Session F30: Quantum Computing Architectures II

Live

In this work, the dynamics of chaotic quantum maps is explored via simulation as a means to test the fidelity of emerging quantum computing hardware. Quantum computers promise to deliver enormous gains in computational power that can potentially be used to benefit Fusion Energy Sciences (FES). Through the quantum-classical correspondence principle, quantum systems of sufficiently large quantum number (or number of qubits) can approximate classical dynamics. Here we study the simplest types of chaotic dynamical systems, defined by classical and quantum maps. It’s been shown that quantum maps of sufficient fidelity can recreate small-scale classical phase space structures in the limit of many qubits [G. Benenti, et al. Phys. Rev. Lett. 87, 227901-1 (2001)]. They can also deviate from the classical dynamics and display dynamical Anderson localization. In this work phase space structures are sought on current hardware, using IBM’s 5-qubit devices and the LLNL Quantum Design and Integration Testbed (QuDIT) platform, with verification from gate set tomography (GST).   

Live

For superconducting quantum computers, most efforts seek to implement a “surface code” architecture, which only couples nearest-neighbor qubits. In such a computer, operations between distant qubits require a large number of nearest-neighbor gates to implement with concomitant increases in gate errors and run time. In contrast, a modular architecture allows for long-range couplings between distant qubits. We have realized a modular quantum state router based on three-wave couplings with all-to-all couplings between 4 modules. We have connected the router to four simple modules consisting of a high Q communication cavity which couples to the router, a single transmon qubit and a readout cavity to demonstrate feasibility of operating the router + module systems as a quantum machine. In this talk, we will demonstrate basic operations in our machine: transferring states and generating entanglement among the modules’ communication modes and qubits. Furthermore, we will discuss the potential for utilizing ancillary modes in the router as ancillary quantum storage, as well as expanding the router system to form a large scale quantum router for an arbitrary number of modules.   

Live

The creation complexity of a quantum state is the minimum number of elementary gates required to create it from a basic initial state. The creation complexity of quantum states is closely related to the complexity of quantum circuits, which is crucial in developing efficient quantum algorithms that can outperform classical algorithms. A major question unanswered so far is what quantum states can be created with a number of elementary gates that scales polynomially with the number of qubits. In this work we first show for an entirely general quantum state it is exponentially hard (requires a number of steps that scales exponentially with the number of qubits) to determine if the creation complexity is polynomial. We then show it is possible for a large class of quantum states with polynomial creation complexity to have common coefficient features such that given any candidate quantum state we can design an efficient coefficient sampling procedure to determine if it belongs to the class or not with arbitrarily high success probability. Consequently partial knowledge of a quantum state’s creation complexity is obtained, which can be useful for designing quantum circuits and algorithms involving such a state.   

Live

An important figure of merit for scalable quantum computing is the cost per channel per unit bandwidth of its control electronics. Fortunately, 5G telecom is a major market force pushing this cost down. These fields share in fact many technical requirements: multiple phase-coherent and wide-band channels for output and input; low noise and low distortion with low cross-talk between channels; easy to reconfigure and field-programmable, with high-speed logic for feedback and feed-forward control.

These requirements have driven the development of a new series of radio-frequency system-on-a-chip (RFSoC) devices from Xilinx. Zynq UltraScale+ RFSoC devices allows for direct digital synthesis and sampling of signals up to 6 GHz, without analog mixers. With 16 channels of input and output coupled to one very large FPGA, these circuits allow for multiple, synchronous signal generation and readout at a cost well below $1.00 / channel / MHz.

We have adapted a RFSoC platform for the characterization and control of superconducting qubit circuits. We describe the hardware modifications and firmware extensions needed for this adaptation. We demonstrate the use of our platform in characterizing a superconducting qubit.   

Live

The next milestone in photonic quantum information processing is to
demonstrate an optical experiment at which the quantum device outperforms
any classical computer. This can be achieved by sending
single photons through a passive linear-optical network. However, this
requires the generation of many identical single photons, which is challenging
to realize. Spectral impurity reduces the visibility of quantum
interference of the photons, which can be mitigated by filtering at the
cost of optical losses. Unfortunately, these losses are also detrimental
to quantum interference.

We recently demonstrated how to analyze the role of imperfections in
multiphoton interference experiments [1]. We now apply these results
to the problem of constructing single-photon sources. In this work
we show that an optimum exists where we can outperform a classical
computer, using off-the-shelf parametric down-conversion photon
sources. These results show that demonstrating a quantum advantage
using photonics is difficult, but possible.
[1] J.J. Renema et al., arXiv: 1809.01953 (2018)   

Live

Efficiently controlling the quantum state of 3D cavity modes is an important ingredient for exploiting their long lifetimes and restricted decoherence channels for quantum information processing. Here we experimentally explore the use of Photon-Number Selective Phase (SNAP) gates and displacements to generate Wigner-negative states useful for continuous variable quantum computing. Our state-preparation protocol consists of a sequence of interleaved SNAP gates and coherent displacements. We use gradient descent algorithm to optimize the parameters of the sequence, and characterize fidelities to the target state by Wigner tomography. It has been shown theoretically that just a few of these blocks can be used to generate highly nonclassical states with high fidelity.   

