Implementing the Quantum von Neumann Architecture with Superconducting Circuits
M. Mariantoni, H. Wang, Tsuyoshi Yamamoto, M. Neeley, Radoslaw C. Bialczak, Yanbin Chen, M. Lenander, Erik A. Lucero, A. D. O’Connell, D. Sank, Martin P. Weides, J. Wenner, Yi Yin, Jie Zhao, Alexander N. Korotkov, Andrew N. Cleland, John M. Martinis
The von Neumann architecture for a classical computer comprises a central processing unit and a memory holding instructions and data. We demonstrate a quantum central processing unit that exchanges data with a quantum random-access memory integrated on a chip, with instructions stored on a classical computer. We test our quantum machine by executing codes that involve seven quantum elements: Two superconducting qubits coupled through a quantum bus, two quantum memories, and two zeroing registers. Two vital algorithms for quantum computing are demonstrated, the quantum Fourier transform, with 66% process fidelity, and the three-qubit Toffoli-class OR phase gate, with 98% phase fidelity. Our results, in combination especially with longer qubit coherence, illustrate a potentially viable approach to factoring numbers and implementing simple quantum error correction codes.
The von Neumann architecture for a classical computer comprises a central processing unit and a memory holding instructions and data. We demonstrate a quantum central processing unit that exchanges data with a quantum random-access memory integrated on a chip, with instructions stored on a classical computer. We test our quantum machine by executing codes that involve seven quantum elements: Two superconducting qubits coupled through a quantum bus, two quantum memories, and two zeroing registers. Two vital algorithms for quantum computing are demonstrated, the quantum Fourier transform, with 66% process fidelity, and the three-qubit Toffoli-class OR phase gate, with 98% phase fidelity. Our results, in combination especially with longer qubit coherence, illustrate a potentially viable approach to factoring numbers and implementing simple quantum error correction codes.
Incompatibility of classical memory for quantum computing is explained. Efficient storage/modification of quantum or classical data is accomplished by using a quantum random-access memory (QRAM). A QRAM applies the principle of quantum superposition for simultaneously accessing all the data stored in the memory cell to assure the superior performance of quantum algorithms. Data reading and writing in a bucket-brigade QRAM are described and the feasibility of its optical implementation is discussed. Quantum computers are understood through layered architectures, viz., 3-, 4-, and 5-layer architectures. Organization of a superconducting quantum computer is illustrated by considering a generic architecture. The superconducting transmon qubits produced inside a dilution refrigerator are manipulated by sequences of microwave pulses. Electronic systems used for producing the control signals and reading the quantum information are described. Quantum computers have been built by several industry giants. Improved quantum processors are being assembled with continuous advancements and refinements. Future computing is likely to follow a hybrid paradigm in which the quantum computers will work in tandem with classical computers in a seamless, integrated environment. Each type of computing will complement the strengths and weaknesses of its counterpart.
Abstract A modern computer system, based on the von Neumann architecture, is a complicated system with several interactive modular parts. It requires a thorough understanding of the physics of information storage, processing, protection, readout, etc. Quantum computing, as the most generic usage of quantum information, follows a hybrid architecture so far, namely, quantum algorithms are stored and controlled classically, and mainly the executions of them are quantum, leading to the so-called quantum processing units. Such a quantum–classical hybrid is constrained by its classical ingredients, and cannot reveal the computational power of a fully quantum computer system as conceived from the beginning of the field. Recently, the nature of quantum information has been further recognized, such as the no-programming and no-control theorems, and the unifying understandings of quantum algorithms and computing models. As a result, in this work, we propose a model of a universal quantum computer system, the quantum version of the von Neumann architecture. It uses ebits (i.e. Bell states) as elements of the quantum memory unit, and qubits as elements of the quantum control unit and processing unit. As a digital quantum system, its global configurations can be viewed as tensor-network states. Its universality is proved by the capability to execute quantum algorithms based on a program composition scheme via a universal quantum gate teleportation. It is also protected by the uncertainty principle, the fundamental law of quantum information, making it quantum-secure and distinct from the classical case. In particular, we introduce a few variants of quantum circuits, including the tailed, nested, and topological ones, to characterize the roles of quantum memory and control, which could also be of independent interest in other contexts. In all, our primary study demonstrates the manifold power of quantum information and paves the way for the creation of quantum computer systems in the near future.
