Session K70: Support and applications for quantum control

Qubit Control Strategies Using Ballistic Fluxons

Fluxons transmitted on long Josephson junctions approximate ballistic particles and have been suggested as a basis reversible computation.  These reversible circuits are attractive from the perspective of quantum computing because they are intrinsically low loss and can be fabricated from superconducting thin films and Josephson junctions—the same building blocks used for superconducting qubits. Here we explore using ballistic fluxons to bias a qubit, a first step in developing an integrated qubit controller. We present measurements of long Josephson junctions for qubit biasing and discuss simulations of measurement and readout circuits based on ballistic fluxons.

 

SNL is managed and operated by NTESS under DOE NNSA contract DE-NA0003525.

    

Single Flux Quantum-Based Digital Control of Superconducting Qubits in a Multi-Chip Module

The single flux quantum (SFQ) digital superconducting logic family has been proposed as a practical approach for controlling next-generation superconducting qubit arrays with more favorable scaling properties compared to conventional microwave-based control. In the initial implementation, the SFQ-based gate fidelity was limited by quasiparticle (QP) poisoning induced by the on-chip dissipative SFQ circuitry. In this work, we introduce a multi-chip module architecture to suppress phonon-mediated QP poisoning, where the SFQ elements and qubits are fabricated on separate chips that are joined with In bump bonds. In this work, we achieve an error per Clifford gate of 1.2(1)%, showing an order of magnitude reduction over the initial realization of SFQ-based qubit control. Additionally, we perform purity benchmarking of the SFQ-based gates to quantify the incoherent error; we measure 0.96(2)%, which can be attributed to photon-mediated QP poisoning through the electromagnetic coupling between the SFQ circuitry and the qubit. With a clear path to further reduce gate errors in future implementations, we thus demonstrate the feasibility of high-fidelity SFQ-based qubit control.   

Characterization and Suppression of Nonlinearities in a Quantum Control System Employing Direct Digital Synthesis

With continual progress in superconducting qubit lifetimes and gate fidelities, the relevance of nonidealities in classical control signals must be reevaluated. Despite established practices for calibrating imperfections in nonlinear microwave components, modern high-fidelity quantum experiments place increased demand on the temporal stability, spectral purity, and channel count of the classical control signals. In recent years, direct digital synthesis has offered a promising solution by removing the mixer from the signal path entirely [1]. However, weak nonlinearities in the digital-to-analog converter (DAC) can produce undesired spectral content which is known to affect qubit gate fidelities [2]. Building on previous efforts in telecommunications, we demonstrate the characterization and suppression of spurious emissions in a modern mixed-signal microwave control system for superconducting qubits. We contextualize our results by examining the operational limits imposed by imperfect control signals both with and without suppression of undesired distortion.

[1] W. Kalfus, et al. IEEE Trans. Quantum Engineering (2020)

[2] J.P.G. van Dijk, et al. Phys. Rev. Appl. (2019)   

A scalable quantum computing control architecture

To support the quantum computing roadmap towards real-world applications, control electronics must advance in concert with device quality, circuit scale, and algorithmic efficiency as the classical interface of quantum information processors. Control system engineering needs to satisfy challenging, and often competing, requirements of excellent signal quality, high-speed signal processing, low cost, and architecture scalability. Here, we show these challenges are addressed with several innovations in our system control solution. Pulsed microwave signals are generated with high spectral purity and stability using double-superheterodyne frequency-conversion circuits. The pulse-level sequencing technique, while providing an intuitive programming interface, enables real-time signal parameter updates and efficient, parallel qubit tune-up. Low-latency communication channels in a centralized real-time feedback network are designed to empower measurement-based quantum error correction. Combined with a performant control software layer, this quantum computing control system is capable of supporting the experimentation of both NISQ-area quantum processors and fault-tolerant quantum information machines.   

