High-derivative DRAG for error reduction in two-qubit and qudit gates

To overcome the challenges posed by the finite coherence time of quantum systems, an important task is devising rapid and precise control schemes. For superconducting qubits, analytical control methods based on the system's Hamiltonian are often favoured over general numerical optimization for practical experimental implementation. In this presentation, we introduce an analytical control framework using multi-derivative pulse shaping, based on the Derivative Removal via Adiabatic Gate (DRAG) technique. This approach provides an efficient, parameterized pulse Ansatz that can simultaneously suppress multiple control errors, including nonperturbative effects and multi-photon dynamics.

In this presentation, we apply this control method both to the Cross-Resonance CNOT gate and to two-level rotations in a Transmon qudit. In both cases, multiple errors are present due to the presence of a much larger Hilbert space than the targeted computational levels, where single-derivative correction brings little help. Correction of errors beyond leakage such as ZZ error is also demonstrated. Experimental testing on IBM's quantum platform results in a two to three-fold improvement for the CNOT gate on several publicly available qubits.