[Paper Review] Multiplicity theorem of singular Spectrum for general Anderson type Hamiltonian

Anish Mallick, Dhriti Ranjan Dolai|arXiv (Cornell University)|Sep 6, 2017

Spectral Theory in Mathematical Physics22 references3 citations

TL;DR

This paper establishes a general multiplicity bound for the singular spectrum of Anderson-type Hamiltonians with finite-rank, non-negative random perturbations. Using spectral theory and matrix-valued Herglotz functions, it proves that almost surely, the multiplicity of the singular spectrum is bounded above by the maximum algebraic multiplicity of eigenvalues of $\sqrt{C_n}(A^\omega - z)^{-1}\sqrt{C_n}$, with the bound being sharp in certain cases.

ABSTRACT

In this work, we focus on the multiplicity of singular spectrum for operators of the form $A^ω=A+\sum_{n}ω_n C_n$ on a separable Hilbert space $\mathcal{H}$, for a self-adjoint operator $A$ and a countable collection $\{C_n\}_{n}$ of non-negative finite rank operators. When $\{ω_n\}_n$ are independent real random variables with absolutely continuous distributions, we show that the multiplicity of singular spectrum is almost surely bounded above by the maximum algebraic multiplicity of eigenvalues of $\sqrt{C_n}(A^ω-z)^{-1}\sqrt{C_n}$ for all $n$ and almost all $(z,ω)$. The result is optimal in the sense that there are operators where the bound is achieved. Using this, we also provide effective bounds on multiplicity of singular spectrum for some special cases.

Motivation & Objective

• To determine the almost sure upper bound on the multiplicity of the singular spectrum for a class of random operators of the form $A^\omega = A + \sum_n \omega_n C_n$.
• To generalize prior results on spectral multiplicity, particularly those of Jakši\'c-Last and Mallick, to a broader class of finite-rank perturbations.
• To unify the treatment of point and singular continuous spectrum by avoiding distinctions between spectral types in the analysis.
• To provide effective bounds on spectral multiplicity for special cases, such as the Anderson dimer/polymer model.
• To establish that the derived bound is optimal by constructing examples where the bound is achieved.

Proposed method

• Define the random operator $A^\omega = A + \sum_n \omega_n C_n$ on a separable Hilbert space $\mathscr{H}$, where $A$ is self-adjoint and $\{C_n\}$ are finite-rank, non-negative operators.
• Use the resolvent $ (A^\omega - z)^{-1} $ and project it onto the range of each $C_n$ to define the operator $ G^\omega_{n,n}(z) = P_n (A^\omega - z)^{-1} P_n $ on $P_n\mathscr{H}$.
• Analyze the matrix-valued Herglotz function $ G^\omega_{n,n}(z) $, which encodes spectral data on the minimal $A^\omega$-invariant subspace containing $P_n\mathscr{H}$.
• Define $ \text{Mult}^\omega_n(z) $ as the maximum algebraic multiplicity of roots of $ \det(C_n G^\omega_{n,n}(z) - xI) $, equivalent to the algebraic multiplicity of $ \sqrt{C_n}(A^\omega - z)^{-1}\sqrt{C_n} $.
• Apply functional calculus and similarity transformations to relate the determinant to the operator $ \sqrt{C_n}(A^\omega - z)^{-1}\sqrt{C_n} $, preserving spectral properties.
• Prove that the singular spectrum multiplicity is almost surely bounded by the essential supremum over $n$ and $z$ of $ \text{Mult}^\omega_n(z) $, using spectral measure and Radon-Nikodym derivative arguments.

Experimental results

Research questions

• RQ1 What is the almost sure upper bound on the multiplicity of the singular spectrum for a general Anderson-type Hamiltonian with finite-rank, non-negative random perturbations?
• RQ2 Can the multiplicity of the singular spectrum be bounded uniformly across all realizations of the random parameters $\omega_n$?
• RQ3 Is the derived bound optimal, and if so, under what conditions is it achieved?
• RQ4 How does the multiplicity of the singular spectrum relate to the spectral properties of the resolvent restricted to the ranges of $C_n$?
• RQ5 Can the method be extended to cases where the perturbations are not orthogonal or have overlapping supports?

Key findings

• The multiplicity of the singular spectrum of $A^\omega$ is almost surely bounded above by the essential supremum over $n$ and $z$ of the algebraic multiplicity of $\sqrt{C_n}(A^\omega - z)^{-1}\sqrt{C_n}$.
• The bound is optimal in the sense that there exist operators for which the singular spectrum multiplicity achieves this upper bound almost surely.
• The result applies uniformly to both pure point and singular continuous spectrum, without distinguishing between them in the analysis.
• The method relies on the spectral properties of matrix-valued Herglotz functions derived from the resolvent restricted to finite-dimensional subspaces.
• The proof uses the Radon-Nikodym derivative of spectral measures and projection onto cyclic subspaces to relate the multiplicity to the eigenvalue structure of the resolvent kernel.
• The result generalizes previous findings on spectral simplicity in rank-one perturbation models and provides a framework for analyzing higher-multiplicity spectra in random operators.