Exponential inequalities for a class of operators (Q1607945)

From MaRDI portal





scientific article; zbMATH DE number 1780423
Language Label Description Also known as
default for all languages
No label defined
    English
    Exponential inequalities for a class of operators
    scientific article; zbMATH DE number 1780423

      Statements

      Exponential inequalities for a class of operators (English)
      0 references
      0 references
      13 August 2002
      0 references
      The paper is a very interesting contribution to the study of exponential-logarithmic type integral operators \[ E_kf(x):=\exp(K\log f)(x), \] where \(K\) is the positive integral operator, given by \[ (Kf)(x)=\int_0^\infty k(x,y)f(y) dy, \] where \(f\) is a non-negative function and the kernel \(k\) is assumed to meet a few quite natural restrictions. Specific examples of such operators have been studied in various situations by many authors, in particular, for example, when \(K\) is the Hardy averaging operator or some of its modifications. It is known (cf., for instance, [\textit{H. P. Heinig, R. Kerman} and \textit{M. Krbec} [Georgian Math. J. 8, 69-86 (2001; Zbl 1009.26014)] and the references therein) that the condition \[ \sup_{y>0}\int_{0}^{\infty}\left[\frac{y}{x}k(x,y)\right]w(x) dx<\infty \leqno {(1)} \] is sufficient for the weighted Knopp-type inequality \[ \int_{0}^{\infty}u(x)(E_kf)(x) dx\leq C \int_{0}^{\infty}v(x)f(x) dx,\leqno {(2)} \] here \(u,v\) are weights and \(w(x)=u(x)\exp (K\log(1/v))(x)\). In specific situations (usually involving operators of Hardy-type), it is also known that the converse implication holds, that is, (1) is also necessary for (2). In the relevant papers, however, it was assumed, sometimes somewhat implicitly, that \(E_k f\) and/or \(E_u(1/v)\) exist almost everywhere and are Lebesgue measurable. The author demonstrates with examples that this is not necessarily the case and gives necessary and sufficient conditions of \(f\) (and \(v\)) for this to be satisfied. He also shows that (1) is equivalent to (2) in fairly general context (the main innovation being, of course, the `only if' part, i.e., the necessity of (2) for (1)). These two main results are then illustrated with non-trivial examples involving the power-weight Hardy averaging operator \[ (P_\beta f)(x):=\beta x^{-\beta}\int_{0}^{x}t^{\beta-1}f(t) dt,\qquad \beta>0, \] the Riemann-Liouville operator \[ (I_\alpha f)(x):=\frac{\alpha}{x^\alpha}\int_{0}^{x}(x-y)^{\alpha-1}f(y) dy,\qquad \alpha\geq 1,\;x>0, \] and the generalized Laplace transform \[ (L_a f)(x):=\frac{1}{x\Gamma(1+1/a)} \int_{0}^{\infty}e^{-(y/x)^a}f(y) dy, \qquad a\geq 1, x\geq 0. . \]
      0 references
      0 references
      exponential inequalities
      0 references
      logarithmic inequalities
      0 references
      weighted inequalities
      0 references
      positive integral operators
      0 references
      weighted Knopp-type inequality
      0 references
      power-weight Hardy averaging operator
      0 references
      Riemann-Liouville operator
      0 references
      generalized Laplace transform
      0 references

      Identifiers