Lovász Hamiltonian path conjecture

Still openEvidence checked

Subjects: Graph theory

Collections: Graphs and coloring

Reading depth

The problem

Imagine a network in which every junction looks the same under a symmetry of the entire network. Can you walk along its links and visit every junction once, without revisiting one? The route may end somewhere different from where it began.[1] [2]

Scope

Hamiltonian paths in all finite connected undirected simple vertex-transitive graphs; no universal Hamiltonian-cycle assertion.

What remains open

Symmetry gives remarkably long routes. The missing step is reaching every last junction. Demanding a round trip is stronger: even the symmetric Petersen graph has no round trip visiting every vertex once.[1] [2]

Why it matters

Symmetry can supply a global organization even when each junction has only a few links. One important example makes group elements into junctions and the allowed operations into links. A route visiting each junction once would order all those elements using only the allowed operations, without repetitions. That connects a local rule for moving between states with an exact listing of the entire state space.[1] [2]

Research directions

1 direction to explore

Definitions

2 terms
vertex-transitive

For any two vertices, a symmetry of the entire graph moves one to the other while preserving all adjacencies. Equal numbers of neighbors alone do not guarantee this global symmetry.[1] [2]

Cayley graph

A graph of a finite group: vertices are group elements, and edges join elements differing by an allowed generator, with inverses also allowed. It is vertex-transitive, but the conjecture covers more than this subclass.[1]

Sources

Partial results
Bowen Li and Abhishek Methuku

Universal cycle lengths reach n^(1−o(1)). Scope: All sufficiently large connected vertex-transitive simple graphs.

Not a spanning path and not an almost-one fraction without further control.

Source date: 2026-09-29Theorem 1.2 gives n^(1−ε) for each fixed positive ε.Read the source ↗
Micha Christoph, Zach Hunter and Benny Sudakov

Almost spanning cycles under polylogarithmic degree. Scope: d=ω(log³n loglogn).

Growing degree is required; o(n) omissions need not vanish.

Source date: 2026-09-24Theorem 1.8 gives (1−o(1))n vertices.Read the source ↗
  1. A nearly linear bound for the Lovász conjecture Bowen Li and Abhishek Methuku · 2026-09-29
    Notes on A nearly linear bound for the Lovász conjecture

    States the path conjecture and universal nearly linear cycle bound; a near-one exponent is not a cycle containing a near-one fraction of vertices.

  2. Thinning and sprinkling: from robust sampling to almost Hamiltonicity Micha Christoph, Zach Hunter and Benny Sudakov · 2026-09-24
    Notes on Thinning and sprinkling: from robust sampling to almost Hamiltonicity

    Theorem1.7 requires t/log³n→∞. Theorem1.8 requires d/(log³n loglogn)→∞; neither theorem gives a spanning cycle at constant toughness or all degrees.

Suggest a correction