A Lightlike Membrane of the Hubble Horizon: Microscopic Theory, Exchange Channel and Consistency Checks (Bilingual Edition)

The Hubble horizon is treated not as a mathematical abstraction but as a physical object — a lightlike (null) viscous membrane. It is shown that this boundary possesses its own tension μ and viscosity ζ; the interaction of matter with this viscous boundary naturally generates the observed accelerated-expansion regime, which in standard cosmology is attributed to dark energy. The dark-energy problem is thereby reformulated: instead of fitting a potential or a parametrization w(z), one computes the properties of an already existing boundary. The model predicts the scale of the exchange coefficient δ ~ 10^{-3} dictated by the microscopic viscosity of the horizon (ζ ∝ hC / ℓ_P^2); δ ≈ 0.004 follows from the normalization closure Π = δ_imp^2. An analysis of DESI DR2 (13 measurements, 7 bins, full covariance) with Planck 2018 priors shows that the exchange is not detected: δ is consistent with zero at the 1σ level, while CPL is preferred by AIC/BIC (Δχ² ≈ −9 relative to ΛCDM, versus −5.7 for the exchange model). However, it is proved that the DESI quadrant (w_0 > −1, w_a < 0) is structurally unreachable for one-directional exchange (Theorem 1): the flexibility of CPL that reaches this quadrant is not tied to the perturbation equations and yields no prediction for the growth rates. The transfer factor Π is not a free normalization: it is fixed by Π = δ_imp^2 closing the chain C → ζ → Π (Sect. 9.3). The model predicts a suppression of structure growth fσ_8 at the level ≈ 0.5% at δ = 0.004 — below the current sensitivity but within reach of DESI DR3 in combination with Euclid/LSST. The running-δ(z) extension gives a suppression of 2.45% at δ = 1/(8π) and predicts a decisive 10σ test of w_0 ≠ −1 with Euclid data. Our model offers a physical alternative to fitting w(z): it sacrifices statistical χ² in favor of physical interpretability and falsifiability.