PaperlayerView exact revision on arXiv

Source-linked AI summary

Efficient tools for quantum metrology with uncorrelated noise

Jan Kolodynski, Rafal Demkowicz-Dobrzanski

arXiv:1303.7271v2quant-ph

TL;DR

Decoherence makes it important to quantify the precision attainable in quantum metrology, where ideal Heisenberg scaling can be difficult to achieve. The paper compares methods for deriving precision bounds and extends the channel extension method to finite particle numbers. Its finite-N bound interpolates between the small-probe and asymptotic regimes and is more accurate than its asymptotic version.

  • Problem

    Imperfections make ideal Heisenberg scaling difficult to achieve, motivating the question of how noise affects attainable precision in quantum metrology.

    Source: §1 · p. 4

  • Method

    The paper extends the channel extension method to finite probes, formulating its Kraus-representation optimization as an efficiently solvable semidefinite program.

    Source: §1 · p. 6, §4.2 · p. 19

  • Results

    The finite-N channel extension bound varies smoothly from N = 1 to N = ∞, is tight at N = ∞, and provides more accurate bounds than its asymptotic version.

    Source: §4.2 · p. 19

  • Takeaways & Limitations

    Classical simulation gives a geometric explanation of inevitable asymptotic SQL-like scaling for ϕ-non-extremal channels.

    Source: §4.1.1 · p. 15

    Because the finite-N CE bound is convex in N, it cannot be naively used to constrain precision by replacing N with the mean photon number.

    Source: §4.2 · p. 20

Abstract

from arXiv · show

Quantum metrology offers an enhanced performance in experiments such as gravitational wave-detection, magnetometry or atomic clocks frequency calibration. The enhancement, however, requires a delicate tuning of relevant quantum features such as entanglement or squeezing. For any practical application the inevitable impact of decoherence needs to be taken into account in order to correctly quantify the ultimate attainable gain in precision. We compare the applicability and the effectiveness of various methods of calculating the ultimate precision bounds resulting from the presence of decoherence. This allows us to put a number of seemingly unrelated concepts into a common framework and arrive at an explicit hierarchy of quantum metrological methods in terms of the tightness of the bounds they provide. In particular, we show a way to extend the techniques originally proposed in Demkowicz-Dobrzanski et al 2012 Nat. Commun. 3 1063, so that they can be efficiently applied not only in the asymptotic but also in the finite-number of particles regime. As a result, we obtain a simple and direct method, yielding bounds that interpolate between the quantum enhanced scaling characteristic for small number of particles and the asymptotic regime, where quantum enhancement amounts to a constant factor improvement. Methods are applied to numerous models including noisy phase and frequency estimation, as well as the estimation of the decoherence strength itself.

1. Introduction · Appendix C RLD-based bound as a special case of asymptotic CE bound 29 · 2. Quantum Fisher Information

The paper orders quantum-metrology precision bounds by predictive power and extends the channel-extension method to finite probes. It also introduces estimation-theory tools, including the quantum Fisher information and related bounds.

1. Introduction

  • Independent decoherence can make quantum-enhanced precision SQL-like at large N, limiting the asymptotic gain to a constant factor.The ideal Heisenberg scaling is therefore difficult to achieve under realistic imperfections.Source: §1 · p. 4
  • The channel-extension method provides more stringent precision bounds than classical or quantum simulation and the other compared approaches.The bounds are guaranteed to be saturable only when an ancilla entangled with the input cannot improve the estimation task.Source: §1 · pp. 5–6
  • Finite-N channel extension optimizes over Kraus representations via an efficiently solvable semidefinite program and applies to several noisy phase and frequency-estimation models.The paper considers dephasing, depolarization, loss, and spontaneous emission, focusing on noise that commutes with the parameterized unitary; correlated decoherence is outside its scope.Source: §1 · p. 6
  • The methods also cover noise-strength estimation and can bound asymptotic scaling when the asymptotic channel-extension method fails.The paper notes that independently acting non-Markovian effects may fit its framework, but correlations between particles’ decoherence processes do not.Source: §1 · p. 6

