Source-linked AI summary
Quantum Computer Systems for Scientific Discovery
Yuri Alexeev, Dave Bacon, Kenneth R. Brown, Robert Calderbank, Lincoln D. Carr, Frederic T. Chong, Brian DeMarco, Dirk Englund, Edward Farhi, Bill Fefferman, Alexey V. Gorshkov, Andrew Houck, Jungsang Kim, Shelby Kimmel, Michael Lange, Seth Lloyd, Mikhail D. Lukin, Dmitri Maslov, Peter Maunz, Christopher Monroe, John Preskill, Martin Roetteler, Martin Savage, Jeff Thompson
TL;DR
Quantum computers face both the challenge of building capable systems and the uncertainty of identifying useful applications. This paper advocates vertically integrated co-design of quantum hardware, software, algorithms, and scientific applications, highlighting opportunities and challenges for scientific discovery over the next decade.
Problem
The useful application scope of entangled quantum systems remains unclear because measurement directly accesses only selected global properties produced by quantum algorithms.
Method
The paper proposes co-designing quantum computers with scientific applications across all levels of the quantum computer stack.
Results
The paper identifies scientific and community needs, opportunities, use cases, and development challenges, including quantum simulations of field theories and quantum dynamics.
Takeaways & Limitations
Scientific applications should help determine hardware platforms, computational modes, gate sets, and software-stack priorities as quantum systems develop.
Takeaways & Limitations
The optimal gate set or computational mode depends on hardware interactions, available controls, and algorithm structure, while hardware candidates face conflicting control, fidelity, initialization, and measurement requirements.
Abstract
from arXiv · showhide
The great promise of quantum computers comes with the dual challenges of building them and finding their useful applications. We argue that these two challenges should be considered together, by co-designing full-stack quantum computer systems along with their applications in order to hasten their development and potential for scientific discovery. In this context, we identify scientific and community needs, opportunities, a sampling of a few use case studies, and significant challenges for the development of quantum computers for science over the next 2--10 years. This document is written by a community of university, national laboratory, and industrial researchers in the field of Quantum Information Science and Technology, and is based on a summary from a U.S. National Science Foundation workshop on Quantum Computing held on October 21--22, 2019 in Alexandria, VA.
Executive Summary
Quantum computing faces intertwined challenges: identifying useful applications while building controllable, scalable hardware. The paper advocates vertically integrated co-design of quantum systems and scientific applications, iterated across technologies and institutions.
- Challenge 1: Quantum applications must extract measurable global properties from entangled states because measurement does not directly reveal their full information.The useful algorithmic task is to guide the quantum state into a simpler form before measurement.
- Challenge 2: Building quantum computers requires isolating many qubits while enabling precise state control and high-accuracy measurement.The technology involves unconventional information carriers and eventually complex quantum error correction.
- Co-design approach: The paper proposes co-designing quantum computers with scientific applications through a vertically integrated approach spanning the complete quantum computer stack.This connects scientific opportunities with device control, native interactions, connectivity, compilation, and error correction.
- Current opportunity: Existing platforms based on trapped atoms and superconducting circuitry are already being built into small quantum systems, opening near-term scientific opportunities.The paper focuses on opportunities from integrating these technologies across the full stack.
- Implementation path: The authors advocate continually developing and operating multiple generations of systems, iterating between building and using devices across universities, national laboratories, and industry.The proposed 2–10-year effort is intended to support scientific applications while training quantum engineers and transitioning programs toward industry.
Introduction
Quantum computation represents information with qubits and entangled superpositions, then uses controlled operations and measurement to obtain useful information. The introduction surveys computational modes, hardware requirements, and error correction while emphasizing that hardware and algorithm structure jointly shape system design.
- Quantum information: An n-qubit register represents an entangled superposition over 2^n basis states whose complex amplitudes determine measurement probabilities.The amplitudes evolve under unitary dynamics, and measuring the register yields definite bit strings probabilistically.
- Quantum algorithms: Quantum algorithms manipulate interference so that one or a few states acquire significant amplitude before measurement provides global information.Known algorithms can offer advantages or speedups over classical approaches, including reductions in complexity class for some problems.
- Computational modes: The universal gate model decomposes algorithms into modular operations on individual qubits, while other computational modes use restricted or global operations for specific routines.Non-universal modes can be useful for simulations or state preparation that do not require full control of the quantum state space.
- Hardware–algorithm matching: The choice of gate set or computational mode depends on native hardware interactions, available controls, and the structure of the algorithm.The figure illustrates rotation and controlled-NOT operations as an example of a universal gate family.
- Error correction: Quantum error correction adds encoded ancilla qubits and feedback measurements to stabilize computations, with fault-tolerant schemes offering arbitrarily long computations with sub-exponential overhead.Successful deployment requires matching correction methods to the asymmetric and time-varying noise profiles of particular qubit systems.
- Hardware requirements: Physical hardware must provide coherent, high-fidelity control together with efficient initialization and measurement, requirements that restrict viable platforms.These requirements are described as seemingly conflicting and currently limit the available hardware candidates.
