---
title: "Does Distance Shorten the Coherence of a Stored Bell Pair?"
subtitle: "A brief for experimenters on a test of a distributed-field prediction"
running-title: "Brief for experimenters"
book: false
toc: false
author:
  - name: Gerd Graßhoff
    email: gerd.grasshoff@hu-berlin.de
    orcid: 0000-0002-4450-357X
    affiliations:
      - name: Humboldt-Universität zu Berlin
        department: Philosophisches Institut
        city: Berlin
        country: Germany
abstract: |
  A recent field-ontology model of Bell correlations (Laidlaw 2026) predicts
  that an entangled pair held at range $r$ loses coherence as if it had been
  held for an extra time $\tau=r/c$. Standard quantum mechanics predicts no such
  term. The test needs two long-lived memory qubits at a known distance, a
  same-hardware short-range control, and a tunable coherence time. It measures
  one number, $\kappa$, which the model puts at $1$ and standard theory at $0$.
  We expect $\kappa\approx0$; a bound is the likely result. This brief states
  the hypothesis, the measurements, the controls, the sample sizes and the
  decision rule.
format:
  arxivai-typst: default
  html:
    toc: false
---

## The claim in one paragraph

Let $C_r(t)$ be the coherence of a two-qubit Bell state held for a time $t$ by two memory qubits
$r$ apart, and $C_0(t)$ the same at negligible distance. Laidlaw's model gives

$$C_r(t)=C_0(t+\kappa\tau),\qquad \tau=r/c\ \text{(geometric light time)},\quad \kappa=1 .$$

Standard theory gives $\kappa=0$ for uncorrelated local noise. For exponential decay
$C_0=e^{-2t/T_\text{coh}}$ the range factor is $R=e^{-2\kappa\tau/T_\text{coh}}$. The delay $\tau$ is the
geometric light time between the qubits, not the optical delay of the heralding fibre.

**Why we expect a null result.** Two systems with local Hamiltonians evolve as a product of local
evolutions, so a Bell state's coherence should not depend on distance. The value $\kappa=1$ is the
model's own transcription of a heuristic, not a derived quantity. A bound on $\kappa$ is the most
likely outcome, and it is worth having: it converts a philosophical position into a bounded parameter.

## Why an existing dataset does not answer it

The effect scales as $\tau/T_\text{coh}$. Published memory-qubit links protect their qubits with echo
sequences ($T_\text{coh}$ of hundreds of $\mu$s to ms), which makes $R$ close to $1$: for a 10 km link at
$T_\text{coh}=5$ ms it is $0.987$. Hensen et al. (2015) might exclude $\kappa=1$ if the relevant
coherence time was below about $60\,\mu$s; that number is in the supplement. **The test therefore needs a
deliberately reduced, known $T_\text{coh}$.** Details are in the companion desk analysis.

## What to measure

Three measurements, all on the heralded pair:

1. **Decay at short range.** On the same hardware in one laboratory, measure $C_0(t)$ for the Bell
   state on a grid of hold times, without assuming a functional form.
2. **The same at range.** At the long-range configuration measure $C_r(t)$ at a working hold time.
   Read out the *equivalent delay* $\delta$ defined by $C_r(t)=C_0(t+\delta)$, with $t$ the actual hold
   time of that configuration, including any added delay. The estimate is $\hat\kappa=\delta/\tau$.
3. **The sum rule.** Prepare both a $\phi^+$-type and a $\psi^-$-type state (they differ by a local flip
   on one node) and measure the coherence of each, plus the single-qubit Ramsey coherences $C_A$, $C_B$
   *with the same pulse sequence*. Form

$$Q=\frac{C_{\phi^+}\,C_{\psi^-}}{(C_A\,C_B)^2}.$$

For jointly Gaussian dephasing noise with any variances and any spatial correlation, standard theory
gives $Q=1$. Under H1, $Q=e^{-4\kappa\tau/T_\text{coh}}$. Correlated noise helps one Bell-state type
exactly as much as it hurts the other, and the product does not see it. This removes the main
confound of a range effect (spatially correlated noise) without measuring the noise.

