Overview
This research paper examines the spectral characteristics of non-Hermitian random matrices that undergo multiplicative deformations. The investigation focuses on matrix products structured as $\mathbf{A}\mathbf{B}$, where $\mathbf{A}$ represents a deterministic $N\times N$ matrix, which is not necessarily Hermitian. The matrix $\mathbf{B}$ is characterized as a rotationally invariant random matrix. The primary finding pertains to the asymptotic behavior of these matrix products, specifically as the dimension $N$ approaches infinity.
Research Context
The study situates itself within the domain of random matrix theory, focusing on multiplicative deformations rather than additive ones. The specific class of matrices under consideration is non-Hermitian, which is a distinction from Hermitian matrices that often have real eigenvalues. The presence of a deterministic matrix $\mathbf{A}$ introduces a structured component to the randomness provided by $\mathbf{B}$, making the product a 'deformation' of the random matrix.
Approach
The research approach involves the analytical study of the spectral properties of the product $\mathbf{A}\mathbf{B}$. The matrices are defined as:
- $\mathbf{A}$: A deterministic $N\times N$ matrix. It is explicitly stated that $\mathbf{A}$ is not necessarily Hermitian.
- $\mathbf{B}$: A rotationally invariant random matrix.
The analysis concentrates on the behavior of these matrices as the dimension $N$ tends to infinity ($N\to\infty$). This asymptotic limit is a common technique in random matrix theory to derive general properties.
Findings
The core finding of the research is that, in the asymptotic limit ($N\to\infty$), the boundary of the complex eigenvalue distribution of the matrix product $\mathbf{A}\mathbf{B}$ is dictated by specific mathematical relationships. These relationships are expressed through equations that incorporate the $\mathcal{R}_1$ and $\mathcal{R}_2$ transforms of $\mathbf{B}$. The paper demonstrates that these transforms play a governing role in defining the outer edge of the eigenvalue spectrum in the complex plane for this class of non-Hermitian random matrices multiplied by a deterministic matrix.
Why This Matters
The explicit discussion of 'Why This Matters' or 'Potential Applications' is not present in the provided source text. The abstract focuses solely on the technical findings of the study.