Date of Award
Doctor of Philosophy
Energy Science and Engineering
Travis S. Humble
Cristian D. Batista, Raphael C. Pooser, Bruce J. MacLennan
Quantum computers are a promising technology expected to provide substantial speedups to important computational problems, but modern quantum devices are imperfect and prone to noise. In order to program and debug quantum computers as well as monitor progress towards more advanced devices, we must characterize their dynamics and benchmark their performance. Characterization methods vary in measured quantities and computational requirements, and their accuracy in describing arbitrary quantum devices in an arbitrary context is not guaranteed. The leading techniques for characterization are based on fine-grain physical models that are typically accurate but computationally expensive. This raises the question of how to extend characterization efficiently to larger scales. We present an empirical-based approach to direct characterization of quantum circuits that reconciles accuracy with scalability by using a reduced set of test circuits that target a chosen application and coarse-graining the noise modeling process to reduce the model complexity. We show that this method performs well in tests with Greenberger-Horne-Zeilinger-state preparation circuits and the Bernstein-Vazirani algorithm, though it does not describe all error present in the system. We benchmark this method with the leading methods of gate set tomography, cycle benchmarking, and Pauli channel noise reconstruction to characterize quantum circuits and we compare the accuracy of these methods in predicting quantum device behavior. We find that our method for empirical direct characterization offers competitive accuracy when compared with finer-grained techniques, while significantly reducing the resources required for characterization. By testing on quantum devices, we quantify the quantum and classical resources required for each characterization method and we monitor the decrease in accuracy as a function of circuit size. We find that these characterization methods can provide an accurate estimate of a quantum computer's performance on a benchmark but the best-performing method varied by test. Our results indicate that these characterization methods perform well in describing the noise of a quantum computer but their performance depends on the size and the context of the application.
Dahlhauser, Megan L., "Characterization and Benchmarking of Quantum Computers. " PhD diss., University of Tennessee, 2021.