Decomplexification of the Capelli identities and holomorphic factorization

TL;DR

This paper introduces a novel decomplexification approach to Capelli identities, revealing new identities with non-trivial correction terms, and extends classical determinant relations to non-commutative settings with applications in representation theory.

Contribution

It presents a new method of decomplexifying Capelli identities, resulting in identities with tridiagonal correction terms and extending classical determinant properties to non-commutative matrices.

Findings

Decomplexification transforms correction terms into tridiagonal matrices.

Determinant of decomplexified matrix equals the square module of the original determinant.

Provides a new perspective linking Capelli identities with non-commutative determinant theory.

Abstract

The Capelli identities claim $det(A)det(B) = det(AB+correction)$ for certain matrices with noncommutative entries. They have applications in representation theory and integrable systems. We propose new examples of these identities, constructed according to the following principle. For several known identities for $n$ by $n$ matrices we construct new identity for $2n$ by $2n$ matrices where each element $z$ of the original matrix is substituted by 2x2 matrix of the form $[real(z) ~imag(z); ~ -imag(z) ~ real(z)]$, i.e. we view the original identity as complex valued, while the new identity is its real form (decomplexification). It appears that "decomplexification" affects non-trivially the "correction term". It becomes tridiagonal matrix, in contrast to the diagonal in the classical case. The key result is an extension to the non-commutative setting of the fact that the determinant of the decomplexified matrix is equal to the square module of the determinant of the original matrix (in non-commutative setting the corrections are necessary). The decomplexified Capelli's identities are corollaries of this fact and standard Capelli identities. We also discuss analogs of the Cayley identity; observe that the Capelli determinant coincides with the Duflo image of the standard determinant; give short proof of the Cayley identity via Harish-Chandra's radial part calculation. The main motivation for us is a recent paper by An Huang (arXiv:1102.2657). From our viewpoint his result is a "dequaternionification" of 1 by 1 Capelli identity. Apparently it can be extended to $n$ by $n$ case, but our approach should be somehow modified for this. The paper aims to be accessible and interesting not only for experts. It gives brief review of Capelli identities, applications, their relation with Wick quantization, Duflo map, some open issues, etc.