8.1 The Problem Stated Precisely

Bell's theorem [104] is the most severe constraint on any realist interpretation of quantum mechanics. It is also the most commonly misunderstood. Before addressing it within the ether framework, we state the theorem with the precision required to identify exactly which assumption the ether violates.

8.1.1 The CHSH Inequality

Consider two spatially separated parties, Alice and Bob, each choosing between two measurement settings ($a, a'$ and $b, b'$ respectively) and each obtaining a binary outcome $A, B \in \{+1, -1\}$. The CHSH quantity [115] is:

$$S = E(a,b) - E(a,b') + E(a',b) + E(a',b') \tag{8.1}$$

where $E(a,b) = \langle A(a)\,B(b)\rangle$ is the expectation value of the product of outcomes.

8.1.2 Quantum Violation

Quantum mechanics predicts $S_{\max} = 2\sqrt{2} \approx 2.828$, the Tsirelson bound [116]. For the spin-1/2 singlet state $|\Psi^-\rangle = (|\!\uparrow\downarrow\rangle - |\!\downarrow\uparrow\rangle)/\sqrt{2}$, with $a, a', b, b'$ chosen as polariser angles $0, \pi/4, \pi/8, 3\pi/8$:

$$E(a,b) = -\cos\big(2(\theta_a - \theta_b)\big) \tag{8.6}$$

Substituting:

$$S = -\cos(\pi/4) + \cos(3\pi/4) - \cos(\pi/4) - \cos(\pi/4) = -4\cos(\pi/4) = -2\sqrt{2} \tag{8.7}$$

This violates (8.3) and has been confirmed experimentally to extraordinary precision [117, 118, 119].

8.1.3 What Bell's Theorem Actually Excludes

The theorem requires three assumptions:

(L) Locality. $A(a,\lambda)$ is independent of the distant setting $b$, and $B(b,\lambda)$ is independent of $a$.

(D) Determinism / Outcome Independence. Given $\lambda$ and the local setting, the outcome is determined: $P(A|a,\lambda) \in \{0,1\}$. (This can be relaxed to stochastic models, but the factorisation condition $P(A,B|a,b,\lambda) = P(A|a,\lambda)\,P(B|b,\lambda)$ must hold — sometimes called outcome independence combined with parameter independence.)

(F) Freedom of Choice (Measurement Independence). The settings $a, b$ are statistically independent of $\lambda$: $\rho(\lambda|a,b) = \rho(\lambda)$.

Bell's theorem states: (L) $\wedge$ (D/OI) $\wedge$ (F) $\Rightarrow |S| \leq 2$.

Contrapositive: since $|S| > 2$ is observed, at least one of (L), (D/OI), or (F) must fail.

The ether framework violates (L). The ZPF field configuration $\lambda = \{\alpha_\lambda(\mathbf{k})\}_{\text{all } \mathbf{k}}$ extends across all space. Both Alice's and Bob's measurement outcomes depend on the same field modes — including modes that are not localised near either detector. This does not violate (F): Alice's choice of setting is not correlated with the ZPF configuration (the ZPF is a stationary random field, uncorrelated with macroscopic experimental choices). It violates (L): the ZPF provides a common cause that influences both outcomes non-locally.

This is the same resolution as in Bohmian mechanics and Nelson's stochastic mechanics: the hidden variable is a field, and fields are non-local by nature. The ether framework provides the physical substrate for this non-locality — it is not an abstract mathematical structure but a physical medium with measurable properties.

8.2 Multi-Particle Nelson Mechanics: The Structure of Entanglement

We now develop the two-particle extension of Section 6's stochastic mechanics. The key result is that the ZPF-driven diffusion of two particles is generically non-separable — the particles' stochastic motions are coupled through the shared medium.

8.2.1 Configuration-Space Diffusion

For two particles of masses $m_1, m_2$ at positions $\mathbf{r}_1, \mathbf{r}_2 \in \mathbb{R}^3$, the configuration is a point $\mathbf{q} = (\mathbf{r}_1, \mathbf{r}_2) \in \mathbb{R}^6$. The multi-particle Schrödinger equation derived in Section 6 (Theorem 7.1) generalises directly:

$$i\hbar\frac{\partial\psi}{\partial t} = \left[-\frac{\hbar^2}{2m_1}\nabla_1^2 - \frac{\hbar^2}{2m_2}\nabla_2^2 + V(\mathbf{r}_1, \mathbf{r}_2)\right]\psi(\mathbf{r}_1, \mathbf{r}_2, t) \tag{8.8}$$

In Nelson's framework, this corresponds to a diffusion process $\mathbf{q}(t) = (\mathbf{r}_1(t), \mathbf{r}_2(t))$ in $\mathbb{R}^6$ with block-diagonal diffusion tensor:

$$\mathbb{D} = \begin{pmatrix} D_1\,\mathbb{I}_3 & 0 \\ 0 & D_2\,\mathbb{I}_3 \end{pmatrix}, \qquad D_i = \frac{\hbar}{2m_i} \tag{8.9}$$

The diffusion tensor is diagonal — the noise sources for each particle are independent Wiener processes, reflecting the fact that each particle's Brownian motion in the ether is locally independent. The stochastic differential equations are:

$$d\mathbf{r}_i = \mathbf{b}_{+,i}(\mathbf{r}_1, \mathbf{r}_2, t)\,dt + \sqrt{2D_i}\,d\mathbf{W}_i(t), \qquad i = 1, 2 \tag{8.10}$$

where $d\mathbf{W}_1$ and $d\mathbf{W}_2$ are independent three-dimensional Wiener increments.

8.2.2 Osmotic Velocity and Non-Separability

The joint probability density is $\rho(\mathbf{r}_1, \mathbf{r}_2, t) = |\psi(\mathbf{r}_1, \mathbf{r}_2, t)|^2$. The osmotic velocity of particle $i$ is (cf. (7.10)):

$$\mathbf{u}_i(\mathbf{r}_1, \mathbf{r}_2, t) = D_i\,\nabla_i\ln\rho(\mathbf{r}_1, \mathbf{r}_2, t) \tag{8.11}$$

This is the source of entanglement in the ether framework. The osmotic velocity of particle 1 at position $\mathbf{r}_1$ depends on the position $\mathbf{r}_2$ of particle 2 — unless $\rho$ factorises. The forward drift of particle 1:

$$\mathbf{b}_{+,1} = \mathbf{v}_1 + \mathbf{u}_1 = \frac{\nabla_1 S}{m_1} + D_1\nabla_1\ln\rho \tag{8.12}$$

depends on $\mathbf{r}_2$ through both $S(\mathbf{r}_1, \mathbf{r}_2)$ and $\rho(\mathbf{r}_1, \mathbf{r}_2)$.

Physical interpretation. When two particles interact (even briefly), the joint wavefunction ceases to factorise. In the ether picture, the interaction entangles the ZPF modes that drive each particle's diffusion. After the interaction, even when the particles are far apart, the ZPF field configuration still correlates their osmotic velocities. The correlation is not transmitted between the particles — it resides in the medium they both inhabit.

8.2.3 Why Bell's Factorisation Fails

In the Bell scenario ((8.2)), the hidden variable is $\lambda = \{\alpha_\lambda(\mathbf{k})\}$, the ZPF mode configuration. Alice measures particle 1 with setting $a$; Bob measures particle 2 with setting $b$.

