PaperlayerView exact revision on arXiv

Source-linked AI summary

Noisy intermediate-scale quantum computers

Bin Cheng, Xiu-Hao Deng, Xiu Gu, Yu He, Guangchong Hu, Peihao Huang, Jun Li, Ben-Chuan Lin, Dawei Lu, Yao Lu, Chudan Qiu, Hui Wang, Tao Xin, Shi Yu, Man-Hong Yung, Junkai Zeng, Song Zhang, Youpeng Zhong, Xinhua Peng, Franco Nori, Dapeng Yu

arXiv:2303.04061v1quant-ph

TL;DR

Quantum computing research is progressing toward fault-tolerant systems, while practical quantum applications remain a central aspiration. This review surveys algorithmic and hardware advances and their challenges; it reports potential variational applications but notes that practical use remains a long-term prospect.

  • Problem

    The review addresses how quantum computing can progress toward fault-tolerant computers and useful applications as the field enters the noisy intermediate-scale quantum era.

    Source: §I · p. 3, §II · p. 7

  • Method

    The review surveys major quantum algorithms, hardware-platform advances, and challenges in the most promising technical routes.

    Source: §I · p. 3, §II · p. 6

  • Results

    The review describes variational algorithms as promising for applications such as quantum chemistry and biological science, while stating that the path to practical applications remains long.

    Source: §II · p. 7

  • Takeaways & Limitations

    Variational algorithms may support future applications including low-energy electronic-structure calculations and biological optimization, with error mitigation perhaps needed on near-term devices.

    Source: §II · p. 7

    Practical variational applications remain distant, and error mitigation may be required for near-term devices.

    Source: §II · p. 7

Abstract

from arXiv · show

Quantum computers have made extraordinary progress over the past decade, and significant milestones have been achieved along the path of pursuing universal fault-tolerant quantum computers. Quantum advantage, the tipping point heralding the quantum era, has been accomplished along with several waves of breakthroughs. Quantum hardware has become more integrated and architectural compared to its toddler days. The controlling precision of various physical systems is pushed beyond the fault-tolerant threshold. Meanwhile, quantum computation research has established a new norm by embracing industrialization and commercialization. The joint power of governments, private investors, and tech companies has significantly shaped a new vibrant environment that accelerates the development of this field, now at the beginning of the noisy intermediate-scale quantum era. Here, we first discuss the progress achieved in the field of quantum computation by reviewing the most important algorithms and advances in the most promising technical routes, and then summarizing the next-stage challenges. Furthermore, we illustrate our confidence that solid foundations have been built for the fault-tolerant quantum computer and our optimism that the emergence of quantum killer applications essential for human society shall happen in the future.

I. INTRODUCTION

This section reviews the development of quantum computing toward fault-tolerant systems, emphasizing hardware and algorithmic progress, emerging applications, and the field’s broader research environment. It surveys major hardware platforms and their advances, capabilities, and challenges.

I. INTRODUCTION

  • Universal gates, precise readout, qubit integration, quantum error correction, and quantum-advantage demonstrations mark progress toward fault-tolerant computing.Examples include Google’s 53-qubit Sycamore randomized-circuit-sampling experiment and practical quantum simulators for chemistry and condensed-matter problems.Source: §I · p. 2
  • Hardware progress includes 50–100-qubit processors, with Google’s 53-qubit processor achieving 99.4% average two-qubit gate fidelity during simultaneous operations.Two-qubit fidelities above 99% were first achieved in a multi-qubit superconducting circuit in 2014.Source: §I · p. 2
  • The field has entered the noisy intermediate-scale quantum stage, where medium-sized systems may support small numbers of logical qubits and demonstrations in chemistry, machine learning, and optimization.The review describes large-scale integrated systems as the eventual setting for universal fault-tolerant computers.Source: §I · p. 2
  • Algorithms highlighted include VQE for quantum chemistry and HHL and QSVT among newer general-purpose approaches.The introduction describes VQE as extending from two-atom calculations toward larger quantum systems.Source: §I · pp. 2–3
  • The review places quantum computing across three development stages and associates potential applications with different stages.Figure 1 distinguishes physical- and logical-level progress and lists stage-specific applications.Source: Figure 1 · p. 3
  • Commercialization, company participation, cloud access, and education have broadened quantum-computing research beyond its earlier academic setting.The section points to cloud quantum computing as a way for users without hardware to explore algorithms.Source: §I · p. 3
  • The review covers superconducting circuits, trapped ions, semiconductors, neutral atoms, NMR, NV centers, and photonics, assessing their quantum-computing elements, advantages, scalability, and challenges.It also discusses recent quantum-algorithm progress and potential applications.Source: §I · p. 3
  • The review presents topological qubits as a possible route to suppressing errors, but Majorana-like trivial states complicate detection and non-locality and braiding remain unproved.A complete discussion of topological quantum computation is outside the review’s scope.Source: §I · p. 3

