PaperlayerView exact revision on arXiv

Source-linked AI summary

Quantum Computing with Continuous-Variable Clusters

Mile Gu, Christian Weedbrook, Nicolas C. Menicucci, Timothy C. Ralph, Peter van Loock

arXiv:0903.3233v1quant-ph

TL;DR

Continuous-variable cluster-state computation requires theoretical and experimental protocols that support universal quantum computation while controlling optical resource costs. This paper develops graph-state representations and measurement rules, implements the cubic phase gate with photon detection, and characterizes offline squeezing. It proves that universal CV cluster states can have squeezing requirements per mode that do not increase with cluster size, and that non-Gaussian resources can support homodyne-only computation.

  • Problem

    The paper addresses how continuous-variable cluster states can provide an experimentally viable framework for universal quantum computation while requiring manageable offline squeezed resources.

  • Method

    The authors develop stabilizer, phase-space, and algebraic graph-state representations, analyze measurement transformations, bound offline squeezing, and adapt photon-counting protocols for the cubic phase gate.

  • Results

    Universal CV cluster states exist whose offline squeezing per mode does not increase with cluster size, and photon detection implements a cubic phase gate within the cluster-state formalism.

  • Takeaways & Limitations

    Suitable non-Gaussian resource states are sufficient for universal computation in the cluster-state framework even when cluster computation uses homodyne detection alone.

Abstract

from arXiv · show

Continuous-variable cluster states offer a potentially promising method of implementing a quantum computer. This paper extends and further refines theoretical foundations and protocols for experimental implementation. We give a cluster-state implementation of the cubic phase gate through photon detection, which, together with homodyne detection, facilitates universal quantum computation. In addition, we characterize the offline squeezed resources required to generate an arbitrary graph state through passive linear optics. Most significantly, we prove that there are universal states for which the offline squeezing per mode does not increase with the size of the cluster. Simple representations of continuous-variable graph states are introduced to analyze graph state transformations under measurement and the existence of universal continuous-variable resource states.

I. INTRODUCTION

Continuous-variable cluster-state computation combines continuous-variable quantum computation with cluster-state methods as an alternative route to quantum computing. The paper develops graph-state representations, measurement rules, resource bounds, and non-Gaussian implementations aimed at universal computation.

  • Motivation: CV cluster-state computation combines continuous-variable quantum computation with cluster-state computation as a potential alternative implementation of a quantum computer.CV modes can also encode qubits, while cluster-state computation uses single-qubit measurements on entangled resource states.
  • Experimental setting: Optical CV cluster states can be generated deterministically through offline squeezing and passive linear optics, with Gaussian cluster transformations requiring only homodyne detection.Some proposals generate large clusters in one step using one optical parametric oscillator and no interferometer.
  • Theoretical framework: The paper introduces phase-space and algebraic representations of CV graph states and uses them to analyze transformations under quadrature measurements.These tools address the unwieldiness of Schrödinger representations for graph states of nontrivial size.
  • Universality: The authors show that universal graph states can serve as resource states for implementing arbitrary continuous-variable circuits.The discussion identifies a CV cluster state usable for arbitrary CV operations.
  • Resource requirements: The paper bounds offline squeezing for arbitrary graph-state generation and shows that squeezing per mode need not grow with universal cluster size.This is presented as a necessary criterion for efficient computation through offline resources.
  • Non-Gaussian operations: Photon counting and homodyne measurements provide a cluster-state implementation of the cubic phase gate, while suitable non-Gaussian resource states can enable universal computation using homodyne measurements alone.A displaced number-state measurement generates a cubic phase state that supports a cubic phase gate of specified strength.

2. Gaussian Transformations

Gaussian transformations are described as linear symplectic actions on quadrature operators and include rotations, displacements, squeezing, and shearing. The section distinguishes readily available operations from transformations requiring squeezing interactions.

