Correlated randomized benchmarking

As quantum circuits increase in size one of the main challenges is how to properly characterize gates in a way that is algorithmically useful. In particular, it is difficult to characterize crosstalk, i.e., unwanted interactions between qubits. These terms would show up in a full tomographic reconstruction, but such techniques are not scalable. The most successful scalable technique is randomized benchmarking (RB). In RB the ground state population is measured after applying a random identity operator constructed from n Clifford gates. The qubit population decays exponentially as the number of Clifford gates n which can be used to infer the average gate fidelity irrespective of preparation and measurement errors. This method has been widely adapted and extended to measure simultaneous gate error, leakage, purity and specific gates via interleaving. Here we will discuss how to measure crosstalk errors by analysing the decay of correlated observables in simultaneous RB data (correlation RB). We show that the decay of the correlated terms is a measure of correlated errors by applying this technique to a four qubit fixed-frequency transmon device.