A Non-Phantom Alternative to Phantom Dark Energy in DESI DR2

The DESI DR2 baryon acoustic oscillation measurements, combined with Planck CMB priors, prefer w0 waCDM over LambdaCDM with the best-fit point in the phantom regime (w < -1)—a trajectory with no healthy field-theoretic realization. We show that a one-parameter Interacting Dark Energy (IDE) model with coupling Gamma(z) = 3A * H(z) * Omega_m(z) * Omega_DE(z) accommodates the DESI signal without phantom crossing, and is BIC-preferred over four-parameter w0 waCDM in every one of four dataset combinations tested: BAO+CMB, BAO+CMB+Pantheon+, BAO+CMB+RSD, and the full BAO+CMB+Pantheon+RSD joint fit. The total Delta BIC vs CPL runs from +1.45 on BAO+CMB to +5.74 on the full joint fit. The interpretation is a parsimony statement, not a fit-quality one: in three of four combinations CPL achieves slightly better raw chi-squared, but its three extra parameters fail the canonical BIC criterion that an extension must reduce chi-squared by more than ln(N). The fitted coupling amplitude |A| is stable at 0.021 to 0.026 across all four combinations—a ~25% spread well inside the profile-likelihood uncertainty of any single fit. The most physically informative single result is the BAO+CMB+RSD combination, where IDE is preferred over LambdaCDM at 2.52 sigma from directly integrated coupled growth, with the same coupling amplitude appearing in a structurally separate physical sector from the BAO expansion-history measurements that produced the original DESI anomaly. In Pantheon+-inclusive combinations, both extensions are BIC-disfavored relative to LambdaCDM due to the well-known DESI-vs-Pantheon+ data tension; the BIC preference for IDE over CPL between the two extensions is robust across all four combinations. Coupled linear perturbation theory yields S8 approximately 0.799, directionally closing roughly half of the approximate Planck-weak-lensing tension; the weak-lensing values themselves assume LambdaCDM, so this comparison is pending native IDE reanalysis of the shear data. The substrate framework that motivates the coupling form predicts |A| of order ~10^-1.5; the fitted value is consistent with this expectation but does not test any specific information bound. The model-comparison result can be read as pure phenomenology, independent of the substrate motivation.