Improving quantum circuits with heterogenous gatesets*

It is well known that a very small set of gates achieves universal quantum computation. This has enabled programmable quantum computers with minimal calibration overhead.

 

In this talk I will present some theoretical as well as experimental evidence that a small expansion of available gatesets will greatly improve the quality of circuits on near-term hardware. These "overcomplete" gatesets are often readily available from the same Hamiltonians that give rise to basic gates such as U and CNOT.

 

First, I will discuss an optimal algorithm for decomposing arbitrary SU(4)s over discrete XX-type basis gates (e.g. cross-resonance or Molmer-Sorensen interactions). I will show that calibrating just 3 discrete angles is enough to nicely approximate the infinite-calibration regime. Second I will discuss XX+YY-type basis gates and optimal decompositions over those, and how Qiskit's compiler picks the best basis for optimal approximation. Depth reductions and experimental fidelity improvements will be discussed for various kinds of circuits, including Linear Functions, Cliffords, QFT, and Quantum Volume.