Arithmetic Renormalization Dynamics for the Syracuse Map: Exact Valuation Cylinders, 3-Adic Transport, and Max-Plus Packet Certificates

This Version 4.0 presents a major extension of an arithmetic-renormalization and operator framework for the accelerated Syracuse map. The manuscript keeps three logically distinct levels of the Collatz problem separate: exact finite arithmetic, averaged transport, and worst-branch deterministic obstruction. A finite valuation word determines a unique odd residue class modulo \(2^{K_m+1}\), and the normalized Haar mass of the corresponding cylinder in the odd \(2\)-adic integers is exactly \(2^{-K_m}\). The geometric valuation law underlying probabilistic Syracuse models is therefore realized as the exact cylinder law of the deterministic valuation tree. Pushing this law through the Syracuse offset recursion produces a finite Markov transfer operator modulo \(3^h\). Along every valuation word realized by an actual positive Syracuse orbit, the finite \(3\)-adic offset coordinate agrees exactly with the current Syracuse state modulo \(3^h\) once the orbit depth exceeds \(h\). For the universal branch problem, the manuscript passes from linear transport to max-plus packet dynamics. It formulates finite Bellman certificates for exact packet paths, proves a no-go theorem for quotients covering all finite terminally bad prefixes, and introduces a normalized least-representative coordinate that separates finite \(2\)-adic compatibility from eventual positive-integer realization. The resulting analysis combines exact packet algebra, scale-adaptive copy collision, Diophantine mantissa holonomy, primitive multiplier recurrences, linear forms in logarithms, \(S\)-unit equations, Chinese-remainder realizability, affine resultants, bridge continuants, reciprocal bridge factorization, and phase-aware return constraints. On the bounded-cycle front, the previously established verified barriers remain in force. Every hypothetical nontrivial positive accelerated Syracuse cycle must satisfy\[m\geq 137{,}528{,}045{,}312\]and\[u-e_2\geq 57{,}079{,}296{,}007.\]Every candidate with length at most\[M_{71}=4{,}909{,}942{,}519{,}757{,}819{,}773{,}359\]is forced into the minimal-excess layer\[K=\left\lceil m\log_2 3\right\rceil,\]while every non-minimal candidate must have more than \(M_{71}\) odd states. The reduced-denominator local-order threshold remains\[Q_{71}=72{,}057{,}431{,}991.\] Version 4.0 substantially extends the cycle analysis beyond these numerical barriers. The manuscript introduces a universal Mersenne-normalized support quotient for every non-all-one valuation word. The complete positive-cycle criterion is compressed to positivity of the cycle denominator together with divisibility of a canonical residual support quotient by the \(3\)-free part of a Mersenne core. This criterion applies without change to arbitrary multi-level words. For two-level valuation words, the manuscript excludes every circularly balanced branch for every valuation gap. It additionally excludes every two-level word with exactly two high letters, every word with exactly two valuation-one letters, and every word with exactly three high letters. Complementary low-gap and high-run support theorems exclude two infinite circularly unbalanced subbranches, while a multi-level extremal-run theorem excludes the corresponding bottom- and top-clustered multi-level configurations. The remaining two-level branch is reduced to a cyclic bidiagonal Cramer-quotient condition together with an exact dyadic state-separation law. The phase-order analysis extracts the multiplicative phase of \(3/2\) modulo the residual Mersenne-compatible quotient. Repeated phases force full-Mersenne state separation with exact dyadic and transverse \(3\)-adic ramps; absence of a phase collision forces a canonical large-order condition. Two-sided phase-defect compression yields quantitative lower bounds for mismatch density and removes the complete normalized-excess-three branch. Spike-compressed support then excludes normalized excesses four, five, and six. A centered multi-rotation theorem eliminates every normalized-excess layer through \(2353\). Consequently every remaining primitive nontrivial cycle candidate satisfies\[E_*\geq 2354,\]where \(E_*\) is the normalized total excess. The full cyclic workload theorem independently excludes every candidate with at most fifteen non-one valuations, so every remaining candidate also satisfies\[r_*\geq 16.