  • General form: A general Gaussian transformation maps the quadrature vector through a 2n × 2n symplectic matrix and a displacement vector.The matrix acts linearly on quadrature operators, while the constant vector represents quadrature displacements.
  • Single-mode transformations: Rotations act as phase shifts in phase space, and a rotation by π/2 implements the Fourier transform.The rotation matrix describes the corresponding linear Heisenberg action.
  • Single-mode transformations: Quadrature displacements translate states in phase space: Z(s) shifts momentum by s, whereas X(s) shifts position by s.The operators use opposite signs in their exponential definitions.
  • Single-mode transformations: Squeezing compresses one quadrature by s while stretching the conjugate quadrature by 1/s.The associated squeeze matrix describes this linear Heisenberg action.
  • Single-mode transformations: Phase-space shearing uses the operator e^{is q^2/2}, which is also called the phase gate.The shearing matrix describes its linear Heisenberg action on quadrature operators.
  • Experimental resources: Phase shifts and coherent-state sources provide readily available single-mode Gaussian transformations, while full single-mode Gaussian control requires squeezing interactions.The paper notes that squeezing interactions require nonlinear optical processes experimentally, despite linear Heisenberg input-output relations.

3. A Universal Gate Set

Universal CV computation combines Gaussian operations with a nonlinear interaction and a two-mode coupling, while graph-state measurements provide a resource-based implementation framework.

  • Universal gate set: Gaussian operations plus any single nonlinear interaction of at least cubic order suffice for universal single-mode quantum computation.The paper gives displacement, shear, and cubic phase gates as examples, together with the Fourier transform.
  • Universal gate set: Adding a nontrivial two-mode interaction, such as the CZ gate, extends the gate set to universal quantum computation.A beamsplitter is identified as another possible two-mode interaction.
  • Scope: Universality statements in this framework do not account for noise, and current error-correction codes require some discretization of CVs.The paper therefore states that CV quantum computation is currently known only for discretized encodings.
  • Cluster-state computation: Universal graph-state families implement arbitrary unitary operations through adaptive choices of local measurements.The measurement bases depend on the algorithm being implemented.
  • CV graph states: CV graph states replace qubit graph-state ingredients with qumodes, quadrature measurements, and CZ interactions between graph-connected modes.Vertices represent qumodes, and edges determine which pairs interact through CZ.
  • Stabilizers and nullifiers: The CV stabilizer and nullifier formalisms specify graph states and analyze their transformations under quadrature measurements.The nullifier space is an n-dimensional vector space for an n-qumode graph state and has a standard graph-derived basis.

B. Wigner Representation

The paper uses nullifiers and Wigner functions to represent CV graph states, then shows how measurement-based circuits implement Gaussian and non-Gaussian operations universally. Gaussian measurements can be parallelized, while non-Gaussian operations require adaptive measurement sequences.

  • Wigner representation: Wigner functions extend nullifier descriptions with explicit phase-space representations that remain useful for non-ideal CV cluster states.Their arguments transform like nullifiers under Gaussian transformations.
  • Measurement-based computation: Any CV unitary U on an input |φ⟩ can be implemented by entangling the input with an appropriate graph state and measuring single qumodes.The remaining modes encode U|φ⟩ up to known single-mode rotations and translations.
  • Measurement-based computation: Measurements in bases Mi = e^-ifi(q)p e^ifi(q) implement unitary-dependent operations, with later bases generally determined by earlier outcomes.The unmeasured output modes retain the computed state modulo known Gaussian corrections.
  • Proof of universality: Concatenating alternating D_q and D_p operations implements arbitrary single-mode operations deterministically.The resulting measurement-dependent Gaussian correction can be tracked as a final basis change.
  • Proof of universality: A sign error is present in the corresponding derivation of Ref..The paper identifies this issue explicitly in a footnote.
  • Proof of universality: Gaussian measurements exhibit parallelism: they may be performed in any order or simultaneously because their outcome adaptation is trivial.This parallels the corresponding property for Clifford measurements in qubit cluster computation.