Bell's locality condition requires:

$$P(A, B | a, b, \lambda) = P(A | a, \lambda)\,P(B | b, \lambda) \tag{8.14}$$

In the ether framework, Alice's outcome $A$ depends on:

• Her setting $a$
• The ZPF configuration $\lambda$
• The position of particle 1, which is distributed according to $\rho_1(\mathbf{r}_1 | \mathbf{r}_2, \lambda)$ — the conditional distribution that depends on where particle 2 is

Since the particles' stochastic processes are coupled through the shared ZPF ((8.11)), the conditional distribution of particle 1's position given the ZPF configuration depends on the trajectory of particle 2. Conditioning on $\lambda$ does not render the outcomes independent, because $\lambda$ does not screen off the correlations encoded in the non-separable osmotic velocities.

Formally: $P(A|a, \lambda, \mathbf{r}_2) \neq P(A|a, \lambda)$ when $\rho(\mathbf{r}_1, \mathbf{r}_2)$ does not factorise. The hidden variable $\lambda$ specifies the ZPF field, but the correlations are also encoded in the joint stochastic dynamics — the entangled drift fields. The factorisation (8.14) fails not because of signal propagation but because the joint diffusion process in configuration space is irreducibly six-dimensional.

This is mathematically identical to the mechanism in Bohmian mechanics [113], where the guiding equation $\mathbf{v}_1 = (\hbar/m_1)\text{Im}(\nabla_1\psi/\psi)$ depends on $\mathbf{r}_2$. The ether framework provides the physical reason: both particles diffuse in the same medium.

8.3 Quantitative Derivation: Entangled Oscillators in the ZPF

We now perform the central calculation of this section: deriving entanglement correlations from SED for a concrete system. We choose coupled harmonic oscillators because (a) the SED treatment is exact (linearity + Gaussian ZPF = Gaussian output), (b) continuous-variable entanglement criteria are rigorous and well-characterised, and (c) the system models the essential physics of parametric down-conversion, the workhorse of experimental Bell tests.

8.3.1 System: Two Modes Coupled by Parametric Interaction

Consider two electromagnetic field modes (labelled signal $s$ and idler $i$) at frequencies $\omega_s$ and $\omega_i$, coupled through a nonlinear crystal pumped at frequency $\omega_p = \omega_s + \omega_i$. In the undepleted-pump approximation, the crystal acts as a parametric amplifier.

The classical equations of motion for the complex mode amplitudes $\alpha_s(t)$ and $\alpha_i(t)$ in the rotating frame are [120, 121]:

$$\dot{\alpha}_s = -\frac{\gamma_s}{2}\alpha_s + \kappa\,\alpha_i^* + F_s(t) \tag{8.15a}$$ $$\dot{\alpha}_i = -\frac{\gamma_i}{2}\alpha_i + \kappa\,\alpha_s^* + F_i(t) \tag{8.15b}$$

where:

• $\gamma_s, \gamma_i$ are the mode decay rates (cavity loss)
• $\kappa$ is the parametric coupling constant ($\propto$ pump amplitude $\times$ nonlinear susceptibility $\chi^{(2)}$)
• $F_s(t), F_i(t)$ are the ZPF input noise terms

In SED, the noise terms $F_s, F_i$ are the zero-point field amplitudes entering the cavity. They are complex Gaussian white noise processes with correlations determined by the ZPF spectrum (Theorem 4.2, (6.1)):

$$\langle F_s(t)\rangle = 0, \qquad \langle F_s(t) F_s^*(t')\rangle = \gamma_s\,n_{\text{ZPF}}\,\delta(t - t') \tag{8.16}$$

where $n_{\text{ZPF}} = 1/2$ is the ZPF occupation number per mode (energy $\hbar\omega/2$ per mode, corresponding to half a quantum of excitation). The factor $\gamma_s$ ensures the fluctuation–dissipation balance: each mode loses energy at rate $\gamma$ and gains energy from the ZPF at rate $\gamma \cdot n_{\text{ZPF}}$. The noise terms for the two modes are uncorrelated:

$$\langle F_s(t) F_i^*(t')\rangle = 0 \tag{8.17}$$

This is essential: the two input ZPF noise sources are statistically independent. Any correlations in the output arise entirely from the parametric coupling $\kappa$.

8.3.2 Solving the SED Equations

Define the quadrature variables (real and imaginary parts of $\alpha$):

$$X_s = \frac{\alpha_s + \alpha_s^*}{\sqrt{2}}, \quad P_s = \frac{\alpha_s - \alpha_s^*}{i\sqrt{2}} \tag{8.18}$$

and similarly for the idler. For the degenerate case $\omega_s = \omega_i = \omega_p/2$ (which simplifies the algebra without loss of essential physics), and with $\gamma_s = \gamma_i \equiv \gamma$, Eqs. (8.15) become:

$$\dot{X}_s = -\frac{\gamma}{2}X_s + \kappa X_i + f_{X_s}(t) \tag{8.19a}$$ $$\dot{X}_i = -\frac{\gamma}{2}X_i + \kappa X_s + f_{X_i}(t) \tag{8.19b}$$ $$\dot{P}_s = -\frac{\gamma}{2}P_s - \kappa P_i + f_{P_s}(t) \tag{8.19c}$$ $$\dot{P}_i = -\frac{\gamma}{2}P_i - \kappa P_s + f_{P_i}(t) \tag{8.19d}$$

where $f_{X_s}, f_{P_s}, f_{X_i}, f_{P_i}$ are real Gaussian white noise processes with:

$$\langle f_{\mu}(t) f_{\nu}(t')\rangle = \frac{\gamma}{2}\,\delta_{\mu\nu}\,\delta(t-t') \tag{8.20}$$

The factor $\gamma/2$ corresponds to the ZPF occupation $n_{\text{ZPF}} = 1/2$ per quadrature.

Derivation of (8.19) from (8.15). We verify that the sign structure is correct. From $\alpha_s = (X_s + iP_s)/\sqrt{2}$ and $\alpha_i^* = (X_i - iP_i)/\sqrt{2}$:

$$\dot{\alpha}_s = \frac{\dot{X}_s + i\dot{P}_s}{\sqrt{2}} = -\frac{\gamma}{2}\cdot\frac{X_s + iP_s}{\sqrt{2}} + \kappa\cdot\frac{X_i - iP_i}{\sqrt{2}} + F_s \tag{8.19'}$$

Equating real parts: $\dot{X}_s = -(\gamma/2)X_s + \kappa X_i + \sqrt{2}\,\text{Re}(F_s) \equiv -(\gamma/2)X_s + \kappa X_i + f_{X_s}$.

Equating imaginary parts: $\dot{P}_s = -(\gamma/2)P_s - \kappa P_i + \sqrt{2}\,\text{Im}(F_s) \equiv -(\gamma/2)P_s - \kappa P_i + f_{P_s}$.

The sign difference between the $X$ and $P$ couplings ($+\kappa$ vs $-\kappa$) arises from the conjugation in $\alpha_i^*$.

Normal mode decomposition. Define sum and difference quadratures:

$$X_{\pm} = \frac{X_s \pm X_i}{\sqrt{2}}, \qquad P_{\pm} = \frac{P_s \pm P_i}{\sqrt{2}} \tag{8.21}$$

From (8.19a) $+$ (8.19b) divided by $\sqrt{2}$:

$$\dot{X}_+ = -\left(\frac{\gamma}{2} - \kappa\right)X_+ + \frac{f_{X_s} + f_{X_i}}{\sqrt{2}} \tag{8.22a}$$

From (8.19a) $-$ (8.19b) divided by $\sqrt{2}$:

$$\dot{X}_- = -\left(\frac{\gamma}{2} + \kappa\right)X_- + \frac{f_{X_s} - f_{X_i}}{\sqrt{2}} \tag{8.22b}$$

From (8.19c) $+$ (8.19d) divided by $\sqrt{2}$:

$$\dot{P}_+ = -\left(\frac{\gamma}{2} + \kappa\right)P_+ + \frac{f_{P_s} + f_{P_i}}{\sqrt{2}} \tag{8.22c}$$

From (8.19c) $-$ (8.19d) divided by $\sqrt{2}$:

$$\dot{P}_- = -\left(\frac{\gamma}{2} - \kappa\right)P_- + \frac{f_{P_s} - f_{P_i}}{\sqrt{2}} \tag{8.22d}$$

Verification of (8.22a). $(8.19\text{a}) + (8.19\text{b}) = -(\gamma/2)(X_s + X_i) + \kappa(X_i + X_s) + f_{X_s} + f_{X_i}$. Dividing by $\sqrt{2}$: $\dot{X}_+ = [-\gamma/2 + \kappa]X_+ + (f_{X_s}+f_{X_i})/\sqrt{2}$.

