Constructing an orthonormal set of eigenvectors for DFT matrix using Gramians and determinants

TL;DR

This paper refines and clarifies Matveev's method for constructing orthonormal eigenvectors of the DFT matrix, compares its computational complexity with traditional methods, and provides a Mathematica implementation.

Contribution

It offers a clearer, corrected version of Matveev's solution and evaluates its computational efficiency relative to Gram-Schmidt.

Findings

Matveev's method is computationally competitive with traditional approaches.

The paper provides a detailed, corrected derivation of the eigenvector construction.

An accessible Mathematica implementation is included.

Abstract

The problem of constructing an orthogonal set of eigenvectors for a DFT matrix is well studied. An elegant solution is mentioned by Matveev in his paper "Interwining relations between the Fourier transfom and discrete Fourier transform, the related functional identities and beyond". In this paper, we present a distilled form of his solution including some steps unexplained in his paper, along with correction of typos and errors using more consistent notation. Then we compare the computational complexity of his method with the more traditional method involving direct application of the Gram-Schmidt process. Finally, we present our implementation of Matveev's method as a Mathematica module.