C. Graph States as Resources

Standard graph states can serve as resources for computation by coupling input modes to designated graph vertices and measuring the input modes. The resulting state reproduces the required computational cluster up to known corrections, yielding CV brickwork states.

  • Graph states as resources: Input modes are coupled to designated graph vertices with CZ operations before the input modes are measured.Each input mode ui is connected to its corresponding graph vertex vi.
  • Graph states as resources: The circuit construction translates into a graph-state representation in which suitable measurements implement the desired single-qumode operations.The corresponding graph states are called CV brickwork states.
  • Graph states as resources: Measuring the coupled input modes produces the desired cluster state up to known single-mode quadrature displacements and rotations.These corrections can be tracked during the remaining computation.

V. UNIVERSAL CLUSTER STATES

The paper establishes universal CV cluster states by analyzing graph-state transformations under quadrature measurements. Computational-basis measurements delete vertices, while momentum-basis measurements shorten linear connections, allowing resource graphs to be carved into computational clusters.

  • Universal cluster states: Universal CV graph states can collapse into the brickwork states required for any given quantum circuit through appropriate quadrature measurements.These universal resources are called CV cluster states.
  • Vertex removal: The nullifier formalism updates graph states after measurements by replacing the measured noncommuting nullifier with the measurement constraint.Other nullifiers receive the measured quadrature value by substitution.
  • Vertex removal: A computational-basis measurement removes the measured vertex and all incident edges, up to known quadrature displacements.This operation can also amputate corrupted parts of a cluster.
  • Wire shortening: Momentum-basis measurements on internal vertices preserve effective neighbor connectivity and shorten linear graph states.Measuring the two inner nodes of a four-node line yields an equivalent two-qumode cluster up to corrections and a phase-space reflection.

B. The Universal Resource State

A sufficiently large two-dimensional square-lattice graph is a universal CV resource that can be reduced to circuit-specific graphs by measurements. Finite squeezing makes these resources physical but introduces measurement-dependent distortions and average noise during teleportation.

  • The universal resource state: A planar square lattice is a CV cluster state because vertex deletion and wire shortening can carve out the graph needed for any circuit.The required finite resource grows linearly with fundamental one- and two-qumode gates and with the number of qumodes.
  • Finite squeezing: Ideal momentum eigenstates are non-normalizable, so practical cluster states replace them with finitely momentum-squeezed vacuum states.These finite-accuracy states are generalized CV cluster states.
  • Finite squeezing: Finite-accuracy cluster states are Gaussian, and their Wigner representations approach the ideal graph-state form as squeezing increases.The Gaussian distributions converge toward uniform and delta-peaked limits.
  • Distortions in state propagation: 1/(2s^2) noise units are added to the q-quadrature on average during teleportation through a finite-squeezing cluster.Repeated teleportation alternates this Gaussian noise between the two quadratures.
  • Distortions in state propagation: Individual teleportation outcomes produce pure conditional states whose Gaussian envelopes depend on the measurement result.Extreme outcomes can shift the envelope enough to cut off substantial portions of the input Wigner-function support.

B. Distortions in Universal Gate Teleportation

Finite squeezing introduces Gaussian noise into measurement-based CV computation, while graph states can be generated using offline squeezing and passive linear optics. Homodyne detection implements Gaussian operations, and photon counting supplies the additional non-Gaussian capability needed for universality.

