Session Y64: Experiments on Current Noisy Quantum Hardware I

Performance of Robust, High-Order Dynamical Decoupling Sequences on Superconducting Quantum Hardware

The performance of today’s quantum hardware is limited by circuit depth and duration due to gate infidelity and decoherence, which adversely constrains the class of experiments achievable without error mitigation. Dynamical decoupling is an error-suppression technique that utilizes carefully timed sequences of pulses inserted during idle operation in order to cancel unwanted interactions with the environment, often allowing higher fidelity circuits to be run. A wide variety of dynamical decoupling sequences exists, ranging from simple first-order protection with uniform pulse intervals to robust, higher-order protection with non-uniform pulse interval sequences. Here, we explore and compare the performance of these various sequences on the Rigetti Aspen-M series of superconducting qubit chips. From this experimental data, we draw conclusions about the relative performance of various dynamical decoupling sequences and offer prognoses about near-term algorithmic capabilities enabled by the improvement in performance, paving the way toward performing deeper circuits with built-in environmental noise protection.   

Characterizing pulse distortions and crosstalks for superconducting qubits

Characterizing pulse distortions and crosstalks is essential to building scalable quantum computers. However, quantum chips are often operated at ultra low temperatures and insulated from the environment. This prevents one from direct characterizing the control electronics at the qubit level. We use superconducting qubits as a probe to estimate pulse distortions and crosstalks. Compared to former methods, ours does not require knowing a priori the qubit response as a function of the control pulse strengths. Moreover, its precision is not limited by the finite widths of the control pulses.   

Characterization and benchmarking of a phase-sensitive two-qubit gate using direct digital synthesis

We implemented an iSWAP gate with two transmon qubits using a flux-tunable coupler.  Precise control of the relative phase of the qubit-control pulses and the parametric coupler drive was achieved with a multi-channel instrument called Presto [1], using direct digital synthesis (DDS).  We describe the process of tuning and benchmarking the iSWAP gate, where the relative phase of the pulses is controlled via software.  Traditional iSWAP gates require multiple local oscillators and mixers with advanced synchronization methods based on external triggering.  Presto synchronizes and generates all outputs using two-step digital mixing, where pulse envelopes are defined as arbitrary waveforms at 1 GS/s. In the first step the envelopes are mixed with a digital carrier where phase (and frequency) can be set arbitrarily each time a pulse is generated. In the second step the signal is upconverted to the sampling rate of the converter and digitally mixed to the target frequency.  With Presto we perform the iSWAP gate in 254 ns, validate it with quantum-state tomography, and evaluate its fidelity with interleaved randomized benchmarking.

[1] Tholén et al., arXiv:2205.15253 [quant-ph] (2022).   

Quantum Computing Systems and Results for Hybrid Quantum Classical Algorithms

In this talk, I will summarize recent advances for hybrid quantum-classical computer design and application benchmarking at Rigetti, including for our 80 qubit scale machines. I will highlight the role of noise tailoring, error mitigation, and speed to achieve performance at scale.   

Robust hardware-efficient implementation of a continuous two-qubit gate set for transmon qubits

Hardware-efficient implementations of continuous gate sets, such as the set of two-qubit controlled-phase gates parameterized by the conditional phase, can improve the performance of noisy intermediate-scale quantum computing by reducing the circuit depth, for example in variational quantum algorithms and quantum machine learning. In a previous experimental demonstration [1], such a continuous gate set was implemented on superconducting transmon qubits, but relied on pulse shapes that are sensitive to distortions and required adjusting multiple control pulse parameters simultaneously. Here, we present an alternative implementation by extending the controlled-phase gate from [2], which is based on the resonant interaction between two flux-tunable transmons. In this implementation, an arbitrary conditional phase can be achieved by tuning a single pulse parameter, and the vanishing time integral of the employed net-zero control pulses [2] provides robustness against memory effects. Furthermore, by activating the gate via flux control of both qubits, we demonstrate that the gate can be performed between far-detuned qubits, strongly suppressing residual interactions. We characterize the continuous gate set with cross-entropy benchmarking for fixed values of the conditional phase and for phases randomly drawn from a uniform distribution, confirming a consistently high gate fidelity over the full range of phases.

