Qubit-Two Level Systems Interaction For Error Mitigation
Quantum system error prevention using qubit and two-level system interactions.
Scientists stabilise noise in superconducting quantum processors to mitigate errors. Researchers have taken a major step towards reliable, large-scale quantum processing by stabilising noise in superconducting quantum processors, boosting error mitigation measures.
Pre-Fault Tolerant Quantum Device Noise Challenge Contemporary quantum computing in the Noisy Intermediate-Scale Quantum (NISQ) era has shown the ability to precisely estimate observable values at sizes exceeding brute-force classical simulation. This capacity is enabled by quantum error mitigation (QEM) approaches, which reduce noise from observable estimates without qubit overhead.
However, probabilistic error cancellation (PEC) and zero-noise extrapolation (ZNE) require a precise, representative device noise model. Learning and maintaining this model is difficult due to unpredictable physical device noise fluctuations.
Superconducting quantum processor fluctuations are caused by qubit-defect two-level system (TLS) resonant interaction. The dispersion of Two Level Systems TLS transition frequencies drives qubit relaxation time fluctuations. Dynamic instabilities are shown to reduce error-mitigation performance or produce unphysical estimates, affecting superconducting quantum computers' stability, uniformity, and throughput.
Controlling Qubit–TLS Interaction
Trials were conducted with a one-dimensional chain of six fixed-frequency transmon qubits with tunable couplers. Noise instability was addressed by placing an electrode above each transmon qubit. A bias applied to this electrode can control the Two Level Systems TLS resonance frequency and qubit–TLS interaction by changing the local electric field at defect locations.
Qubit levels could change by nearly 300% in 60 hours. Researchers aimed to manipulate the qubit–TLS interface to reduce noise instabilities and improve QEM performance.
The group studied two noise stabilisation methods:
Optimisation Noise Strategy: Tracking the temporal snapshot of the TLS landscape, this active technique finds a value that maximises the excited state population after a predefined delay time. Avoiding qubit–TLS interactions with high strength is the goal.
Averaged Noise Strategy: This passive strategy reduces oscillations by averaging randomly selected Two Level Systems surroundings per shot. This is done by utilising sinusoidal (or triangular) amplitude modulation that varies slowly at 1 Hz, far lower than the shot repetition rate of 1 kHz. Because it samples a new quasi-static TLS environment for each shot, this approach doesn't need constant monitoring.
Overhead Reduction and Noise Model Stabilisation
These approaches were tested for noise stability connected to layers of concurrently entangling two-qubit gates as well as stabilisation. Accurate noise characterisation is essential for PEC, which uses the Sparse Pauli–Lindblad (SPL) model.
Total sampling overhead, which measures the multiplicative factor that magnifies estimator variation during error mitigation, linked noise severity to experiment cost.
In the control experiment (no modulation), noise model parameters and sampling overhead varied greatly and often correlated with high qubit–TLS interaction.
Optimised sampling overhead was lowest. The model coefficients were steady overall, save for slight transient fluctuations between optimisation cycles.
Importantly, the averaged noise approach outperformed control and optimised tests in stability over time. Smoothing out minor changes and creating more stable device noise models minimises the need for regular active monitoring.
More reliable error mitigation
The approaches were benchmarked using a 6-qubit mirror circuit, which should provide an identity operator. The ideal anticipated value of the weight-6 Pauli-Z observable is 1.
Variations from 0.341 to 0.446 in unmitigated observable values were far from the ideal value of 1. After error mitigation (PEC) with noise models:
The mitigated observable values fluctuated more in the control setting during heavy Two Level Systems interaction. Unlike the control experiment, the optimised and averaged noise solutions stabilised error mitigation findings, reducing observable estimation volatility.
The investigation showed a substantial correlation between the mitigation error and the anticipated deviation (based on noise model fluctuation), showing that time fluctuation considerably spreads the error-mitigated observable. These modulation methods stabilised QEM performance temporal fluctuation, as shown by the optimised and averaged channels' narrower distributions.
In conclusion
Controlling the Two Level Systems TLS interaction stabilises superconducting quantum processor device noise. Even if the optimised strategy yielded the best results (lowest sampling overhead), it is still subject to re-optimization adjustments. The averaging technique creates a relatively robust device noise model at non-trivial scales, making error-mitigated quantum computation on solid-state processors reliable despite a little increase in sampling overhead and bias.