PaperlayerView exact revision on arXiv

Source-linked AI summary

Continuous-variable quantum computing in the quantum optical frequency comb

Olivier Pfister

arXiv:1907.09832v1quant-ph

TL;DR

This review surveys CV quantum computing in the quantum optical frequency comb, including measurement-based approaches and routes toward fault tolerance. It discusses large-scale cluster-state demonstrations and temporal wires reaching one million sequentially accessible qumodes.

Abstract

from arXiv · show

This topical review introduces the theoretical and experimental advances in continuous-variable (CV) --- i.e., qumode-based in lieu of qubit-based --- large-scale, fault-tolerant quantum computing and quantum simulation. An introduction to the physics and mathematics of multipartite entangled CV cluster states is given, and their connection to experimental concepts is delineated. Paths toward fault tolerance are also presented. It is the hope of the author that this review attract more contributors to the field and promote its extension to the promising technology of integrated quantum photonics.

I. INTRODUCTION

The introduction argues that quantum scalability depends on increasing the number of logic units, not merely enlarging the Hilbert space of a fixed system. It presents CV quantum information in optical frequency combs as a scalable research direction and frames the review for readers familiar with qubit-based quantum information.

A. Quantum computing: revolutionary promise and daunting challenge

  • Scalability requires increasing the number N of logic units; enlarging the Hilbert space of a fixed number of qudits is not a genuine scalability claim.The text identifies growth in N, regardless of each unit’s Hilbert-space dimension, as the relevant meaning of scalability.Source: A. Quantum computing: revolutionary… · p. 2
  • Quantum simulation is motivated by problems ranging from spin arrays and quantum field theory to quantum chemistry and machine learning.Examples include the Bose-Hubbard model, chemical reactions such as nitrogen fixation, and envisaged machine-learning applications.Source: A. Quantum computing: revolutionary… · p. 2
  • Practical quantum computing requires enough qubits that are resilient to decoherence, while large-scale systems remain an outstanding challenge for leading qubit platforms.The introduction distinguishes progress in fidelity from having large-scale systems and notes that the NISQ concept does not capture the whole field.Source: A. Quantum computing: revolutionary… · p. 3

B. Quantum computing over continuous variables: a premium on scalability

  • Experiments have produced thousands of entangled frequency-domain modes and a million temporal modes, motivating CV information encoded in dense qumodes.The review focuses on the frequency-domain implementation in a single-OPO quantum optical frequency comb.Source: B. Quantum computing over continuous… · p. 3
  • The review introduces CV quantum information to readers familiar with qubit-based information, with particular attention to integrated quantum photonics.It focuses on the frequency-domain implementation in a single OPO and notes the extension of integrated photonics to the CV domain.Source: B. Quantum computing over continuous… · pp. 3–4

II. CONTINUOUS-VARIABLE QUANTUM COMPUTING

The review introduces qumode-based quantum information through continuous quadrature eigenstates and their relation to oscillator observables. It then maps qubit concepts to qumode operations, including Fourier transforms, phase-space displacements, and EPR entanglement.

A. Continuous-variable quantum information: quadrature eigenstates

  • In optical CV information, amplitude and phase quadratures are analogous to position and momentum observables of a quantum harmonic oscillator.Qumodes use a continuous amplitude eigenbasis, with eigenvalues in the real numbers.Source: A. Continuous-variable quantum… · p. 4

Conjugate basis

  • The qubit-to-qumode correspondence replaces discrete sums with integrals and maps the qubit Hadamard transform to the qumode quantum Fourier transform.Table I is identified as presenting the correspondence between qubit- and qumode-based quantum computing.Source: Table I · p. 5, Conjugate basis · p. 6
  • Qubit Pauli operations correspond to Weyl-Heisenberg phase-space displacements generated by quadrature translations.The text describes the Weyl-Heisenberg group as the qumode counterpart of the Pauli group.Source: Conjugate basis · pp. 5–6
  • Qumode EPR states are joint eigenstates of commuting operators Q1 − Q2 and P1 + P2, whose measurement noise is zero.These operators are also called variance-based entanglement witnesses or nullifiers.Source: Conjugate basis · p. 6

B. Realistic CVQI: squeezed states

The section explains how finite-squeezing states approximate unphysical EPR states and describes squeezing through its parameter and associated uncertainties. It also outlines why non-Gaussian operations matter for CVQC’s computational power.

