Author ORCID Identifier

https://orcid.org/0009-0007-3431-8382

Date of Award

Fall 2026

Document Type

Thesis (Ph.D.)

Department or Program

Physics and Astronomy

First Advisor

Lorenza Viola

Abstract

Current quantum processors, at the intermediate scale of tens to hundreds of qubits, remain error-limited. This thesis studies two related sources of error. The first is environmental noise, which may have temporal and spatial correlations and nonclassical components. The second is state-preparation and measurement (SPAM) error, which arises in the operations used to characterize this noise, a prerequisite for boosting operational fidelities. Neither can be characterized alone. Noise spectroscopy techniques use imperfect preparation and readout, while SPAM characterization is affected by qubit decoherence. Our methods vary measurement depth, drive duration, or sequence repetition so each source changes the measured signal differently. Regression separates them without assuming either is negligible. We introduce quantum SPAM (QSPAM) protocols. These separate state-preparation and measurement errors in parallel across a register using only single-qubit gates and repeated measurements at a fixed cost per qubit. Tests on a 127-qubit IBM Quantum device validate the protocols. SPAM-mitigation bias grows linearly with register size and preparation error. SPAM-robust spin-locking noise spectroscopy reconstructs multiaxis spectra, including nonclassical components, without assuming ideal SPAM. On hardware, neglecting SPAM shifts spectra upward by up to 26.4% and can make quantum spectra unphysical. Two-qubit frequency-comb spectroscopy supplies spectra for noise-tailored entangling-gate design. The idling and entangling gates can outperform generic protocols employing dynamical decoupling by factors of 23.7 and 9.8, respectively, in appropriate parameter regimes. For our product pair–spectator inputs, the worst-case leading-order gate error is pair-local when distinct two-body neighbor channels are mutually uncorrelated. For bounded connectivity and sequential pair operation, characterization and design costs grow only linearly with qubit number. Spectra accessible through two-qubit spectroscopy also guide entanglement storage. When leakage invalidates the two-level SPAM model, a qutrit framework recovers all eight parameters in closed form and tests physicality. It distinguishes the coherent post-measurement rotation caused by a non-diagonal measurement from the classical state update of an inefficient detector. Quantum fault-tolerance estimates depend on assumptions about the locality and independence of physical-layer errors, which must be tested on each device. Taken together, our results replace assumptions about device errors with measurements that can guide control design, providing tools for this challenge.

Download

Share

COinS