2.1. Classical Cramér-Rao bound

  • The classical Fisher information is additive across independent repetitions and supports local estimation; the classical Cramér–Rao bound is saturable in the many-repetition limit.The classical estimation procedure infers the parameter from outcomes sampled from a probability distribution.Source: §2.1 · p. 7

2.2. Quantum Cramér-Rao bound

  • The quantum Cramér–Rao bound depends on the parameterized state rather than the chosen POVM, and an SLD-eigenbasis measurement can saturate it as repetitions grow.The bound’s information quantity is the quantum Fisher information, defined through the symmetric logarithmic derivative.Source: §2.2 · pp. 7–8

2.3. Purification-based definition of QFI

  • For mixed states, purification-based definitions avoid diagonalizing the state; the quantum Fisher information equals the minimum over purification quantum Fisher informations.The paper builds on an alternative purification definition whose form facilitates minimization.Source: §2.3 · pp. 8–9

2.4. Right logarithmic derivative (RLD)-based upper bound on QFI

  • Relaxing the logarithmic derivative’s Hermiticity yields an RLD-based upper bound on quantum Fisher information when the state derivative lies within the state’s support.This bound is tight only when the right and symmetric logarithmic derivatives coincide; the discussion excludes multiparameter estimation.Source: §2.4 · p. 9

3. Estimation of a quantum channel

Channel QFI is the maximum output-state QFI over pure inputs, giving an operational precision bound for estimating a parameter encoded by a quantum channel. Purification and input-space extension provide ways to bound or evaluate it, with extension improving single-channel precision only for certain noise models.

3.1. Channel QFI

  • Channel QFI maximizes output QFI over pure inputs, and the Quantum Cramér-Rao bound applies to the resulting channel output.QFI convexity motivates restricting the optimization to pure inputs; Figure 2a illustrates the definition.Source: §3.1 · p. 9

3.2. Purification-based definition of channel QFI

  • Purifying a channel makes its QFI depend on a local environment rotation, equivalent to choosing a Kraus representation, optimized to hide parameter information.For unitary time evolution, the minimum occurs at h = ⟨ψin|H|ψin⟩.Source: §3.2 · p. 11

3.3. Extended channel QFI

  • Because optimizing pure-input channel QFI requires a difficult max–min optimization, an auxiliary space yields an extended QFI that can be evaluated efficiently by semidefinite programming.The extension permits mixed system inputs and allows the maximization and minimization to be exchanged; if its optimum is pure, auxiliary entanglement is unnecessary.Source: §3.3 · pp. 11–12
  • Extension improves single-channel precision only for depolarization and spontaneous-emission channels, where inputs maximally entangled with an auxiliary space are optimal.The comparison covers phase estimation with depolarization, loss, and spontaneous emission; no numerical values are supplied here.Source: dephasing · p. 12

dephasing, · Depolarization

The RLD-based bound depends only on a channel’s Choi-Jamiołkowski representation, but its applicability and tightness are limited. Its applicability condition has a geometric interpretation, and the extended-channel setting does not imply that mixed inputs outperform pure inputs for an unextended channel.

3.4. RLD-based upper bound on extended channel QFI

  • The Choi-Jamiołkowski representation of a channel map is defined as Ω_Λϕ = Λ_ϕ ⊗ I applied to |I⟩, with |I⟩ = Σ_i |i⟩⊗|i⟩.The C-J representation is introduced in the discussion of the RLD-based bound, using the definition in.Source: §3.4 · p. 12
  • The RLD-based upper bound depends solely on the channel’s C-J representation and contains no additional information about the input states accepted by the map.This differs from the purification-based unextended and extended channel QFIs and facilitates analytic results in Table 1.Source: §3.4 · p. 12
  • The RLD bound is valid only when the derivative of the channel’s C-J matrix is fully supported by the C-J matrix [37].The passage states this support condition as a restriction on applicability.Source: §3.4 · p. 12
  • The RLD-based bounds have limitations in both applicability and tightness.The passage flags these restrictions while noting that the bounds shown for relevant channels in Table 1 appear far above an unspecified reference.Source: §3.4 · p. 12
  • The support condition is equivalent to the channel being ϕ-non-extremal, giving the C-J-only bound a geometric interpretation.The equivalence is proved in Appendix B.Source: §3.4 · p. 12
  • For an unextended channel, mixed inputs never outperform pure-state inputs because the QFI is convex.This clarifies that the extended-channel QFI in (12) does not represent using mixed inputs for an unextended channel.Source: §3.4 · p. 12

