LATTICE

When does a measurement advantage survive the experiment?

LATTICE examines phase-estimation protocols after accounting for state preparation, finite samples, decoherence, and imperfect readout. The research question is how much usable information remains after those costs are included.

The central comparison is resource-matched: a candidate quantum protocol and an appropriate classical baseline should be evaluated under the same acquisition budget. The analysis maps where a measurement advantage can persist, where it disappears, and which assumptions control that boundary.

1. 01Prepare
2. 02Interrogate
3. 03Read out
4. 04Reset

Compare estimation error at a fixed total acquisition time, with state-preparation and readout costs included.

Observation model

From a physical signal to an estimate.

A useful reference is a binary phase measurement with finite contrast. Let N denote the number of independent readouts, V the visibility, and φ the accumulated phase. The observed count is modeled as:

k ∼ Binomial(N, p)p(1 | φ) = ½ [1 + V cos(φ)]

The phase may encode a field through φ = γBτ, where γ is the coupling, B the field, and τ the interrogation time. Near quadrature, the classical Fisher-information bound for an unbiased field estimator is:

Var(B̂) ≥ 1 / [N V² γ² τ²]

This reference assumes independent outcomes, a known contrast, and a locally identifiable phase. Unknown visibility, phase wrapping, or a changing calibration must be treated as part of the inference problem.

The resource constraint changes the comparison

Increasing interrogation time can improve phase accumulation but reduce visibility and the number of observations that fit in the budget. Preparation and readout consume that same budget:

T_cycle = T_prepare + τ + T_read + T_resetN = ⌊T_budget / T_cycle⌋

The relevant comparison therefore includes the acquisition schedule, nuisance parameters, estimator choice, and the noise channel. An ideal scaling law is only one component of that analysis.

Application context

A weak field in a drifting environment.

Consider repeated field measurements while the sensor’s contrast and phase offset change with temperature or time. A shift in the output could reflect the field, the instrument, or both. The research task is to separate those contributions and quantify what remains identifiable within the available measurement window.

This connects the estimation problem to calibration cadence, reference measurements, and the decision to extend acquisition or report insufficient information.

Method

Simulate phase estimation across a specified noise model, then sweep sample count, interrogation time, and readout error. Keep the baseline and candidate budgets explicit.

Evaluation

Report estimator bias, mean squared error, interval coverage, and sensitivity to noise-model mismatch. Separate theoretical bounds from achieved estimator performance.

Output specification

A reproducible comparison protocol, a resource accounting specification, and a documented set of failure regimes.

Evaluation protocol and limitations +

Start with a bounded problem

Use a single-parameter estimation task with declared priors, parameter range, and measurement access. Evaluate repeated independent runs rather than selecting a favorable realization.

Test the assumptions

Vary noise strength and calibration error independently. A protocol that improves under one assumed noise channel may lose that improvement when the channel is misspecified.

Boundary of the study

A simulation result does not establish an advantage on a physical device. Hardware validation requires a specified platform, calibration records, and measured acquisition overhead.

Methodological reference: Giovannetti, Lloyd & Maccone, Advances in Quantum Metrology (2011).

Further reference: Degen, Reinhard & Cappellaro, Quantum sensing, Reviews of Modern Physics 89, 035002 (2017).