Bounds on the spectral shift function and the density of states (Q852380)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Bounds on the spectral shift function and the density of states
scientific article

    Statements

    Bounds on the spectral shift function and the density of states (English)
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    29 November 2006
    0 references
    The authors study spectral properties of Schrödinger operators of the form \(H=H_A+V=(-i\nabla -A)^2+V\) acting on \(L^2(\mathbb R^d)\), where the magnetic potential \(A\) belongs to \(L^2_{\text{loc}}\), the positive part of the electric potential \(V_+\) is element of \(L^1_{\text{loc}}\) and the negative part \(V_-\) belongs to the Kato class. Let \(H_1\) be a Schrödinger operator of the above type and let \(H_2=H_1 + u\), where the electric potential \(u\) obeys the same assumptions as above and is compactly supported. The authors show that the singular values \((\mu_n)_{n \in \mathbb N}\) of the corresponding semigroup difference \(V_{\text{eff}}=e^{-H_1}-e^{-H_2}\) satisfy \(\mu_n \leq C e^{-c n^{1/d}}\). Here, the constant \(C\) depends on the Kato class norms of \(V_-\) and \(u_-\) and on the diameter of the support of \(u\), whereas the constant \(c\) depends only on the dimension. The same result holds for \(V_{\text{eff}}^l=e^{-H_1^l}-e^{-H_2^l}\), where \(H_k^l\), \(k=1,2\), is the Dirichlet restriction of \(H_k\) to the cube \(\Lambda_l=[-l/2,l/2]^d\). The proof of both results uses lower bounds on the eigenvalues of \(H_k^l\) and the Feynman--Kac--Itô formula. The decay of the singular values is used to proof integral bounds on the spectral shift function \(\xi\) of the pairs \(H_1, H_2\) and \(H_1^l,H_2^l\), respectively. The authors show that \[ \int\limits_{-\infty}^T \int\limits_0^{|\xi(\lambda)|} \left( \exp( t y^{1/d})-1 \right) dy\,d\lambda \leq K_1(t) e^T \] and \[ \int f(\lambda) \xi(\lambda) \, d\lambda \leq K_1 e^b + K_2 \left\{ \log \left( 1 + \| f\|_\infty \right)\right\}^d \| f\|_1, \] where \(b=\sup\text{supp}(f)\) and the constants \(K_1\) and \(K_2\) depend only on \(d\), the diameter of \(\text{supp}(u_+)\), and on the Kato class norms of \(V_-\) and \(u_-\). A random Schrödinger operator \(H_\omega= H_0 +V_\omega\) is called an alloy type model if \(H_0= H_A +V_{\text{per}}\), with a periodic potential \(V_{\text{per}}\) and \(V_\omega(x)= \sum_{k\in\mathbb Z^d}\omega_k u(x-k)\), where the coupling constants \(\omega_k\), \(k \in \mathbb Z^d,\) form a sequence of bounded random variables which are independent and identically distributed with distribution \(\mu\). The single site potential \(u \neq 0\) has compact support. Denote \[ s(\mu,\varepsilon)=\sup\{\mu([E-\varepsilon,E+\varepsilon]):E\in\mathbb R\}, \quad \varepsilon>0. \] Supposing that \(u \geq \kappa \cdot \chi_{\Lambda_1}\), \(\kappa>0\), it is shown that for each \(E_0 \in \mathbb R\) there exists a constant \(C_W\) such that for all \(E \leq E_0\), \(\varepsilon \leq 1/2\), \[ \mathbb E\{ \text{Tr}[ \chi_{[E-\varepsilon,E+\varepsilon]}(H_\omega^l)]\} \leq C_W s(\mu,\varepsilon) \left(\log \frac{1}{\varepsilon}\right)^d |\Lambda_l|. \] Here, \(H_\omega^l\) is the Dirichlet restriction of the alloy type model \(H_\omega\) to \(\Lambda_l\) and \(\mathbb E\) denotes expectation with respect to the product measure \(\bigotimes_{k \in \mathbb Z^d}\mu\). This result is a refinement of a result in [\textit{P.\,Stollmann}, Arch.\ Math.\ 75, No.\,4, 307--311 (2000; Zbl 1068.82523)]. An estimate of the above type is called a Wegner estimate and it is used to show a continuity result for the integrated density of states, which is defined as the limit of the distribution functions \[ N_\omega^l(E)=|\Lambda_l|^{-1}\#\{\text{eigenvalues of } H_\omega^l\text{ not greater than }E\} \] as \(l\) tends to infinity. This limit is almost surely independent of \(\omega\) and it follows that \[ |N(E_1)-N(E_2)| \leq C_I s(\mu,|E_1-E_2|)\left(\log\frac 1{|E_1-E_2|}\right)^d, \] where the constant \(C_I\) may be chosen uniformly for \(E_1\) and \(E_2\) varying in a compact interval \(I\) such that \(|E_1-E_2| \leq 1/2\).
    0 references
    Schrödinger operators
    0 references
    Kato-class norms
    0 references
    alloy type model
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references

    Identifiers

    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references