Session R68: Control and Calibration Tools for Scalable Quantum Computing

Estimation of statistical significance in the quantum supremacy experiment with the Sycamore processor

Google's quantum supremacy experiment is based on sampling of output bitstrings of random quantum circuits [1]. To demonstrate quantum supremacy, it is critical to establish that the hardware fidelity at sampling time is not degraded to zero due to errors and the sampled bitstrings correspond to the expected noisy distribution. In addition, since the run time of classical approximate simulations is proportional to fidelity, it is important to verify that the fidelity is above a threshold at which classical simulation is estimated to be hard. In this talk, I will describe the methodologies for the statistical analyses and show that the bitstring distributions are extremely unlikely to be explained by noise alone and the fidelity is significantly above the threshold.

[1] Quantum supremacy using a programmable superconducting processor, Google AI Quantum and collaborators. Nature 574, 505-510 (2019).   

How OpenSuperQ is planning to fully calibrate and characterize a 100 qubit superconducting QPU over the weekend

The current methodology of designing control pulses for superconducting circuits often results in an absurd situation: simplified analytic models which do not predict gate fidelities to high accuracy, and calibrated pulse shapes which achieve good fidelities, but do not correspond to the model. For large number of qubits the QPU calibration is long and difficult, and determining a detailed error budget is nearly impossible.

To rectify the situation, we have implemented a novel procedure of Combined Calibration and Characterization (C3): An interative combination of open-loop pulse optimization, model-free tune-up and iterative model fitting and refinement, utilizing a high-performance TensorFlow simulator. It allows for a rapid, and largely automated bring-up process of QPUs. The result is a high-fidelity model, comensurate high-fidelity gates and a detailed error budget.

The above components are then utilized to implement machine-learning capabilities such as adversarial system characterization and automated experiment design, to further accelerate the process of gaining insight into the behaviour of our systems.

C3 software will be made available as an open-source project.   

Advanced Control Calibration for NISQ QPUs and Quantum Devices - Theory and Experiments

As the era of NISQ is dawning, calibration and characterization of QPUs is becoming an increasingly complex task due to the growing amount of qubits and high fidelity requirements. To tackle this problem, we developed a machine learning driven approach for Combined Calibration and Characterization procedure, C^3. This talk focuses on the gate calibration task, where optimization algorithms are used to find the best pulse parameter values in a multidimensional space.
We present a strategy which is capable of optimizing dozens of parameters, optimizing both an entangling gate and the single qubit gates of the two qubits involved, and to do so significantly faster than previously possible.
To succeed in this endeavor we engaged in an in-depth study of the multitudes of gradient-free algorithms available. The result is a unique portfolio of algorithms, for fast initial convergence and high fidelities.
Finally, we present experimental results by several labs that applied the above methodology. We show how our method takes entire gate sets all the way from 0.85 fidelities to state-of-the-art fidelities, using our open-source software.   

Automatic Calibration and Characterization of Quantum Devices - Experimental Results on NISQ QPUs and Quantum Memory devices

The practical application of optimal control techniques, especially in superconducting settings, requires extensive calibration to reach high fidelities. By controlling both experiment and a high performance numerical simulation, the C3 procedure provides a framework to systematically design and apply even intricate open loop optimal control pulses.

We demonstrate the characterization and tune-up of a quantum computing device with a few qubit, as well as a quantum memory device composed of a transmon with a 3D microwave cavity. The memory experiment presents different challenges, as the control protocol involves high-power sideband transitions, leading to system dynamics that are incompatible with simple models for qubit-cavity interactions.

In both cases, we numerically simulate the measurement of gate sequences and, by comparing the results to experimental data, improve the model of the experiment, including both simple parameters such as qubit frequencies, as well as non-trivial aspects such as line responses and bandwidth limitations of the electronic equipment that distort and imposes constraints on control signals.   

