Dérivabilité du plus grand exposant caractéristique des produits de matrices aléatoires indépendantes à coefficients positifs. (Derivability of the greatest characteristic exponent of products of independent random matrices with positive coefficients) (Q757957)

From MaRDI portal
Revision as of 01:09, 5 March 2024 by Import240304020342 (talk | contribs) (Set profile property.)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
scientific article
Language Label Description Also known as
English
Dérivabilité du plus grand exposant caractéristique des produits de matrices aléatoires indépendantes à coefficients positifs. (Derivability of the greatest characteristic exponent of products of independent random matrices with positive coefficients)
scientific article

    Statements

    Dérivabilité du plus grand exposant caractéristique des produits de matrices aléatoires indépendantes à coefficients positifs. (Derivability of the greatest characteristic exponent of products of independent random matrices with positive coefficients) (English)
    0 references
    0 references
    1991
    0 references
    Let \(X_ 1(\lambda),...,X_ n(\lambda),..\). be an i.i.d. sequence of random matrices in SL(d,\({\mathbb{R}})\) depending on a parameter \(\lambda\). Under suitable integrability conditions the upper Lyapunov exponent \(\gamma\) (\(\lambda\)) of the product of these matrices is well defined. Then one can look to the problem of the smoothness of the function \(\lambda\hookrightarrow \gamma (\lambda)\). This question has already been studied by different authors, in particular: \(\bullet\) By Le Page when the probability distribution of the matrices has a density on SL(d,\({\mathbb{R}})\) and the mapping \(\lambda\hookrightarrow X(\lambda)\) is analytic. In this case \(\gamma\) (\(\lambda\)) is \(C^{\infty}.\) \(\bullet\) By Ruelle when the distribution of the matrices has a compact support, the matrices map a cone in its interior and the mapping \(\lambda\hookrightarrow X(\lambda)\) is analytic. In this case \(\gamma\) (\(\lambda\)) is analytic. This paper is a generalization of this last work to the case where the matrices are nonnegative, but without assuming either strict positivity of all the matrices or an analytic dependence from the parameter \(\lambda\). Hence the technics used are much more related to the general theory of products of random matrices, namely: uniqueness of the invariant measure, action of a Markov chain on spaces of Hölder continuous functions, spectral theory for positive quasicompact operators. Roughly speaking the result is the following: Let assume that the matrices are nonnegative, that the mapping \(\lambda\hookrightarrow X(\lambda)\) is \(C^{k+2}\) and that there exist positive matrices with a nonzero probability. Then \(\lambda\hookrightarrow \gamma (\lambda)\) is \(C^ k.\) One of the essential facts used in this work is the contractive action of the matrices on the positive cone \(\bar C\) of the projection space of \({\mathbb{R}}^ d\) endowed with the Hilbert distance. This yields a uniform (with respect to \(\lambda\)) spectral decomposition of the quasi-compact Markov operators on the space of Hölder functions on \(\bar C\) which in turn implies the smoothness of the resolvent operators (with respect to the parameter \(\lambda\)).
    0 references
    0 references
    positive matrices
    0 references
    perturbation theory
    0 references
    random matrices
    0 references
    Lyapunov exponent
    0 references
    invariant measure
    0 references
    quasi-compact Markov operators
    0 references

    Identifiers

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