Experimental Foundations Roadmap

A useful experimental-foundations roadmap starts from a falsifiable hypothesis and works backward to an observable and its controls. Increasing size or precision is valuable only when it improves discrimination between specified alternatives. This page organizes stable test strategies; it is not a list of promised discovery dates or a ranking of current records.

Match the question to the observable

Bell locality. Bell tests compare spatial correlations with a Bell-local bound under explicit setting and sampling assumptions. Their goal is not to detect a controllable faster-than-light message. Marginal no-signaling tests answer the separate communication question.

Temporal definite-value models. Leggett–Garg tests and macrorealism tests probe temporal consistency, with intervention disturbance central to the inference.

Modified dynamics. Collapse-model tests compare measured coherence, heating, or radiation with parameterized localization noise. Linearity tests search for specified state-dependent dynamics.

Probability rules. Born-rule tests use quantities such as higher-order interference, with detector response and aperture-dependent propagation included.

Mediation. Gravity-mediated entanglement aims to distinguish interaction models through an entanglement witness, provided other channels are controlled.

Build a chain from theory to counts

A test requires four connected objects: a null or alternative model, a predicted observable, a detector model, and an inference procedure. In a counting experiment, write schematically

$$n_k\sim\operatorname{Poisson}[\mu_k(\theta,\eta)],$$

where $\theta$ contains the physics parameters and $\eta$ the nuisance parameters. The same change in expected counts may arise from either group. Controls must break those degeneracies; collecting more counts alone need not do so.

For example, a visibility model $V(t)=A e^{-(\Gamma_{\rm env}+\Gamma_{\rm new})t}$ cannot identify $\Gamma_{\rm new}$ if $\Gamma_{\rm env}$ is arbitrary. Independently varied pressure, temperature, separation, or mass can give the two mechanisms different predicted dependence. An independently measured contrast $A$ prevents preparation error from being silently absorbed into the desired effect.

A staged investigation

First demonstrate the relevant preparation and readout. For a cat state, show coherence-sensitive data, not only two populated branches. Next calibrate the nuisance effects with control settings that retain sensitivity to them. Then acquire interleaved blinded or predefined analysis data for the test statistic. Finally report a model-conditioned exclusion or discrepancy with uncertainty and reproducible assumptions.

Each stage is useful scientific progress, but they support different claims. Ground-state cooling supports state preparation; it does not by itself demonstrate a separated superposition. A fringe visibility constrains dephasing within a model; it does not resolve the measurement problem. A small statistical error does not eliminate a shared calibration bias.

Compare platforms by sensitivity

Matter waves directly expose spatial coherence, superconducting cats expose controlled oscillator branches, and optomechanics can probe diffusion without preparing a large spatial cat. The relevant comparison is the predicted signal relative to independently constrained noise for a specified alternative.

For a toy extra-dephasing parameter measured through $r=e^{-\Gamma t}$, local propagation of an uncertainty $\sigma_r$ gives $\sigma_\Gamma\simeq\sigma_r/(tr)$. Longer time helps the derivative but can reduce $r$ and detected contrast, increasing uncertainty. Optimizing duration therefore requires the full noise and count model.

Exercise

With ordinary decoherence $V=e^{-\gamma t}$ and a fixed absolute visibility error $\sigma_V$, minimize the approximate sensitivity scale $\sigma_\Gamma=\sigma_V e^{\gamma t}/t$.

Solution

The logarithmic derivative is $\gamma-1/t$, so the optimum is $t=1/\gamma$. This toy optimum changes when the acquisition cost, errors, or preparation fidelity depend on duration; it illustrates why maximizing the hold time alone is not an experimental strategy.

References

• Bassi, A., et al., “Models of wave-function collapse, underlying theories, and experimental tests,” Reviews of Modern Physics 85, 471–527 (2013), doi:10.1103/RevModPhys.85.471 — model-to-observable experimental strategies.
• Emary, C., Lambert, N., and Nori, F., “Leggett–Garg inequalities,” Reports on Progress in Physics 77, 016001 (2014), arXiv:1304.5133 — why temporal tests require a distinct control strategy.