The Quantum Computer Stack
The quantum computer stack spans algorithms, software, control engineering, and qubit technology, with performance opportunities and challenges at every level and interface. The paper advocates vertically co-designing applications with hardware and intervening layers.
- Co-design: Vertical co-design links quantum applications to specific hardware and all intervening layers because the stack is not yet cheap or commoditized.The paper compares this early development model with application-specific integrated circuits for intensive computations.
- Stack architecture: The stack ranges from algorithms and quantum software to control engineering and physical qubit technology, with opportunities at each level and their interfaces.Its levels include problem mapping, native-gate compilation and compression, error-correction strategy, and Hamiltonian control.
- Quantum algorithms: Quantum algorithms must be assessed for near-term devices or future fault-tolerant architectures, while quantum advantage remains a central field-wide challenge.Fault-tolerant work emphasizes asymptotic performance at large numbers of qubits and gates.
- Quantum algorithms: Quantum simulations target molecules, materials, and quantum field theories whose models are intractable classically, with some less-universal simulators easier to realize.Applications include physical properties such as energy levels, phase diagrams, and thermalization times.
- Quantum algorithms: Variational algorithms use classical control parameters to optimize objectives on entangled quantum states, but their heuristic convergence requires testing on hardware.VQE and QAOA are explored for machine learning and combinatorial optimization, and can be relatively insensitive to systematic gate errors.
- Quantum software: Quantum software includes compilers, simulators, verifiers, benchmarking protocols, and operating systems, while optimal compilation is provably intractable.Heuristic resource optimization is therefore often necessary, and quantum compilers generally must be developed from scratch.
- Control engineering: Scalable control requires application-specific processing, system characterization, and noise-aware engineering beyond laboratory racks of test equipment.Characterization methods trade information against speed for small systems, while the best approach for larger systems remains unclear.
- System scale: Networked quantum computers may provide an optimal route to the necessary system size, with scientific applications motivating the engineering needed to expose further challenges and solutions.The paper presents full-stack case studies as illustrations of how future applications may be realized and co-designed.
Quantum Computer Case Studies
The paper illustrates the full-stack approach through case studies spanning optimization, programmable quantum simulations, error-correcting codes, and textbook algorithms. These examples show how problems can be mapped onto quantum-computing modes and specific hardware platforms.
- Case-study scope: The case studies cover optimization, programmable quantum simulations, quantum error-correcting codes, and textbook quantum algorithms.They are illustrative rather than exhaustive examples of future quantum applications.
- Full-stack mapping: Mapping problems onto particular quantum-computing modes and hardware platforms illustrates the need to translate applications across the full stack.
I. Gate-Based Quantum Simulation
Gate-based quantum simulation targets difficult quantum dynamics and equilibrium properties, while co-design links algorithms, hardware, and scientific goals to address resource and noise constraints.
- Classical simulation of n interacting qubits requires solving 2^n complex differential equations, limiting arbitrary dynamics to about 50 qubits.
- Hamiltonian simulation evolves an initial state under a model Hamiltonian while minimizing gate count as system size, evolution time, and precision increase.
- Theoretical evidence indicates that sampling measured outcomes after Hamiltonian evolution cannot be performed by any polynomial-time classical algorithm.
- Variational Estimation of Ground States: VQE maps an electronic Hamiltonian onto qubits, then measures parameterized-state expectations while classical optimization minimizes the Hamiltonian function.
- Simulating Dynamics in Quantum Field Theories: Quantum-field-theory simulations target phenomena across condensed matter, high-energy, and nuclear physics, but larger gauge theories require substantially more qubits and control.
- Simulating Dynamics in Quantum Field Theories: Near-term QFT simulations are constrained by qubit coherence, communication-fabric time dependence, and measurement-error mitigation costs at larger scales.
- Simulating Dynamics in Quantum Field Theories: Co-design and collaboration among theorists, quantum scientists, developers, and engineers are presented as necessary to close the gap between QFT resources and hardware capabilities.
- Physical Platform: Trapped-ion platforms offer high gate fidelities, long-range connectivity, and flexible gate expression; native Ising XX decompositions can reduce errors in variational circuits.
II. Combinatorial Optimization with QAOA
QAOA applies quantum optimization to graph problems, with neutral-atom platforms offering direct encodings for Unit Disk Graph instances and opportunities for larger experiments.
- Combinatorial optimization minimizes a cost function over a combinatorially large solution set, with applications including routing, finance, and machine learning.
- Physical Platform: QAOA’s simple dynamics may enable near-term implementation using direct techniques on optically addressed Rydberg atoms rather than universal gate-based systems.
- Maximum Independent Set Problem: Maximum Independent Set seeks the largest set of mutually nonadjacent vertices and is NP-hard; some graphs with N above 300 defeat classical exact algorithms.
- Maximum Independent Set Problem: In the Hamiltonian encoding, each graph vertex becomes a qubit, while U penalizes simultaneous excitation of connected vertices; for U ≫ 0, the ground state encodes MIS.