B. Realistic CVQI: squeezed states

  • EPR states are unphysical because they have infinite energy, so laboratory implementations use approximations based on two-mode squeezed states.An EPR state corresponds to an infinitely squeezed two-mode squeezed state; the squeezing parameter is r.Source: B. Realistic CVQI: squeezed states · p. 6
  • For vacuum inputs, finite squeezing gives nonzero standard deviations, while the text presents entanglement witnesses as generalizable to multipartite settings.The review highlights signatures of multipartite entanglement from multimode squeezing, particularly for cluster states.Source: B. Realistic CVQI: squeezed states · p. 7

1. Exponential speedup: the Gottesman-Knill theorem and its CVQC version

  • CV algorithms restricted to Gaussian operations admit efficient classical simulation; non-Gaussian operations are therefore crucial for exponential speedup.The review describes Gaussian operations as quadratic-or-lower Hamiltonian evolutions and non-Gaussian operations as cubic or higher order.Source: §1 · p. 8
  • Proposed laboratory routes to non-Gaussian resources include photon-number-resolved detection, photon subtraction, and photon addition.The text also describes PNR detection as a proposed sole non-Gaussian resource for implementing a cubic phase gate.Source: §1 · p. 8

2. Fault tolerance

Finite squeezing creates a fault-tolerance question for CVQC, but GKP encoding provides a route to a fault-tolerance threshold without requiring infinite squeezing. The review also discusses the practical challenge of realizing GKP states and related non-Gaussian resources.

Non-Gaussian resources

  • Gaussian resources alone cannot implement qumode Bell-inequality violation, entanglement distillation, or error correction; non-Gaussian resources remedy these restrictions.Examples include Fock states, photon-number-resolving measurements, and field-cubic Hamiltonians.Source: §2 · p. 9

Finite squeezing

  • Fuzzy CV wavefunctions raise the question of whether small shifts can be distinguished and corrected, unlike orthogonal qubit states.The passage contrasts shifts δq ≪ ∆Qψ with distinguishable qubit basis states.Source: §2 · p. 9
  • nmax = exp(2r) teleportation steps before fidelity reaches the 50% classical limit for TMS channels; this is 2 steps at 3 dB and 10 at 10 dB.This illustrates how finite squeezing constrains concatenated CV teleportation.Source: §2 · p. 9

GKP fault tolerance

  • GKP encoding applied to CVQC proves the existence of a fault-tolerance threshold, so infinite squeezing is not required.The encoding stores a qubit in an oscillator to address small CV drifts.Source: §2 · p. 9

GKP resources

  • Optical GKP-state generation remains an arduous endeavor: proposals have been made for optical GKP states [77], and an experimental realization over the phononic vibration field of a trapped ion was recently performed.Source: §2 · p. 9

Related CV approaches

  • Gaussian boson sampling and CV instantaneous quantum computing are described as subuniversal approaches proven hard classically.The review also notes that non-Gaussian operations are considered critical to fault tolerance in current understanding.Source: §2 · pp. 8, 10

III. MEASUREMENT-BASED, ONE-WAY QUANTUM COMPUTING

One-way quantum computing uses a multipartite cluster state as a resource, with computation proceeding through measurements and feed-forward. The review presents scalable generation of a CV cluster state in a quantum optical frequency comb as the approach under discussion.

III. MEASUREMENT-BASED, ONE-WAY QUANTUM COMPUTING

  • In one-way QC, measurement outcomes inform feed-forward operations on neighboring qubits, using the cluster as a substrate containing the entanglement needed by an algorithm [83].The choice of measurement basis determines the gate applied to the teleported state [86].Source: A. Introduction. Cluster states · p. 10
  • Cluster states are multipartite entangled resources distinct under LOCC from GHZ and W states for four or more qubits.The passage notes that the GHZ and W examples are themselves not LOCC equivalent.Source: A. Introduction. Cluster states · p. 10
  • A cluster state is initialized in |+⟩ states and entangled by CZ gates in a two-dimensional pattern, typically a square lattice.A two-dimensional lattice permits the two-qubit gates needed for a universal gate set [83, 86, 87].Source: A. Introduction. Cluster states · pp. 10–11
  • In the illustrated teleportation sequence, the input is fused with the simplest cluster, a neighbor is measured in the |±⟩ basis, and its result is fed forward to place that neighbor in the input state.The measurement result is m = ±1.Source: A. Introduction. Cluster states · p. 11
  • If the whole cluster state can be generated scalably, computation proceeds using single-qubit measurements and feed-forward in the QOFC-based CVQC approach.The review also mentions closely related time-domain implementations.Source: A. Introduction. Cluster states · p. 11

B. The CV cluster state

The review develops CV cluster states from qubit cluster states, then compares bottom-up generation with a single-OPO frequency-comb approach. It also describes graph-based observables that can certify cluster-state preparation.