Frequency Up-Conversion Schemes for Controlling Superconducting Qubits

High-fidelity control of superconducting qubits requires the generation of microwave-frequency pulses precisely tailored on nanosecond timescales. These pulses are most commonly synthesized by up-converting and superimposing two narrow-band intermediate-frequency signals referred to as the in-phase (I) and quadrature (Q) components. While the calibration of their DC-offsets, relative amplitude and phase allows one to cancel unwanted sideband and carrier leakage, this IQ mixing approach suffers from the presence of additional spurious frequency components. Here, we experimentally study an alternative approach based on double frequency conversion, which overcomes this challenge and circumvents the need for IQ-calibration. We find a spurious-free dynamic range of more than 70 dB and compare the quality of pulse generation against a state-of-the-art IQ mixing scheme by performing repeated single-qubit randomized benchmarking on a superconducting qubit.   

Scalable and low-latency communication of classical information for advanced quantum computing and networking in a highly distributed quantum control stack

Realizing fault-tolerant quantum computers and quantum networks requires large-scale, low-latency data communication across a large number of control and computational resources. In the most stringent applications, this requires distribution in just 100s of nanoseconds. 

For these purposes Qblox has developed a highly distributed control architecture where modules execute control and readout operations on a small number of qubits by incorporating processors capable of sequencing pulses, their parameters, and measurement operations in real time. Upon executing qubit measurements, these modules transfer their results into an extended star network that allows for the distribution of these results fast and scalable. We here demonstrate fast feedback protocols between 120 control nodes, for instance capable of controlling and reading out 30 superconducting qubits.   

Signal Processing Perspectives on Pulse Design for Two-Qubit Gates in Superconducting Circuits

Although there has been tremendous progress towards achieving low error rates with superconducting qubits, error-prone two-qubit gates remain a bottleneck in realizing large-scale quantum computers. To boost the two-qubit gate fidelity to the highest attainable levels, given limited coherence times, it is essential to develop a systematic framework for optimizing pulse design. In this talk, we formulate the problem of pulse design for two-qubit gates in superconducting qubits within the context of classical signal processing. We take advantage of filter design techniques, including window functions and the Parks-McClellan algorithm to approach the problem. Our research indicates that pulse classes popular in signal processing applications – the Chebyshev pulses and pulses given by the Parks-McClellan algorithm – have the potential to outperform the Slepian pulse, which is currently widely used in quantum experiments.   

Q-Profile: Profiling Tool for quantum control stacks applied to the Quantum Approximate Optimization Algorithm

Current variational quantum algorithm implementation runtimes are dominated by classical overhead. Profiling quantum control stacks is an essential step towards mitigating these bottlenecks. However, existing benchmark suites only provide highly abstracted runtime assessment. In this work, we present Q-Profile, an open-source and hardware-agnostic tool to profile quantum control stacks. It uses direct access to the control stack, providing high accuracy in identifying performance bottlenecks in the steps of classical optimization, compilation, communication and quantum circuit execution. We demonstrate the use of our tool by profiling the execution of QAOA on a Qblox Cluster for a simulated 5 to 14-qubit transmon system. Our results identify the major execution bottlenecks in the communication and qubit reset. We provide both demonstrated and expected performance gains by implementing parallel initialization of the hardware modules and implementing active qubit reset. Furthermore, we predict the performance scaling up to 400 qubits. The tool is applicable to other benchmarks and it is included in the open-source quantify-scheduler quantum control software which supports multiple hardware back-ends.   

QICK (Quantum Instrumentation Control Kit): Progress on an open-source qubit controller

The latest improvements to QICK (Quantum Instrumentation Control Kit) leverage the success of the open-source QICK qubit controller introduced in April 2022. QICK now supports three off-the-shelf Xilinx RFSoC boards, with up to 16 output channels per board capable of direct digital synthesis of control pulses with carrier frequencies of up to 10 GHz. The companion RF board can be used to replace the rest of the warm electronics used for qubit control. We show how multiple QICK boards can be synchronized by a sub-picosecond phase lock and ultra-fast communication links. The latest QICK's new timed processor speeds up pulse generation, reduces latency and increases feedback capabilities. We demonstrate these new features with one and two-qubit randomized benchmarking examples at U. Chicago (Schuster Lab) and Princeton (Houck Lab). The Advanced Quantum Testbed (AQT) at LBNL collaborated with the QICK team on the development and testing of QICK as an AQT user project, including contributions to multiplexed readout, software integration, motherboard and daughterboard design, and cold-qubit tests.   