- Physical Platform: Unit Disk Graph MIS maps directly onto neutral atoms without encoding overhead, with Rydberg blockade interactions supplying the edge penalty.
- Physical Platform: Reconfigurable tweezer arrays encode graph structure through atom positions, while global laser pulses can reduce control complexity.
- Physical Platform: Experiments have demonstrated coherent evolution in 51-atom chains, with arrays of 100–1000 atoms described as within current technological reach.
- Open challenges include generalizing unit-disk MIS sampling, selecting QAOA parameters and H0, and extending QAOA to entangled-state preparation.
III. Quantum Error Correction and Architectures
The paper proposes co-designing quantum architectures and error correction across the stack, using heterogeneous and virtualized qubits to balance operational speed, memory stability, and scalability. Universal fault-tolerant operations remain a particularly difficult challenge.
- Virtualization and memory: Current qubit limits motivate virtualizing qubits through quantum error-corrected memories.The architecture separates volatile operational qubits from stable memory qubits, trading simpler device design for more serial operations.
- Virtualization and memory: Separating fast-operation and long-lifetime qubits enables performance studies as the active-to-memory qubit ratio changes.The paper proposes estimating the operation and memory fidelities needed for scientific goals such as molecular simulation.
- Hardware architectures: A vertically integrated project would combine physical superconducting qubits into fault-tolerant virtualized qubits, while treating the chosen architecture and platform as illustrative examples.The design space includes transferring information between transmons and cavity qubits with longer memory times.
- Hardware architectures: Heterogeneous superconducting circuits could assign different roles to transmons, cavity-encoded qubits, 0 −π qubits, and other designs, with device- and software-level protection.Effective error correction requires improvements in both underlying physics and control-system engineering.
- Open challenges: Universal fault-tolerant operations are harder than memory, requiring hardware-dependent approaches such as bosonic-code gates or lightweight gate sets on block codes.Realizing two universal fault-tolerant qubits is identified as a grand challenge for building large, reliable systems.
IV. Standard Quantum Algorithms
The paper treats Shor’s and Grover’s algorithms as full-stack benchmarks whose implementation must be co-designed with hardware, compilation, connectivity, error mitigation, and modular software. These algorithms can expose system-wide integration challenges while informing architecture and error-control choices.
- Algorithm benchmarks: Shor’s and Grover’s algorithms provide textbook cases for mapping practically relevant quantum algorithms onto computing platforms.The paper frames their implementation as a co-design problem spanning the quantum computing stack.
- Implementation challenges: Classical oracles must be realized with quantum gates, while classical optimizations and post-processing can trade against quantum circuit complexity.Examples include Bennett’s pebbling game for Grover’s oracle and windowed arithmetic for Shor’s algorithm.
- Implementation challenges: Near-term implementations depend strongly on matching algorithms to available qubit connectivity and native gate sets, despite unchanged asymptotic scaling.High connectivity can provide significant implementation advantages, while constrained geometries remain an open research area.
- Error management: Testing textbook algorithms across architectures and error-mitigation settings can identify which realistic errors are critical and guide algorithm-specific mitigation.The paper distinguishes mitigation from correction and calls for hardware, software, and algorithm expertise to be integrated.
- Software modularity: Modular libraries and recurring programming patterns can encapsulate optimized quantum circuits and support scalable development of complex algorithms.Examples include quantum Fourier transforms, multiple-control gates, arithmetic libraries, QPE, amplitude amplification, and period finding.
- Algorithm benchmarks: Implementing textbook algorithms can benchmark both complete quantum systems and individual stack components because hardware and software must work in concert.The paper notes that no single benchmark captures every relevant system aspect.
Outlook and Paths Forward
The paper argues that full-stack quantum-system development is especially promising for scientific discovery and should be supported by broad user engagement and coordinated facilities. It proposes QCLabs to unite science, engineering, computer science, and industrial systems while iterating across generations of devices.
- Outlook: Full-stack consideration is presented as a way to accelerate building, using, and optimizing quantum computers for scientific discovery.The paper connects qubit physics, control engineering, algorithm optimization, and applications across science.
- Community engagement: Challenge competitions and quantum information centers are proposed to broaden participation, but existing centers currently play a limited role in coordinating collaborative programs.The centers are described as growing faculty, researchers, students, and public engagement.
- QCLabs: QCLabs are proposed as scientific user facilities that bring the science, computer science, and engineering of the quantum stack together.The workshop community consensus presents them as a response to both application-finding and machine-building challenges.
- QCLabs: QCLabs would support continual co-design by iterating device design, software optimization, use cases, and the construction of next-generation systems.Each facility could specialize in a qubit system, scaling architecture, or use-case family while maintaining broad stack collaboration.
- Industry interaction: Industrial cloud quantum computers offer growing variation in platforms, qubit counts, gate depths, and connectivity, but generally provide limited access to low-level controls.QCLabs are envisioned to leverage and benchmark these services while pursuing scientific applications that may lack obvious commercial value.