B. The CV cluster state

  • CV cluster states use qumodes prepared in phase-quadrature eigenstates and entangled by CZ/CPHASE gates along a square lattice, paralleling the qubit construction.The ideal states are infinitely squeezed and therefore unphysical; laboratory states have finite phase squeezing.Source: B. The CV cluster state · pp. 11–12

C. Bottom up scalability

  • For vacuum inputs, Bloch–Messiah decomposition reduces Gaussian cluster-state generation to N single-mode squeezers followed by one N-mode interferometer.This removes the otherwise irrelevant first interferometer and simplifies the protocol; a more compact generation method is also noted.Source: C. Bottom up scalability · p. 12

D. Top-down scalability: CV cluster state generation in the QOFC

  • Rather than scaling to N qumodes with N OPOs, it is possible—and experimentally more tractable—to use the whole QOFC of a single OPO.Proposals progressed from N-mode GHZ states to cluster-state generation and a square cluster lattice in a single OPO; experimental realizations followed in the QOFC.Source: D. Top-down scalability: CV cluster… · pp. 12–13

1. Qubits

  • Qubit cluster-state stabilizer generators combine X on a vertex with Z on its graph neighbors, providing observables for certifying the state and reconstructing its graph.With noiseless measurements on many copies, a state that is an eigenstate of all generators is a cluster state.Source: §1 · pp. 13–14

2. Qumodes

The review characterizes CV cluster states through nullifiers and connects experimentally motivated Hamiltonian graphs to canonical graphs. Its GHZ example illustrates a mode-number squeezing boost and contrasts GHZ connectivity with the local structure of a QC cluster.

2. Qumodes

  • CV stabilizer generators are unitary, and their exponentiated Hermitian operators yield nullifiers that are quadrature observables measurable with homodyne detection and RF networks.The vector nullifier equation uses the cluster graph’s adjacency matrix V.Source: §2 · pp. 14–15

B. H(amiltonian) graph states and their connection to canonical graph states

  • The Hamiltonian graph’s adjacency matrix G yields equations of motion; diagonalizing G gives the squeezing parameters and squeezed multimode observables.The discussion assumes an initial vacuum state and addresses the relation between G and the canonical graph matrix V at infinite and finite squeezing.Source: B. H(amiltonian) graph states and… · p. 15

C. An example: the GHZ graph

  • The GHZ example has a mode-number squeezing boost, and a Fourier transform of all but one qumode yields nullifiers for the canonical GHZ graph.The GHZ graph is nonlocal, unlike the local nearest-neighbor structure of a QC cluster; it is also described as too entangled for one-way QC.Source: C. An example: the GHZ graph · pp. 16–17

D. Finite squeezing

The finite-squeezing formalism relates canonical graph adjacency G to a complex adjacency matrix Z = V + iU. For self-inverse matrices, the section confirms equivalence between the real part V′ of the finite-squeezing matrix and G, while noting this condition is sufficient but not necessary.

D. Finite squeezing

  • The section asks how canonical graph adjacency G relates to V when finite squeezing is represented by Z = V + iU.The complex-adjacency formalism incorporates finite-squeezing effects.Source: D. Finite squeezing · p. 18
  • Earlier work established a general, non-bijective relation between G and V, with G = V in the particular case G = G−1 [95, 99].The section traces this development from multimode squeezing to the self-inverse case.Source: D. Finite squeezing · p. 18
  • For self-inverse matrices, the finite-squeezing real-part matrix V′ equals G, but the condition is sufficient rather than necessary.The GHZ example demonstrates that self-inverseness is not necessary.Source: D. Finite squeezing · p. 18

E. Fault tolerance, high-dimensional lattices, and topological qumodes

The review surveys CV fault-tolerance thresholds and routes to higher-dimensional cluster states, alongside a polarization-based square-lattice proposal. It notes both promising threshold results and unresolved resource requirements and implementation demands.