Live

In addition to communication networks, routers and repeaters, actively under development in the context of quantum information processing, multi-core architectures are a critical component of distributed computing. As quantum processors scale and quantum networks are realized, motherboards will be required to coherently route information between chips, enabling entanglement between cores in a quantum processor or possibly between quantum memory or processing modules deriving from different physical architectures. Furthermore, for superconducting qubit integrated circuits, a modular processor composed of several smaller individual chips mitigates the impact of exponentially decreasing chip yield as the number of qubits per die increases. Towards these aims, we present here experimental results from a multi-core quantum processing unit: a 32-qubit platform formed from short-range interconnects across four 8-qubit chips. We show that this inter-chip coupling does not significantly impact single-qubit or two-qubit performance, and conclude with examples of high fidelity multi-qubit algorithms across these inter-chip edges.   

Live

For decades, Quantum Computing has been seen as a compelling possibility as a novel form of computing offering the potential of speedups large enough to make useful intractable problems tractable. In recent years, near-term quantum computers have been built demonstrating the basics of the approach. A huge gap exists, however, between the resource requirements of “useful” QC algorithms, and the resources available on current near-term prototypes.

Computer Architecture research seeks to help close this gap by bringing expertise on optimizations and abstractions developed in decades of advances in classical computing. For example, our group has worked on techniques for (i) portable noise-aware compilation, (ii) circuit cutting to fit large codes into smaller prototypes, and (ii) algorithmic optimizations well-suited to today’s near-term implementations. Offering 3X-50X benefits, these techniques offer the potential to close the algorithms-prototype gap years sooner than otherwise. Also interestingly, I will note ways in which computer architecture explorations for QC are in turn also influencing how classical computer systems are designed and run.   

Live

Quantum optimal control is a powerful tool for the robust realization of quantum information processing tasks such as preparation of nonclassical quantum states and implementation of unitary maps. We studied quantum optimal control of the the spin I=9/2 nucleus of 87Sr, an alkaline earth atom that has attracted substantial recent attention for metrology, quantum simulation, and quantum computing. By employing nuclear spin magnetic resonance in the presence of a laser-induced nonlinear AC Stark shift, the system is controllable; we can design any SU(10) unitary matrix acting on the d=10 dimensional manifold of nuclear magnetic sublevels. We design control waveforms that generate the fundamental gates required for universal qudit logic gates. We also study experimental trade-offs including the affects of decoherence and robustness to imperfections.   

Live

In current architecture of quantum computing, interaction between qubits needs to be turned on and off efficiently in a multi-qubit system. Therefore, to perform single-qubit gates the interaction between the target qubit and the other qubits (spectators) must be turned off completely. While to perform two-qubit gates the interaction between the targeted two qubits must be turned on, and the interactions with the spectators must be remained off. A large on/off ratio of the tunable interaction is favorable. However, in fact, there is always residual interactions which introduces errors. We develop a framework including analytical protocol together with numerical method to dynamically mitigate such errors. Our method could be used to find robust control pulses to perform both single qubit gates and two qubit gates at the presence of unwanted interactions with other qubits. Further more, using our method one could perform perfect single-qubit rotations to a target qubit with strong always-on interaction to spectator qubits. This gives an alternative solution for quantum computing with fixed frequency qubits.   

Live

Two-qubit gates activated via a parametrically driven, flux-tunable coupler are a promising approach to engineering fast, high-fidelity entanglement without the potential sacrifice in qubit coherence that comes with a fully tunable architecture. Flux-modulated couplers can also enable a universal hardware-native gateset — an invaluable resource for reducing circuit depth and improving algorithm performance on noisy intermediate scale quantum (NISQ) devices. However, realizing a coherence-limited gate fidelity remains a challenge. In this work, we experimentally investigate the sources of coherent error that limit the parametric gate fidelity in a multi-qubit device. Through identifying these error mechanisms, we explore a physically motivated, systematic approach to calibrating a continuous hardware-native gateset for efficient near-term algorithms.   

Live

In circuit quantum electrodynamics, systems with many types of energy spectra can be engineered, by coupling superconducting qubits and resonators. Arranging the components in certain geometries can lead to the presence of degenerate eigenstates, which -- in a large chain -- forms a dispersionless (or “flat”) band. In a flat band with nonlinearity (provided by transmon qubits), an external drive can populate strongly correlated quantum states. In this talk, we discuss experimental and theoretical progress towards realizing a circuit QED Lieb chain, a system with a gapped flat band that can be driven into a density wave state. We also discuss driven-dissipative schemes that may allow us to stabilize strongly correlated quantum states, in particular one that makes use of a SNAIL (Superconducting Nonlinear Asymmetric Inductive eLement).   

On Demand

The ability to prepare sizeable multi-qubit entangled states with full qubit control is a benchmark for the development of near-term quantum computers. We demonstrate two forms of entanglement within IBM Quantum devices: entanglement in the sense of inseparability with fixed qubit bipartitions, and the stricter genuine multipartite entanglement (GME) which is the inability to express the state as a mixture of only product pure states. A graph state was prepared along a line on all 20 qubits of the ibmq_poughkeepsie device. Each neighbouring pair of qubits was found to be inseparable, indicating entanglement across the device. On the 65-qubit ibmq_manhattan device, a native graph state was prepared and entanglement was detected for all neighbouring pairs of qubits. A parity verification method was implemented for Greenberger-Horne-Zeilinger states on the 27-qubit ibmq_montreal device and an average fidelity increase of up to 0.056+-0.023 was observed. Additionally, a fidelity of 0.61+-0.05 was measured on a 27-qubit state demonstrating GME across the entire device.