II. QUANTUM ALGORITHMS

This section surveys quantum algorithms from foundational query methods to Hamiltonian simulation, quantum linear algebra, and variational approaches. It describes their algorithmic ideas and potential, while noting input-assumption and practical-application caveats.

Early quantum algorithms

  • Early algorithms combined quantum parallelism with interference to reduce query counts, while Simon’s algorithm later inspired Shor’s factoring and discrete-log algorithms.The section cautions that the Deutsch–Jozsa speedup vanishes with a small error probability.Source: §II · pp. 4–5

Quantum Fourier transform

  • The quantum Fourier transform is a circuit-efficient Fourier transform primitive used in Shor’s period-finding approach to integer factoring and in algorithms for Abelian stabilizer problems.The section connects these problems to the hidden subgroup framework.Source: §II · p. 5

Quantum search

  • Grover’s algorithm searches an unstructured database with a quadratic query speedup, and amplitude amplification extends the quadratic improvement to algorithms succeeding with probability p.A fixed-point version can handle an unknown number of targets.Source: §II · p. 5
  • Quantum walks provide another search framework; Ambainis’s element-distinctness algorithm matched a theoretical lower bound and later work applied quantum walks to problems including triangle finding.Szegedy developed a general framework for quantizing classical Markov chains.Source: §II · p. 5

Hamiltonian simulation

  • Hamiltonian simulation approximates e^-iHt; product formulas alternate local Hamiltonian terms, while LCU, block encoding, and QSVT provide other approaches with strong precision or parameter scaling.The review notes that product formulas can have competitive gate-complexity scaling.Source: §II · pp. 5–6

Quantum linear algebra and quantum machine learning

  • HHL uses Hamiltonian evolution and phase estimation to prepare |x⟩ from |b⟩ for Ax = b, with exponential speedup claims conditional on state preparation and not needing the full solution description.Related quantum machine-learning speedup claims remain debated: strong input assumptions and classical dequantization results qualify them.Source: §II · p. 6

Variational quantum algorithms

  • Variational quantum algorithms prepare parameterized quantum states and use classical optimization; VQE targets ground states and energies, while QAOA addresses combinatorial optimization and quantum neural networks address machine learning.The section presents variational approaches as promising for practical applications, but says the path to such applications remains long and error mitigation may be needed for near-term devices.Source: §II · pp. 6–7

III. SUPERCONDUCTING QUBITS

Superconducting qubits are nonlinear Josephson-junction circuits used for quantum information processing. The section reviews their construction and operation alongside challenges in coherence, error correction, and scaling.

Qubit construction

  • A large shunt capacitor suppresses transmon charge-noise sensitivity, while Josephson nonlinearity creates the anharmonicity needed to define a qubit using its two lowest energy levels.The transmon’s effective charging energy is lowered to the regime EJ/EC > 50.Source: §III · p. 7

Readout and initialization

  • Dispersive readout infers the qubit state from a state-dependent shift in a detuned resonator, while Purcell filters and parametric amplifiers can improve readout speed and fidelity.Conditional and unconditional reset techniques can accelerate initialization between measurements.Source: §III · p. 8

Gates

  • Single-qubit microwave drives can cause leakage and phase errors, while frequency crowding complicates two-qubit control as connectivity grows; tunable couplers have supported fidelities approaching 99.9%.DRAG adds pulse quadratures to combat coherent errors, and virtual Z gates shift XY-rotation phases.Source: §III · pp. 8–9

Decoherence

  • Qubit lifetimes improved six orders of magnitude, from nanoseconds to milliseconds, in about 20 years, while spurious two-level systems remain a suspected major decoherence source and dynamical decoupling can correct coherent noise.Optimized compiling can also shorten native-gate sequences, though planar geometry and limited connectivity make compilation challenging.Source: §III · p. 9

Quantum error correction

  • Surface-code experiments are approaching or partly exceeding threshold, but substantial, reproducible error-correction gains at larger scales remain challenging; bosonic codes alone have reached break-even in experiments.A logical qubit may require hundreds to a thousand physical qubits, while cosmic rays can cause chip-wide failure and threaten surface codes.Source: §III · pp. 9–10

