Convolution processes of Fejér type and the divergence almost everywhere of a pointwise comparison (Q1310435)

The paper is concerned with divergence assertions in connection with a pointwise comparison of multivariate convolution processes \(\{T_ n^ \mu\}\), \(\{T_ n^ \nu\}\) of Fejér's type in spaces of continuous functions. While H. S. Shapiro established the uniform estimate \(\| T_ n^ \mu f\|\leq M\). \(\| T_ n^ \nu f\|\) under the assumption of the global divisibility of the Fourier transforms of the generating measures \(\mu\), \(\nu\), it is now shown that a pointwise analogue fails almost everwhere if the Fourier transforms of the linearly independent measures \(\mu\), \(\nu\) have the same asymptotic behaviour at the origin. The proof is based upon a lemma of A. P. Calderón and a quantitative resonance principle, dealing with the comparison of two families of sublinear functionals, depending on rather arbitrary index sets. Specific examples considered include the approximation processes of Bochner-Riesz and Gauss-Weierstrass.