Hamilton decompositions of the directed 3-torus: a return-map and odometer view

Abstract: We prove that the directed 3-torus D_3(m), or equivalently the Cartesian product of three directed m-cycles, admits a decomposition into three arc-disjoint directed Hamilton cycles for every integer m >= 3. The proof reduces Hamiltonicity to the m-step return maps on the layer section S=i+j+k=0. For odd m, five Kempe swaps of the canonical coloring produce return maps that are explicitly affine-conjugate to the standard 2-dimensional odometer. For even m, a sign-product invariant rules out Kempe-from-canonical constructions, and a different low-layer witness reduces after one further first-return map to a finite-defect clock-and-carry system. The remaining closure is a finite splice analysis, and the case m=4 is handled separately by a finite witness. A Lean 4 formalization accompanies the construction.

• The paper proves that the directed 3-torus D₃(m) can be decomposed into three arc-disjoint Hamilton cycles for all integers m ≥ 3.
• The methodology uses a reduction to return maps and models each color class as an odometer dynamic to simplify the Hamiltonicity analysis.
• The work distinguishes odd and even m cases, applying Kempe swaps and arithmetic splicing, with formal verification provided by Lean 4.

Hamilton Decompositions of the Directed 3-Torus: A Return-Map and Odometer Perspective

Graph-Theoretic Framework and Problem Statement

The paper establishes that the directed three-dimensional torus $D_3(m)$, defined as the Cayley digraph $\mathrm{Cay}((\mathbb{Z}_m)^3,\{e_1,e_2,e_3\})$ or equivalently the Cartesian product $\vec C_m\square\vec C_m\square\vec C_m$ of three directed $m$-cycles, admits a partition of its $3m^3$ arcs into three arc-disjoint directed Hamilton cycles for every integer $m\ge 3$. The result synthesizes longstanding conjectures and partial results on Hamilton decompositions of Cayley graphs and products of cycles, solidifying the status of $D_3(m)$ as a natural boundary case in directed graph decompositions.

Methodological Innovations: Return Maps and Odometer Models

The central methodological innovation is the reduction to return maps via the layer function $S = i + j + k \pmod m$. Every arc increases $S$ by $1$, so repeated application of any color map $f_c$ yields a deterministic $m$-step return to the layer-$0$ plane $P_0 = \{S = 0\}$, which has only $m^2$ vertices. The cycle structure of $f_c$ on $D_3(m)$ is thus completely determined by the induced return map (an $m$-step section) on $P_0$. This reduction is structurally analogous to a Poincaré section or first-return map, and isolates the nontrivial Hamiltonicity from ambient layer drift.

Each color class is further modelled as a dynamical system akin to a two-dimensional odometer:

$O(u, v) = (u + 1,\ v + \mathbf{1}_{u=0}),$

where $u$ is a clock coordinate and $v$ records the carry. The odometer’s orbit structure yields a single cycle on $m^2$ vertices, providing the structural target for the return maps.

Odd $m$ Case: Kempe Swap Construction and Affine Odometry

For odd $m$, the approach uses Kempe-chain recoloring. The canonical coloring partitions arcs according to the generator directions $(0,1,2)$ at every vertex, yielding $m^2$ disjoint directed $m$-cycles per color. A sequence of five Kempe swaps, supported on explicit affine substructures (lines or planes on $P_0$ and $P_1$), produces a coloring whose return maps are affine-conjugate to the odometer dynamics. This conjugacy is made explicit via closed-form formulas and affine coordinate changes. The construction is verifiably Hamiltonian: each color map is a single cycle on $V$, as proven via bijection with the odometer on $P_0$ and formalized in Lean 4.

Strong numerical results: for odd $m$, each color’s return map yields a primitive $m^2$-cycle, and the three Kempe-modified color classes are arc-disjoint Hamiltonian cycles covering all arcs.

Even $m$ Case: Parity Obstruction and Low-Layer Repair

For even $m$, the sign-product invariant of Kempe swaps imposes a parity barrier: the canonical coloring's sign product is $+1$ while any Hamilton decomposition’s sign product is $-1$. Thus a Kempe-from-canonical derivation is impossible. The construction instead uses an explicit direction assignment (Route~E) on the layers $S \in \{0,1,2\}$, tailored via affine defect families and arithmetic splicing. The induced return maps are finite-defect perturbations of odometer-type dynamics, and the cycle structure is resolved first by further reduction to lane transversals (explicit one-dimensional coordinate slices) and then by analysis of arithmetic family-blocks and splice permutations.

The Route~E mechanism repairs the clock-and-carry dynamic via local layer modifications, ensuring that each color’s return map is Hamiltonian. The proof is granular, separating the different mod-$6$ congruence classes of $m$ and classifying the arithmetic family-blocks and their splicing via explicit combinatorics. For $m=4$, a finite table-based witness is produced, corroborated by machine verification.

Detailed Numerical Results and Explicit Cycle Structures

• Hamilton cycles for all $m \geq 3$: Each color map produces a single directed cycle covering all $m^3$ vertices (verified for odd $m$ via affine conjugacy and for even $m$ via first-return reductions and finite splice permutations).
• Explicit cycle structure: Arithmetic family-block decompositions and splice graphs are constructed for all congruence classes, with cycle counts matching the expected primitive cycles (e.g., $m^2$ for return map cycles on $P_0$, $m^3$ for global cycles).
• Parity distinctions: Odd and even $m$ cases exhibit fundamentally different behavior due to sign-product obstructions, confirmed both theoretically and via Lean formalization.

Bold/contradictory claims: The paper directly proves that no sequence of Kempe swaps on the canonical coloring yields a Hamilton decomposition for even $m$ (parity barrier), necessitating a fundamentally different geometric repair mechanism.

Implications and Speculative Directions

Theoretical Implications

• Abelian Cayley digraphs: The result strengthens the position of symmetric directed tori as archetypal test cases for Hamilton decompositions, suggesting that layer-function reductions and odometer-like models will be broadly applicable in higher dimensions and more general Cayley graph settings.
• Return map dynamics: The reduction from $m^3$ to $m^2$ via layer sections provides a powerful conceptual tool, hinting at further simplification for nested product structures or other toroidal digraphs.

Practical and Computational Implications

• Formalization and verification: The proofs are fully formalized in Lean 4, and accompanied by machine-readable datasets and audit scripts, supporting computational verification of large-case Hamiltonicity and coloring permutations.

Future Developments

• Higher-dimensional tori: The paper speculates on the applicability of finite-defect odometer models in directed 4-torus and higher-dimensional cases, potentially leading to new critical-lane theorems and fully nested return map frameworks.
• Parity obstructions: The sign-product invariant hints at deeper parity phenomena in Kempe chain recoloring and Hamilton decomposition, especially as dimensions increase or generator sets diversify.
• Combinatorial block splicing: The arithmetic family-block/decomposition principle suggests new combinatorial methods for understanding splice dynamics and cycle closure in complex graph products.

Conclusion

This work resolves a longstanding decomposition problem for the directed 3-torus $D_3(m)$, establishing for all $m\ge 3$ a decomposition into three arc-disjoint directed Hamilton cycles. It does so via a conceptual reduction to odometer dynamics and explicit repair for even $m$, with rigorous formal and computational validation. The return-map methodology is likely to generalize to higher dimensions and other Cayley digraphs, and the parity-based obstruction theory promises further structural insights in graph decomposition and coloring. The formalization in Lean and transparent verification mechanisms provide a reproducible foundation for further research in algebraic and combinatorial graph theory.