[1] Lacroix et al., PRX Quantum 2020

[2] Negirneac et al., PRL 2021   

Experimental implementation and characterization of a virtual two-qubit gate

Quantum computers seem to be on a promising path to realize the capabilities which could prove to be more fast, secure, and efficient than their classical counterparts. However, the size of current quantum devices, termed as Noisy intermediate-scale quantum (NISQ) devices, is strictly limited. The other major challenges faced by NISQ devices are finite noise and limited coherence times. In relation to the limited size issue, several techniques have been proposed to increase the effective size of the devices with additional classical processing costs. One such technique was proposed by K. Mitarai and K. Fujii [New J. Phys. 23 023021 (2021)], which constructs a general two-qubit gate from only single-qubit operations or referred here as a "virtual two-qubit" gate. This virtual two-qubit gate allows us to, for example, simulate the quantum circuit of 2N qubits by using only N physical qubits with sampling overhead when the goal of the quantum circuit is expectation value estimation. Hence, it enables us to expand the computing capabilities of NISQ devices in certain algorithms. In this work, we present the experimental demonstration and characterization of the "virtual CZ" gate. While local operations on each qubit involved in a virtual gate consist of single-qubit gates and measurements, measurement errors are usually much larger than single-qubit gate errors in experiments. Thus, we have also implemented measurement error mitigation to improve virtual gate fidelity. As a result, we experimentally achieved virtual CZ gate with average gate fidelity of 0.9975±0.0028. This technique helps us to obtain expectation values in "scaled-up" quantum circuits, which are used in many quantum algorithms such as variational quantum eigensolver and others.   

Quasiparticle cooling via engineered dissipation in Floquet quantum circuits

We introduce an approach to realise non-trivial steady states in systems of many superconducting qubits out of equilibrium. The approach combines periodic application of unitary gates, which implements Floquet evolution of the system, and ancilla qubits that are periodically reset. The coupling of ancillas to the system realises a dissipative channel engineered to extract quasiparticle excitations from the system. The quasiparticles are defined in a general case by the Floquet circuit, with Hamiltonian evolution being a particular example. This drives the system to a steady state where quasiparticle occupation numbers are low, despite the uncontrollable weak noise inevitably present in the device. We demonstrate applications ranging from many-body states with low energy density to quantum transport in correlated systems. Effect of external weak noise on the steady state properties, including quasiparticle occupation numbers and entanglement, is analyzed. This work provides a scalable experimental path to realising long-lived entangled states in noisy intermediate-scale quantum devices.   

Quantum Zeno effect in presence of engineered dissipation.

The quantum Zeno effect manifests itself as the freezing of the quantum state of a system subject to repeated measurements. Decay processes, like the spontaneous emission of a photon from a cavity into the environment, are generally modelled using a Markovian environment, and , in that framework, they should not be submitted to the quantum Zeno effect. A decay process, however, can in principle be frozen if the system is measured fast enough such that the Markov approximation breaks. In this talk, we present experiments to demonstrate how the decay of a photon from a cavity into an engineered environment can be reduced by exploiting the non-Markovian nature of the bath. The experiment is implemented in a circuit QED architecture, where a photon is initially placed in a 3-D microwave cavity, and the bath is engineered such that the photon decays through another cavity into which it is swapped using a transmon qubit. Experimental investigation of the "no-click" evolution of the system when a photodetector is placed in this engineered loss channel will be presented.   