  • Distortions: Finite squeezing universally adds Gaussian noise that blurs momentum and position quadratures, with magnitude inversely proportional to cluster accuracy and linear in cluster length.Redundant rails may reduce this noise but require more squeezing resources.
  • Optical Advantages: Optical CV cluster-state generation is deterministic, unlike discrete-variable implementations that rely on challenging nondeterministic entangling operations.The paper notes that CV clusters had been experimentally realized for up to four qumodes and could potentially be generated in a single step with frequency-encoded modes.
  • Universal Gate Teleportation: Multi-qumode Gaussian operations require only quadrature measurements, or homodyne detection, once the cluster state is prepared.This measurement-only implementation supports Gaussian transformations and graph-state computation.
  • Universal Computation: Photon counting adds the non-Gaussian operation required for universal quantum computation.The cubic phase gate is implemented through photon detection together with homodyne measurements.
  • Cluster State Generation: Any CV graph state can be prepared from squeezed vacuum states using passive linear optics, avoiding online squeezing.The generation matrix and its singular-value decomposition provide an explicit recipe for this decompositional method.

1. A simple example

The two-mode example compares canonical CZ-based preparation with a decompositional scheme that transfers squeezing into offline resource states. The decompositional method reduces the squeezing cost while preserving the target graph-state accuracy.

  • 1. A simple example: The decompositional method shifts the CZ operation’s online squeezing into stronger initial resource squeezing.The largest singular value is generally greater than the target accuracy parameter s.
  • 1. A simple example: A factor of 2 is the squeezing overhead needed in the usual high-accuracy case to generate the same two-mode graph-state accuracy.This factor is identified as the squeezing overhead.
  • 1. A simple example: The decompositional method remains advantageous even when offline squeezing is treated as costly as online squeezing.The paper revisits this comparison because online and offline squeezing may have different experimental costs.
  • 1. A simple example: The canonical method requires two online squeezers for the CZ gate, whereas the decompositional method uses offline squeezing and passive linear optics.The decompositional scheme is obtained from the generation matrix through singular-value decomposition.
  • 1. A simple example: 2.36 dB of squeezing is saved for all values of s in the two-mode decompositional method.The saving is calculated as (2 modes) × (4.18 dB/mode − 3 dB/mode).

2. Resource Requirements for General Graph States

The required offline squeezing for a graph state is determined by the underlying graph and can be bounded using its adjacency structure. For universal cluster states, the squeezing overhead per mode remains bounded as the cluster grows.

  • 2. Resource Requirements for General Graph States: The squeezing required for each resource mode depends linearly on target accuracy s and on the adjacency matrix’s singular values.The theorem gives the resource requirement in the large-squeezing limit.
  • 2. Resource Requirements for General Graph States: The graph structure completely determines the squeezing overhead, allowing exact resource calculations for arbitrary graph states.Upper bounds suffice in situations where exact resources are unnecessary.
  • 2. Resource Requirements for General Graph States: Universal cluster states with fixed accuracy have bounded squeezing overhead because their maximum vertex degree is 4.The paper gives a corresponding overhead of 17 and an additional 12.31 dB of squeezing per mode.
  • 2. Resource Requirements for General Graph States: 5 is the maximum overhead for quantum wires, corresponding to an additional 6.99 dB of squeezing.This provides a separate resource bound for the quantum-wire graph family.
  • 2. Resource Requirements for General Graph States: 4.41N^2 dB of squeezing is saved for an N × N square lattice using the decompositional method instead of two CZ gates per vertex.The comparison uses 12.31 dB per vertex for decomposition versus 16.72 dB for the canonical method.
  • 2. Resource Requirements for General Graph States: Multiple-rail encoding may reduce finite-squeezing errors, but its resource requirements and effectiveness are only bounded rather than completely proved.For m rails, the excess-noise reduction scales as 1/(2s^2m), while the stated lower bound scales as roughly 1/(2s^2m^2).

B. Optical Cluster-State Computation

Optical CV cluster computation uses quadrature measurements for Gaussian operations and photon counting to generate a measurement-dependent cubic phase resource. Adaptive squeezing then realizes arbitrary cubic phase gates, completing universal computation.

