PaperlayerView exact revision on arXiv

Source-linked AI summary

Optimal Hamiltonian Parameter Estimation in the Presence of Nuisance Parameters

Zhiyao Hu, Haidong Yuan, Liang Jiang, Zain H. Saleem

arXiv:2609.12857v1quant-phcs.ITmath.IT

TL;DR

Nuisance parameters make target precision depend on correlations between target and nuisance responses, while joint design of the probe and control remains challenging for general time-dependent Hamiltonians. The paper introduces an effective generator to characterize the target precision limit and constructs protocols that attain it.

  • Problem

    With nuisance parameters, target precision depends on the full QFIM and correlations between target and nuisance responses, while jointly optimizing probe and control remains an open challenge for general time-dependent Hamiltonians.

    Source: p. 1

  • Method

    The framework removes from the target generator the response attributable to nuisance variations, forming an effective generator for target estimation.

    Source: p. 1

  • Results

    The authors derive the fundamental target-precision limit and construct probe, control, and measurement protocols that saturate it.

    Source: p. 5

  • Takeaways & Limitations

    The effective-generator approach is presented as a general framework for target estimation when not all parameters are known.

    Source: p. 5

    The nuisance-coordinate choice is a convenience and does not affect target-parameter precision.

    Source: p. 2

Abstract

from arXiv · show

In many sensing applications, the quantity of interest is not the only unknown, there are also additional unknown parameters, known as nuisance parameters, that affect the precision of estimation. While the ultimate local precision limit for a target parameter is well understood in the absence of nuisance parameters, the problem becomes significantly more challenging when they are present. In this work, we develop a framework for optimal Hamiltonian parameter estimation in the presence of nuisance parameters. We introduce an effective generator that captures the influence of nuisance parameters on the target precision, providing an explicit characterization of the ultimate precision limit for estimating the target parameter. Finally, we provide explicit optimal protocols, including probe state, control, and measurement that saturate this fundamental limit.

END MATTER

The end-matter examples apply the framework to frequency-drift estimation with an unknown offset and AC-frequency estimation with an unknown phase. They compare optimal nuisance-aware protocols with a conventional strategy and specify controls for both settings.

Example 1: Frequency-drift estimation with an unknown offset

  • A single π pulse cancels the integrated response to α while retaining a nonzero response to ω; the optimal probe and binary measurement attain the limit.The probe is a balanced superposition of the σz eigenstates, and the measurement is target-specific.Source: END MATTER · p. 8
  • Collinear drift and offset generators remain distinguishable because the drift response grows linearly with time while the offset response remains constant.The changing relative magnitude prevents the drift from being absorbed into the static offset.Source: END MATTER · p. 8

Example 2: Frequency estimation with an unknown phase

  • For AC-frequency estimation, the optimal control applies π pulses when (t−β⋆)sin(ωt+ϕ) changes sign, with β⋆ a weighted median.The weight is |sin(ωt+ϕ)|.Source: END MATTER · p. 8
  • For T ≫ 2π/|ω|, repeating the procedure ν times gives total effective QFI ≃ νB²T⁴; this is the stated scaling in the long-interrogation regime relative to ω.Source: END MATTER · p. 8
  • A single interrogation duration makes the conventional AC-strategy QFIM singular, so that strategy combines experiments with two different durations.The two-duration design divides the total time between repeated evolutions of durations τ1 and τ2.Source: END MATTER · p. 8

Example 1: Frequency-drift estimation with an unknown offset

  • The conventional frequency-drift strategy’s effective QFI is four times smaller than the optimal effective QFI.The comparison is reported for estimating drift rate ω with unknown static offset α.Source: END MATTER · pp. 8–9

ADAPTIVE ESTIMATES

The adaptive procedure first estimates the operating point, then implements the locally optimal target-parameter protocol and updates the target estimate.

Adaptive procedure

  • Use an asymptotically negligible fraction of measurements, such as √ν, to obtain preliminary estimates of the target and nuisance parameters.These estimates localize the operating point for the subsequent protocol.Source: ADAPTIVE ESTIMATES · p. 9
  • Prepare the optimal probe, apply the optimal control, and perform the optimal measurement using the preliminary estimate and β⋆.The optimal choices generally depend on the unknown operating point and are therefore locally optimal.Source: ADAPTIVE ESTIMATES · p. 9
  • Nuisance parameters need only be estimated accurately enough to localize the operating point, not to their ultimate precision.Measurements can periodically update nuisance estimates when needed.Source: ADAPTIVE ESTIMATES · p. 9

EFFECTIVE QUANTUM FISHER INFORMATION WITH NUISANCE PARAMETER

The section expresses target precision with nuisance parameters through an effective QFI and an equivalent generator-based variance minimization. This formulation also covers singular nuisance blocks.

Effective QFI

  • The effective QFI is the Schur complement of the nuisance block, using a generalized Schur complement when that block is singular.The singular-block definition uses the Moore–Penrose pseudoinverse.Source: EFFECTIVE QUANTUM FISHER INFORMATION… · p. 10
  • When the nuisance QFI block is zero, positivity forces the off-diagonal block to vanish, and the effective QFI reduces to the target QFI.This is the extreme singular case described in the section.Source: EFFECTIVE QUANTUM FISHER INFORMATION… · p. 10

Effective generator

  • The generator formulation minimizes the variance of Sθ1(T)−βSθ2(T) over the nuisance weight β.For positive nuisance-generator variance, the minimizing weight is given by the covariance-based expression in the section.Source: EFFECTIVE QUANTUM FISHER INFORMATION… · p. 10
  • The minimized generator variance equals the effective QFI, including in the singular case where the variance is independent of β.The section combines the nonsingular and singular cases in a unified variational representation.Source: EFFECTIVE QUANTUM FISHER INFORMATION… · p. 10

CHOICE OF NUISANCE PARAMETER

The effective QFI does not depend on which linearly independent nuisance coordinate completes the target parameter. The orthogonal coordinate choice is a convenience, not a constraint on target precision.