\]For \(r_*=16\), the certified joint bound strengthens to\[E_*\geq 17{,}808,\]with further explicit bounds through \(r_*=34\). The spike-support family is then organized as an integer cyclic module. Its bidiagonal operator has Smith normal form\[\operatorname{diag}(1,\ldots,1,\Delta),\]and the universal support quotient becomes the exact order of a canonical coefficient class in a cyclic cokernel. On a hypothetical cycle, the normalized supports form a positive odd integer orbit with an exact conjugacy to the original Syracuse states. A canonical two-three power-gap theorem excludes the exact affine-geometric workload extremizer and places the remaining workload reserve on an explicit rational lattice. Version 4.0 culminates in a stability-to-phase-resultant theorem for the quantitatively near-sharp branch. The full selector slack is decomposed into an exact positive weighted sum, while the deviations from the affine-geometric fan form a \(\gcd(K,m)\)-divisible zero-sum integer cocycle. Every hypothetical primitive nontrivial cycle therefore satisfies the improved reserve\[\Gamma_*-\Gamma_{(r_*,E_*)}(q)\geq\frac{\gcd(K,m)}{m^2S_{r_*}(q)}.\] The phase analysis gives an arithmetic dichotomy. Either distinct phase-defect classes force a multiplicity-amplified workload reserve, or two distinguished spike rotations have the same workload defect and the same phase. In the second case they determine an exact zero-fan segment, two complementary positive odd segment resultants, and a canonical odd phase resultant. The exact residual cokernel order is trapped between an exponential support-height lower bound and the canonical phase resultant:\[\frac{(2^g-1)(1-\exp(-\Lambda))}{r_*}\,2^{m\sigma}<Q\leq\gcd(M_{3'},3^d-2^d)<3^d.\]Thus the formerly qualitative near-sharp cycle problem is reduced to a precise universal order-versus-height incompatibility for one distinguished rotation pair. On the unbounded terminal front, Version 4.0 strengthens the primitive-phase event-or-scale analysis to a joint phase-scale-contraction law. It proves exact current-enhanced factorization for event-free phase returns, an exact transverse \(2\)-adic/\(3\)-adic ramp inside primitive copied blocks, and literal congruence-depth consumption. Joint closure modulo \(2^\rho\) and \(3^h\) requires at least\[h+\left\lceil\frac{\rho-1}{2}\right\rceil\]copied levels and forces an explicit exponential scale floor. The corresponding modulus-height audit shows that all available powers of two and three are already exhausted by the explicit copied-state difference. The remaining quotient is exactly a nonzero primitive unit multiplier. The manuscript therefore identifies a unit-multiplier frontier: another count of \(2\)-adic or \(3\)-adic depth cannot provide the missing strict inequality. On all but at most one cofinal thin primitive copy, the admissible collar capacity grows stretched-exponentially. A complementary dyadic trace theorem determines the exact \(2\)-adic valuation of the normalized multi-bridge return discrepancy. For a dyadic bridge count \(L=2^k\) below the initial dyadic depth, the discrepancy is divisible by \(L\) but not by \(2L\). The forbidden-residue route is therefore reduced to a literal one-extra-bit target: any independent actual-orbit relation forcing the additional factor \(2\) would contradict the exact trace. These results sharply narrow both universal fronts, but they do not complete them. On the bounded front, it remains to prove that the canonical phase-resultant collar is universally impossible: no distinguished pair may simultaneously satisfy the phase-order divisibility, both complementary segment-resultant constraints, and the support-height inequality. On the unbounded front, one still needs a cofinal one-extra-bit relation, a unit-prime or forbidden-unit-residue obstruction, an independent upper law for the unit collar capacity, or an equivalent absolute return contradiction sufficient to force nonpositive maximum Bellman cycle mean. This manuscript is an open-problem research preprint and does not claim a proof of the full Collatz conjecture.