Thread by @kenbrownquantum

Natalie Brown (@GTPhys), Andrew Cross (#IBMQ), and I posted a new paper on leakage errors in the surface code to the arXiv today. arxiv.org/abs/2003.05843 @DukeEngineering #DukeQuantum

Natalie recently defended her thesis on leakage errors. First, she compared whether it was better to have a magnetic field sensitive qubit or a leaky qubit using the standard depolarizing error model for the leaky qubit interaction. arxiv.org/abs/1803.02545

Then Natalie, Mike Newman ,and I realized the physical model of how leaked states and qubit states interact for trapped ion gates is nicer than the standard depolarizing error model. arxiv.org/abs/1903.03937 arxiv.org/abs/1904.10724

What about the leakage interaction models for other physical systems? Natalie was visiting Andrew at IBM as part of the @NSF QISE-Net triplet program. @jaygambetta mentioned to Natalie that for cross-resonance two-qubit gates, one qubit is more likely to leak than the other.

From her work on ions, Natalie knew that leakage on syndrome qubits was more damaging than leakage on data qubits because of how error spread. Taking advantage of the asymmetry of leakage after a cross-resonance gate, she could design improved gate compilations for the code.

Again we see that the physical errors of the system inform the best choice for circuit compilation.

For those allergic to device specific error models, Natalie went one step further and designed a new leakage reduction circuit that works for the standard leakage error model and saves gates by fixing leakage on syndrome and data qubits separately.