Loss

Bound (13) is additive for the stated maps, and the resulting bound limits QFI growth to at most linear in the number of parallel extended channels. Because extension can only improve precision, the bound also applies to unextended channels, whose asymptotic precision scaling is limited to a constant-factor improvement over the SQL.

Loss

  • Bound (13) is additive for any maps Λ(1).Source: Loss · p. 13
  • The bound restricts the QFI of N extended channels used in parallel to scale at most linearly with N.Source: Loss · p. 13
  • Because extension can only improve precision, the bound also applies to N unextended channel uses, limiting asymptotic precision scaling to a constant-factor improvement over the SQL.Source: Loss · p. 13

4. Estimation of N independent quantum channels

The section compares bounds for estimating parameters encoded in N independent quantum channels, then introduces a finite-N channel-extension method for tighter precision limits between the single-channel and asymptotic regimes.

4. Estimation of N independent quantum channels

  • The probe evolves through N identical independent channels acting on a possibly entangled pure input; linear or quadratic QFI scaling corresponds to SQL or HL precision scaling.Source: §4 · p. 13

4.1. Standard quantum limit (SQL)-like bounds on precision in the asymptotic N limit

  • The asymptotic methods bound maximal quantum precision enhancement, but computing the N-channel QFI directly is generally infeasible because its complexity rises exponentially with N.CS and RLD bounds are unavailable for extremal maps such as lossy and spontaneous-emission interferometry, while QS also covers loss and is as accurate as CE there.Source: §4.1 · p. 14

4.1.1. Classical simulation (CS) method

  • Classical simulation bounds the N-channel QFI by N times a classical Fisher information for non-extremal channels, giving an intuitive account of SQL-like scaling but weaker bounds than CE.Full-rank channels are among those covered by this approach.Source: §4.1.1 · p. 15

4.1.2. Quantum simulation (QS) method

  • The QS method applies to dephasing and depolarization with bounds coinciding with CE, also matches CE for lossy interferometry, but fails for spontaneous emission.QS cannot outperform CE because its constraint is stricter; for spontaneous emission, the β̃_K = 0 condition forces α̃_K to be non-proportional to identity.Source: §4.1.2 · pp. 16–17

4.1.3. Channel extension (CE) method

  • For asymptotic SQL-bound channels, the CE method gives the tightest bounds across the broadest class, and its bounds are attainable for dephasing and lossy interferometry.The CE bounds can be determined by semidefinite programming; for dephasing and loss, squeezing strategies attain the asymptotic bounds.Source: §4.1.3 · pp. 17–18

4.2. Finite-N channel extension (CE) method

  • The finite-N CE bound is efficiently computable, tighter than its asymptotic version, and connects the extended-channel QFI at N=1 to the asymptotic CE bound as N→∞.It remains applicable when the asymptotic CE condition β̃_K = 0 cannot be met, including cases where the asymptotic precision may exceed SQL-like scaling.Source: §4.2 · p. 19
  • For dephasing, low-N GHZ probes attain the finite-N bound, while in lossy interferometry it is tight only at moderate and very large N.Figure 3 compares finite-N and asymptotic CE bounds with spin-squeezed and GHZ strategies for dephasing and optimized N-photon states for loss; the finite-N bound cannot be naively applied at mean photon number.Source: Figure 3 · p. 20, §4.2 · p. 20

5. Frequency estimation in atomic models

The section applies quantum Cramér–Rao and channel-QFI methods to Ramsey frequency estimation with uncorrelated noise, optimizing each shot duration. It reports finite-N and asymptotic bounds and quantum enhancements, while noting that some enhancements are not guaranteed to be attainable.

5. Frequency estimation in atomic models

  • The Ramsey setup estimates oscillator detuning using N identical two-level atoms over total time T, repeating shots of duration t; for fixed t, the bound is saturable as T→∞.The independent-particle noise is modeled by a Lindblad master equation.Source: §5 · p. 21
  • Because ω=ϕ/t and the unitary and decoherence maps commute, the phase-estimation results apply after rescaling QFI and substituting master-equation-dependent noise strengths.The considered noise maps are dephasing, depolarization, loss, and spontaneous emission; the effective QFI per shot duration is optimized over t.Source: §5 · p. 21
  • Table 3 reports t-optimized channel QFIs, finite-N and asymptotic CE bounds, and maximal enhancements; depolarization, loss, and spontaneous-emission bounds are newly derived, while dephasing results are recovered.The dephasing enhancement factor √e had already been reported.Source: Table 3 · p. 22, §5 · p. 22
  • Enhancement factors for depolarization and spontaneous emission are bounds rather than guaranteed attainable gains; unextended versions may not be achievable.Source: §5 · p. 22

6. Estimation of decoherence strength

Section 6 applies the Section 4 methods to estimate decoherence strength η, finding that quantum estimation strategies improve precision only for depolarization among the four noise models considered.

6. Estimation of decoherence strength

  • Only depolarization allows quantum estimation strategies to enhance precision; for all other considered channels, F[Λη] = F[Λη ⊗ I] = Fas.The geometrical CS bounds coincide with the asymptotic extended-channel QFIs for decoherence-strength estimation.Source: Table 4 · p. 23
  • Because channel-space motion away from the boundary cannot be amplified N-fold, entangled probes yield constant-factor rather than scaling enhancements, and geometric classical simulability tightly bounds asymptotic precision.For the four noise models considered, F^CS_as = F[Λη ⊗ I].Source: §6 · p. 23
  • For extended channels, factorizable inputs are optimal: channels are uncorrelated with one another, though each particle may be entangled with its ancilla.This had been noted for low-noise and generalized Pauli channels, including dephasing and depolarization.Source: §6 · p. 23
  • For dephasing, F[Λη] = F[Λη ⊗ I] = 1/(1 − η^2), achievable with unentangled qubits in any pure equatorial-plane Bloch-sphere state.The single-channel extension is unnecessary, and the CS bound is attainable classically.Source: §6 · p. 23
  • Spontaneous emission is fully classical, while loss has F^CS_as = F[Λη] = F[Λη ⊗ I] = 1/(η(1 − η)).For spontaneous emission, the asymptotic CS bound coincides with the extended and unextended channel QFIs derived in. For loss, a photon in any mixed state attains the bound without an ancilla; photon-number fluctuations limit precision, and Fock states can attain the bound.Source: §6 · p. 24
  • For depolarization, extending the qubit channel increases precision, F[Λη] < F[Λη ⊗ I], leaving scope for entangled probes across unextended channels.Such enhancement was observed for two depolarization channels used in parallel.Source: §6 · p. 24

7. Further discussion · 8. Summary and outlook

The paper develops efficient bounds for quantum metrology with uncorrelated noise and extends them to finite-particle regimes and related estimation problems. It also discusses the scope of SQL-like bounds, candidate optimal states, and future directions.