E. Fault tolerance, high-dimensional lattices, and topological qumodes

  • The section reports a 1.4% error-probability threshold for depolarizing errors and describes other architectures as potentially reaching fault tolerance with about 10 dB squeezing.The current optical squeezing record is 15 dB for a single mode; excess technical noise above the reciprocal of the squeezing level is reported not to affect the QC outcome.Source: E. Fault tolerance, high-dimensional… · pp. 18–19
  • Higher-valence n-hypercubic CV cluster states can be generated by interfering n identical OPOs using a fractal construction.The section leaves open how the expected non-Gaussian error-correcting resource would be incorporated in this setting.Source: E. Fault tolerance, high-dimensional… · pp. 18–19
  • Topological qumode work includes theoretical study of entanglement entropy for structures such as the toric code.The review also describes topological error encoding over cluster states as yielding a fault-tolerance threshold.Source: E. Fault tolerance, high-dimensional… · pp. 18–19
  • The GKP-based CV fault-tolerance work established that squeezing thresholds exist and specified squeezing requirements for resource states at given thresholds.Table II reports upper-bound squeezing thresholds for different error-rate thresholds and encodings.Source: E. Fault tolerance, high-dimensional… · p. 19, Table II · p. 19
  • The review concludes that fault-tolerant one-way CVQC has no fundamental limits, while substantial theoretical and experimental work remains.It also identifies non-Gaussian resources beyond GKP states, including Fock states, as an avenue for further work.Source: E. Fault tolerance, high-dimensional… · p. 19

A. Toroidal square lattice proposal

  • Adding polarization as another degree of freedom to the frequency label of qumodes overcame the no-go theorem for linear-chain and square-lattice cluster states in the QOFC, enabling universal CV cluster states.Source: A. Toroidal square lattice proposal · p. 20
  • The proposal required a doubly resonant OPO with an engineered KTP crystal phase-matched to three pump/signal polarization sets with equal coupling strengths.The crystal was successfully designed and demonstrated.Source: A. Toroidal square lattice proposal · p. 21
  • Its 15-mode pump field and nontrivial frequency spacings required sophisticated phase modulation, making the approach impractical and prompting exploration of smaller cluster graphs.The architecture remains compact and might still be implementable in the future.Source: A. Toroidal square lattice proposal · p. 21

B. Many squares

The review describes frequency-comb experiments that scale CV cluster-state generation in copies and wire length. Demonstrations range from many small square clusters to a million-qumode quantum wire, with measurements spanning 60 modes and a broader estimated generation bandwidth.

B. Many squares

  • The QOFC generated and verified 15 simultaneous copies of a four-qumode square cluster state, the review’s first large-scale CV cluster-state demonstration.Each state was a 2 × 2 cluster.Source: B. Many squares · p. 21

C. Dual-rail quantum wire

  • One million qumodes were achieved in the dual-rail quantum wire, accessible sequentially two at a time.The earlier wire reached 104 qumodes; the sequential access is described as compatible with QC.Source: C. Dual-rail quantum wire · p. 23
  • The frequency-domain wire experiment measured a 60-qumode CV cluster state and found nullifier noise squeezed by 3.2(2) dB across the measurement range.The measurement range was limited by the 15 GHz electro-optic modulator used to create the local-oscillator fields.Source: C. Dual-rail quantum wire · pp. 23, 25
  • In the frequency-domain scheme, frequency-shifted EPR-pair sets are combined through balanced beam splitting to form a dual-rail wire; the frequency shift is lossless.The pairs occupy two orthogonal polarizations and many frequencies before the polarization rotation and polarizing beam splitter.Source: C. Dual-rail quantum wire · pp. 23–24
  • The OPO’s nonlinear interaction had constant strength across 3.2 THz, corresponding to 6,700 qumodes, although the measured span did not cover the full phasematching range.The authors expected a 4–5 THz range, or on the order of 10^4 modes; their measurement was limited by a diode laser’s tuning range.Source: C. Dual-rail quantum wire · p. 25
  • Dispersion can chirp the QOFC free spectral range and shift distant qumodes out of OPO resonance; a slightly broadened pump field is described as a remedy.This limitation may arise before the phasematching bandwidth is exhausted.Source: C. Dual-rail quantum wire · p. 25

D. Dual-rail quantum wires in the QOFC

Integer-multiple pump detuning produces equivalent H-graph descriptions, and a single OPO experimentally generated two independent quantum wires of 30 measured qumodes each. Spectral scaling avoids fiber-delay losses but is bounded by phase-matching bandwidth.

E. Square and hypercubic cluster states in the QOFC

The review surveys proposals for scaling QOFC cluster states across dimensions and frequency-time domains, alongside other entanglement implementations. It also identifies integrated quantum photonics as a hoped-for future direction amid unresolved challenges such as decoherence.

E. Square and hypercubic cluster states in the QOFC

F. Hybrid temporal and spectral approach

G. Other implementations of entanglement in the QOFC

  • Synchronous pumping yielded qumode-resolved multipartite entanglement, and photon addition and subtraction induced non-Gaussian features with counterintuitive propagation along a graph.Source: G. Other implementations of… · p. 29

VI. CONCLUSION

  • The review hopes CV quantum information will extend to integrated photonics, while noting that real-life quantum technology still faces daunting challenges, especially decoherence.Source: §VI · p. 29
Loading 1907.09832v1…