Scalable Software Defined Radio Platform for Precision Control of Nuclear Spin Qubits

Neutral atom quantum computing systems require precise control over the duration, frequency, amplitude, and phase of laser pulses at nanosecond timescales for cooling, trapping, state preparation, and the execution of quantum algorithms. In this talk, we will describe a custom software defined radio (SDR) hardware platform that generates long, repetitive sequences of arbitrary RF waveforms in the 10 MHz-10 GHz range to selectively address individual qubits and actuate sequences of 1- and 2-qubit gates. The control system must perform classical computations such as loops, branches, and subroutine callouts in real time and within the coherence time of the qubits in order to support quantum error correction and other iterative operations. We have implemented a lightweight, highly efficient flow control mechanism for these classical computations by means of Lua scripts that execute on the embedded processor of the SDR and feed CPU-like instructions to the input buffers of the arbitrary waveform generator.   

Efficient autonomous system-wide gate calibration

An important step in the path towards full-fledged quantum computation is the ability to maintain high fidelity gates across a full device and across time. Current NISQ devices are error-prone and fragile, suffering from a wide range of errors, including dephasing, leakage and cross-talk, which may lead to low quality gates and drift in gate performance over time. In this talk we present a quantum control pipeline, based on Clifford circuits, to efficiently perform a system-wide calibration through parallel and automated-closed loop optimization of microwave pulses. Widely used methods such as randomized benchmarking tend to underestimate coherent errors, while methods that are based on process tomography require significant resources. Our method maintains high sensitivity to both coherent and non coherent errors while being machine efficient, allowing full device calibration in a matter of minutes. We show that our implemented single and two qubit gates yield fidelities near the T1 limit, and demonstrate our methods across different hardware providers and across devices, showing over 7x improvement in fidelity compared to default gates.   

Simulating self-consistent reference states on a quantum computer using optimized custom gates

Self-consistent reference states are a critical starting point for computing many phenomena in physics. The Hartree-Fock (HF) ground state is a ubiquitously used self-consistent reference state for advanced methods to describe both static and dynamic properties of quantum many-body systems. It has been shown that the HF method can be implemented on a superconducting quantum computer by realizing the basis rotation using Givens rotations on a parametrized quantum circuit. In this work, we optimize this procedure by employing optimal control pulse theory to find customized microwave pulses that realize the Givens rotation gate. This leads to a significant reduction in the implementation time of the quantum HF algorithm and a high-fidelity Givens gate. Furthermore, we use machine learning to generalize our method to find optimal pulses for a family of Givens gates parametrized by a continuous variable of the rotation angle. We test the accuracy of our method on exactly solvable nuclear models and observed a good agreement. Thereby, this work contributes towards transformative calculations of quantum simulations on near-term quantum devices.   

Continuous Quantum Gate Sets and Pulse-Class Meta-Optimization

In the context of quantum computing gate-synthesis and control problems present a vast range of external parameter dependencies, both physical and application specific. These parameters may represent e.g. Hamiltonian-specific terms, which drift in time or vary significantly from one system to another, or alternatively different gate configurations and even specific experimental parameters such as those describing e.g. filter and transfer functions. In this work, we address the possibility of learning families of optimal-control pulses that depend adaptively on various parameters, in order to obtain a global optimal mapping from the space of potential parameter values to the control space and hence to produce continuous classes of gates. We test our algorithms on different experimentally relevant quantum gates in the context of superconducting quantum circuits and show how we can construct high-fidelity pulses even in the presence of multiple variables or uncertain parameters with wide ranges. We also discuss and simulate different possible direct applications of our methods to experiments.