Characterization of non-Markovian noise effects in superconducting qubits using pseudo-identity gates

In superconducting quantum computers available today, interactions between qubits and two-level system (TLS) defects in the device are known to be a significant source of noise. Qubit-defect interaction can manifest itself as non-Markovian noise in the dynamics of the qubit subsystem. Existing methods to identify such effects involve low-level noise spectroscopy experiments. We develop a method based on mirrored pseudo-identity gates to characterise qubit-TLS interactions, and include them in a noise model to describe the effect of the TLS defects on the quantum circuits. We run experiments on superconducting quantum computers and find that our method is well suited to characterize the interactions between TLS defects and a qubit, and that their presence is an important source of noise. We also use the method to characterize residual cross-talk interactions between neighbouring qubits. Including the non-Markovian components within our noise model allows us to significantly improve the accuracy of the predictions of the noise model when compared to experiment.   

Towards Testing Quantum Mechanics on NISQ Devices and Characterizing non-Markovian noise on IBM Quantum Computers

We work towards testing the prediction of the emergent quantum mechanics (EmQM) by benchmarking a recently proposed protocol [1] on IBM's quantum computers. EmQM hypothesizes that quantum mechanics is not fundamental but instead emerges from theory with less computational power. The protocol involves simulating a non-Clifford quantum circuit with random 1-qubit and 2-qubit gates, followed by its inverse, and then measuring the fidelity. We also characterize the non-Markovian noise in the quantum devices by examining the fidelity decay rate. The presence of non-Markovian noise is indicated by a significant deviation from exponential decay at small circuit depth.   

Waveguide Variational Quantum Algorithms

Variational Quantum Algorithms (VQAs) [1], which make use of a classical optimizer to train a parametrized quantum circuit (PQC), have emerged as the leading proposal to leverage state-of-the-art quantum computers. However, they are still subjected to important constraints, such as scalability problems, undesired interactions between qubits and a limited connectivity, which motivate the search of new resources when designing PQCs. In this work we capitalize on the long-range interactions naturally appearing between quantum emitters coupled to waveguide-QED setups [2-4] as a resource for designing a novel class of PQCs, which we benchmark against other hardware-agnostic PQCs. Remarkably, these long-range interactions introduce entanglement between qubits in a controlled way, which allows an efficient classical training while reducing the gate count and noise impact. Our results suggest that such waveguide-mediated interactions can be a useful resource when designing VQAs.

[1] M. Cerezo et al., Nature Review Physics 3, 625-644 (2020).

[2] D. E. Chang et al., Rev. Mod. Phys. 90, 031002 (2018).

[3] Eunjong Kim et al., Physical Review X 11, 011015 (2021).

[4] Tamara Dordevic et al., Science 373 , 1511-1514 (2021).   

A co-design superconducting quantum circuit for simulating nanoscale-NMR systems

The co-design strategy, where application-specific algorithms and hardware are developed in parallel, is a viable strategy for reaching quantum advantage with NISQ computation devices. For efficiently simulating problems requiring all-to-all interactions, star-type-architectures require a minimal number of SWAP operations to map the algorithm connectivity onto the physical qubit topology. At IQM, we have designed and fabricated a star-type superconducting quantum circuit, where a distributed-element resonator is used as centralcomponent. To operate the central resonator as an effective qubit, we have developed two types of qubit-resonator operations: a SWAP operation for shuffling around excitations and a CZ-gate for algorithmic execution. In this talk, we present the experimental progress towards digitally simulating a nanoscale NMR system [1] – such as an NV center coupled to multiple nuclear spins – on such a circuit.

[1] Manuel G. Algaba et. al., arXiv:2202.05792 [quant-ph] (2022).   

Quantum reservoir computing with superconducting nonlinear oscillators

Quantum reservoir computing (QRC) is an emerging approach to quantum machine learning that harnesses the dynamics of a quantum system in order to perform computation. It is especially attractive in the era of small and noisy quantum devices as it is inherently hardware-efficient and relaxes the requirements on precise control of the system state. While many approaches to building a practical QRC have been proposed, the strong nonlinearities acheivable with superconducting quantum devices make them an attractive platform for implementing a quantum reservoir. Here, we present our development of a small quantum reservoir based on superconducting Kerr oscillators. We focus on theoertical and experimental investigations of its computational capacity, and discuss challenges in scaling up this technology towards application-relevant performance.