[Paper Review] Scalable in-situ qubit calibration during repetitive error detection

J. Kelly, R. Barends+20 more|arXiv (Cornell University)|Mar 9, 2016

Quantum Information and Cryptography4 citations

TL;DR

This paper introduces Active Detection Event Parameter Tuning (ADEPT), a scalable, in-situ method for optimizing qubit control parameters during repetitive error detection in superconducting quantum processors. By using error detection outcomes as feedback, ADEPT enables real-time calibration of single- and two-qubit gates and compensates for independent frequency drifts across all qubits without interrupting computation, achieving O(1) scalability for large-scale fault-tolerant quantum computing.

ABSTRACT

We present a method to optimize qubit control parameters during error detection which is compatible with large-scale qubit arrays. We demonstrate our method to optimize single or two-qubit gates in parallel on a nine-qubit system. Additionally, we show how parameter drift can be compensated for during computation by inserting a frequency drift and using our method to remove it. We remove both drift on a single qubit and independent drifts on all qubits simultaneously. We believe this method will be useful in keeping error rates low on all physical qubits throughout the course of a computation. Our method is O(1) scalable to systems of arbitrary size, providing a path towards controlling the large numbers of qubits needed for a fault-tolerant quantum computer

Motivation & Objective

• To develop a scalable method for in-situ calibration of qubit control parameters during continuous error detection in large-scale quantum processors.
• To address the challenge of time-varying control parameters and parameter drift in physical qubits that degrade gate fidelity over time.
• To enable real-time optimization without interrupting error detection or requiring qubit state tomography or randomized benchmarking.
• To demonstrate O(1) scalability of calibration across arbitrary qubit arrays using error detection outcomes as feedback.
• To maintain low error rates throughout long computations by dynamically compensating for drift in real time.

Proposed method

• ADEPT uses error detection outcomes—specifically, the rate of detection events (e.g., bit-flip errors)—as a feedback metric to optimize control parameters.
• The method partitions qubits into fixed-size groups (e.g., data and measurement qubits in a repetition code) to allow independent calibration of gate parameters within each group.
• Control parameters such as Rabi frequency, detuning, and gate angles are adjusted based on the error rate observed in each group’s detection outcomes.
• A hardware pattern interleaving strategy is used to cycle through different qubit groupings, enabling simultaneous calibration of all qubits without overlap or interference.
• A frequency-following algorithm is combined with ADEPT to track and compensate for independent frequency drifts on each qubit in real time.
• The technique is implemented on a nine-qubit superconducting processor using the surface code’s repetition code, with calibration performed during active error detection.

Experimental results

Research questions

• RQ1 Can qubit control parameters be optimized in real time during active error detection without interrupting computation?
• RQ2 Can ADEPT scale to arbitrary qubit array sizes while maintaining O(1) computational overhead per qubit?
• RQ3 Can independent frequency drifts on each qubit be tracked and compensated for simultaneously during computation?
• RQ4 Does using detection event rates as a feedback metric enable stable, low-error operation over long durations?
• RQ5 Can ADEPT be applied to various quantum error correction codes beyond the repetition code?

Key findings

• ADEPT successfully optimized single- and two-qubit gate parameters in parallel on a nine-qubit superconducting processor during active error detection.
• The method compensated for independent frequency drifts of ±10 MHz on each of nine qubits over 48 million detection rounds, equivalent to ~42 seconds of real-time operation.
• All measurement qubits maintained a stabilized error rate metric (ζ), indicating successful compensation and stabilization below the randomization limit.
• The error rate metric ζ remained stable throughout the experiment, demonstrating that drift was effectively tracked and corrected in real time.
• The technique achieved O(1) scaling with system size, enabling calibration of all qubits without increasing overhead per qubit.
• ADEPT does not require qubits to operate below logical threshold and can be applied to any error correction code with bounded group size and fixed number of group memberships per qubit.