Integrality and Specialized Symmetric Functions

Allan Berele · 2025

Read Integrality and Specialized Symmetric Functions by Allan Berele (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-08018.txt.

Introduction

Allan Berele, a solo author, proves trace identities for matrix-like structures in a 2025 paper. Given a set of weights in a field, he defines a weighted trace that adds up scaled inputs. He shows this space satisfies two kinds of identity. In one, a high power of a variable equals a mix of lower powers. In the other, a product of traces equals a mix with fewer traces. The proofs run through specialized symmetric functions.

Catalog record

CREATORAllan Berele
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE14 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2508-08018

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Integrality and Specialized Symmetric Functions Allan Berele Abstract 1 Introduction In [ 3 ] , together with S. Catoiu, we introduced the specialized symmetric functions. These can be regarded from two points of view, either symmetric function theory or trace idenitities. Given (d)=(d_{1},\ldots,d_{k}) we defined the specialized power function as p^{(d)}_{n}(x_{1},\ldots,x_{k}) to be d_{1}x_{1}^{n}+\cdots+d_{k}x_{k}^{n} . In the special case in which the d_{i} are all positive integers, p^{(d)}_{n} can be gotten from the ordinary power symmetric functions by specializing certain groups of variables to be equal, hence the name. We will sometimes surpress the superscript (d) since we have no need to investigate the classical case corresponding to (d)=(1,\ldots,1) . There are corresponding analogues of the other classical symmetric functions: The elementary symmetric functions, the complete symmetric functions and the Schur functions, and these satisfy most of the relations satisfied by their more classical counterparts, although our focus in the current work will only be on the specialized power symmetric functions. As is standard in symmetric function theory, given a partition \lambda=(\lambda_{1},\lambda_{2},\ldots) we denote the product p_{\lambda_{1}}p_{\lambda_{2}}\cdots as p_{\lambda} . Our main interest will be in the algebraic relations satisfied by the p_{\lambda} . From the trace identity point of view, let A=F^{k} , be k copies of the field, sometimes considered as diagonal k\times k matrices. Since A is commutative, the trace functions on A are simply the linear functionals, and so every non-degenerate trace function on A is of the form tr(a_{1},\ldots,a_{k})=d_{1}a_{1}+\cdots d_{k}a_{k}, for some non-zero d_{1},\ldots,d_{k} in the field. We use the notation D{(d)} for the corresponding algebra of diagonal matrices with trace. The study of the trace identities of D{(d)} was initiated by Ioppolo, Koshlukov and La Mattina in [ 4 ] , in which they found bases for the ideal of identities when k\leq 2 . If X=(x_{1},\ldots,x_{k}) , then the trace of X^{n} is d_{1}x_{1}^{n}+\cdots d_{k}x_{k}^{n} , which is p^{(d)}_{n}(x_{1},\ldots,x_{k}) . Let \Lambda{(d)}\subset F[x_{1},\ldots,x_{k}] be the subring generated by the p^{(d)}_{n} . Again, we will surpress the (d) when we don’t feel that it is necessary. We will also write \Lambda_{n} or \Lambda_{n}{(d)} for the algebra generated by \{p_{1},\ldots,p_{n}\} . We showed in [ 3 ] that \Lambda{(d)} has transcendence degree k , so any set of more than k of the p_{n} will satisfy algebraic relations. In the classical case in which each d_{i}=1 , the Cayley-Hamilton theorem says that X is monic algebraic over \Lambda of degree n . From the point of view of trace identities, the Cayley-Hamilton theorem is a mixed trace identity (i.e., and identity involving both trace and non-trace terms) in one variable. We would like to know what the trace identities are in this more general case. In the current work we will be focusing mainly on trace identities in one variable. So, pure trace identities in one variable are are relations between the p_{\lambda} ; and an analogue of Cayley-Hamilton would be a mixed trace identity in one variable. We define absolute identities to be ones which depend on k , but not otherwise on (d) . In [ 3 ] we proved that all of the pure, absolute one variable identities are consequences of the