7. Further discussion

  • SQL-like bounds apply to indefinite-particle-number states using mean particle number and avoid controversies arising in decoherence-free scenarios.For small parameter fluctuations, QFI bounds are saturable in a single shot by using independent groups of finite but sufficiently large size.Source: §7 · p. 24
  • Asymptotically optimal inputs need not have long-range particle correlations, so searches can be restricted to short-range states such as low-bond-dimension matrix product states.The discussion connects this restriction to grouped-particle inputs that approach the ultimate asymptotic SQL bound.Source: §7 · pp. 24–25
  • The authors conjecture that Bayesian and information-theoretic methods recover bounds compatible with the local approach regardless of prior knowledge; this is known for optical interferometry with losses, but remains open more generally.Source: §7 · p. 25

8. Summary and outlook

  • The methods derive precision bounds directly from single-particle channel Kraus operators, without educated guesses or numerical searches for optimal input states.They compare CS, QS, RLD, and CE for asymptotic phase estimation under different decoherence models.Source: §8 · p. 25
  • The generalized CE method provides tighter finite-particle bounds and can be formulated as a semidefinite program.The methods are also applied to frequency estimation.Source: §8 · p. 25
  • For estimating decoherence strength, the simplest CS approach typically provides the tightest bounds.Although the bounds avoid searching for optimal many-particle inputs, constructing an explicit optimal strategy may require such a search; reaching the fundamental bound for a given input certifies optimality.Source: §8 · p. 25
  • A natural future direction is extending these methods to multiparameter estimation, where the most fruitful approach and possible new bounds are unclear.Source: §8 · p. 25

Appendix A. Channels considered

The appendix specifies the noisy channel models used for phase, frequency, and decoherence-strength estimation, together with the Kraus-representation purification used to determine the extended-channel QFI.

Appendix A. Channels considered

  • The phase parameter ϕ is a z-axis Bloch-ball rotation generated by Uϕ = exp[iσ3ϕ/2], and the considered noise maps Dη commute with that rotation.The notation uses Id for the d×d identity matrix and σi for the Pauli operators.Source: Appendix A · p. 26
  • For each channel, the appendix gives its Kraus operators and specifies the purification for extended-channel QFI through the optimal Kraus-rotation generator h.The generator h is the one introduced in equation (10), and the extended-channel QFI is identified as F[Λϕ ⊗ I] in Table 1.Source: Appendix A · p. 26
  • For frequency estimation, effective one-particle Kraus operators follow by substituting ϕ→ωt and η→η(t); the appendix also gives each particle’s Liouvillian and effective η(t).For decoherence-strength estimation, η is the estimated parameter in the noise map Λϕ=η = Dη.Source: Appendix A · p. 26

Appendix A.1. Dephasing · Appendix A.2. Depolarization

Appendix A.1 characterizes dephasing by an equatorial Bloch-ball contraction, while Appendix A.2 characterizes depolarization by isotropic contraction; each gives an optimal purification for extended channel QFI (12) and a one-particle Liouvillian for frequency estimation.

Appendix A.1. Dephasing

  • Dephasing parameter η represents the final equatorial radius of the Bloch ball after uniform shrinking in the xy plane.Source: Appendix A.1 · p. 26
  • The appendix gives an optimal purification determining the extended channel QFI (12) for dephasing.Source: Appendix A.1 · p. 26
  • The appendix presents a one-particle Liouvillian for dephasing in frequency estimation tasks.Source: Appendix A.1 · p. 26

Appendix A.2. Depolarization

  • Depolarization parameter η represents the final radius of the Bloch ball after isotropic shrinking by the channel.Source: Appendix A.2 · p. 26
  • The appendix gives an optimal purification determining the extended channel QFI (12) for depolarization.Source: Appendix A.2 · p. 27
  • The appendix presents a one-particle Liouvillian for depolarization in frequency estimation tasks.Source: Appendix A.2 · p. 27

Appendix A.3. Loss · Appendix A.4. Spontaneous emission (amplitude damping)

The appendices define loss and spontaneous-emission channel parameters and describe their use in extended-channel-QFI calculations. The loss discussion also identifies broader models to which the methods apply while focusing on a simple equal-loss model.

