Source-linked AI summary
Challenges and Opportunities of Near-Term Quantum Computing Systems
A. D. Corcoles, A. Kandala, A. Javadi-Abhari, D. T. McClure, A. W. Cross, K. Temme, P. D. Nation, M. Steffen, J. M. Gambetta
TL;DR
Near-term quantum systems provide cloud-accessible platforms for research despite noise and limited fault tolerance. The article reviews the coordinated hardware, access, software, benchmarking, compilation, application, and error-mitigation considerations involved. It reports growing use of IBM’s systems and identifies integrated system design as necessary for progress toward practical quantum calculations.
Problem
Near-term devices are noisy and non-fault-tolerant, while practical quantum systems require many interacting design components and must operate despite rapidly accumulating errors.
Method
The article reviews near-term quantum systems through their cloud access, software, benchmarking, compilation, applications, and error mitigation and correction techniques.
Results
IBM’s Quantum Experience hosted thirteen distinct quantum systems, with 28 million user executions and more than 180 research papers enabled.
Takeaways & Limitations
Near-term quantum progress depends on integrating hardware, hardware access, software, benchmarking, applications, and error mitigation into a user-oriented system.
Abstract
from arXiv · showhide
The concept of quantum computing has inspired a whole new generation of scientists, including physicists, engineers, and computer scientists, to fundamentally change the landscape of information technology. With experimental demonstrations stretching back more than two decades, the quantum computing community has achieved a major milestone over the past few years: the ability to build systems that are stretching the limits of what can be classically simulated, and which enable cloud-based research for a wide range of scientists, thus increasing the pool of talent exploring early quantum systems. While such noisy near-term quantum computing systems fall far short of the requirements for fault-tolerant systems, they provide unique testbeds for exploring the opportunities for quantum applications. Here we highlight the facets associated with these systems, including quantum software, cloud access, benchmarking quantum systems, error correction and mitigation in such systems, and understanding the complexity of quantum circuits and how early quantum applications can run on near term quantum computers.
I. INTRODUCTION
Near-term quantum systems are noisy, non-fault-tolerant platforms that require coordinated hardware, software, access, benchmarking, and error-management choices. The article reviews these design considerations and emphasizes short-depth applications such as quantum machine learning and quantum chemistry.
- I. INTRODUCTION: NISQ systems comprise tens or hundreds of qubits and are being developed as general-purpose platforms rather than only one-off demonstrations.Available or prototyped devices range from 5 to 79 qubits.
- I. INTRODUCTION: A complete cloud-accessible quantum system must accommodate diverse users through adaptable functionality and multiple access interfaces.Interfaces can span pulse-level and gate-level control through application-level systems with different connectivities.
- I. INTRODUCTION: A general-purpose system requires an SDK for developing experiments and applications, including tools for circuit composition, simulation, error characterization, and algorithm deployment.Qiskit is organized around Terra, Aer, Ignis, and Aqua.
- I. INTRODUCTION: No single community-wide benchmark yet captures the multiple metrics used to compare quantum systems and track progress.Quantum Volume is proposed as a potential aggregate benchmark incorporating factors such as qubit count, connectivity, and gate performance.
- I. INTRODUCTION: Error mitigation seeks to improve measured observables without full fault-tolerant codes, but remains exploratory for shallow-depth circuits.Near-term algorithms are limited by rapidly accumulating errors, while their circuit outcomes can still be computationally hard for classical machines to reproduce.
- I. INTRODUCTION: Hardware stability is essential because external influences and internal noise can drive parameter drift, making repeated calibration prohibitively costly.The article explicitly places detailed superconducting-hardware challenges outside its scope.
- I. INTRODUCTION: The article treats quantum machine learning and quantum chemistry as application examples because they can map to short-depth circuits believed hard for classical simulation.Developing all system ingredients simultaneously is described as a substantial engineering challenge.
II. CLOUD QUANTUM SYSTEMS AND USER ACCESS LEVELS
Cloud quantum systems expanded from an initial five-qubit IBM device to diverse public and network-accessible backends, enabling research across user roles and device configurations. Their usefulness depends on matching access levels and circuit designs to hardware connectivity, while managing the resulting fidelity and spectator-error trade-offs.
- II. CLOUD QUANTUM SYSTEMS AND USER ACCESS LEVELS: IBM’s cloud offering grew from a five-qubit processor introduced in 2016 to thirteen distinct quantum systems offered publicly or through the IBM Q Network.The platform accumulated 28 million user executions and supported more than 180 research papers.
- II. CLOUD QUANTUM SYSTEMS AND USER ACCESS LEVELS: Different backend connectivity maps and two-qubit error distributions support exploration of user preferences and device performance.The devices shown include five-, sixteen-, and several twenty-qubit systems.