Choice of nuisance parameter

  • Changing nuisance coordinates shifts the target generator by a nuisance-generator term and rescales the nuisance generator.The transformation preserves the effective QFI because the corresponding minimization weights map bijectively.Source: CHOICE OF NUISANCE PARAMETER · pp. 11–12
  • The effective QFI is invariant under the nuisance-coordinate transformation, so the orthogonal choice is only a convenient convention.The alternative nuisance coordinate need only be linearly independent of the target parameter.Source: CHOICE OF NUISANCE PARAMETER · pp. 11–12

CONSISTENCY OF THE OPTIMAL NUISANCE COEFFICIENT

The optimal nuisance coefficient has consistent variational meanings: it minimizes integrated instantaneous spectral width and, for the optimal probe, accumulated-generator variance.

CONSISTENCY OF THE OPTIMAL NUISANCE COEFFICIENT

  • The balanced probe built from the initial extremal eigenstates saturates the accumulated effective-generator spectral-width variance bound.The balance condition makes the nuisance-generator expectations equal on the two extremal states, yielding zero covariance with the effective generator.Source: CONSISTENCY OF THE OPTIMAL NUISANCE… · pp. 12–13
  • When the extreme eigenvalues are degenerate, the construction remains valid if the extremal states are chosen within their eigenspaces to satisfy the balance condition.Source: CONSISTENCY OF THE OPTIMAL NUISANCE… · p. 13
  • The two definitions of β⋆ coincide: minimizing integrated instantaneous spectral width also minimizes accumulated-generator variance for the optimal probe.This makes the control optimization and nuisance-coefficient minimization consistent whichever is performed first.Source: CONSISTENCY OF THE OPTIMAL NUISANCE… · pp. 12–13

OPTIMAL SCHEME

For one nuisance parameter, an extremal-eigenstate probe, aligned control, and effective-observable measurement jointly attain the optimized local precision bound.

OPTIMAL SCHEME

  • The optimal control transports the extremal eigenstates so they remain extreme eigenstates of the accumulated effective generator, whose variance reaches the spectral-width bound.Source: OPTIMAL SCHEME · p. 13
  • The eigenbasis of the effective optimal observable defines the optimal measurement, which can be completed as a binary POVM.Source: OPTIMAL SCHEME · p. 14
  • The measurement has zero first-order nuisance response while retaining the target's effective information, including when the nuisance QFIM block is singular.Source: OPTIMAL SCHEME · pp. 14–15
  • The probe, control, and measurement together attain the optimized local precision for the target parameter with one nuisance parameter.The construction achieves the bound for locally unbiased estimators, including along the nuisance direction.Source: OPTIMAL SCHEME · pp. 13, 15

EXTENSION TO MULTIPLE NUISANCE PARAMETERS

For multiple nuisance parameters, the framework constructs a probe, control, and measurement that attain the target's optimal effective QFI while removing first-order nuisance sensitivity.

EXTENSION TO MULTIPLE NUISANCE PARAMETERS

ROBUSTNESS OF THE ADAPTIVE PROTOCOL

Across the examples, the adaptive protocol attains the nuisance-aware optimum exactly for the linear-drift model and is locally insensitive to first-order estimation errors in the frequency model.

Toggling-function model

  • In the toggling-function model, the stated control can be implemented by a single instantaneous π pulse at T/2, and the resulting effective generator has a spectral width.The control is described through a toggling function y(t) = ±1.Source: ROBUSTNESS OF THE ADAPTIVE PROTOCOL · pp. 19–20

Linear-drift model

  • In the linear-drift model, the adaptive protocol attains the optimal nuisance-aware information exactly because its optimal control is independent of the unknown parameter values.The adaptive implementation replaces unknown parameters in the control with their current estimates.Source: ROBUSTNESS OF THE ADAPTIVE PROTOCOL · pp. 19–20

Frequency-estimation model

  • For frequency estimation with ω as target and ϕ as nuisance phase, the optimal coefficient is the weighted median of t with weight |sin(ωt + ϕ)|..Source: ROBUSTNESS OF THE ADAPTIVE PROTOCOL · p. 20
  • The frequency model's controlled generators can be aligned in the effective generator, and β⋆ ≃ T/2 in the long-time regime.The long-time statement uses averaging over rapid oscillations.Source: ROBUSTNESS OF THE ADAPTIVE PROTOCOL · p. 21
  • When |ω|T ≫ 1, |δϕ| ≪ 1, and |δω|T ≪ 1, adaptive information loss is second order in δϕ and δωT, giving first-order insensitivity to estimation errors.In this regime, β⋆ and its estimated counterpart approach T/2, and accumulated control mismatch remains small over the interrogation time.Source: ROBUSTNESS OF THE ADAPTIVE PROTOCOL · p. 21
Loading 2609.12857v1…