Candelas, de la Ossa and van Straten conjecture that the p-adic Frobenius structure constants of an order-four Calabi–Yau operator satisfy α₁ = α₂ = 0 with α₃ a rational multiple of ζₚ(3), and Beukers and Vlasenko prove this for the simplicial and hyperoctahedral families. The equal-mass banana integrals supply a tower of open cases because at ℓ loops the Newton polytope is the A_ℓ root polytope, whose facets include products of simplices. I measure the tower. At four loops, the Hulek–Verrill family, operator 34 of the Almkvist–van Enckevort–van Straten–Zudilin list (AESZ 34), I find α₁ = α₂ = 0 and α₃ ≡ −(2/3)ζ₅(3) to twelve base-5 digits, with the same rational independently at p = 7, 11 and 13, the published quintic value reproduced through the identical code path as the control. At six, eight and ten loops, at p = 5, 7, 11, 13, every certified measurement is consistent with the single closed law U = exp( Σk≥1 −(2/(2k+1)) ζₚ(2k+1) N^(2k+1) ), uniform in the loop order and the prime. The even slots are generated by the exponential, the ζₚ(3)-slot is the same p-adic number across operators, and at p = 5 the law holds through the Kubota–Leopoldt pole branch. The rationals are derived, not fitted. The four-loop ratio is −2/3 in the normalization in which the quintic reads −40, and newly computed topological data of the six-loop fivefold gives the six-loop ratios −2/3 and −2/5 classically. Every computation is exact, every reading carries its modulus, and three falsifiers run with every measurement. The four-loop row and the tower law are verified on the stated ranges and not proved.
No takes yet. Share an insight, caveat, or question.
Eric Yaw (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: