On the rate of almost everywhere convergence of certain classical integral means (Q1293268)

Let \(m\in L^\infty(0,\infty)\) and \(m_t(u)=m(tu) \) for \( t>0, u>0\). The multiplier operator \(T_{m_t}\) generated by \(m\) is defined by \[ \widehat{T_{m_t}f}(\xi)=m_t(|\xi|)\widehat{f}(\xi), f\in L^2({\mathbb R}^n). \] The question discussed in this paper is: Under which smoothness conditions (in the \(L^p\) norm) on \(f\) the convergence \[ T_{m_t}f(x)-f(x)=o_x(w(t)) \text{ a.e. as } t\to 0^+ \] is valid with a prescribed rate \(w\)? The paper consists of two parts. In the first part the authors describe first some potential spaces to which the multiplier generator \(m\) will belong to. Then some known results of boundedness of the maximal operators \(\sup_{t>0} t^{-s}|T_{m_t}|\) (\(s\geq 0\)) and related results of a.e. convergence obtained by A.Carbery and others are stated. The second part is devoted to the investigation of a.e. convergence of certain operators \(K_t*f\) of convolution type (Theorem 2.1) and of certain multiplier operator \(T_{m_t}f\) (Theorem 2.2) for \(f\) in Hölder spaces of one single variable functions. The authors establish nearly optimal results for certain integral means including Gauss-Weierstrass, Abel-Poisson, and Bochner-Riesz means of one dimensional integrals. The sharpness of the results is announced to be discussed in a sequel to this paper [\textit{A. Kamaly, A. M. Stokolos} and \textit{W. Trebels}, J. Approximation Theory 101, No. 2, 240-264 (1999; following review)]. In the proof some techniques developed by Oskolkov are applied.