Active Learning of Hamiltonians

Querying a quantum system produces measurement outcomes which originate from the Hamiltonian dynamics of the system and its environment. Learning a Hamiltonian from a class that best fits these observations is the Hamiltonian tomography problem. Prior work has focused on estimation and offline optimal experiment design. Here, we consider an active learner that is given an initial set of training examples and the ability to interactively query the quantum system to generate new training data. The goal is to then minimize the training data required for Hamiltonian learning. To this end, we present an efficient active learning algorithm based on Fisher information and assess its performance on recalibrating superconducting qubit systems based on the cross-resonance (CR) gate. The CR Hamiltonian has up to nine parameters and admits queries involving input state, measurement operator and interaction time. Practical challenges include exponential growth of number of parameters with number of qubits, and modeling different noise sources such as readout, and imperfect pulse shapes. We show that we can achieve a constant 30% reduction in queries compared to a uniformly random approach. We also describe a regime where we achieve Heisenberg limited convergence rate during learning.   

Optimizing Quantum Gate Frequencies for Google’s Quantum Processors

A crucial component of operating a quantum processor is mitigating computational errors from energy-relaxation, dephasing, leakage, and control imperfections. In superconducting qubits, these sources of error can arise from control-electronics noise, control-pulse distortions, and the parasitic coupling of qubits to other qubits, two-level system defects, spurious microwave modes, and the control and readout circuitry. In frequency-tunable qubit architectures, it is possible to mitigate these sources of error by choreographing qubit gate frequencies over the course of quantum algorithms. This choreography maps to constructing and optimizing a high-dimensional, high-constraint, non-convex, and time-dependent objective over a search space that significantly exceeds the Hilbert-space dimension of the processor. In this talk, I will introduce the frequency optimization problem and the Snake optimizer that we developed to solve it for Google’s flagship quantum processors [1].

[1] Quantum supremacy using a programmable superconducting processor, Google AI Quantum and collaborators. Nature 574, 505-510 (2019).   

Quantum Orchestration Platform Integrated hardware and software for design and execution of complex quantum control protocols

The incredible progress in designing quantum systems, engineering their environment, and controlling them effectively, has led to significant improvements in coherence times, gate fidelities and the ability to integrate more qubits into a single quantum processor. While development of quantum processors remains the number one challenge, many bottlenecks exist in the classical control hardware layer as well as the software layer, where optimizations can play a critical role for near term quantum computing. Some examples include (1) feedback for error correction and repeat until success protocols, (2) complex calibrations, and (3) hybrid quantum classical algorithms.
Here we present a new platform for designing quantum control protocols, executing them on a wide range of quantum hardware, and optimizing performance. The platform, called Quantum Orchestration Platform incorporates a unique FPGA architecture and integrates classical hardware and software in a novel way that allows for combining the most cutting-edge real-time control capabilities with an intuitive programming environment and programing language. We show examples and present results of several protocols, including multiqubit feedback and complex control flow.   

Quantum Orchestration Platform Integrated hardware and software for design and execution of complex quantum control protocols

Quantum computing holds a great promise for immense computational power. In last years, it was suggested that noisy intermediate-scale quantum (NISQ) processors may already demonstrate an advantage over classical hardware for particular computational tasks which may be of commercial use. Nonetheless, significant additional technological progress, at every layer of the quantum computer, is required in order to advance from NISQ processors to ones capable of running fault-tolerant-quantum-computation (FTCC). In particular, for FTCC, multi-qubit real-time feedback capabilities have to be developed far beyond those we know today.
Here we present the recently developed Quantum Orchestration Platform (QOP) allowing for general multi-qubit protocols to be run. The QOP is a scalable system designed to support hybrid quantum-classical algorithms with the outlook to FTCC, and is optimized for driving multiple qubits simultaneously as well as performing complex readout schemes and branching based on readout results with ultra-low latency feedback. In this talk, we will show examples for quantum-error-correction protocols and demonstrate how they are programmed and run with the QOP.   

