Real positive definite counterexamples to a monotonicity conjecture for the $q$-permanent
Carlos M. da Fonseca, Dev D. Goyal
math.CO
Nov 10, 2026 · v1
TL;DR
The main counterexample theorems, small-order minimality result, gapped-family identities and bounds, and witnesses are formalized in Lean 4 with Mathlib.
Abstract
The $q$-permanent $P_q$ of an $n\times n$ matrix weights the term of each permutation $σ$ by $q^{\inv(σ)}$, where $\inv(σ)$ is the number of inversions of $σ$. It interpolates between the determinant ($q=-1$) and the permanent ($q=1$). Extending a conjecture of Bapat on $[-1,1]$, it was conjectured in 2010 that the $q$-permanent of every non-diagonal real symmetric positive definite matrix is strictly increasing on an interval $(\varepsilon,+\infty)$ with $\varepsilon<-1$. In this note, we disprove this conjecture. An explicit real symmetric positive definite integer matrix of order five, whose $q$-permanent has only nine nonzero terms, satisfies $P_{14}(A)>P_{15}(A)$. A sparse family shows that for every real $r>1$ some real symmetric positive definite matrix has a $q$-permanent with negative derivative at $q=r$. The counterexamples can be chosen to be correlation matrices without zero entries or, separately, to have integer entries. We also show that the smallest order of a counterexample is four, and that matrices supported on three coordinates separated by identity blocks fail much closer to $q=1$: in order five already near $q=1.7$, and in order $O(1/(r-1))$ at any prescribed $q=r>1$.
Problem
A 2010 conjecture extending Bapat's conjecture states that the q-permanent of every non-diagonal real symmetric positive definite matrix is strictly increasing on an interval (ε, ∞) with ε < -1. The q-permanent weights each permutation term by q raised to its number of inversions.
Approach
The authors build explicit sparse positive definite matrices whose q-permanents have few nonzero terms and compute those polynomials exactly. A parametrized sparse family uses passive identity coordinates to make a negative derivative term dominate at any prescribed q = r > 1. Exact arithmetic settles the minimal order, and 'gapped' matrices are used to get failures near q = 1. The main results are formalized in Lean 4 with Mathlib.
Results
An order-5 integer matrix satisfies P_14(A) > P_15(A). For every r > 1 some positive definite matrix has negative derivative at q = r. The smallest order of a counterexample is four, and order-five gapped matrices already fail near q ≈ 1.7.