Functions with piecewise linear Fourier transforms (Q808364)

Let \(1\leq p<\infty\), let \((\alpha_ n)\) be a strictly increasing sequence of positive numbers, and let \((\nu_ n)\) be the sequence of piecewise linear functions on \((0,\infty)\) given by \(\nu_ n(x)=\max (\alpha_ n-| x|,0)\) if n is even, \(\nu_ n(x)=0\) if n is odd and \(x\geq \alpha_ n,\) and \(\nu_ n(x)=(\alpha_ n)^{-1/p}\) if n is odd and \(x<\alpha_ n.\) Let \(PL(p,\alpha)\) denote the Banach space of functions f in \(L^ p(0,\infty)\) which are piecewise linear in the sense that on \((0,\alpha_ 1)\) and upon each interval \([\alpha_{n- 1},\alpha_ n)\) the restriction of f is a polynomial of degree at most 1. It is proved that \((\nu_ n)\) is a basis for \(PL(p,\alpha)\) if and only if \((\alpha_ n)\) is a lacunary (Hadamard) sequence, and in this case the basis is equivalent to the standard basis for \(\ell^ p\). When \((\alpha_ n)\) is lacunary, this leads to a characterization of functions which have piecewise linear Fourier transforms, in the following sense: if \(D_ n\) and \(F_ n\) denote the Dirichlet and Fejér kernels on \((-\pi,\pi),\) respectively, then \(\sum^{\infty}_{n=1}\alpha_ n^{-1/2}(c_ nD_{\alpha_ n}-d_ nF_{\alpha_ n})\) converges if and only if \(\sum^{\infty}_{n=1}| c_ n|^ 2\) and \(\sum^{\infty}_{n=1}| d_ n|^ 2\) converge, and in this case, the functions which are the sums of such series are precisely those whose Fourier transforms are in \(PL(2,\alpha).\) This result is a type of analogue for series of Dirichlet and Fejér kernels of the classical result in Fourier analysis that if the Fourier transform of a continuous function on \((-\pi,\pi)\) is non-zero only upon a lacunary set of integers, then the Fourier series of the function converges uniformly to the function.