- II. CLOUD QUANTUM SYSTEMS AND USER ACCESS LEVELS: Greater connectivity can shorten entangling circuits but may reduce gate fidelities or introduce spectator errors.Spectator errors affect passively connected qubits not directly involved in an operation.
- II. CLOUD QUANTUM SYSTEMS AND USER ACCESS LEVELS: IBM’s cloud-accessible devices illustrate how system scale and entangling-gate connectivity evolved across successive backends.The first five-qubit device was followed by a device with additional entangling gates, then a sixteen-qubit device and multiple twenty-qubit variants.
- II. CLOUD QUANTUM SYSTEMS AND USER ACCESS LEVELS: Near-term systems therefore require co-design of desired quantum circuits and physically implemented connectivity.This design direction responds directly to the performance effects of connectivity and imperfections.
- II. CLOUD QUANTUM SYSTEMS AND USER ACCESS LEVELS: Cloud access is organized around quantum physicists, quantum information scientists, and quantum developers with different technical goals.The physicist seeks device-level control, the information scientist studies circuit implementation and correction primitives, and the developer focuses on application outcomes.
- II. CLOUD QUANTUM SYSTEMS AND USER ACCESS LEVELS: CNOT error distributions improved over successive IBM devices as gate fidelities, coherence times, and understanding of spectator-qubit errors improved.The figure compares distributions across earlier devices and the device represented in Fig. 4.
- II. CLOUD QUANTUM SYSTEMS AND USER ACCESS LEVELS: Each user class needs appropriately matched data, from device specifications and calibrated gate errors to application-level computation results.Relevant specifications include coherence times, qubit frequencies, crosstalk, and several gate and readout error metrics.
III. QISKIT AND COMPILATION
Near-term quantum computing research depends on cloud access and software that supports users across abstraction levels. The paper focuses especially on compilation strategies tailored to noisy systems.
- Cloud quantum computers have accelerated research and development by widening access to early quantum systems.
- Qiskit is presented as a software suite for near-term quantum computing, supporting quantum information science through its software stack.
- The paper gives special attention to compilers tailored to near-term noisy quantum systems rather than fault-tolerant machines.
- The software stack must serve users ranging from quantum developers to physicists through interfaces at different abstraction levels.
A. Qiskit architecture
Qiskit uses modular components to create, simulate, characterize, compile, and execute quantum programs across devices and abstraction levels.
- Qiskit combines Terra, Aer, Ignis, and Aqua to support quantum programming, simulation, error characterization, and application development.
- Terra provides circuit and pulse representations and compiles and optimizes them for specific machines.
- Qiskit’s modular architecture supports extensions such as new optimization passes, simulation noise models, algorithms, and error-mitigation methods.
- The compiler consists of a transpiler for circuit rewriting and a scheduler that converts circuits into device-executed pulse sequences.
- Transpiler passes can unroll gates, allocate ancillas, route interactions to device topology, merge gates, cancel gates, and analyze circuit depth.
B. Compiling for near-term machines
Compiling for near-term quantum machines prioritizes computation fidelity under device constraints rather than speed alone. Effective compilation uses hardware information, optimization, and sometimes approximation, while verification remains difficult.
- Near-term quantum compilers primarily combat rapidly accumulating errors by translating programs into efficient equivalent circuits.
- Compilation must synthesize high-level routines, satisfy connectivity and native-gate constraints, and translate circuits into control pulses.
- A 15% reduction in two-qubit entangling-gate count can yield dramatic improvements in final computational fidelity, although general compilation is NP-hard.
- Circuit depth and gate count are pseudo-objectives because hardware fidelity also depends on crosstalk and coherent-error mitigation.
- Compiler performance improves with device information including topology, native gates, error rates, latencies, lifetimes, and crosstalk.
- Approximate compilation can trade unitary precision for shorter circuits when the resulting error reduction exceeds the approximation loss.
- Verification of general circuit transformations becomes impractical at roughly 50 or more qubits, motivating smaller-case testing or formal verification.
IV. BENCHMARKING NEAR-TERM DEVICES
Benchmarking near-term quantum devices requires metrics that reflect complete system performance rather than isolated hardware properties. Quantum Volume is presented as an accessible system-level benchmark incorporating qubits, gates, connectivity, errors, and software.
- Benchmarking is necessary to compare quantum systems and measure progress, but no universally accepted rigorous metric yet exists.
- Relevant benchmark factors include qubit number, connectivity, error rates, gate sets, compiler performance, and variations in gate fidelities and spectator errors.