A quantum computer will use the properties of quantum physics to solve certain computational problems much faster than otherwise possible. One promising potential implementation is to use superconducting quantum bits in the circuit quantum electrodynamics (cQED) architecture. There, the low energy states of a nonlinear electronic oscillator are isolated and addressed as a qubit. These qubits are capacitively coupled to the modes of a microwave-frequency transmission line resonator which serves as a quantum communication bus. Microwave electrical pulses are applied to the resonator to manipulate or measure the qubit state. State control is calibrated using diagnostic sequences that expose systematic errors. Hybridization of the resonator with the qubit gives it a nonlinear response when driven strongly, useful for amplifying the measurement signal to enhance accuracy. Qubits coupled to the same bus may coherently interact with one another via the exchange of virtual photons. A two-qubit conditional phase gate mediated by this interaction can deterministically entangle its targets, and is used to generate two-qubit Bell states and three-qubit GHZ states. These three-qubit states are of particular interest because they redundantly encode quantum information. They are the basis of the quantum repetition code prototypical of more sophisticated schemes required for quantum computation. Using a three-qubit Toffoli gate, this code is demonstrated to autonomously correct either bit- or phase-flip errors. Despite observing the expected behavior, the overall fidelity is low because of decoherence. A superior implementation of cQED replaces the transmission-line resonator with a three-dimensional box mode, increasing lifetimes by an order of magnitude. In-situ qubit frequency control is enabled with control lines, which are used to fully characterize and control the system Hamiltonian.
The equivalence between the instructions used to define programs and the input data on which the instructions operate is a basic principle of classical computer architectures and programming. Replacing classical data with quantum states enables fundamentally new computational capabilities with scaling advantages for many applications, and numerous models have been proposed for realizing quantum computation. However, within each of these models, the quantum data are transformed by a set of gates that are compiled using solely classical information. Conventional quantum computing models thus break the instruction-data symmetry: classical instructions and quantum data are not directly interchangeable. In this work, we use a density matrix exponentiation protocol to execute quantum instructions on quantum data. In this approach, a fixed sequence of classically-defined gates performs an operation that uniquely depends on an auxiliary quantum instruction state. Our demonstration relies on a 99.7% fidelity controlled-phase gate implemented using two tunable superconducting transmon qubits, which enables an algorithmic fidelity surpassing 90% at circuit depths exceeding 70. The utilization of quantum instructions obviates the need for costly tomographic state reconstruction and recompilation, thereby enabling exponential speedup for a broad range of algorithms, including quantum principal component analysis, the measurement of entanglement spectra, and universal quantum emulation.
Quantum computing is a novel paradigm for information processing that promises computational speedup for certain computational tasks while not being driven by Moore's law. Superconducting circuits based on Josephson junctions are one of the leading approaches for implementing the quantum bits, or qubits, in a quantum processor. Most superconducting qubit architectures involve circuit quantum electrodynamics, where the qubits are analogous to artificial atoms that are coupled to superconducting linear resonant cavities. This scheme allows for the readout of individual qubits, the manipulation of single-qubit states, and the generation of entanglement between qubits. The performance of current state-of-the-art systems is limited by decoherence of the qubits and associated gate errors. Nonetheless, experiments have been performed that already demonstrate a significant computational speedup compared to classical processing. Future implementations of fault-tolerant quantum computers will require the development of quantum error correction. Initial tests of elements of error correction in superconducting qubit arrays are underway.
As the size of quantum systems becomes bigger, more complicated hardware is required to control these systems. In order to reduce the complexity, I discuss the amount of parallelism required for a fault-tolerant quantum computer and what computation speed can be achieved in different architectures. To build a large-scale quantum computer, one can use architectural principles, from classical computer architecture, like multiplexing or pipelining. In this document, a Quantum von Neumann architecture is introduced which uses specialized hardware for the different tasks of a quantum computer, like computation or storage. Furthermore, it requires long qubit coherence and the capability to move quantum information between the different parts of the quantum computer. As an example, a Quantum von Neumann architecture for trapped ions is presented which incorporates multiplexing in the memory region for large-scale quantum computation. To illustrate the capability of this architecture, a model trapped ion quantum computer based on Quantum von Neumann architecture, the Quantum 4004, is introduced. Its hardware is optimized for simplicity and uses the classical Intel 4004 CPU from 1971 as a blueprint. The Quantum 4004 has only a single processing zone and is structured in 4 qubit packages. Its quantum memory can store up to 32768 qubit ions and its computation speed is 10 μs for single qubit operations and 20 μs for two-qubit operations.