Scalability

  • Monolithic processors have integrated dozens of qubits, but scaling requires simultaneous high-fidelity operations, denser qubits, better wiring, and reduced crosstalk; packaging, multiplexing, and cryogenic electronics address related constraints.Qubit control wires scale with chip perimeter while qubit density scales with chip area.Source: §III · p. 10
  • Modular links offer an alternative to monolithic designs constrained by wafer size, device yield, and crosstalk; improved connections have raised inter-module state-transfer fidelity to about 90%.Earlier cable-based transfers reached up to about 80%, while flip-chip modular approaches have also achieved high fidelities.Source: §III · p. 10

IV. TRAPPED-ION QUBITS

Trapped-ion systems combine long-lived, precisely controlled qubits with high-fidelity operations, while entangling-gate speed and scaling remain central challenges. The section reviews linear-chain, QCCD, and remote-entanglement approaches alongside integration advances and their limitations.

Qubit coherence

  • One-hour coherence time was approached in a 171Yb+ ion qubit, allowing systems to execute millions of gate operations before losing quantum features.The reported coherence time progressed from 50 s for 43Ca+ to 10 minutes for 171Yb+ with dynamical decoupling and sympathetic cooling, then approached one hour.Source: §IV · p. 11

Quantum gates

  • Single-qubit error rates below 10^-4 were reached for optical and Raman transitions, and 10^-6 for microwave-driven hyperfine qubits.The section identifies entangling operations, rather than single-qubit rotations, as the main limit on current processor quality.Source: §IV · p. 11

Entangling gates

  • A light-shift gate on two 40Ca+ ions approached 6 × 10^-4 infidelity, described as the best entangling gate achieved at the time.The review also reports error rates below 8 × 10^-4 for a Mølmer–Sørensen gate on two 9Be+ ions and below 9 × 10^-4 for a light-shift gate on two 43Ca+ ions.Source: §IV · pp. 12–13

Fast gates

  • A 1.6 µs entangling gate retained 99.8% fidelity, whereas reducing duration to 480 ns lowered fidelity to around 60%.An ultrafast-pulse demonstration prepared a Bell state with 76% fidelity in 1.96 µs; reported infidelities included imperfect kick control and off-resonant coupling.Source: §IV · p. 13

Scalability

  • Linear ion chains offer full connectivity within the same potential, but larger chains require more laser power and face imperfect cooling and greater noise sensitivity.The review reports GHZ-state demonstrations with up to 24 ions and programmable processors expanded from 5 to 11 qubits.Source: §IV · p. 13
  • QCCD links multiple ion chains by shuttling ions between trap regions, but scaling requires complex traps and remains challenged by anomalous heating, junction barriers, and low trap depth.Fast shuttling methods have been investigated to avoid making transport a processing-speed bottleneck; microfabricated chip traps address the control complexity of multiple regions.Source: §IV · pp. 13–14

Remote entanglement

  • Remote-ion entanglement connects ions across separate vacuum systems, but its best reported fidelity was 94% and its generation rates were slower than direct gates.Reported generation rates rose from 4.5 Hz to 182 Hz; photon collection limits the practical rate, while photon conversion to telecom wavelengths was demonstrated with low efficiency.Source: §IV · p. 14

System integration

  • Integrated waveguides, gratings, detectors, and voltage sources support scalable ion control, while truly fault-tolerant encoding and preserving control quality as systems grow remain open challenges.Waveguide-delivered beams have supported single-qubit rotations and two-qubit gates with robustness against vibrational noise; on-chip photon detectors have also been demonstrated.Source: §IV · p. 14

V. SEMICONDUCTOR SPIN QUBITS

The review surveys semiconductor spin-qubit platforms and their progress in gates, readout, and integration. It highlights silicon’s scaling prospects alongside remaining fidelity and engineering challenges.

