Around the Merino--Welsh conjecture: improving Jackson's inequality
Péter Csikvári
TL;DR
The paper addresses the Merino–Welsh conjecture and its matroid generalization by establishing sharper multiplicative inequalities for the Tutte polynomial via the permutation Tutte polynomial. It develops a framework around the local basis exchange graph and Harris’ inequality to convert local bounds into global ones, yielding an improved constant x=2.355 for which T_M(x,0)T_M(0,x) ≥ T_M(1,1)^2 holds for all loopless and coloopless matroids, and further shows that matroids with prescribed circuit-length ranges (in M and M^*) satisfy the conjecture. The results substantially broaden the known cases where Merino–Welsh holds and provide a systematic, computation-backed approach that could extend to other polynomial inequalities. This approach bridges combinatorial polynomial inequalities and probabilistic inequalities on bipartite graphs, with potential implications for related conjectures and matroid classes.
Abstract
The Merino-Welsh conjecture states that for a graph $G$ without loops and bridges we have $$\max(T_G(2,0),T_G(0,2))\geq T_G(1,1).$$ Later Jackson proved that for any matroid $M$ without loop and coloop we have $$T_M(3,0)T_M(0,3)\geq T_M(1,1)^2.$$ The value $3$ in this statement was improved to $2.9242$ by Beke, Csáji, Csikvári and Pituk. In this paper, we further improve on this result by showing that $$T_M(2.355,0)T_M(0,2.355)\geq T_M(1,1)^2.$$ We also prove that the Merino--Welsh conjecture is true for matroids $M$, where all circuits of $M$ and its dual $M^*$ have length between $\ell$ and $(\ell-2)^4$ for some $\ell\geq 6$.