  • B. Optical Cluster-State Computation: Rotated-quadrature measurements followed by rescaling implement shearing transformations and other multi-qumode Gaussian operations.These operations can be performed with homodyne detection on sufficiently connected graph states.
  • B. Optical Cluster-State Computation: The cubic phase gate is more challenging because its Hamiltonian is nonquadratic and requires a nonlinear optical resource or photon counting.The paper considers embedding the resource in the cluster or generating it by measuring an existing Gaussian graph state.
  • B. Optical Cluster-State Computation: Photon counting on a displaced two-mode squeezed resource approximately collapses the unmeasured qumode into a cubic phase state dependent on the measurement result n.The resulting phase coefficient is γ(n) = (6√2n + 1)^-1.
  • B. Optical Cluster-State Computation: A cubic phase gate e^(iaq^3) is implemented by combining the measurement-dependent gate e^(iγ(n)q^3) with two squeezers depending on n and a.The required squeezing parameter is t(n) = [a/γ(n)]^(1/3).
  • B. Optical Cluster-State Computation: The squeezers must be applied after photon counting because their strength depends on the measurement outcome, making the protocol adaptive.The order of measurements therefore matters.

VIII. DISCUSSION AND CONCLUSION

The paper refines CV cluster-state computation through new graph-state representations, resource analyses, and an optical cubic-phase-gate implementation. It establishes universal resource states and shows that offline squeezing per mode need not grow with cluster size, while identifying scalability challenges from error accumulation and error correction.

  • Graph-state representations: The paper introduces Heisenberg nullifier and Wigner-function representations to analyze graph-state transformations under measurement and establish universal CV cluster resource states.These representations address the unwieldiness of the earlier Schrödinger representation for graph states of nontrivial size.
  • Universal computation: A displaced number-state measurement generates a cubic phase state, which combines with squeezing-correction subclusters to implement a cubic phase gate of any specified strength.The cubic phase gate is implemented within the cluster-state formalism through photon counting and homodyne measurements.
  • Offline resources: The paper proves that offline squeezing overhead per mode does not grow with cluster-state size for a given accuracy, alleviating concerns about excessive resource overhead.This result applies to constructing graph states through passive linear optics and includes universal cluster states.
  • Universal computation: Suitable non-Gaussian resource states, combined with Gaussian operations such as homodyne detection, suffice for universal quantum computation in the cluster-state framework.The paper suggests that difficult nonlinear measurements could generate these resources offline for later use by consumers limited to simpler measurements.
  • Scalability: Scalability remains constrained by measurement-step errors that grow linearly with cluster length, making efficient CV error correction and practical non-Gaussian integration the main challenges.Increasing cluster accuracy can compensate for accumulated errors but makes squeezing per mode depend on cluster and computation size.

APPENDIX A: THE NULLIFIER FORMALISM FOR QUADRATURE MEASUREMENTS

The appendix develops an efficient nullifier-based formalism for quadrature measurements on CV graph states. It classifies measurements by commutation structure, updates the nullifiers accordingly, and illustrates the procedure on a three-qumode linear cluster.

  • Measurement formalism: Quadrature measurement transformations have an efficient nullifier description based on whether the measured operator commutes with all nullifier basis elements.The formalism reduces the analysis to two distinct commutation cases.
  • Measurement formalism: If the measured momentum commutes with every nullifier, the state is already an eigenstate of that momentum and measurement leaves it undisturbed.The measurement result becomes the eigenvalue associated with the measured mode.
  • Measurement formalism: When exactly one nullifier fails to commute, replacing that nullifier with the measured operator minus its outcome yields the transformed nullifier algebra.The measured mode then becomes disentangled and can be discarded after choosing a suitable basis.
  • Measurement formalism: General quadrature measurements of the form p + s q are handled by applying the unitary exp(i s q^2/2) before a standard momentum measurement.Computational-basis q measurements can be analyzed analogously.
  • Three-qumode example: For a linear three-qumode cluster, sequential p-measurements on the first two modes leave the remaining mode in the momentum eigenstate |m1⟩p.The result follows by updating the cluster nullifiers after outcomes m1 and m2 and discarding measured modes.
Loading 0903.3233v1…