Qubit platforms

  • Spin qubits use electron or hole spins in quantum dots or silicon dopants; platforms include Si-MOS, Si/SiGe, Ge/SiGe, nanowires, and donor devices.The review focuses on spin qubits while briefly introducing other types.Source: §V · p. 15, Figure 5 · p. 16

Decoherence

  • Silicon’s zero-nuclear-spin isotopes and long T1 support long spin coherence, whereas fluctuating nuclear fields limit GaAs spin dephasing to tens of nanoseconds.The review relates the long-T1 prerequisite to 2T1 ≥ T2.Source: §V · p. 15

Gates

  • A fixed-exchange Si/SiGe two-qubit gate reached 99.8% fidelity; tunable Si/SiGe CZ gates reached F_CZ = 99.65% and F_CZ = 99.81%, while a donor CZ gate reached F_CZ = 99.37%.Single-qubit gate fidelities also exceeded 99% in donor, Si-MOS, and Si/SiGe systems.Source: §V · pp. 15, 17

Challenges and conclusion

  • Charge noise and fidelity remain challenges; the review identifies low-noise interfaces, improved materials and fabrication, and attention to crosstalk, heating, and frequency crowding as important.The conclusion positions silicon spin qubits for scaling and intermediate-scale applications while noting these engineering overheads.Source: §V · pp. 17–18

Readout and initialization

  • Spin readout uses state-to-charge conversion and charge sensing; reported fidelities include 98.4% for singlet-triplet qubits and FM = 99.8% for single spins.Single-lead rf-reflectometry readout showed fidelities above 98% and a 6 µs readout time.Source: §V · pp. 17–18

Scalability

  • Cryo-CMOS, hot qubits, 300 mm processing, and multi-qubit demonstrations advance silicon integration, while remote spin coupling and networks remain a next step.Demonstrations include three- and six-qubit Si/SiGe devices and a four-qubit Ge/SiGe GHZ state.Source: §V · p. 18

VI. NV CENTERS

NV centers combine an optically addressable electron spin with nearby nuclear-spin memories, supporting quantum processing, sensing, and networking. The section reviews control, coherence, and application advances while identifying fabrication and multi-qubit control as scalability challenges.

Qubits and coherence

  • Dynamical decoupling extends NV electron-spin dephasing time from microseconds to milliseconds by suppressing quasi-static noise.Nearby nuclear spins provide memory qubits, with coherence up to T2 = 2 s at room temperature and carbon spin pairs reaching T2 = 1 min and T1 > 6 min at 4 K.Source: §VI · p. 19

Initialization and readout

  • Optical pumping initializes the electron spin, while resonant excitation enables high-fidelity single-shot readout at low temperature.Preparation and readout fidelities reached 99.9% and 98%, respectively; projective measurement-based initialization and single-shot readout have also been demonstrated for several nuclear-spin systems.Source: §VI · p. 19

Gates

  • Composite pulses with modified GRAPE yielded single- and two-qubit gate fidelities of 0.99995 and 0.992, respectively.The two-qubit gate lasted 700 ns, three orders of magnitude shorter than the coherence time.Source: §VI · p. 19

Multi-qubit control

  • Dynamical-decoupling-based control has enabled universal gates on weakly coupled nuclear spins and entanglement of up to seven nuclear spins in a ten-qubit register.The DDrf phase-compensation scheme removes dependence of interpulse delay on hyperfine parameters and allows delay optimization to protect electron coherence.Source: §VI · pp. 19–20

Gates

  • Fault-tolerant operations on the logical-qubit level have been achieved on a seven-qubit NV quantum processor.The work is described as a major step toward fault-tolerant quantum information processing.Source: §VI · p. 20

Quantum simulation and quantum algorithm

  • NV simulators have explored topological phases, non-Hermitian dynamics, open-system behavior, time crystals, and emergent hydrodynamics.Demonstrated algorithms include Deutsch-Jozsa, adiabatic factorization, Grover search, quantum-enhanced machine learning, and resonant quantum principal component analysis.Source: §VI · pp. 20–21

Quantum network

  • Remote NV nodes with memory qubits have been entangled over distances greater than 1 km, and a three-node network with long-lived memory qubits has been realized.Source: §VI · p. 21

Quantum sensing

  • NV centers support sensing of magnetic fields, electric fields, temperature, and pressure, including under low-field, high-pressure, and high-temperature conditions.Quantum optimization and error correction have also been incorporated into metrology to improve sensitivity in the presence of noise.Source: §VI · p. 21

VII. NMR SYSTEM

The NMR system section reviews liquid-state nuclear spins as a quantum-computing platform, its control and measurement methods, and applications in algorithms, simulation, and machine learning. It also identifies state preparation, suitable registers, and scalable high-fidelity control as constraints.

