Write a PREreview

We prove strict log-concavity of the Riemann-Jacobi kernel, establish hyperbolicity of the Jensen polynomials J_{d,n}(x) for d <= 22, n <= 14, and prove that the global Riemann Hypothesis is algebraically equivalent to a subluminal condition on the Wronskian components. Part I (Sections 2-5) proves the kernel is strictly log-concave (TP_2) with curvature kappa >= 19.24, via the convex potential decomposition and a perturbation bound using only 4.3% of the log-concavity budget. Part II (Sections 6-8) establishes K_{d,n}(x) < 0 for x >= 0 (all d, n) purely analytically, and for x < 0, d <= 22, n <= 14 by interval-arithmetic certification (330/330 cases, Bernstein-basis enclosure with double-double precision on GPU). This extends the Griffin-Ono-Rolen-Zagier result (d <= 8) to d <= 22 with full coverage of all real x. Part III (Section 9) introduces the even-odd decomposition: setting t = y^2, the condition K_{d,n}(-y) < 0 is equivalent to the Lorentz norm P(t) = A(t)^2 - t B(t)^2 > 0, where A and B are the even- and odd-indexed coefficient polynomials. Global RH (unconditional). Section 11 proves D_r(n) > 0 for all r and n. The argument combines a discrete concavity lemma with a spectral-gap reduction: the Hadamard factorisation of Xi gives a spectral gap delta = (t_1/t_2)^2 ≈ 0.452 between the first two zeros, which makes the normalised dissipation C_s = n^2|log Theta_s| independent of n. The two-variable unitarity condition reduces to a single-variable bound S = sum C_s < a*n, verified by certified computation (S <= 19.41, a >= 1.31). The proof combines: (A) Borell log-concavity (L_1 > 1, all n); (B) 10,822-point interval certification; (C1) the dissipation bound for n >= 100; (C2) DJ log-space certification plus dominant-pole tail for n <= 99. By Edrei-Schoenberg, Xi is in the Laguerre-Polya class and Lambda = 0.