Some new classes of almost periodic functions (Q910656)

For complex-valued f on the real line \({\mathbb{R}}\) and \(1\leq p<\infty\), \[ \| f\|_{D^ p}:=\sup_{x\in {\mathbb{R}}}\sup_{\pi}(\sum^{n}_{m=1}| (D)-\int^{x_ m}_{x_{m-1}}f(t)dt|^ p)^{1/p} \] with the sup over all partitions \(\pi =\{x=x_ 0<x_ 1<...<x_ n=x+1\}\) and (D)-\(\int\) =special Denjoy-integral (=Perron integral); \[ \| f\|_{D^{\infty}}:=\sup_{x\in {\mathbb{R}},\quad 0<h\leq 1}| (D)- \int^{x+h}_{x}fdt|; \] \(\| \|_{D^ p}\) are seminorms. A Lebesgue-measurable \(f: {\mathbb{R}}\to {\mathbb{C}}\) is called \(D^ p\)-almost- periodic \((=a.p.)\) if to each \(\epsilon >0\) there is a in \({\mathbb{R}}\) relatively dense set \(T_{\epsilon}\) with \(\| \tau_ sf- f\|_{D^ p}<\epsilon\) for \(s\in T_{\epsilon}\), and \(F\in V^ p_{loc}\); here \((\tau_ sf)(x):=f(x+s)\), \(F(x):=(D)- \int^{x}_{0}fdt\), assumed to exist for \(x\in {\mathbb{R}}\), and \(F\in V^ p_{loc}\) means to each compact interval [a,b] and \(\epsilon >0\) there is \(\delta >0\) such that \(\sum^{n}_{1}| F(x_ m)-F(x_{m- 1})|^ p<\epsilon^ p\) for all \(a=x_ 0<x_ 1<...x_ n=b\) with \(\sum^{n}_{1}| x_ m-x_{m-1}|^ p<\delta^ p.\) For these \(D^ p\)-a.p. functions linearity, \(\| \|_{D^ p}\)- closedness, the approximation theorem and a summation theorem are demonstrated (since \(D^ p\)-a.p. f have \(\| \|_{D^ p}\)- continuous \(\tau_ tf\), these results follow from the Bochner-von Neumann theory of vector-valued a.p. functions, because for continuous g: \(G\to M\) Bohr's, Bochner's and Maak's definitions of a.p. coincide, M metric space, G locally compact group, with intervals of length L replaced by \(x\bar U\), U open relatively compact). \(D^ 1\)- a.p.=Stepanoff \(S^ 1\)-a.p. since \(\| \|_{D^ 1}=\)Stepanoff- \(\| \|_{S^ 1}\); for \(p>1\) one gets however new function classes, not \(\| \|_{D^ p}\)-complete. One has a Bohl-Bohr-analogon: If f is \(D^ p\)-a.p. and \(\| F\|_{D^{\infty}}<\infty\), then F is \(V^ p\)-a.p. ([\textit{E. R. Love}; J. Lond. Math. Soc. 26, 14-25 (1951; Zbl 0042.091)], more than Bohr-a.p.). If f is \(D^ p\)-a.p., \(f'\) exists everywhere and \(\tau_ tf'\) is \(\| \|_{D^ p}\)-continuous, then \(f'\) is \(D^ p\)-a.p. (an example of Bochner's theorem for vector-valued a.p. functions). \(f\cdot g\) is \(D^ p\)-a.p. if f is \(D^ p\)-a.p. and g is Bohr-a.p. (plus additional conditions if \(p>1)\). If \(1<p<q\), f is \(D^ q\)-a.p. and \(\tau_ tf\) is \(\| \|_{D^ p}\)-continuous, then f is \(D^ p\)- a.p. (always \(\{D^ p\)-a.p.\(\}\) \(\subset \{D^ q\)-a.p.\(\})\). If \(1<p<q<r\), f is \(D^ r\)-a.p. and \(\| f\|_{D^ p}<\infty\), then f is \(D^ q\)-a.p. (but generally not \(D^ p\)-a.p.). Various examples show that these results are in general sharp.