Graph Puzzles III.1: A Proof of Sabidussi's Compatibility Conjecture
Sabidussi's compatibility conjecture asks the following. Take an Eulerian multigraph with minimum degree at least four and fix an Euler tour. Can the edges be partitioned into circuits so that no circuit contains two edges used consecutively in the tour?
The graph problem is reduced to colouring the gaps of a cyclic word with elements of F_2^2. The colouring must give consecutive gaps different colours, and each colour must occur an even number of times at every letter. Two parity lemmas are used: a four-colour parity criterion via a quadratic form, and a three-state balancing lemma proved by a Chevalley–Warning-style counting argument. These lemmas yield the required colouring. The result is also formalized in Lean 4.
Every such graph has an edge four-colouring in which consecutive tour edges get different colours and every colour class is even. Splitting each colour class into circuits proves the conjecture. As a corollary, a cubic graph with a dominating circuit C has a 5-cycle double cover containing C.