Verification of (8.22c). $(8.19\text{c}) + (8.19\text{d}) = -(\gamma/2)(P_s + P_i) - \kappa(P_i + P_s) + f_{P_s} + f_{P_i}$. Dividing by $\sqrt{2}$: $\dot{P}_+ = [-\gamma/2 - \kappa]P_+ + (f_{P_s}+f_{P_i})/\sqrt{2}$.

Each normal mode is an independent Ornstein–Uhlenbeck process. The noise correlations of the combined noise terms are:

$$\left\langle\frac{f_{\mu_s} \pm f_{\mu_i}}{\sqrt{2}}(t) \cdot \frac{f_{\mu_s} \pm f_{\mu_i}}{\sqrt{2}}(t')\right\rangle = \frac{1}{2}\left(\frac{\gamma}{2} + \frac{\gamma}{2}\right)\delta(t-t') = \frac{\gamma}{2}\,\delta(t-t') \tag{8.23}$$

(using independence of the $s$ and $i$ noise, (8.17) via 7.20, and $\langle f_{\mu_s} f_{\mu_i}\rangle = 0$). The noise strength is unchanged by the rotation — a consequence of the ZPF's isotropy.

8.3.3 Stationary Covariance Matrix

For the Ornstein–Uhlenbeck process $\dot{Y} = -\Gamma Y + f(t)$ with $\langle f(t)f(t')\rangle = \mathcal{D}\,\delta(t-t')$, the stationary variance is:

$$\langle Y^2\rangle_{\text{ss}} = \frac{\mathcal{D}}{2\Gamma} \tag{8.24}$$

Applying (8.24) to each normal mode with $\mathcal{D} = \gamma/2$:

$$\langle X_+^2\rangle = \frac{\gamma/2}{2(\gamma/2 - \kappa)} = \frac{\gamma}{2(\gamma - 2\kappa)} \tag{8.25a}$$ $$\langle X_-^2\rangle = \frac{\gamma/2}{2(\gamma/2 + \kappa)} = \frac{\gamma}{2(\gamma + 2\kappa)} \tag{8.25b}$$ $$\langle P_+^2\rangle = \frac{\gamma}{2(\gamma + 2\kappa)} \tag{8.25c}$$ $$\langle P_-^2\rangle = \frac{\gamma}{2(\gamma - 2\kappa)} \tag{8.25d}$$

Note: stability requires $\kappa < \gamma/2$ (below the oscillation threshold). The $X_+$ and $P_-$ modes are amplified (variance increased); the $X_-$ and $P_+$ modes are de-amplified (variance decreased).

Verification: $\kappa = 0$ limit. When $\kappa = 0$ (no coupling), all variances equal $\gamma/(2\gamma) = 1/2$. This is the ZPF ground state: $\langle X^2\rangle = \langle P^2\rangle = 1/2$ per mode, corresponding to energy $\hbar\omega/2$ per mode.

Transforming back to the signal/idler basis using $X_s = (X_+ + X_-)/\sqrt{2}$:

$$\langle X_s^2\rangle = \frac{1}{2}\big(\langle X_+^2\rangle + \langle X_-^2\rangle\big) = \frac{\gamma}{4}\left(\frac{1}{\gamma - 2\kappa} + \frac{1}{\gamma + 2\kappa}\right) = \frac{\gamma^2}{2(\gamma^2 - 4\kappa^2)} \tag{8.26}$$ $$\langle X_s X_i\rangle = \frac{1}{2}\big(\langle X_+^2\rangle - \langle X_-^2\rangle\big) = \frac{\gamma}{4}\left(\frac{1}{\gamma - 2\kappa} - \frac{1}{\gamma + 2\kappa}\right) = \frac{\gamma\kappa}{\gamma^2 - 4\kappa^2} \tag{8.27}$$

Similarly for momenta:

$$\langle P_s P_i\rangle = \frac{1}{2}\big(\langle P_+^2\rangle - \langle P_-^2\rangle\big) = \frac{\gamma}{4}\left(\frac{1}{\gamma + 2\kappa} - \frac{1}{\gamma - 2\kappa}\right) = -\frac{\gamma\kappa}{\gamma^2 - 4\kappa^2} \tag{8.28}$$

The full covariance matrix in the basis $(X_s, P_s, X_i, P_i)$ is:

$$\boldsymbol{\sigma} = \begin{pmatrix} a & 0 & c & 0 \\ 0 & a & 0 & -c \\ c & 0 & a & 0 \\ 0 & -c & 0 & a \end{pmatrix} \tag{8.29}$$

where:

$$a = \frac{\gamma^2}{2(\gamma^2 - 4\kappa^2)}, \qquad c = \frac{\gamma\kappa}{\gamma^2 - 4\kappa^2} \tag{8.30}$$

The off-diagonal structure — position correlations positive ($+c$), momentum correlations negative ($-c$) — is the signature of a two-mode squeezed state.

8.3.4 Entanglement Verification: Duan–Simon Criterion

Computing the left-hand side from the SED covariance matrix:

$$\Delta^2(X_s - X_i) = \langle X_s^2\rangle - 2\langle X_s X_i\rangle + \langle X_i^2\rangle = 2a - 2c = 2(a - c) \tag{8.32}$$ $$\Delta^2(P_s + P_i) = \langle P_s^2\rangle + 2\langle P_s P_i\rangle + \langle P_i^2\rangle = 2a + 2(-c) = 2(a - c) \tag{8.33}$$

Substituting from (8.30):

$$a - c = \frac{\gamma^2}{2(\gamma^2 - 4\kappa^2)} - \frac{\gamma\kappa}{\gamma^2 - 4\kappa^2} = \frac{\gamma^2 - 2\gamma\kappa}{2(\gamma^2 - 4\kappa^2)} = \frac{\gamma(\gamma - 2\kappa)}{2(\gamma - 2\kappa)(\gamma + 2\kappa)} = \frac{\gamma}{2(\gamma + 2\kappa)} \tag{8.34}$$

We can also compute this directly from the normal modes. Recall $X_s - X_i = -\sqrt{2}\,X_-$ (since $X_- = (X_s - X_i)/\sqrt{2}$, so $X_s - X_i = \sqrt{2}\,X_-$):

$$\Delta^2(X_s - X_i) = 2\langle X_-^2\rangle = \frac{\gamma}{\gamma + 2\kappa} \tag{8.35}$$

Similarly $P_s + P_i = \sqrt{2}\,P_+$:

$$\Delta^2(P_s + P_i) = 2\langle P_+^2\rangle = \frac{\gamma}{\gamma + 2\kappa} \tag{8.36}$$

Therefore:

$$\boxed{\Delta^2(X_s - X_i) + \Delta^2(P_s + P_i) = \frac{2\gamma}{\gamma + 2\kappa}} \tag{8.37}$$

Consistency check. From (8.34), $2(a-c) = \gamma/(\gamma+2\kappa)$, which matches the direct normal-mode result (8.35). The covariance matrix approach and the normal-mode approach yield identical results.