Appendix A.3. Loss

  • In the loss model, η is each particle’s survival probability, and the single-particle channel maps a two-dimensional system to a three-dimensional system with a vacuum mode.The vacuum mode represents particle loss.Source: Appendix A.3 · p. 27
  • For loss, phase acquisition and particle loss commute, making Uϕ and ˜Uϕ physically indistinguishable; in a Mach–Zehnder setup, η is equal-arm effective power transmittance.It accounts for preparation and transmission loss and detector inefficiencies.Source: Appendix A.3 · p. 27
  • The paper presents an optimal purification for determining the extended channel QFI (12).Source: Appendix A.3 · p. 27
  • For frequency-estimation tasks under loss, the appendix identifies a one-particle Liouvillian and transition generators from qubit basis states to the vacuum mode.The generators are written as m,+ = |vac⟩⟨m| for |0⟩ and |1⟩.Source: Appendix A.3 · p. 27
  • The methods also apply to unequal arm losses and distinguished preparation, transmission, and detection losses, though the paper restricts its analysis to the simplest loss model.Source: Appendix A.3 · p. 28

Appendix A.4. Spontaneous emission (amplitude damping)

  • For spontaneous emission, η is the radius of the disk obtained by projecting the channel’s deformed Bloch-ball output onto the xy plane.Source: Appendix A.4 · p. 28
  • The cited passage identifies an optimal purification for determining the extended channel QFI (12); the accompanying parameter specification associates ξ and h with equation (A.11).Source: Appendix A.4 · p. 28

Appendix B. Equivalence of RLD-based bound applicability and local classical simulability of a channel · Appendix C. RLD-based bound as a special case of asymptotic CE bound · Appendix D. Optimal local QS of a channel

These appendices characterize when RLD-based bounds apply, show how they fit within the asymptotic CE bound, and give conditions for locally quantum-simulable channels.

Appendix B. Equivalence of RLD-based bound applicability and local classical simulability of a channel

  • ϕ-non-extremality is characterized by a nonzero Hermitian matrix μ_ij; substitution proves one direction, while resolving derivatives into channel support and null-space establishes the converse.Source: Appendix B · pp. 28–29
  • RLD-bound applicability is equivalent to the stated local classical-simulability condition for a channel.Source: Appendix B · pp. 28–29

Appendix C. RLD-based bound as a special case of asymptotic CE bound

  • For channels with an RLD-based bound, the authors choose a Kraus gauge satisfying the CE constraint β̃_K = 0.Source: Appendix C · p. 29
  • The asymptotic CE bound is at least as tight as the RLD-based bound on the extended-channel QFI, with the RLD bound arising from a possibly suboptimal Kraus representation.Source: Appendix C · p. 29

Appendix D. Optimal local QS of a channel

  • A rank-r channel must satisfy the stated condition to be locally quantum simulable near the parameter value.Source: Appendix D · p. 30
  • The local simulation yields purified-state QFI 4b^2 and requires finite b; locality lets the remaining unitary entries be chosen to satisfy unitarity.Source: Appendix D · p. 30
  • Local quantum simulation is possible when the channel has a Kraus representation satisfying main-text conditions (21), generated via a Hermitian r×r matrix h.Source: Appendix D · p. 30

Appendix E. Finite-N CE method as a semi-definite programming task

The appendix formulates the finite-N CE bound as a semi-definite program over locally equivalent Kraus representations. A CVX implementation evaluates the program from a channel’s Kraus operators and their derivatives, with modifications for extended-channel QFI bounds.

Appendix E. Finite-N CE method as a semi-definite programming task

  • The finite-N CE bound can be evaluated as an SDP by minimizing over locally equivalent Kraus representations generated by Hermitian h, with the β~K = 0 constraint.This extends the approach in.Source: Appendix E · p. 31
  • A CVX program for Matlab efficiently evaluates the SDP from a channel’s Kraus operators and their derivatives; only these inputs are needed because the QFI is a local quantity.Source: Appendix E · p. 31
  • Slight modifications to the program also evaluate the extended-channel QFI at N = 1 and the asymptotic extended-channel QFI with N = 1 and β~K = 0.Source: Appendix E · p. 31
Loading 1303.7271v2…