[Paper Review] On the maximum nilpotent orbit intersecting a centralizer in M(n,K)

Roberta Basili|arXiv (Cornell University)|Feb 15, 2012

Advanced Algebra and Geometry10 references3 citations

TL;DR

This paper resolves a conjecture by Polona Oblak by providing a constructive algorithm to determine the maximum nilpotent orbit intersecting the centralizer of a given nilpotent matrix in $M(n,K)$. It proves that the maximal orbit type is determined by recursively decomposing the partition of the original matrix into almost rectangular components and applying a rank-based transformation, yielding a unique maximal partition that is idempotent under the map $Q$.

ABSTRACT

To any pair of commuting n x n nilpotent matrices it is associated a pair of partitions of n. We describe a maximal nilpotent subalgebra of the centralizer of a given nilpotent n x n matrix and prove a conjecture of Polona Oblak which consists in an algorithm for the determination of the maximum partition which forms with a given partition a pair with the previous property.

Motivation & Objective

• To resolve Polona Oblak's conjecture on the structure of the maximal nilpotent orbit intersecting the centralizer of a nilpotent matrix in $M(n,K)$.
• To characterize the largest possible Jordan type (partition) of nilpotent matrices within the centralizer of a given nilpotent matrix $J$.
• To establish a constructive algorithm for computing the maximum orbit $Q(B)$ associated with any partition $B$ of $n$.
• To prove that the map $Q$ is idempotent and that its image consists of partitions with parts differing by at least 2.
• To generalize results on commuting nilpotent matrices and their associated algebras using Hilbert function and Gorenstein algebra properties.

Proposed method

• Define $Q(B)$ as the maximum partition corresponding to nilpotent elements in the centralizer $ ext{Cent}(J)$ of a nilpotent matrix $J$ with Jordan type $B$.
• Decompose the partition $B$ into $r_B$ almost rectangular components $B_1, ar{B}_p$, each of which corresponds to a Jordan block of size $n_i$.
• Apply a recursive transformation: for each component, compute the sum of block sizes weighted by their multiplicity to form a new partition $\widetilde{B} = (n_1, \dots, n_{r_B})$.
• Use the rank condition $\text{rank}(A^{s_B})^m \leq \text{rank}(J^m)$ to bound the possible orbit types in the centralizer.
• Leverage irreducibility of the nilpotent cone in the centralizer and open dense subsets to ensure existence of a maximal orbit type.
• Prove that the resulting partition $Q(B)$ is the unique maximal one via induction and properties of $\mathfrak{sl}_2$-triples and grading.

Experimental results

Research questions

• RQ1 What is the maximal nilpotent orbit type that intersects the centralizer of a given nilpotent matrix in $M(n,K)$?
• RQ2 How can the maximal orbit $Q(B)$ be algorithmically computed from the partition $B$ of $n$?
• RQ3 Why is the map $Q$ idempotent, and what structural properties of $B$ ensure this?
• RQ4 Under what conditions does the centralizer of a nilpotent matrix intersect all nilpotent orbits?
• RQ5 How do the Hilbert functions of the algebras $K[A,J]$ relate to the orbit structure in the centralizer?

Key findings

• The maximum nilpotent orbit intersecting the centralizer of a nilpotent matrix with partition $B$ is given by $Q(B)$, which is computed via a recursive decomposition into almost rectangular components.
• The map $Q$ is idempotent: $Q(Q(B)) = Q(B)$, and its image consists of partitions whose parts differ by at least 2.
• For $B = (15,13,5,4,3^2,2,1)$, the maximal orbit is $Q(B) = (16,13,11,5,1)$, computed by successive reduction of the partition.
• If $B$ is almost rectangular (i.e., $ ext{max} - ext{min} \ leq 1$), then $Q(B) = (n)$, the full Jordan block.
• The centralizer $ ext{Cent}(J)$ intersects all nilpotent orbits if and only if $J^2 = 0$, provided $n > 3$.
• The algorithmic procedure for computing $Q(B)$ is based on the minimal number $r_B$ of almost rectangular components needed to decompose $B$, and the maximal rank condition in the centralizer.