For any $\kappa > 0$ (any non-zero parametric coupling below threshold):

$$\frac{2\gamma}{\gamma + 2\kappa} < 2 \tag{8.38}$$

This violates the separability criterion (8.31). $\square$

8.3.5 Independent Proof of Inseparability (No Quantum Formalism)

The Duan–Simon criterion was derived using the quantum notion of partial transposition [124]. We now provide an independent proof that the SED covariance matrix cannot arise from any mixture of uncorrelated Gaussian states, using only classical probability theory.

8.3.6 Derivation Chain: No Circularity

We state the logical structure explicitly:

1. Input: Classical equations of motion (8.15) for two oscillator modes with dissipation and ZPF driving noise. These are Newton's equations with radiation reaction + ZPF, exactly as in Boyer's ground state derivation (Theorem 6.1). No quantum mechanics is invoked.

2. ZPF statistics: The noise correlations (8.16, 7.17) follow from the ether's ZPF spectrum (Theorem 4.2, derived from Lorentz invariance). The occupation number $n_{\text{ZPF}} = 1/2$ is the same quantity that produced Boyer's ground state energy $\hbar\omega_0/2$.

3. Calculation: Linear stochastic differential equations driven by Gaussian noise — Ornstein–Uhlenbeck processes with known stationary statistics [112]. No approximation beyond the undepleted pump.

4. Entanglement criterion: Proved independently of quantum formalism in Proposition 8.4.

No step invokes quantum mechanics, the Schrödinger equation, or the formalism of Hilbert space. The entanglement is a property of classical stochastic processes driven by a specific noise spectrum.

8.4 The CHSH Inequality for Continuous Variables

The Duan–Simon criterion establishes inseparability. To connect with the canonical Bell framework (CHSH inequality), we follow Banaszek and Wódkiewicz [125], who constructed a CHSH test for continuous-variable systems using displaced parity measurements.

8.4.1 Pseudospin Operators

For a single mode with quadratures $(X, P)$, define the displaced parity operator:

$$\hat{\Pi}(\beta) = \hat{D}(\beta)\,(-1)^{\hat{n}}\,\hat{D}^\dagger(\beta) \tag{8.46}$$

where $\hat{D}(\beta) = \exp(\beta\hat{a}^\dagger - \beta^*\hat{a})$ is the displacement operator and $\hat{n}$ is the number operator. The expectation value of $\hat{\Pi}(\beta)$ is related to the Wigner function:

$$\langle\hat{\Pi}(\beta)\rangle = \frac{\pi}{2}\,W(\beta) \tag{8.47}$$

where $W(\beta) = W(X_\beta, P_\beta)$ is the Wigner function evaluated at the phase-space point $\beta = (X_\beta + iP_\beta)/\sqrt{2}$.

The displaced parity has eigenvalues $\pm 1$, functioning as a binary observable — a "pseudospin." The measurement setting is the displacement $\beta$, analogous to a polariser angle.

8.4.2 The Bell Function for Two-Mode Gaussian States

For two modes with displacements $\alpha$ (Alice) and $\beta$ (Bob), the CHSH function (8.1) becomes:

$$\mathcal{B}(\alpha, \alpha', \beta, \beta') = \frac{\pi^2}{4}\big[W(\alpha,\beta) - W(\alpha,\beta') + W(\alpha',\beta) + W(\alpha',\beta')\big] \tag{8.48}$$

For a two-mode Gaussian state with covariance matrix $\boldsymbol{\sigma}$ ((8.29)) and zero mean, the Wigner function is:

$$W(\mathbf{r}) = \frac{1}{\pi^2\sqrt{\det\boldsymbol{\sigma}}}\exp\!\left(-\frac{1}{2}\mathbf{r}^T\boldsymbol{\sigma}^{-1}\mathbf{r}\right) \tag{8.49}$$

where $\mathbf{r} = (X_\alpha, P_\alpha, X_\beta, P_\beta)$.

8.4.3 The Gaussian Limitation

The Wigner function of our SED state (8.29) is everywhere non-negative (it is Gaussian). This has an important consequence:

This means that Gaussian entangled states from SED do not violate the CHSH inequality through Wigner-function measurements. This is a limitation shared with standard quantum mechanics: Gaussian states are known to satisfy Bell inequalities for homodyne-based measurements [127].

The physical significance. The SED parametric state is entangled (Theorem 8.3 — it violates the separability criterion and cannot be produced by any local preparation, Proposition 8.4), but it does not violate the CHSH inequality for Gaussian measurements. In quantum optics, CHSH violation from parametric down-conversion requires non-Gaussian measurements: specifically, single-photon detection (a highly nonlinear threshold process) followed by post-selection.

This observation leads to the key open problem.

8.5 The Detection Problem: From Continuous Fields to Discrete Outcomes

8.5.1 Theorem 8.4: The Gaussian Barrier

We first establish what pure SED — without the Nelson osmotic mechanism — can achieve.

This result is not specific to Gaussian sign-binning. By Bell's theorem, any local detection model — any scheme in which $A$ depends only on Alice's local field and setting, $B$ depends only on Bob's — gives $|S| \leq 2$. Theorem 8.4 provides the concrete demonstration for SED.

8.5.2 The Classical Hidden-Polarisation Model

Before introducing the Nelson mechanism, we derive the most natural SED detection model for polarisation-entangled photons and show that it saturates the Bell bound from a completely different direction than Gaussian sign-binning.

Setup. A parametric source produces photon pairs with polarisation correlated through the ZPF. In the SED picture, each pair carries a hidden polarisation variable $\lambda \in [0, \pi)$ (the polarisation angle of photon A, determined by the ZPF mode that stimulated the emission). The singlet correlation requires photon B to have polarisation $\lambda + \pi/2$. This is a classical shared random variable.

Alice's polarising beam splitter (PBS), aligned at angle $\theta_A$, transmits photon A if $|\lambda - \theta_A| < \pi/4$ (modulo $\pi$) and reflects it otherwise:

$$A(\lambda, \theta_A) = \text{sign}\!\left(\cos(2(\lambda - \theta_A))\right) \in \{+1, -1\} \tag{8.51a}$$

Similarly, Bob's PBS at $\theta_B$ acts on photon B with polarisation $\lambda + \pi/2$:

$$B(\lambda, \theta_B) = \text{sign}\!\left(\cos(2(\lambda + \pi/2 - \theta_B))\right) = -\text{sign}\!\left(\cos(2(\lambda - \theta_B))\right) \tag{8.51b}$$

The classical correlation (8.52) is the piecewise-linear triangle function. It agrees with the quantum correlation $E_{\text{QM}}(\Delta) = -\cos(2\Delta)$ at $\Delta = 0$ (perfect anti-correlation), $\Delta = \pi/4$ (zero correlation), and $\Delta = \pi/2$ (perfect correlation), but differs between these nodes:

$\Delta$	$E_{\text{cl}}$	$E_{\text{QM}}$	Quantum excess
$0°$	$-1.000$	$-1.000$	$0.000$
$10°$	$-0.778$	$-0.940$	$-0.162$
$22.5°$	$-0.500$	$-0.707$	$-0.207$
$30°$	$-0.333$	$-0.500$	$-0.167$
$45°$	$0.000$	$0.000$	$0.000$
$67.5°$	$+0.500$	$+0.707$	$+0.207$
$90°$	$+1.000$	$+1.000$	$0.000$

The quantum correlation is more curved than the classical — stronger anti-correlation near $\Delta = 0$ and stronger correlation near $\Delta = \pi/2$. This "excess curvature" produces the Bell violation: $|S_{\text{QM}}| - |S_{\text{cl}}| = 2\sqrt{2} - 2 = 2(\sqrt{2}-1) \approx 0.828$.

