[Paper Review] Spectral Asymptotics for Krein-Feller-Operators with respect to Random Recursive Cantor Measures

Lenon Minorics|arXiv (Cornell University)|Sep 21, 2017

Advanced Mathematical Modeling in Engineering3 references3 citations

TL;DR

This paper establishes the spectral asymptotics for Krein-Feller operators associated with random recursive Cantor measures, deriving the almost sure spectral dimension as the unique solution to a random moment equation. It proves that the spectral exponent for random recursive measures exceeds that of random homogeneous Cantor measures unless the measure structure is perfectly balanced, with a numerical example yielding γ_r ≈ 0.3964 for a 1/3–1/5 recursive Cantor set.

ABSTRACT

We study the limit behavior of the Dirichlet and Neumann eigenvalue counting function of generalized second order differential operators $\frac{d}{d μ} \frac{d}{d x}$, where $μ$ is a finite atomless Borel measure on some compact interval $[a,b]$. We firstly recall the results of the spectral asymptotics for these operators received so far. Afterwards, we give the spectral asymptotics for so called random recursive Cantor measures. Finally, we compare the results for random recursive and random homogeneous Cantor measures.

Motivation & Objective

• To determine the almost sure spectral asymptotics of Krein-Feller operators with respect to random recursive Cantor measures.
• To compare the spectral dimension of random recursive Cantor measures with that of random homogeneous Cantor measures.
• To establish conditions under which the spectral exponents for both types of measures coincide.
• To derive explicit formulas for the spectral exponent in terms of the random measure's geometric and probabilistic parameters.

Proposed method

• The spectral asymptotics are derived using the eigenvalue counting function and the generalized $ L^2 $ weak derivative with respect to the measure $ \mu $, defined via $ \frac{df}{d\mu} $.
• The analysis relies on the eigenvalue counting function $ N_D^\mu(x) $ and $ N_N^\mu(x) $, which count eigenvalues below $ x $, and their asymptotic behavior as $ x \to \infty $.
• The spectral exponent $ \gamma_r $ is defined as the unique solution to $ \mathbb{E}\left(\sum_{i=1}^{N_{U_\emptyset}} (r^{(U_\emptyset)}_i m^{(U_\emptyset)}_i)^{\gamma_r}\right) = 1 $, characterizing the growth rate of eigenvalues.
• The method applies Jensen’s inequality to compare $ \gamma_r $ and $ \gamma_h $, the spectral exponents for recursive and homogeneous measures, respectively.
• The eigenvalue counting function is rescaled via $ x = \log t $, linking it to a periodic function $ G(t) $, which enables the derivation of the asymptotic law.
• The comparison between recursive and homogeneous measures is based on the equality condition in Jensen’s inequality, showing $ \gamma_h \leq \gamma_r $, with equality iff all local measure structures are balanced.

Experimental results

Research questions

• RQ1 What is the almost sure spectral asymptotics for Krein-Feller operators with respect to random recursive Cantor measures?
• RQ2 How does the spectral exponent $ \gamma_r $ for random recursive Cantor measures compare to $ \gamma_h $, the spectral exponent for random homogeneous Cantor measures?
• RQ3 Under what conditions does $ \gamma_r = \gamma_h $, and what does this imply about the measure’s structure?
• RQ4 What is the explicit formula for the spectral exponent in the case of i.i.d. branching with equal scaling and mass distribution?
• RQ5 How does the spectral dimension depend on the randomness in the recursive construction of the Cantor set and measure?

Key findings

• The spectral exponent $ \gamma_r $ for random recursive Cantor measures is the unique positive solution to $ \mathbb{E}\left(\sum_{i=1}^{N_{U_\emptyset}} (r^{(U_\emptyset)}_i m^{(U_\emptyset)}_i)^{\gamma_r}\right) = 1 $, governing the asymptotic growth of eigenvalues.
• For the $ \frac{1}{3} $–$ \frac{1}{5} $-recursive Cantor set with $ p = \frac{3}{5} $, the spectral exponent is numerically $ \gamma_r \approx 0.396403 $, satisfying $ \left(\frac{1}{6}\right)^{\gamma_r} + \left\{\frac{1}{15}\right\}^{\gamma_r} = \frac{5}{6} $.
• The spectral exponent for random homogeneous Cantor measures is given by $ \gamma_h = \frac{\mathbb{E}\log N}{\mathbb{E}\log(N/r)} $, where $ N $ is the number of intervals and $ r $ the scaling factor.
• It is proven that $ \gamma_h \leq \gamma_r $, with equality if and only if $ \sum_{i=1}^{N_j} (r_i^{(j)} m_i^{(j)})^\alpha = 1 $ for all $ j \in J $ and some $ \alpha > 0 $, indicating structural balance.
• The inequality $ \gamma_h < \gamma_r $ holds almost surely unless the local measure structures are perfectly balanced across all levels of the recursive construction.
• The eigenvalue counting function satisfies $ N_D(x) \leq N_N(x) \leq N_D(x) + 2 $, ensuring that the Neumann and Dirichlet counting functions have the same asymptotic growth rate.