## Controls

| Control | Purpose | Expected |
|---|---|---|
| Short-range $Q$ | checks the Gaussian assumption and the sequences | $Q=1$ within errors |
| Added heralding delay at short range (fibre spool or electronic delay) | separates elapsed time from distance | no shift beyond the extra hold time; van Leent et al. (2022) did this with electronic delay |
| Different $T_\text{coh}$ (weaker echo or injected noise), 3 values | the lag adds a term proportional to $1/T_\text{coh}$ and independent of $t$ | $\hat\kappa$ constant across $T_\text{coh}$ |
| Different hold times $t$, 3 values | separates the lag from a change of the decay rate (bond fragility) | intercept shift, not slope |
| Single-qubit coherence measured on the same day, same sequence | $T_\text{coh}$ enters the prediction | — |

## Sample sizes

Five-standard-deviation detection of the H1 deficit in $Q$, per Bell state, assuming Gaussian
dephasing, exponential decay, independent binomial readout and abundant single-qubit shots. These are
lower bounds on effort, computed from published numbers with a short script (available on request); your noise
budget should replace them.

| Platform and range | $T_\text{coh}$ | $\tau$ | $Q$ under H1 | Heralded events per Bell state |
|---|---|---|---|---|
| NV, $1.3$ km | $13\,\mu$s | $4.3\,\mu$s | 0.26 | 440 |
| NV, $1.3$ km | $30\,\mu$s | $4.3\,\mu$s | 0.56 | 360 |
| NV, $1.3$ km | $96\,\mu$s | $4.3\,\mu$s | 0.83 | 710 |
| NV link of $10$ km, ideal start contrast | $300\,\mu$s | $33\,\mu$s | 0.64 | 2,300 |
| same, start fidelity $0.80$ | $300\,\mu$s | $33\,\mu$s | 0.64 | 4,500 |
| same, start fidelity $0.53$ | $300\,\mu$s | $33\,\mu$s | 0.64 | 17,000 |

With the fully heralded rate of $0.022\,\mathrm{s^{-1}}$ reported by Stolk et al. (2024) for the 10 km
link, 2,300 to 4,500 events take 1 to 2.5 days per Bell state and configuration; at the measured
fidelity of $0.53$ it would be about nine days. Contrast, rate and $T_\text{coh}$ are the entries that
matter; please give yours.

## Decision rule (fixed before unblinding)

Commit to the predicted values of $\delta$ and $Q$ from the independently measured $t$, $\tau$ and $T_\text{coh}$.

| Observation | Reading |
|---|---|
| $\hat\kappa\approx1$, $Q<1$ with $Q=1$ at short range, no shift with added delay | supports the lag term on this platform |
| $\hat\kappa\approx0$, $Q=1$ | bounds the lag term; refutes it as formulated for this platform |
| shift for $\psi^-$ only, $Q=1$ | correlated noise (standard physics) |
| shift with the added short-range delay | the effect follows the heralding path, not distance |
| $Q\neq1$ at short range | non-Gaussian noise or a defect in the sequence; the sum rule does not apply |

Report $\hat\kappa\pm\sigma$ in all cases. A null result is a bound, not a refutation of every field ontology.

## What we ask of you

1. Your $T_2^*$, echo and dynamical-decoupling coherence times, in the sequences you would use.
2. The delivered state contrast, the heralded rate, and how far $T_\text{coh}$ can be tuned down.
3. Whether a same-hardware short-range configuration is available (Stolk et al. describe one).
4. Your view on which systematics we have missed: phase drift of the heralding link, differing magnetic
   environments at the sites, timing offsets.

## Caveats

The lag term is not derived from a dynamical law; the source model has free parameters. The
transfer to circuit qubits is not argued. Sample sizes are idealised. The sum rule fails for
non-Gaussian noise; the short-range control diagnoses that. The companion paper is *Causal Boundaries
of Bell's Inequality* (in preparation); the source model is Laidlaw, *Academia Quantum* 3(2), 2026.