NMR quantum processor

  • NMR has achieved universal control of up to 12 qubits and supports quantum simulation, control, tomography, and machine-learning studies.The section presents these advances as enabled by established experimental technologies.Source: §VII · pp. 22–23

Qubit

  • Room-temperature NMR uses spin-1/2 nuclei as qubits, selecting molecular spins with distinguishable precession frequencies for individual addressability.Heteronuclear spins differ by gyromagnetic ratio; homonuclear spins can be distinguished by nuclear shielding shifts.Source: §VII · p. 22, Figure 7 · p. 23

Initialization

  • Because room-temperature NMR ensembles have low polarization, pseudopure states provide a convenient small-scale substitute for pure initial states, but their achievable purity scales unfavorably with qubit number.Algorithmic cooling and parahydrogen-induced polarization have produced spin systems with purities above the entanglement threshold.Source: §VII · p. 22

Operation

  • NMR implements single-qubit rotations with selective radio-frequency pulses and two-qubit gates through natural J-coupling, with refocusing pulses suppressing unwanted couplings.Reported control estimates include a 4.7 ± 0.3 × 10−3 average error rate for one- and two-qubit gates on a three-qubit system and 55.1% average experimental Clifford-gate fidelity on a seven-qubit processor.Source: §VII · p. 22

Measurement

  • NMR detects free-induction decay and Fourier-transforms it to recover spectral information, while tomography requires a complete set of basis measurements whose size grows exponentially.Unlike projective readout, the ensemble directly measures the expectation value of a single coherent Pauli observable.Source: §VII · p. 22

Quantum algorithm

  • Liquid-state NMR has implemented Deutsch-Jozsa and Grover search on two-qubit samples, QFT on three qubits, and Shor factoring on seven qubits.Source: §VII · pp. 23–24

Quantum simulation

  • NMR simulations span many-body physics and chemistry, including phase behavior, information scrambling, molecular energies, reaction dynamics, and topological order.Examples include a four-qubit simulator probing out-of-time-order correlators in a chaotic Ising chain and simulations of hydrogen and water molecules.Source: §VII · p. 24

Quantum machine learning

  • NMR implementations include handwritten 9-versus-6 recognition with a four-qubit quantum support vector machine and small-scale face recognition with quantum principal component analysis.An image-processing algorithm was also used experimentally to detect boundaries separating image regions.Source: §VII · p. 24

VIII. NEUTRAL ATOM ARRAYS

Neutral-atom arrays combine flexible tweezer rearrangement with long-range Rydberg interactions for digital gates and analog simulation. The review highlights scaling progress alongside coherence, control, and readout challenges.

Scalability

  • Cryogenic tweezers have trapped Rb atoms for up to 6000 s, and arrays exceeding 300 atoms reached a 37% defect-free preparation probability; earlier arrays reached 256 defect-free atoms.Rearrangement can achieve filling fractions up to 99%, while the authors anticipate scaling from hundreds to thousands of qubits and coupling processors with atom-photon interconnects.Source: §VIII · pp. 24–25

Initialization and readout

  • Atom loss in state-selective ejection readout prevents measurements during circuit execution; lossless fluorescent detection has been demonstrated, with atom-heating losses still needing reduction for repetitive QEC.Single-alkali preparation fidelity is estimated above 99.5% and typical measurement fidelity above 98%.Source: §VIII · p. 25

Gates

  • Rydberg blockade enables two-qubit gates in 0.4–2 µs; reported results include 87Rb entanglement fidelity F2 ≥97.4%, 171Yb gate fidelity F2 = 83%, and a 88Sr Bell state above 99.5%.Rydberg excitation can require briefly turning off tweezers to mitigate anti-trapping, whereas 174Yb Rydberg states have been stably trapped in red-detuned tweezers.Source: §VIII · pp. 25–26, Figure 8 · p. 25

Coherence

  • Magic-intensity trapping mitigates differential light shifts, enhancing coherence to 225 ms; alkaline-earth-like nuclear-spin qubits are estimated to reach T′2 = 40 s in spin echo.Single 87Sr atoms are estimated at T1 ≫10 s and T*2 = 21 s; the cited 40 s value is for T′2 in spin echo.Source: §VIII · p. 26

Digital quantum operations

  • Digital neutral-atom processors demonstrated GHZ states up to 6 qubits and chemistry and QAOA circuits, while coherent transport supported a 12-qubit cluster state and a 7-qubit Steane code state.The transport architecture also generated topological surface and toric code states and enabled hybrid analogue-digital evolution.Source: §VIII · p. 26