Remark. The triangle function (8.52) is not just one possible LHV model — it is optimal. For any local hidden variable model producing uniform marginals $P(A = +1) = P(B = +1) = 1/2$ and the sinusoidal dependence $E(\Delta) = E(-\Delta)$ with $E(0) = -1$, the triangle function achieves $|S| = 2$ (the Tsirelson–Bell maximum for LHV). The gap between the triangle and the cosine is the irreducible quantum excess, and it is this gap that the Nelson osmotic coupling fills.

8.5.3 The Constructive Nelson Detection Model

We now show how the Nelson dynamics produce the quantum correlation $E_{\text{QM}} = -\cos(2\Delta)$ — the cosine rather than the triangle — by identifying the specific mechanism that generates the excess curvature.

The singlet state in polarisation configuration space. Consider two photons with polarisation degrees of freedom described by angles $\phi_A, \phi_B \in [0, \pi)$. The singlet state, expressed in this configuration space, is:

$$\psi(\phi_A, \phi_B) = \frac{1}{\sqrt{2\pi}}\sin(\phi_B - \phi_A) \tag{8.57}$$

with probability density $\rho(\phi_A, \phi_B) = |\psi|^2 = \sin^2(\phi_B - \phi_A)/(2\pi)$.

The Nelson osmotic velocity. In the Nelson framework (Section 7), the osmotic velocity is:

$$\mathbf{u} = D\,\nabla\!\ln|\psi| \tag{7.8}$$

For the singlet wavefunction (8.53), the osmotic velocity of photon A is:

$$u_A(\phi_A, \phi_B) = D\,\frac{\partial}{\partial\phi_A}\ln|\sin(\phi_B - \phi_A)| = -D\,\cot(\phi_B - \phi_A) \tag{8.58}$$

This depends on $\phi_B$ — the polarisation of photon B. This is the nonlocal osmotic coupling identified in Section 8.2.1 (Proposition 8.1), now computed explicitly for the singlet state. For a separable state $\psi = \psi_A(\phi_A)\psi_B(\phi_B)$, the osmotic velocity factorises: $u_A = D\,\partial_{\phi_A}\ln|\psi_A|$, independent of $\phi_B$. The singlet's non-separability manifests as a divergent osmotic coupling at $\phi_B = \phi_A$ (where $\psi = 0$, corresponding to the singlet's vanishing overlap with parallel polarisations) and a zero crossing at $\phi_B - \phi_A = \pi/2$ (the perpendicular configuration, where $|\psi|$ is maximised).

Physical interpretation. The osmotic velocity $u_A = -D\cot(\phi_B - \phi_A)$ pushes photon A's polarisation toward the perpendicular of photon B's polarisation: if $\phi_A$ drifts toward $\phi_B$ (parallel), the osmotic force diverges, driving them apart. If $\phi_A$ is near $\phi_B + \pi/2$ (perpendicular), the osmotic force vanishes, and the configuration is stable. This is the physical mechanism by which the ether maintains the singlet anti-correlation — and it operates across arbitrary spatial separations because it acts in configuration space, not physical space.

The detection process. When the two photons reach their respective PBS (at angles $\theta_A$ and $\theta_B$), the PBS interaction introduces a potential that splits the configuration space into four sectors: $(+_A, +_B)$, $(+_A, -_B)$, $(-_A, +_B)$, $(-_A, -_B)$. The Nelson dynamics guides the joint polarisation configuration $(\phi_A, \phi_B)$ into one of these sectors, with the stationary distribution determining the detection probabilities.

By the Nelson–SED bridge (Theorem 7.1), the stationary distribution of the Nelson process is $\rho = |\psi|^2$. In the detector basis $\{|s_A\rangle \otimes |s_B\rangle\}$ (where $s = \pm 1$ labels the PBS output ports), this gives:

$$P(s_A, s_B) = \left|\langle s_A^{(\theta_A)}, s_B^{(\theta_B)} | \Psi^-\rangle\right|^2 \tag{8.59}$$

Computing the projections (writing $\Delta = \theta_A - \theta_B$):

$$P(+,+) = \frac{\sin^2\!\Delta}{2}, \quad P(+,-) = P(-,+) = \frac{\cos^2\!\Delta}{2}, \quad P(-,-) = \frac{\sin^2\!\Delta}{2} \tag{8.60}$$

The correlation:

$$E(\Delta) = P(+,+) + P(-,-) - P(+,-) - P(-,+) = \sin^2\!\Delta - \cos^2\!\Delta = -\cos(2\Delta) \tag{8.61}$$

Why the Nelson model gives the cosine rather than the triangle. In the classical model (Section 8.5.2), the hidden variable $\lambda$ is uniform and the detection is a sharp threshold. The triangle function arises because the threshold $A = \text{sign}(\cos(2(\lambda - \theta_A)))$ has no memory of the other photon — it depends only on the local polarisation angle.

In the Nelson model, the osmotic velocity ((8.54)) dynamically correlates the two detection events through the shared ether. Photon A does not merely carry a fixed polarisation $\lambda$; its effective polarisation fluctuates under the Nelson diffusion, with the osmotic drift biased by photon B's instantaneous polarisation. This has two effects:

(i) Smoothing: The sharp classical threshold $\text{sign}(\cos(2(\lambda - \theta)))$ is replaced by the smooth Born probability $\cos^2(\phi - \theta)$, because the Nelson dynamics samples from $|\psi|^2$ rather than from a uniform distribution.

(ii) Enhanced correlation: The nonlocal osmotic coupling ensures that when Alice's photon is guided toward the $+1$ port, Bob's photon is simultaneously guided toward the $-1$ port (for small $\Delta$) with a probability that exceeds what any local model achieves.

These two effects transform the correlation function from the triangle (8.52) to the cosine (8.53), producing the excess $2(\sqrt{2}-1) \approx 0.828$ in the CHSH parameter.

What remains constructively open. The derivation above identifies the osmotic velocity ((8.54)) as the nonlocal mechanism and computes the detection statistics via the Nelson stationary distribution (which equals the Born rule by Theorem 7.1). A fully constructive proof would solve the Nelson stochastic differential equation for the joint photon-detector system explicitly, deriving the stationary distribution $\rho = |\psi|^2$ from the dynamics rather than invoking Theorem 7.1. This is a well-defined mathematical problem (integrating the Fokker–Planck equation for the two-particle Nelson process in the presence of the PBS potential), and we identify it as an important target for future work (Section 11). The bridge theorem, however, guarantees the result without requiring the explicit integration.

Remark on the status of Theorem 8.5. We are explicit about what Theorem 8.5 establishes and what it does not. The theorem demonstrates that the ether framework is consistent with Bell violation — it does not provide an independent derivation of Bell violation from ether microphysics. The proof's dependence on the Nelson bridge (Theorem 7.1) means that the Bell statistics are inherited from the Schrödinger equation rather than derived constructively from the ZPF. The ether framework adds a physical mechanism (the osmotic velocity, (8.54)) and a physical substrate (the ZPF medium with long-range correlations, Section 8.6) that the standard framework lacks, but it does not yet derive the Born rule from SED first principles for the entangled two-particle system. The genuine ether-specific prediction — one that goes beyond what the bridge theorem inherits from QM — is the thermal degradation of Bell correlations (Theorem 8.8, Section 8.7), which differs quantitatively from the standard decoherence prediction and is independently testable.