- Quantum Volume is described as a system-agnostic metric for comparing devices across physical implementations while measuring qualities needed for practical quantum computing.
- Quantum Volume measures the largest model circuits a system can successfully run, using compiled random two-qubit circuits and a success criterion based on ideal output probabilities.
- Quantum Volume benchmarks the complete system, including qubits, gates, control electronics, and software compilation.
- 16 was the largest Quantum Volume measured at the time, first on Johannesburg and later on Boeblingen.
V. ERROR MITIGATION AND CORRECTION
Near-term systems combine hardware and environmental error reduction with error correction and mitigation because fault-tolerant resources remain unavailable. The paper reviews general-purpose and problem-specific mitigation methods while emphasizing that these techniques remain exploratory and do not replace fault tolerance.
- Near-term devices are unlikely to combine sufficiently low error rates with enough qubits for fault-tolerant quantum computing.
- Zero-noise extrapolation recombines measurements from amplified-noise circuit runs to estimate expectation values at the zero-noise limit.Increasing noise strengths can support Richardson extrapolation to suppress noise contributions.
- Probabilistic error cancellation samples noisy circuits according to inverse-noise-map coefficients to estimate error-mitigated outcomes.Its experimental challenge is characterizing noisy gates, with demonstrations for up to two-qubit experiments on superconducting and trapped-ion platforms.
- Problem-specific methods mitigate variational energy estimates by measuring additional excitation operators or exploiting conserved quantities such as particle number.
- Error mitigation remains in its infancy, while continued error-correction experiments are presented as critical for understanding long-term protection of encoded information.The paper expects fault-tolerant design and quantum error correction to remain necessary.
VI. QUANTUM APPLICATIONS ON NEAR-TERM QUANTUM SYSTEMS
Near-term quantum applications must be selected for problems where quantum circuits are not efficiently classically simulable. The paper therefore advocates beginning application development from complexity-theoretic circuit hardness, while noting that shallow-circuit advantage depends on the circumstances and assumptions involved.
- Quantum speedup is problem-dependent: theoretical guarantees exist for some tasks, but not every classically difficult problem benefits from quantum computation.
- Efficient classical simulation of a quantum algorithm prevents computational advantage, making circuit complexity central to application design.
- Small noisy devices make quantum advantage difficult to observe immediately, motivating questions about when shallow-depth circuits can outperform classical circuits.
- A systematic application-development path starts with complexity-theoretic hardness of quantum circuits and lets the circuit determine the application.
A. Quantum Machine Learning
Quantum machine-learning proposals use hard quantum feature maps within hybrid classifiers related to support vector machines. However, classical hardness of the feature map is necessary rather than sufficient for practical quantum advantage.
- Quantum machine-learning classifiers use a hard feature-map circuit to embed data into a space where it can be linearly classified.
- Two classifier constructions either follow the feature map with a variational circuit or estimate a kernel matrix directly on the quantum computer.
- Some quantum circuits produce feature maps that are hard to evaluate classically under complexity-theoretic assumptions.
- A hard feature map is only a necessary condition for advantage; additional circuits must be connected to complex real-world classification problems.
B. Quantum Chemistry
Near-term quantum chemistry uses hybrid algorithms such as the variational quantum eigensolver to estimate molecular ground-state energies. Its quantum role is preparing trial states and measuring expectation values, but demonstrated implementations remain small and require careful ansatz and mapping choices.
- The variational quantum eigensolver targets estimates of ground-state energies for chemistry or general many-body Hamiltonians.
- The quantum computer prepares trial states and measures associated expectation values, which can be difficult to perform classically depending on the trial state.
- Most VQE implementations have focused on small molecular Hamiltonians with fewer than 10 qubits.
- Implementations use hardware-efficient circuits or unitary coupled cluster ansätze, trading native hardware interactions against a structured ansatz that preserves physical symmetries.
VII. CONCLUSIONS
Near-term quantum systems require coordinated progress across hardware, access, software, benchmarking, applications, and error mitigation. Their full potential remains unknown, but user-oriented tools and stable hardware are presented as foundations for continued exploration.
- Practical quantum systems require a unique interplay among hardware, hardware access, software, benchmarking, applications, and error mitigation.
- The article presents a heavily user-oriented software approach and reports appreciable demand for access levels, SDKs, and Qiskit.
- Integrating user tools with stable hardware remains a significant effort while researchers develop systems that have not yet been built.
- The full potential of near-term quantum systems is presently unknown, and their application range may lag behind enthusiastic expectations.
- The article identifies fault-tolerant computing, detailed superconducting-qubit hardware, and other quantum topics as areas beyond its discussion.