determinental identities, |p_{a_{i}+b_{j}}|=0 , for any a_{1},\ldots,a_{k+1} and b_{1},\ldots,b_{k+1} . We also proved that X was algebraic over \Lambda , but we didn’t prove that it was monic. [ 3 ] posed six questions or conjectures. In this paper we will address the following two of them: Conjecture 1: X is integral over \Lambda . Conjecture 3: The algebra of k\times k diagonal matrices satisfies an identity of the form tr(y_{1})\cdots tr(y_{m})= a linear combination of trace terms each a product of fewer than m traces, for a certain specified m . We will prove that Conjecture 1 is true generically and that Conjecture 3 is true. By generically we mean that diagonal matrices satisfy an identity of the form \alpha_{n}X^{n}=\sum\alpha_{\lambda}p_{\lambda}X^{n-|\lambda|}, in which the \alpha are in F[d_{1},\ldots,d_{k}] . However, for certain values of the d_{i}\in F , the coefficient \alpha_{n} of X^{n} may become zero. Specifically, if \sum\{d_{i}|i\in I\} is zero for some I\subseteq\{1,\ldots,k\} , then X will not be integral over \Lambda . We now conjecture that X will be integral in all othe cases. Conjecture 3 has the consequence that any product of m or more traces can be written as a linear combination of terms with at most m-1 . Conjecture 1 has a somewhat similar application: If X is integral of degree m over \Lambda_{m} , then \Lambda_{m}=\Lambda . Because if X^{m}=\sum a_{\lambda}p_{\lambda}X^{m-|\lambda|} where a_{\lambda}\in F[d_{1},\ldots,d_{k}] , then if we multiply by X^{i} , i\geq 1 , and take the trace of both sides of the equation, we get p_{m+i}=\sum a_{\lambda}p_{\lambda}p_{m-|\lambda|+i} so p_{m+i}\in\Lambda_{m+i-1} , and by induction p_{m+i}\in\Lambda_{m} . There is also a Conjecture 2 in [ 3 ] that we have not succeeded in proving, but it seems germain to mention it. Let (d) have t parts with multiplicities m_{1},\ldots,m_{t} and let m=\prod(m_{i}+1) . Conjecture 2 says that the degree of integrality in Conjecture 1 should be m-1 and the number of traces in Conjecture 2 should be m . We gratefully acknowledge the useful conversations we had with S. Catoiu in the course of this investigation. 2 Non-Integrality The ideal of (pure or mixed) trace idenities is homgeneous. Henceforth, whenever we prove the existence or non-existence of identities we will assume homogeneity. We first show that if there is a set I\subseteq\{1,\ldots,k\} such that \sum_{i\in I}d_{i}=0 , then X cannot be integral over \Lambda . Theorem 2.1 . If \sum_{i\in I}d_{i}=0 , then X cannot be integral over \Lambda . Proof. Assume to the contrary that X^{N}+\sum a_{\lambda}p_{\lambda}X^{i}=0, where each i<N , and so each \lambda in the sum has degree N-i>0 . If we specialize x_{i}=\delta(i\in I)x , where \delta(i\in I) is 1 or 0, depending on whether i\in I is true, then each p_{j}=\sum_{m}d_{m}x_{m}^{j} becomes 0, but X^{N} does not. ∎ In the case in which each d_{i} is either plus or minus 1, the specialized symmetric functions are related to the representations of the general linear Lie algebra. If (d) has k ones and \ell minus ones then the diagonal matrices D{(d)} are a trace subalgebra of the Kemer algebra, M_{k,\ell} . The trace identities of M_{k,\ell} were found in [ 2 ] , although Kantor and Trishin had already found the one variable identities in [ 5 ] . Translating their results to the language of specialized symmetric functions, given a partition \sigma of cycle type (a_{1},\ldots,a_{k}) let T(\sigma) be p_{a_{1}}\cdots p_{a_{k}} . Theorem 2.2 . Let \lambda be the partition (k+1,\ldots,k+1) of n=(k+1)(\ell+1) . Then the group algebra FS_{n} has a two-sided ideal I_{\lambda} , and for every e\in I_{\lambda} , T(e) is an identity in \Lambda , and all identities for \Lambda are consequences of these. 3 Initialization Step Definition 3.1 . We sill say that a homogeneous polynomial f\in F[d_{1},\ldots,d_{k};\allowbreak p_{1},p_{2},\ldots] is sufficiently monic if homogeneous in total degree and if it is of the form \sum_{\lambda\vdash N}\alpha_{\lambda}p_{\lambda} , where each \alpha_{\lambda} is in F[d_{1},\ldots,d_{k}] , \alpha_{(N)}\neq 0 and \alpha_{(1^{N})}=1 . I.e., f is of the form \alpha(d_{1},\ldots,d_{k})p_{N}+\cdots+p_{1}^{N}. We will say that \Lambda{(d)} satisfies such an

Read the complete text