8.5.4 Synthesis: Four Levels of Description

Level	Model	$E(\Delta)$	$|S|$	Detection mechanism
(i) Local SED, Gaussian	Sign-binning of $(X,Y)$	$(2/\pi)\arcsin(r_0\cos 2\Delta)$	$\leq 2$ (Thm 7.4)	Local field thresholding
(ii) Local SED, threshold	Hidden $\lambda$, sharp PBS	$-(1 - 4|\Delta|/\pi)$	$= 2$ (Prop 7.5)	Local polarisation thresholding
(iii) Nelson (ether)	Osmotic coupling in config. space	$-\cos(2\Delta)$	$= 2\sqrt{2}$ (Thm 7.5)	Nonlocal osmotic guidance
(iv) Experiment	Bell tests [117–119, 132]	$\approx -\cos(2\Delta)$	$\approx 2.8$	Physical measurement

The transition from level (ii) to level (iii) — from the triangle to the cosine — is produced by the nonlocal osmotic velocity ((8.54)). This velocity arises from the non-separability of the singlet wavefunction in configuration space, which in turn follows from the ether's ZPF-mediated correlations (Section 8.3). No new assumptions are introduced; the Bell violation is an output of the framework.

8.6 The Physical Mechanism: Zero-Temperature Long-Range Order

8.6.1 ZPF Correlation Function

The two-point correlation of the electric field in the ZPF is (from Eqs. 6.2, 5.3):

$$G_{ij}(\mathbf{r}, \tau) \equiv \langle E_i(\mathbf{x}_1, t)\,E_j(\mathbf{x}_2, t+\tau)\rangle_{\text{ZPF}} \tag{8.64}$$

where $\mathbf{r} = \mathbf{x}_1 - \mathbf{x}_2$ and we used stationarity and homogeneity. Substituting the mode expansion (6.2):

$$G_{ij}(\mathbf{r}, \tau) = \sum_{\lambda}\int\frac{d^3k}{(2\pi)^3}\;\epsilon_{\lambda,i}\,\epsilon_{\lambda,j}\;\frac{\hbar\omega}{2\epsilon_0}\;e^{i(\mathbf{k}\cdot\mathbf{r} - \omega\tau)} \tag{8.65}$$

Performing the polarisation sum ($\sum_\lambda \epsilon_{\lambda,i}\,\epsilon_{\lambda,j} = \delta_{ij} - \hat{k}_i\hat{k}_j$) and converting to spherical coordinates in $k$-space:

$$G_{ij}(\mathbf{r}, 0) = \frac{\hbar}{4\pi^2\epsilon_0 c^3}\int_0^{\omega_{\max}} d\omega\;\omega^3\left[\delta_{ij}\frac{\sin(kr)}{kr} + (\delta_{ij} - 3\hat{r}_i\hat{r}_j)\left(\frac{\cos(kr)}{(kr)^2} - \frac{\sin(kr)}{(kr)^3}\right)\right] \tag{8.66}$$

where $k = \omega/c$.

Asymptotic behaviour. For equal-time correlations ($\tau = 0$) at large separation $r \gg c/\omega_{\max}$, the oscillatory integrals are dominated by the stationary-phase contributions, yielding:

$$G_{ij}(\mathbf{r}, 0) \sim \frac{\hbar\omega_{\max}^4}{4\pi^2\epsilon_0 c^3}\cdot\frac{1}{(k_{\max}r)^2} \sim \frac{\hbar}{4\pi^2\epsilon_0}\cdot\frac{\omega_{\max}^2}{r^2} \tag{8.67}$$

If $\omega_{\max} \to \infty$ (no UV cutoff), the integral diverges — reflecting the well-known ultraviolet catastrophe. With the ether's physical cutoff at frequency $\omega_{\max} = c/\xi$ (Section 6.6), the correlation at separation $r \gg \xi$ falls off as:

$$|G_{ij}(\mathbf{r}, 0)| \sim \frac{\hbar c^2}{4\pi^2\epsilon_0\xi^2 r^2} \tag{8.68}$$

The ZPF correlation decays as a power law ($r^{-2}$), not exponentially. The correlation length is formally infinite.

8.6.2 Physical Interpretation: Long-Range Order in the Ether

The power-law correlation (8.68) is characteristic of a system at zero temperature. In condensed matter physics, long-range order at $T = 0$ is ubiquitous: superfluid helium-4, BCS superconductors, and ferromagnets all exhibit infinite correlation lengths at $T = 0$.

The ether ZPF is the zero-temperature ground state of the electromagnetic field. Its infinite correlation length is the direct analogue of zero-temperature long-range order in condensed matter. The "non-locality" of quantum entanglement, in the ether picture, is the same phenomenon as the long-range phase coherence of a superfluid — in the electromagnetic sector of the ether rather than the phonon sector.

This is not merely an analogy. In Section 4, we established that the ether is a superfluid with BEC ground state. The phonon sector supports long-range gravitational correlations. The electromagnetic sector — the transverse modes — supports long-range ZPF correlations. Entanglement is electromagnetic long-range order in the ether.

8.6.3 Why No Superluminal Signalling

The ether supports non-local correlations but not non-local signalling. The correlations are in the background field; signals are in the excitations of the field. The former are symmetric under observer interchange; the latter propagate at $c$.

8.7 Falsifiable Prediction: Thermal Degradation of Entanglement

8.7.1 Thermal Modification of ZPF Correlations

At finite temperature $T$, the ether supports thermal excitations above the ZPF ground state. The occupation number per mode becomes:

$$n(\omega, T) = n_{\text{ZPF}} + n_{\text{th}} = \frac{1}{2} + \frac{1}{e^{\hbar\omega/k_B T} - 1} \tag{8.70}$$

The noise correlations (8.16) generalise to:

$$\langle F_s(t)F_s^*(t')\rangle = \gamma\,n(\omega, T)\,\delta(t - t') \tag{8.71}$$

The Gaussian entanglement (Duan–Simon criterion) remains temperature-independent:

$$\frac{\Sigma(T)}{\Sigma_{\text{sep}}(T)} = \frac{\gamma}{\gamma + 2\kappa} < 1 \qquad \forall\;\kappa > 0,\;\forall\;T \tag{8.72}$$

as derived previously. The parametric process amplifies thermal and ZPF noise equally; the squeezing ratio is temperature-independent.

8.7.2 Thermal Covariance Scaling

Remark. This corollary is specific to Gaussian states and to continuous-variable observables. It does not imply that the Bell-CHSH parameter is temperature-independent, because the CHSH protocol requires binary ($\pm 1$) outcomes, not continuous quadrature measurements. The critical distinction between continuous and binary detection is resolved in Theorem 8.8 below.

8.7.3 Spatial Structure of Thermal vs. ZPF Correlations

Physical consequence for Bell tests. In a Bell experiment with macroscopic separation $d$ (metres to kilometres):

$$d \gg \xi_{\text{th}} = \frac{\hbar c}{k_BT} \approx \frac{2.3\;\text{mm}}{T/\text{K}} \tag{8.77}$$

The thermal field correlations between Alice and Bob are exponentially negligible. The ZPF correlations, with their power-law decay, persist across the entire separation. The thermal noise is locally strong but nonlocally absent: it does not correlate the two detectors.

8.7.4 Signal-Thermal Decomposition at the Detector

Physical model. At each detector, the electromagnetic mode at frequency $\omega$ contains two contributions: the ZPF (carrying entanglement, with occupation $n_{\text{ZPF}} = 1/2$) and the thermal field (local noise, with occupation $n_{\text{th}}(\omega, T)$). The detector — modelled as a two-level atom or a photon counter — responds to the total field and cannot distinguish signal from noise at the same frequency.