High-Fidelity, Scalable Quantum-Classical Control Interface using Photonics

Quantum computing (QC) has been hailed as the next big leap for the digital age; however, state-of-the-art QC devices have yet to surpass classical computers. One bottleneck is in the number of quantum operations that can be done within a qubit lifetime, also known as the "circuit depth". The circuit depth on current hardware is limited to 10-100 operations due to uncontrolled coupling to the classical environment, infidelities in the qubit operations and the gate operation time.
Effective control signal generation is essential to extending the circuit depth by improving classical control over the quantum system. To address this need, we propose an RF-photonic implementation of a QC control interface, via high-fidelity, scalable quantum drive signal generation and fiber optic transport of this signal to the qubit. By increasing the signal dynamic range and bandwidth, the time per quantum gate operation can be reduced, thus widening the circuit depth bottleneck. Additionally, replacing RF cabling with optical fiber reduces waste heat and thermalization issues, improving scalability to larger qubit systems. We show experimental results from a single-channel system and discuss its extension to a multi-channel system via simulations.   

Automatic single qubit characterization with QubiC

Initial setup a physical qubit is a complex procedure which involve multiple instruments and multiple experiments and optimization. After the initial setup, they still need to be carefully calibrated routinely because the qubit is sensitive to the environment and the qubit itself can have slow drift so as to impact the gate fidelity.
The Qubit Control system (QubiC) combines the high-speed pulse generation, data acquisition, parallel processing, and deterministic high-throughput data exchange together through the FPGA. Based on this feature, we propose a single qubit characterization and gate optimization protocol which automatically find and tune the qubit. The automatic calibration executes a full suite of serialized and recursive tasks from initial device bring-up to qubit setup to gate optimization in few hours.   

Optimal Control of Superconducting Qubits

Fast and accurate two-qubit gates are a key requirement to perform complex algorithms on current quantum computers. Ideally, the duration of the gate should be much shorter than the coherence time of the system. However, shorter gates can result in unwanted loss of states from the computational subspace. Optimal control theory aims to design fast control pulses suppressing such side effects of the driving field. Even with an accurately calibrated system model, control pulses require a tune-up to accommodate for parameter-drifts and model-inaccuracies. Here we present our work on optimal control algorithms, using a closed loop approach with direct experimental feedback to design complex pulses. This approach avoids errors from an inaccurate initial system model and uses information gained during the pulse optimization to update the model. We improve the interplay of control instruments and multidimensional optimization algorithms to speed up the tune-up of feedback-loops, reducing evaluation times from several minutes to a few seconds . With these measures, we achieve a reduction of gate errors by more than a factor of five for short pulses.   

Quantum Supremacy: computational complexity and applications

The promise of quantum computers is that certain computational tasks might be executed exponentially faster on a quantum processor than on a classical processor.
A fundamental challenge is to build a high-fdelity processor capable of running quantum algorithms in an exponentially large computational space. Here we report the use of a processor with programmable superconducting qubits to create quantum states on 53 qubits.
We create ergodic dynamics by applying a sequence that alternates between a randomly chosen single-qubit gates and an entangling two-qubit gate. Measuring the output of these quantum circuits amounts to sampling from the underlying probability distribution and produces a set of bitstrings. We verify that the quantum processor is working properly using a method called cross-entropy benchmarking, which compares how often each bitstring is observed experimentally with its corresponding ideal probability computed via simulation on a classical computer. We discuss the computational complexity of the sampling problem and show how the results of computation in a $10^{16}$ dimensional Hilbert space can be verified. Furthermore, we demonstrate how our processor is programmable by showing how various algorithms can be implemented.   

Calibration of flux crosstalk in flux qubit based quantum annealers with persistent current readout devices

Quantum annealing with superconducting flux qubits requires reliable control of externally applied magnetic fluxes. We present results on the calibration of flux crosstalk in a device consisting of two flux qubits coupled by a chain of seven couplers. The device has 27 superconducting loops and 27 bias lines. The method relies on the use of flux detectors formed of RF-SQUID terminated waveguide resonators, attached to each qubit and coupler. An iterative version of the method is demonstrated, which is more tolerant to errors arising from persistent currents in the complete circuits. The ability to perform reliable crosstalk calibration and compensation allows maintaining high coherence qubit designs and eliminates overhead in designs that are robust against crosstalk. We discuss future extensions of this work, relevant for large scale annealing control.