Researchers Link Random Circuit Errors to Specific Noise Sources

Difficulty in connecting overall quantum processor performance metrics to specific noise sources has long hampered error interpretation; now researchers at Sandia National Laboratories and JILA have established a method for directly relating randomized benchmarking decay rates to elementary error types. The team showed that Pauli-stochastic errors contribute linearly to RB parameters while Hamiltonian errors follow an explicitly defined quadratic form using three coordinates describing single qubit behaviour.

They developed a method to connect overall quantum computer performance scores from randomized benchmarking with specific types of errors; previously these scores were difficult to interpret as it was unclear if they reflected widespread minor issues or just a few major ones.

Random “flips” in qubit behaviour predictably contribute to benchmark results, while consistent drifts follow a defined mathematical relationship involving three key characteristics of single qubits. The team has demonstrated that randomized benchmarking, essentially a standardized test giving multiple-choice questions to assess accuracy across different areas, can be used to pinpoint elementary error types. This research focuses on linking these benchmarks to the underlying causes, rather than measuring an overall score.

Disentangling Qubit Errors via First Order Gauge Invariant Coordinates

A mean-length normalized Clifford randomized benchmarking (RB) error of 0.009 was achieved, an improvement over previous limitations where disentangling noise sources at this scale proved impossible. Pinpointing contributions from specific error types like qubit over-rotation and axis tilt is now possible through RB results, moving past reliance on assuming average gate infidelity. This advancement stems from modelling errors as deviations from ideal operations and utilising first-order gauge invariant coordinates: a method for simplifying complex quantum behaviour by focusing on key characteristics rather than detailed pathways.

Qubit over-rotation and axis tilt were isolated using the new approach, exceeding simple measurements of overall gate infidelity. Errors are modelled as small deviations from perfect operations; Pauli-stochastic errors impact the RB decay parameter linearly while Hamiltonian errors manifest predictably via a quadratic relationship. For example, analysis of the single-qubit native gate set comprising Xπ/2 and Yπ/2 gates revealed explicit formulas relating these errors to two intrinsic over-rotations and one relational axis-tilt error, allowing accurate prediction of how direct Randomized Benchmarking (RB) responds to different sampling distributions.

This validation against numerical simulations showed discrepancies only at higher orders. A key challenge for the team is extending these findings to highly noisy devices exhibiting non-Markovian behaviour.

Error decomposition reveals diagnostic potential despite simplified noise modelling

The research clarifies how specific errors impact randomized benchmarking results, moving beyond simply quantifying overall fidelity; it relies on a simplified picture where quantum noise is small and localized. The limitation raises questions about whether their neat separation of Pauli and Hamiltonian contributions holds true when error characteristics become more complex or correlated in real-world systems.

Acknowledging assumptions about well-behaved, small noise, and its limitations when applied to today’s most challenging quantum computers, this work provides a step forward in diagnosing errors. Establishing a direct link between elementary quantum error rates and randomized benchmarking decay was previously difficult due to assuming average performance across all gates; now there are specific targets for improvement beyond reducing total error rates. This allows separation of Pauli-stochastic errors (affecting benchmarks linearly) from Hamiltonian errors that follow a defined mathematical relationship involving the same coordinates, enabling targeted improvements in future hardware.

This research demonstrated how individual gate errors contribute to the overall decay observed in randomized benchmarking experiments. It establishes a connection between experimentally measurable parameters and fundamental quantum error rates, specifically, two over-rotations and one axis-tilt error for the Xπ/2 and Yπ/2 single-qubit gates, allowing researchers to interpret benchmark results with greater precision.

The study shows that Pauli-stochastic errors affect benchmark decay linearly while Hamiltonian errors are described by a quadratic form using these identified characteristics. These findings provide a means of diagnosing specific sources of noise beyond simply quantifying total error, which may be useful as they extend their work toward more complex, highly noisy devices.

More like this

Quantum Research NewsResearch Cuts Quantum Simulation Costs by 61% with ArcTN

Quantum Research NewsResearchers Cut Simulation Errors Using Hamiltonian Norms

Quantum Research NewsResearchers Cut Quantum Computing Costs with New Scheduler

Quantum Research NewsPavia Team Finds Link Between Qubit Resources and State Overlap