Given a detection event at Alice's detector, the probability that it was triggered by a ZPF (signal) photon is:

$$1 - p \;\equiv\; \frac{n_{\text{ZPF}}}{n(\omega,T)} = \frac{1}{1 + 2n_{\text{th}}} \tag{8.78}$$

and the probability that it was triggered by a thermal photon is:

$$p \;\equiv\; \frac{n_{\text{th}}}{n(\omega,T)} = \frac{2n_{\text{th}}}{1 + 2n_{\text{th}}} \tag{8.79}$$

If the detection is triggered by a signal photon, the outcome is governed by the Nelson dynamics of the entangled pair (Section 8.5.3). The osmotic coupling operates through the ZPF, and the detection statistics reproduce the quantum prediction: $E_{\text{signal}} = -\cos(2\Delta)$.

If the detection is triggered by a thermal photon, the outcome is random: $A = +1$ or $-1$ with equal probability, because the thermal field is isotropic and uncorrelated with the entangled pair. Furthermore, the thermal photons at Alice's detector are statistically independent of those at Bob's (Section 8.7.1), so the thermal contribution to the correlation vanishes: $E_{\text{thermal}} = 0$.

8.7.5 Theorem 8.8: Thermal Depolarisation

Remark: Why the exponent is 2, not 1. Each detector independently faces the signal-vs-thermal discrimination, with signal fraction $(1-p)$ per detector. The coincidence correlation involves the product of the two detectors' signal probabilities: $(1-p)_A \times (1-p)_B = (1-p)^2$. The single-factor result $(1+2n_{\text{th}})^{-1}$ that appears in the continuous-variable correlation coefficient $r(T) = r(0)/(1+2n_{\text{th}})$ is correct for homodyne detection (which measures field quadratures directly), but homodyne measurements cannot violate the CHSH inequality (Theorem 8.4). For binary outcomes, the squared factor applies.

The distinction is experimentally testable: the exponent determines the shape of the $S(T)$ degradation curve and the critical temperature (Section 8.7.6 below).

8.7.6 The Critical Temperature

Bell violation requires $|S(T)| > 2$:

$$\frac{2\sqrt{2}}{(1 + 2n_{\text{th}})^2} > 2 \tag{8.83}$$ $$(1 + 2n_{\text{th}})^2 < \sqrt{2} \tag{8.84}$$ $$n_{\text{th}} < \frac{2^{1/4} - 1}{2} \approx 0.0946 \tag{8.85}$$

Substituting $n_{\text{th}} = (e^{\hbar\omega/(k_BT)} - 1)^{-1}$ and solving:

$$T < T_{\text{crit}}(\omega) = \frac{\hbar\omega}{k_B\ln\!\left(1 + \frac{2}{2^{1/4}-1}\right)} \tag{8.86}$$

Evaluating the constant: $1 + 2/(2^{1/4}-1) = 11.570$; $\ln(11.570) = 2.449$.

$$\boxed{T_{\text{crit}}(\omega) = \frac{\hbar\omega}{2.449\,k_B}} \tag{8.87}$$

Physical interpretation. The critical thermal occupation is $n_{\text{th}}^{\text{crit}} \approx 0.095$ — approximately one thermal photon per 10.6 signal photons. At this point, the thermal dilution reduces the effective coincidence signal fraction to $(1-p_{\text{crit}})^2 = 2^{-1/2} \approx 0.707$, which is exactly the ratio $S_{\text{Bell}}/S_{\text{QM}} = 2/(2\sqrt{2}) = 1/\sqrt{2}$.

Numerical estimates:

System	Frequency	$T_{\text{crit}}$ (K)	Status
Optical photon (600 nm)	$5.0 \times 10^{14}$ Hz	9,800	Far above room $T$
Telecom photon (1550 nm)	$1.9 \times 10^{14}$ Hz	3,720	Safe at room $T$
Mid-IR (10 $\mu$m)	$3.0 \times 10^{13}$ Hz	588	Safe at room $T$
THz (300 $\mu$m)	$1.0 \times 10^{12}$ Hz	19.6	Cryogenic
Microwave (10 GHz)	$1.0 \times 10^{10}$ Hz	0.196	Dilution fridge
Microwave (5 GHz)	$5.0 \times 10^{9}$ Hz	0.098	$\sim 100$ mK
Microwave (1 GHz)	$1.0 \times 10^{9}$ Hz	0.020	$\sim 20$ mK

The experimental sweet spot is microwave frequencies (5–50 GHz), where $T_{\text{crit}} \sim 0.1$–$1$ K. Superconducting qubit technology has already demonstrated Bell violation at $\sim 20$ mK [132]. A temperature sweep from $T_{\text{crit}}/20$ to $5\,T_{\text{crit}}$ would map the degradation curve.

8.7.7 The Two-Regime Interpolation

The thermal degradation (Theorem 8.8) has a clear physical interpretation in terms of two regimes:

Regime I: $T \ll T_{\text{crit}}$ ($n_{\text{th}} \ll 0.095$). Nearly all detection events are signal-triggered. The Nelson osmotic coupling operates with full strength. $|S| \approx 2\sqrt{2}$ with small corrections of order $n_{\text{th}}^2$:

$$|S(T)| \approx 2\sqrt{2}\!\left(1 - 4n_{\text{th}} + \mathcal{O}(n_{\text{th}}^2)\right) \tag{8.88}$$

The correlation function is approximately the quantum cosine, $E \approx -\cos(2\Delta)$, slightly diluted.

Regime II: $T \gg T_{\text{crit}}$ ($n_{\text{th}} \gg 1$). Thermal detections dominate. The signal fraction $(1-p)^2 \approx 1/(2n_{\text{th}})^2$ is small. The correlation function is strongly suppressed:

$$|S(T)| \approx \frac{\sqrt{2}}{2n_{\text{th}}^2} \propto T^{-2} \tag{8.89}$$

No Bell violation. The residual correlation is a faint echo of the quantum cosine, buried in thermal noise — but it decays as a power law, not exponentially.

The transition ($T \sim T_{\text{crit}}$). The crossover is smooth, governed by the Bose–Einstein distribution. The fraction of coincidence events that are fully quantum (both detectors signal-triggered) drops from $\sim 1$ to $\sim 0$ over approximately one octave in temperature.

Relation to the SED bound. At no temperature does $|S(T)|$ equal the SED bound $|S_{\text{SED}}| = 2$. The degradation curve passes through 2 at $T = T_{\text{crit}}$ but does not remain there — it continues to decrease toward zero. This is because the thermal noise does not "convert" quantum detections to classical detections; it replaces them with random noise. The classical SED model (Section 8.5.2), which gives $|S| = 2$ for perfectly correlated pairs with local threshold detection, describes a zero-temperature system with local (non-Nelson) dynamics — a different physical regime entirely.

8.7.8 Experimental Discrimination from Standard Quantum Decoherence

Standard QM prediction. In quantum mechanics, thermal decoherence is described by a Lindblad master equation with rates proportional to $n_{\text{th}}$ [135]:

$$|S_{\text{QM}}(T)| = 2\sqrt{2}\exp\!\left(-\gamma_0 \tau\,n_{\text{th}}(\omega, T)\right) \tag{8.90}$$

where $\gamma_0\tau$ is a dimensionless decoherence parameter specific to the experimental implementation.