Analog quantum operations

  • Tunable Rydberg arrays explored Ising dynamics from chains to graphs, demonstrated Zn-ordered phase transitions and antiferromagnetic order, and observed many-body scars avoiding rapid thermalization.Other reported studies include topological phases, an XXZ model, and optimization for the maximum independent set problem.Source: §VIII · p. 26

Outlook

  • The platform’s stated primary challenges are higher two-qubit gate fidelity, quantum nondemolition measurement, and low crosstalk in large arrays.Proposed approaches include further cooling alkali atoms and using a second atomic element for auxiliary-qubit monitoring and lower-crosstalk manipulation.Source: §VIII · p. 27

IX. PHOTONIC QUANTUM COMPUTING

Photonic quantum computing uses light’s weak environmental coupling and networking potential, with approaches ranging from boson sampling to continuous-variable processing. The review surveys experimental advances while emphasizing loss and weak photon interactions as scaling hurdles.

Boson sampling

  • Boson sampling is a nonuniversal linear-optics model that is easier to build than universal quantum computing, while Gaussian boson sampling replaces single-photon inputs with squeezed states and offers potential graph and chemistry applications.The review reports that photonic advances in sources, networks, and detectors enabled large entangled states and quantum computational advantage through boson sampling.Source: §IX · pp. 27, 30

Quantum light sources

  • Quantum-dot single-photon sources reached 57% overall efficiency, compared with typical SPDC single-photon efficiency of about 1%; the review anticipates efficiency above 70%.The reported efficiency benefits from polarization-dependent Purcell enhancement in a microcavity.Source: §IX · p. 28

Linear optical networks

  • Rectangular interferometers reduce optical depth to n −1 versus 2(n −1) −1 for triangular designs; micro-optics offers highest transmission, while programmable time-bin and integrated circuits suffer serious losses.The review identifies reducing losses while retaining programmability as a longstanding goal.Source: §IX · p. 28

Boson sampling experiments

  • A quantum-dot source and low-loss circuit produced five-photon boson sampling at a rate 24,000 times faster than previous experiments; a later setup used 20 input photons and a 60-mode interferometer.In the latter experiment, at most 14 photons were detected, with an output Hilbert-space dimension of 3.7×10^14.Source: §IX · p. 29

Gaussian boson sampling

  • Jiuzhang’s Gaussian boson sampling used 50 squeezed-state inputs and a 144-mode interferometer, detecting up to 113-click coincidences and yielding a sampling rate 10^10 faster than the stated classical simulation strategy.The setup routes 144 output modes to superconducting nanowire single-photon detectors.Source: §IX · p. 29, Figure 9 · p. 29

Programmability

  • Time-bin boson sampling permits real-time adjustment of each bin’s splitting ratio and phase; a time-bin Gaussian boson sampler used 216 squeezed-state inputs and detected up to 219 photons.That latter machine was partially programmable and partially connected, with most matrix elements zero.Source: §IX · p. 30

Scalability

  • Photonic scaling faces both loss and ultra-weak independent-photon interactions; one cited photon-photon gate had 4.8% efficiency and 76.2% fidelity, while storage and retrieval later reached 39%.The review says loss can be handled with relatively large thresholds, while high-fidelity and high-efficiency photon-photon gates remain a future goal.Source: §IX · p. 30

X. OUTLOOK AND CONCLUSION

The conclusion identifies platform-specific engineering needs and shared scaling overheads as ongoing challenges. It anticipates that combining hardware approaches and hybrid algorithms may support future applications.

Platform challenges

  • Quantum platforms differ in maturity and limitations, with challenges including materials and fabrication for solid-state systems, atom-photon interconnectors, scalable NMR control, and controllable NV-center production.The review also identifies low-charge-noise interfaces and foundry-level fabrication as important for semiconductor qubits.Source: §X · p. 30

Cross-platform integration

  • Across platforms, crosstalk, gate heating, and frequency crowding must be considered; the review calls for high-fidelity control that scales without compromising quality.It also suggests that integrating platforms could allocate computation, communication, and storage to different units, while hybrid quantum-classical algorithms are described as critical for near-term applications.Source: §X · pp. 30–31

Future applications

  • The authors anticipate fault-tolerant operations and quantum algorithms in medium-sized systems, quantum advantage in areas such as chemistry and machine learning, and broader real-world applications as platforms scale.They also identify specialized machines, quantum clouds, sensing, and simulations as possible future developments.Source: §X · p. 31
Loading 2303.04061v1…