summer 7-28-2026

Abstract

This dissertation studies problems at the interplay of classical and quantum computation in the settings of optimization, complexity, and sampling. In most chapters, we start from a classical combinatorial or algorithmic notion and develop its quantum counterpart or apply it to solve a quantum problem. We do this either by introducing a new model, proving a new hardness result, designing a new algorithm or proving that a known algorithm is efficient. The contributions of this dissertation split along two broad directions.   The first concerns the optimization and complexity of ground state energy problems for physically motivated quantum Hamiltonians, most of which arise as natural quantum extensions of classical combinatorial problems. We introduce quantum generalizations of Vertex Cover, characterize their computational hardness, and obtain a deterministic linear time approximation algorithm by adapting the classical local ratio method to the quantum setting for the first time. We then define a fermionic version of the Independent Set problem and prove a computational hardness result about the problem. A small modification of the same proof resolves an open conjecture about a natural problem in topological data analysis, namely the hardness of detecting topological holes in graph structured data. Next, we consider the Quantum Max-Cut problem and design a family of convex relaxations adapted to its natural symmetry. We prove results about exactness of these relaxations on several new families of graphs for Quantum Max-Cut that include well known condensed matter spin models.   The second direction develops sampling based methods for extracting information about quantum systems. We propose a Bayesian framework for measurements on a quantum device that incorporates the physical constraints relating expectation values of different observables, and demonstrate numerically that it gives substantially tighter uncertainty bars than standard estimators. We also prove that a widely used operator-loop Quantum Monte Carlo algorithm, a workhorse classical algorithm for quantum simulation that has enjoyed decades of empirical success, runs in provably polynomial time on the full class of stoquastic XY Hamiltonians with arbitrary longitudinal and stoquastic transverse fields, providing the first such guarantee in the transverse field regime.

Degree Name

Physics

Level of Degree

Doctoral

Department Name

Physics & Astronomy

First Committee Member (Chair)

Milad Marvian

Second Committee Member

Ojas Parekh

Third Committee Member

Tameem Albash

Fourth Committee Member

Akimasa Miyake

Language

English

Keywords

Hamiltonian Complexity, Approximation algorithms, Markov chains

Document Type

Dissertation

Available for download on Friday, July 28, 2028

Share

COinS