UC Berkeley

Quantum computing is a relatively new computing paradigm that leverages entanglement to implement algorithms beyond the reach of classical hardware. Superconducting quantum circuits have emerged as a platform to build quantum processors, with demonstrations of quantum supremacy over classical supercomputers and quantum error correction beyond fault tolerance. However, current noisy devices require significant improvements to unlock the full potential of quantum algorithms.This thesis investigates intrinsic noise channels in superconducting circuits and the design of noise-resilient qubits. By characterizing high-quality superconducting resonators, we identify lossy dielectric materials at the interfaces as the dominant loss mechanism, which we mitigate via targeted surface treatments. Subsequently, we leverage Josephson junction nonlinearity to engineer noise-biased qubits utilizing the Kerr parametric oscillator (KPO). By analyzing the efficiency of strong, far-off-resonance driving to facilitate three- and four-wave mixing processes, we demonstrate high-fidelity operations on Fock, Kerr-cat, and dual-rail qubits. For the Kerr-cat qubit, we introduce a novel 2D architecture that enables strong light-matter coupling with suppressed Purcell decay, yielding millisecond-scale bit-flip lifetimes. Within this same architecture, we demonstrate a longitudinal qubit readout with a $99.6\%$ quantum nondemolition score, alongside a high-fidelity universal single-qubit gate set, all while maintaining bit-flip error suppression. Finally, we exploit the efficient wave-mixing of strongly coupled, far-detuned KPOs to realize a dynamically protected dual-rail qubit exhibiting both erasure bias and high-fidelity single-qubit gates.