TL;DR
This paper introduces a novel, automated, and circuit-model-independent method called CISCIQi for calibrating flux crosstalk in large-scale superconducting quantum circuits. This iterative approach significantly improves the accuracy of controlling individual flux lines, reducing errors to below 0.17%, which is crucial for building reliable quantum computers.
What Problem Does This Paper Address?
Superconducting quantum computers rely heavily on precisely controlling magnetic flux to tune qubit frequencies and implement quantum gates. However, in complex circuits with many control lines and interacting components (like qubits and couplers), these control signals inevitably "crosstalk," meaning a signal intended for one component affects others. This crosstalk is a major hurdle for scaling up quantum computers, as it makes independent control of individual elements extremely difficult. Existing methods often struggle with strong interactions between circuit elements or require detailed circuit models, which are hard to obtain for large, complex devices. This paper tackles the challenge of accurately calibrating this flux crosstalk in large, interconnected superconducting circuits.
Methods
The authors propose an iterative calibration method called CISCIQi. It's designed to be "circuit model independent," meaning it doesn't rely on knowing the exact physical layout or precise values of inductances and capacitances. Instead, it leverages fundamental symmetry properties of superconducting circuits.
The core idea is an iterative process:
- Initial Estimate (CISCIQ): The first step, CISCIQ, involves a four-stage procedure that measures various coupling coefficients and flux offsets. This stage makes some simplifying assumptions about interactions.
- Iterative Refinement (CISCIQi): Subsequent iterations use the results from the previous step to refine the estimates. Instead of controlling voltages directly, the method controls "flux coordinates" derived from the current estimates. This iterative refinement gradually accounts for the strong inductive couplings between circuit elements that were initially ignored.
- Error Characterization: A separate measurement technique is introduced to quantify the remaining crosstalk errors after calibration. This involves checking how well individual flux changes can be isolated.
The data analysis relies on advanced image processing techniques, specifically recurrence plots and KAZE feature detection, to extract symmetries from experimental data.
Datasets
The methods were tested on two custom-built quantum annealing devices (Device A and Device B) with up to 27 control loops. These devices are not presented as new benchmarks but rather as experimental platforms to validate the proposed calibration technique.
Key Insights and Ideas
- Iterative Refinement is Key: The most significant insight is that a single-pass calibration is insufficient for complex, strongly interacting circuits. Iteratively refining the crosstalk matrix and flux offsets is essential to accurately capture these interactions.
- Symmetry is Your Friend: The method cleverly exploits inherent symmetries in superconducting circuits (periodicity and mirror symmetry) to extract calibration parameters without needing detailed circuit models.
- Automated and Scalable: The procedure is automated, making it suitable for calibrating the large number of control lines required for scalable quantum processors.
- Error Characterization is Crucial: Developing a method to quantify the remaining errors is as important as the calibration itself, providing confidence in the achieved accuracy.
Key Results
- The CISCIQi method successfully reduced crosstalk errors. For Device A, the errors in measured crosstalk coefficients were all below 0.17%.
- The iterative process showed clear convergence, with coupling coefficients approaching identity matrices and flux offsets approaching zero after a few iterations. Device B, with more complex interactions, required more iterations to reach convergence.
- The authors demonstrated that their calibration accuracy is comparable to or better than previous methods, especially in the presence of strong circuit interactions.
- Flux offsets were found to drift over time (on the order of mΦ₀), but the calibration remained stable for several days, which is sufficient for many quantum annealing experiments.
Limitations and Future Work
- Measurement Time: The data acquisition process is time-consuming, taking many hours for the first iteration, although further iterations are faster. The authors suggest that optimizing readout schemes and software could significantly speed this up.
- Circuit Interactions: While the iterative approach mitigates errors from circuit interactions, it doesn't eliminate them entirely. The authors acknowledge that systematic errors due to these interactions are the primary limitation.
- Future Work: The authors suggest exploring optimization algorithms for a more direct approach to crosstalk calibration and investigating the impact of multi-tier architectures on reducing crosstalk at the design level. Further work is also needed to understand and mitigate the sources of flux drift.
Summary updated