Abstract
In this paper, it is shown that the $\alpha$-permanent in algebra is closely related to loop soup in probability. We give explicit expansions of $\alpha$-permanents of the block matrices obtained from matrices associated to $*$-forests, which are a special class of matrices containing tridiagonal matrices. It is proved in two ways: one is the direct combinatorial proof, and the other is the probabilistic proof via loop soup.
Introduction
The $\alpha$-permanent is a generalization of the standard matrix permanent (where $\alpha=1$) and is related to the determinant (where $\alpha=-1$). These functions are of significant interest in fields like spatial statistics and quantum physics because they appear as coefficients in the Taylor expansion of certain determinantal forms. However, they are notoriously difficult to compute and expand for general matrices.
While previous research has found expansions for simple matrix types, a general explicit expansion for more complex structures remained elusive. This paper addresses this gap for matrices associated with $*$-forests. These are matrices where the non-zero entries correspond to the edges of a graph that contains no cycles (other than self-loops). This class includes tridiagonal matrices, which are common in many physical and numerical applications.
Methodology
The authors employ two distinct frameworks to prove their main result (Theorem 1.1):
1. Combinatorial Approach
The authors use a "vertex-splitting" technique and induction on the number of vertices in the graph. By treating each vertex as having multiple "copies" (based on the block sizes in the matrix), they show that the $\alpha$-permanent can be decomposed into contributions from cycles and "bridges" within the permutation orbits. On a forest (or $*$-forest), the path-counting and cycle-counting required for the $\alpha$-permanent simplify drastically. Because there are no cycles in the underlying graph, the only way for a permutation to form a cycle is to move back and forth along the same edge or stay at a self-loop.
2. Probabilistic Approach (Loop Soups)
The authors utilize the theory of loop soups—Poisson ensembles of Markov loops. They relate the $\alpha$-permanent to the "occupation time field" (the total number of times a random walk visits a vertex or crosses an edge). By analyzing the statistics of these visits on tree-like structures using negative binomial and multinomial distributions, they derive the same expansion formula. This reinforces the idea that the $\alpha$-permanent is the natural tool for describing the law of occupation times for random loops with intensity $\alpha$.
Key Results
- Theorem 1.1: Provides a specific, closed-form formula for the $\alpha$-permanent of block matrices $A[\mathbf{q}]$ associated with $*$-forests. The formula uses Gamma functions and factorials to weight the entries of the matrix.
- Theorem 1.3: Derives simple expressions for the probability distribution of occupation times on both edges ($N$) and vertices ($\theta$) for loop soups on $*$-forests.
- Uniqueness on Forests: The authors show that if the matrix is associated with a simple forest (no self-loops), there is exactly one "non-trivial" way to assign powers to the monomials in the expansion, making the probability of a specific occupation field equal to the probability of a specific edge-crossing configuration.
Discussion and Future Work
The explicit formulas derived in this work are strictly limited to $*$-forests. The authors note that these results cannot be generalized to graphs with cycles (like a triangle graph) for $\alpha \neq 1$, as the dependencies between different paths become significantly more complex. While the connection to loop soups is robust for $\alpha > 0$, the algebraic definition of the $\alpha$-permanent holds for any $\alpha \in \mathbb{R}$, suggesting that the probabilistic interpretation is less straightforward when $\alpha$ is non-positive.
A natural next step for this research would be exploring whether "near-tree" graphs—those with a small number of cycles—allow for perturbative expansions based on the formulas provided in this paper.
Summary updated