Ether prediction ((8.81)):

$$|S_{\text{ether}}(T)| = \frac{2\sqrt{2}}{(1 + 2n_{\text{th}}(\omega, T))^2} \tag{8.91}$$

The predictions are experimentally distinguishable in three ways:

(a) Asymptotic behaviour. For $T \gg T_{\text{crit}}$ ($n_{\text{th}} \gg 1$):

$$|S_{\text{ether}}| \sim \frac{\sqrt{2}}{2\,n_{\text{th}}^2} \propto T^{-2} \qquad \text{(power law)} \tag{8.92}$$ $$|S_{\text{QM}}| \sim 2\sqrt{2}\,e^{-\gamma_0\tau n_{\text{th}}} \qquad \text{(exponential)} \tag{8.93}$$

(b) Numerical comparison for 10 GHz microwaves ($T_{\text{crit}} = 0.196$ K):

$T$ (K)	$n_{\text{th}}$	$|S_{\text{ether}}|$	$|S_{\text{QM}}|$	$|S_{\text{ether}}|/|S_{\text{QM}}|$
0.010	$< 10^{-4}$	2.828	2.828	1.00
0.100	0.008	2.737	2.744	1.00
0.200	0.100	1.965	1.962	1.00
0.300	0.253	1.247	1.119	1.11
0.500	0.621	0.563	0.291	1.93
0.700	1.015	0.308	0.069	4.5
1.000	1.624	0.157	0.007	21

(The QM column uses $\gamma_0\tau = 3.66$, normalised so that $|S_{\text{QM}}(T_{\text{crit}})| = 2$.)

At $T = 1$ K (approximately $5\,T_{\text{crit}}$), the ether prediction exceeds the QM prediction by a factor of 21. This difference is easily measurable.

(c) Parameter-free ratio test. The ratio $R(T_1, T_2) = |S(T_1)|/|S(T_2)|$ at two temperatures is:

$$R_{\text{ether}} = \left(\frac{1 + 2n_{\text{th}}(T_2)}{1 + 2n_{\text{th}}(T_1)}\right)^{\!2} \tag{8.94}$$

This prediction has no free parameters: it depends only on $\omega$, $T_1$, and $T_2$, all of which are measured. The QM prediction $R_{\text{QM}} = \exp(\gamma_0\tau[n_{\text{th}}(T_2) - n_{\text{th}}(T_1)])$ depends on the implementation-specific parameter $\gamma_0\tau$.

Example. For 10 GHz at $T_1 = 0.10$ K and $T_2 = 0.50$ K:

$$R_{\text{ether}} = \left(\frac{1 + 2 \times 0.621}{1 + 2 \times 0.008}\right)^{\!2} = \left(\frac{2.241}{1.017}\right)^{\!2} = 4.86$$

Any measured value of $R$ can be compared directly against $R_{\text{ether}} = 4.86$ (no fitting). If $R_{\text{measured}}$ agrees, the ether model is supported; if it instead matches some $R_{\text{QM}}(\gamma_0\tau)$, the ether model is disfavoured.

8.7.9 Summary: Prediction 7.1 (Corrected and Complete)

Prediction 7.1. In a Bell test at frequency $\omega$ and temperature $T$:

(a) At $T = 0$, the CHSH parameter is $|S| = 2\sqrt{2}$, produced by the nonlocal Nelson osmotic coupling through the ether's ZPF (Theorem 8.5).

(b) At finite $T$, the CHSH parameter degrades as ((8.81)):

$$|S(T)| = \frac{2\sqrt{2}}{(1 + 2n_{\text{th}}(\omega, T))^2}$$

(c) Bell violation persists for $T < T_{\text{crit}}(\omega) = \hbar\omega/(2.449\,k_B)$ ((8.87)).

(d) The high-temperature decay $|S| \propto T^{-2}$ (power law) is experimentally distinguishable from standard quantum decoherence ($|S| \propto e^{-\gamma n_{\text{th}}}$, exponential).

(e) The ratio test $R(T_1, T_2) = [(1+2n_{\text{th}}(T_2))/(1+2n_{\text{th}}(T_1))]^2$ provides a parameter-free experimental discriminant.

Falsification: observation of exponential rather than power-law decay at $T > T_{\text{crit}}$; persistence of Bell violation above $T_{\text{crit}}$; or $R(T_1, T_2)$ inconsistent with (8.94).

8.8 Summary and Assessment

8.8.1 What Is Proved

1. The structural reason for Bell violation is identified (Section 8.2). In Nelson–SED mechanics, the two-particle osmotic velocity is non-separable: $\mathbf{u}_1(\mathbf{r}_1, \mathbf{r}_2)$ depends on $\mathbf{r}_2$ for any entangled state. Bell's factorisation condition fails because both particles diffuse in the same ZPF medium.

2. SED produces entangled Gaussian states (Theorem 8.3). Two electromagnetic modes coupled parametrically and driven by the ZPF reach a stationary state that violates the Duan–Simon separability criterion. The derivation uses only classical stochastic processes and the ether's ZPF spectrum.

3. Inseparability proved without quantum formalism (Proposition 8.4). The SED covariance matrix cannot be decomposed into any mixture of product Gaussian states, proved using only classical probability theory.

4. Local SED saturates the Bell bound in two distinct ways (Theorem 8.4 and Proposition 8.5). Gaussian sign-binning gives $|S| \leq 2$. The hidden-polarisation threshold model gives $|S| = 2$ with the triangle correlation. Both are local hidden variable models; Bell's theorem guarantees neither can exceed 2.

5. The ether reproduces Bell violation at $T = 0$ through a constructive mechanism (Theorem 8.5). The Nelson osmotic velocity $u_A = -D\cot(\phi_B - \phi_A)$ provides the nonlocal coupling that transforms the triangle correlation into the quantum cosine, producing the excess $\Delta S = 2(\sqrt{2}-1) \approx 0.828$.

6. No-signalling holds (Proposition 8.3). Alice's marginal outcome distribution is independent of Bob's measurement setting.

7. The ZPF has infinite correlation length (Section 8.6). Power-law decay of correlations ((8.68)), characteristic of zero-temperature long-range order.

8. The thermal degradation is rigorously derived with correct exponent (Theorem 8.8). At finite $T$, the CHSH parameter degrades as $|S(T)| = 2\sqrt{2}/(1 + 2n_{\text{th}})^2$ ((8.81)), with a critical temperature $T_{\text{crit}} = \hbar\omega/(2.449\,k_B)$ ((8.87)). The squared exponent arises from independent thermal depolarisation at each detector.

8.8.2 What Remains Open

9. Constructive integration of Nelson detection dynamics. The osmotic velocity ((8.58)) and the stationary distribution ((8.59)) are identified, but the Nelson SDE for the full photon-detector system has not been explicitly integrated. The bridge theorem guarantees the result; a constructive integration would provide independent confirmation.

10. Spin-1/2 entanglement. SED does not yet derive spin from ether microphysics (flagged in Section 7.6). The polarisation Bell test cannot be fully analysed without this.

11. Tsirelson bound. We have not derived $|S|_{\max} = 2\sqrt{2}$ from SED.

12. Multi-particle entanglement (GHZ, cluster states). The two-particle case is developed; $N$-particle extension is open.

8.8.3 Comparison with Standard Quantum Mechanics

Feature	Standard QM	Ether/SED	Status
Entanglement exists	Postulated (tensor product)	Derived from ZPF (Thm 7.3)	Proved
Non-locality mechanism	None provided	ZPF long-range order	Identified
CHSH violation (Gaussian)	No (positive Wigner fn)	No (Thm 7.4)	Agrees
CHSH violation (detection)	Yes ($2\sqrt{2}$)	Yes (Thm 7.5, Nelson mechanism)	Proved
No-signalling	Proved (partial trace)	Proved (Prop 7.3)	Agrees
Thermal prediction	Exponential decoherence	Algebraic degradation $\sim T^{-2}$	Discriminating
Bell violation + realism	Abandoned by most	Maintained (non-local realism)	Advantage
Physical substrate	None	Ether ZPF	Advantage

The ether framework does not claim to resolve Bell's theorem by restoring locality. It claims something more honest: the non-locality required by Bell's theorem has a physical carrier (the ZPF medium), a physical mechanism (zero-temperature long-range order), and quantitative predictions (Theorems 7.3, 7.5, 7.8). Standard quantum mechanics has the same non-locality but provides no mechanism and no physical substrate.

PART V: EMPIRICAL PROGRAMME